Mutations in the Yeast MRF1 Gene Encoding Mitochondrial Release Factor Inhibit Translation on Mitochondrial Ribosomes*

Although the control of mitochondrial translation in the yeast Saccharomyces cerevisiae has been studied ex-tensively, the mechanism of termination remains ob-scure. Ten mutations isolated in a genetic screen for read-through of premature stop codons in mitochondrial genes were localized in the chromosomal gene encoding the mitochondrial release factor mRF1. The mrf1–13 and mrf1–780 mutant genes, in contrast to other alleles, caused a non-respiratory phenotype that correlated with decreased expression of mitochondrial genes as well as a reporter ARG8 m gene inserted into mitochondrial DNA. The steady-state levels of several mitochondrially encoded proteins, but not their mRNAs, were dramatically decreased in mrf1–13 and mrf1–780 cells. Structural models of mRF1 were constructed, allowing localization of residues substituted in the mrf1 mutants and offering an insight into the possible mechanism by which these mutations change the mitochondrial translation termination fidelity. Inhibition of mitochondrial translation in mrf1–13 and mrf1–780 correlated with the three-dimensional localization of the mutated residues close to the PST motif presumably involved in the recognition of stop codons in mitochondrial mRNA. The mitochondrial genetic system of the yeast Saccharomyces MB143-3D/13 Ura Gly integrant the MRF1 mrf1–13 allele. Gly (cid:3) isogenic allele R780 pRS306. mrf1 structure alignment, was based on the results of secondary structure prediction and tertiary-fold recognition carried out with the Meta Server (24) using the consensus alignment approach and three-dimen-sional assessment (25). In addition, a C (cid:3) trace model of mRF1 bound to the ribosome was derived using the low resolution structure of E. coli RF2 in a complex with the ribosome (26).

The mitochondrial genetic system of the yeast Saccharomyces cerevisiae has been a subject of intensive studies. Seven genes in the yeast mitochondrial DNA (mtDNA) 1 encode integral membrane proteins that are assembled together with subunits encoded in the nuclear DNA into multimeric respiratory complexes. Maintenance and expression of mtDNA require approximately 200 proteins encoded in the nuclear DNA and imported to mitochondria from the cytosol (1,2). Among them are mitochondrial ribosomal proteins, general translation factors, and translational activators. Current analysis of the proteome of yeast mitochondria allowed for identification of 65 mitoribosomal proteins (2); however, only ϳ60% of these have recognizable homology with bacterial ribosomal proteins (3). Multiple mitoribosome-specific proteins in yeast and mammals demonstrate the considerable divergence of mitochondrial ri-bosomes from their prokaryotic ancestors (4,5). The mechanisms that control mitochondrial translation are difficult to study because no in vitro system for the expression of translation products encoded in mtDNA has yet been established. In contrast to the biochemical difficulties, the yeast S. cerevisiae, being a facultative anaerobe, offers distinct advantages for mitochondrial genetic studies. Several hundred point mutations or small deletions called mit Ϫ have been genetically mapped and assigned to genes in mtDNA. We have long been involved in a search for suppressors of mit Ϫ mutants that act at the level of mitochondrial translation. Analogous suppressor approach applied to chromosomal mutants allowed identification of proteins responsible for translation fidelity in the yeast cytosolic system. Ribosomal proteins from the decoding center of the ribosome and translation termination factors represented two classes of proteins that control decoding of the stop signal (6,7).
An ochre mit Ϫ mutation cox2-V25, localized in the coding region of the mitochondrial COX2 gene, was used for suppressor screen. One dominant suppressor, NAM9 -1 (8), and 10 recessive suppressors (9) were identified. The cloning of the NAM9 -1 gene identified a structural component of the mitochondrial ribosome homologous to the S4 ribosomal protein from bacteria and eukaryotes that controls translation fidelity (8,10). Here we show that all of the remaining recessive nuclear suppressors are mutations in a gene specifying the mitochondrial translation termination factor mRF1 (11). mRF1 is the only release factor that recognizes stop codons in mitochondrial mRNAs (11,12). In eukaryotic cytosol, also a single factor, eRF1, recognizes all three stop codons. In contrast, prokaryotes have two codon-specific factors, RF1 for UAA/UAG and RF2 for UAA/UGA. Eukaryotic eRF1s and prokaryotic RFs show no detectable similarity at the level of amino acid sequences, but their shapes mimic tRNA as seen from their crystal structures (13,14). Functional mimicry of release factors and tRNA is supported by biochemical and genetic studies that point to tripeptide "anticodons" in bacterial RF1 and RF2 (15,16).
The yeast mrf1 mutants described here are, to our knowledge, the first identified informational suppressors of mit Ϫ mutations that act by changing mitochondrial translation termination fidelity. We constructed mRF1 structural models that allowed localization of the amino acids substituted in the mutants and offered a clue to the possible mechanism of suppression. Interestingly, molecular analysis showed that some mrf1 mutations block translation of mitochondrially encoded proteins. 2% glycerol (YPGly). Minimal medium SC (0.67% yeast nitrogen base without amino acids) was supplemented with 2% glucose, amino acids, uracil, and adenine as required (17). All of the reagents were from Difco. For determination of the respiratory (Gly) phenotype, yeast strains were grown on YPD plates, replica-plated on YPGly, and incubated for 3 days at 30°C. To estimate accumulation of [rho Ϫ ] mutations, initially [rho ϩ ] strains were plated for single colonies on YPD and replicacrossed with a [rho 0 ] tester followed by scoring of diploids on YPGly. For in vivo labeling and isolation of mitochondria, yeast were grown at 30°C in liquid YPGly containing 0.3% glucose and harvested right after the point of glucose exhaustion as determined with Glucostix (Bayer).
Yeast strains, Plasmids, and DNA Manipulations-S. cerevisiae strains used in this study are listed in Table I. Cytoduction, generation of [rho 0 ] derivatives, and standard genetic manipulations were according to published procedures (17). As reported before, R13 and R780 had the Gly Ϫ phenotype with wild type mtDNA but leaky growth on YPGly with the mit Ϫ mutant mtDNA (9). Genetic analysis was done to confirm the phenotype of R13, referred to here as the mrf1-13 mutant. 10 full tetrads from the cross-R13 [rho 0 ] ϫ W303-1B were analyzed. Clear 2:2 segregation of the mrf1-13 phenotype (Gly Ϫ ) was observed. All of the ascospores were made [rho 0 ] and then crossed back with the target mit Ϫ mutation cox2-V25 (strains CD111 and CD112). Only two clones per tetrad, those identified previously to carry mrf1-13, yielded Gly ϩ/Ϫ because of suppression of cox2-V25.
A yeast genomic library in a pRS315 vector (18) was kindly supplied by M. Wysocka (Institute of Biochemistry and Biophysics, Polish Academy of Sciences). Yeast cells were transformed by the lithium acetate method (19). MRF1 was subcloned as the SacI-SalI fragment in pRS306, pRS314, and pRS315, resulting in pRS306-MRF1, pRS314-MRF1, and pRS315-MRF1, respectively. For chromosomal integration in the MRF1 locus, pRS306-MRF1 was cut with EcoRI before transformation.
The isogenic wild type MRF1 strain was generated by two-step gene replacement (20). pRS306-MRF1 was integrated in the MB143-3D/13 genome, resulting in the Ura ϩ Gly ϩ phenotype. Homologous recombination, forced by growth of the integrant on SC medium supplemented with 5-fluoroorotic acid (2 mg/ml), resulted in excision of the URA3 plasmid and loss of either the MRF1 or mrf1-13 allele. The Gly ϩ clone containing MRF1 is referred to as MB143-3D. An analogous approach was subsequently applied to generate the isogenic mrf1-780 mutant (MB143-3D/780) using the PCR-amplified mrf1-780 allele of R780 subcloned in pRS306.
In Vivo Labeling of Mitochondrial Translation Products-Cells grown as described above were suspended in buffer P (40 mM KPi, pH 7.0, 0.45% glucose, 0.0077% Complete supplement mixture-methionine (BIO101)) and incubated on a shaker for 30 min at 30°C. Cycloheximide was added to a final concentration of 100 g/ml. After 10 min at 30°C, the reactions were supplemented with 40 Ci of [ 35 S]methionine (1000 Ci/mM, ICN). At time points indicated, the labeling reaction was stopped by rapid chilling of the cells on ice and simultaneous addition of unlabeled methionine to a final concentration of 20 mM. After centrifugation, the cells were washed in CM buffer (40 mM KPi, pH 7.0, 100 g/ml cycloheximide, 20 mM methionine) and collected. To measure [ 35 S]methionine incorporation, cells were filtered under vacuum on GF/A glass microfiber filters (Whatman), washed with 10% trichloroacetic acid, 95% ethanol, and acetone, and dried. Radioactivity was measured in a scintillation counter. To determine mitochondrial trans-lation products, total protein was extracted from collected cells as described by Yaffe and Schatz (21). Mitochondrial translation products were analyzed by 12% SDS-PAGE followed by autoradiography.
Analysis of Mitochondrial Protein and RNA-Yeast were grown as described above. Mitochondria were purified according to Glick and Pon (22). The Nycodenz purification step was omitted. Protein concentration was estimated with a Bio-Rad protein assay using bovine serum albumin as standard.
Mitochondrial proteins were separated on SDS-PAGE, transferred to Hybond C-Extra (Amersham Biosciences), and hybridized with antibody according to instructions. The monoclonal antibodies against Cox2 and Cox3 were from Molecular Probes. All of the other antibodies were from the collection of G. Schatz (Biozentrum, Basel, Switzerland). Secondary antibody coupled to horseradish peroxidase (DAKO) was visualized by chemiluminescence using the ECL detection kit (Amersham Biosciences) followed by autoradiography. Films were quantified using ImageQuant, version 1.1 with local average background correction (Molecular Dynamics, Sunnyvale, CA).
RNA was isolated from mitochondria using the TRIzol reagent (Invitrogen). 10 g of each RNA sample was resolved by electrophoresis in 1% agarose gel containing 0.925% formaldehyde in NBC buffer (0.5 M boric acid, 10 mM sodium citrate, 50 mM NaOH). The RNAs were transferred from gel to a Hybond N ϩ membrane (Amersham Biosciences) and hybridized with probes according to instructions. 5Ј-TATAAGCCCACCGCAGGTTCCCCTACGGTAACTGTA-3Ј and 5Ј-GT-TCATTTAATCATTCCAAAAATTTAGGTAATGATACTGCTTCGATCT-TA-3Ј oligonucleotides labeled with [␥-32 P]ATP and T4 polynucleotide kinase were used as probes for 16 S rRNA and COX2 mRNA, respectively. PCR fragment synthesized with 5Ј-TATGCCTTCACCATGAC-CTA-3Ј and 5Ј-CATCAGTAGAAGACTACGTA-3Ј primers and labeled with [␣-32 P]ATP using random-prime system served as a probe for COX3 mRNA. After hybridization, blots were washed 15 min in 6ϫ SSC at room temperature and exposed to film or to a PhosphorImager plate. RNA was quantified using the laser densitometer GelScan XL (Pharmacia). COX2 and COX3 mRNA levels, measured for two independent blots, were corrected for 16 S rRNA levels.
Cytochrome Spectra-Cells were grown on YPGal plates. Cytochrome absorption spectra were recorded for whole cells in liquid nitrogen as described previously (23).
Three-dimensional Models of mRF1-A three-dimensional model of S. cerevisiae mRF1 protein in the conformation unbound to the ribosome was generated with the MODELER program (24) using the Escherichia coli RF2 (1GQE) crystal structure (14) as template. The Nterminal mitochondrial targeting sequence was not considered in the model. The critical step in the modeling, generation of the sequence-tostructure alignment, was based on the results of secondary structure prediction and tertiary-fold recognition carried out with the Meta Server (24) using the consensus alignment approach and three-dimensional assessment (25). In addition, a C␣ trace model of mRF1 bound to the ribosome was derived using the low resolution structure of E. coli RF2 in a complex with the ribosome (26).

Identification of mit Ϫ Suppressors as Mutations in the MRF1 Gene Specifying the Mitochondrial Translation
Termination Factor-Several recessive suppressors of the mit Ϫ nonsense mutation cox2-V25 were isolated previously and characterized genetically (9) but not identified up to now. Specificity studies involving a collection of several hundred point mutations or small deletions in mtDNA revealed that the suppressors act on Ͻ20 mutations localized in different mitochondrial genes. The action spectra and growth phenotypes of the suppressors were similar but not identical, indicating that they are not identical mutations. 2 of 10 suppressors were temperaturesensitive, and these two mutants (designated previously as R13 and R780) did not grow on a medium containing non-fermentable carbon when combined with wild type mtDNA (9). To clone the gene able to complement the Gly Ϫ phenotype of the R13 mutant, the wild type yeast single copy genomic library was introduced into MB143-3D/13 cells (see Table I and "Experimental Procedures"). Screening of 12,000 Leu ϩ transformants for complementation of the respiratory defect resulted in two clones that showed a plasmid-dependent phenotype. Further analysis revealed two overlapping plasmids that have the MRF1 gene in common. MRF1 is known to encode the mRF1 release factor involved in decoding of stop signals during termination of protein synthesis on mitochondrial ribosomes (11).
To confirm that R13 and possibly other recessive suppressors are alleles of MRF1, we first mapped these mutants by genetic recombination. The wild type strain W301-1B with the URA3 marker integrated into the chromosomal MRF1 locus was crossed with R13 and several other suppressors, including the CK311/B145 strain carrying the nam3-1 mutation (27). Diploids were sporulated, and Ͼ10 tetrads from each cross were analyzed for segregation of Ura ϩ and suppressor phenotypes. In none of the crosses did these traits co-segregate, providing strong evidence that the examined suppressor mutations are alleles of MRF1.

Localization of Suppressor Mutations in mRF1 Domains-
The strong sequence conservation between the release factors from prokaryotes and mitochondria and the known crystal structure of E. coli RF2 (14) allowed us to build a reliable three-dimensional model of the S. cerevisiae mitochondrial mRF1 protein (Fig. 1A). Moreover, on the basis of a low resolution structure of the E. coli RF2 termination complex (22), a C␣ trace model for mRF1 in a complex with the ribosome was also derived (data not shown). mRF1 is composed of four domains (Fig. 1), largely corresponding to those predicted by Pel (11). The N-terminal helical domain 1, shown in brown, is apart from the compact structure formed by the remaining domains. The highly conserved domain 2 (blue) is supposed to be responsible for stop codon recognition. Domain 4 (magenta) is considered as a subdomain of domain 2. Domain 3 (green) contains the absolutely conserved GGQ motif responsible for hydrolysis of the peptidyl-tRNA bond (9). The highly conserved Arg-313 in domain 3 is located close to the GGQ loop and possibly interacts directly with the tRNA molecule. Mutating this charged residue to the hydrophobic isoleucine in mrf1-16 may destabilize the interactions of domain 3 with tRNA, decreasing catalytic activity of mRF1 toward the peptidyl-tRNA bond.
The PST motif in domain 2 of mRF1 corresponds to the tripeptide anticodons identified in the RFs of E. coli (15), resembling rather the PAT motif of RF1 than the SPF motif of RF2. This is consistent with the fact that the RF2-specific UGA codon is not a stop signal in yeast mitochondria (28). We believe that the PST directly binds to the stop codons in mitochondrial mRNA. In addition, in the model of mRF1 in complex with the ribosome, the PST motif contacts helix 44 of 16 S rRNA. Helix 44 is involved in the interaction between the ribosomal subunits, possibly important for the signaling from the stop codon to the peptidyl transferase center (29). As shown in Fig. 1, domain 2 contains the majority of the amino acid substitutions caused by our mrf1 mutations. These mutations can be divided into three groups. The first comprises mrf1-780 (G164C) and mrf1-13 (R231G), which cause lack of growth on media requiring respiration and inhibit mitochondrial translation (see results below). Interestingly, the substituted residues, Gly-164 and Arg-231, are located most closely to the PST motif (Fig. 1). Gly-164 is one of the two conserved glycines that form a tight loop from the middle ␤-strand in domain 2 to the central ␣-helix of mRF1. This structural feature seems to be important for the tight packing of mRF1 onto the ribosome as well as for PST loop maintenance and orientation. Replacement of Gly-164 with a bulky amino acid side chain probably causes several steric clashes and destabilizes the correct orientation of the PST loop. The positively charged side chain of the highly conserved Arg-231 seems to stabilize the conformation of the PST loop and, in addition, could be directly involved in contact with 16 S rRNA. The R231G substitution would destroy these interactions.
The S216Y substitution, in a position occupied by small amino acids within homologous RFs, represents the second group of mutations in mRF1 domain 2. Ser-216 is buried in the mRF1 structure in a tight pocket formed by surrounding residues. Conversion of the serine side chain into the bulkier tyrosine could destroy internal packing, leading to a local conformational change of the backbone, possibly affecting the linker between domains 1 and 2 and, consequently, the relative orientation of these domains.  The substitutions representing mutations from the third group, V163D, V226F, R228G, and A247P, cluster on the surface of the ␤-sheet of domain 2. Interestingly, in the structure of mRF1 complexed with the ribosome, these usually (when unbound to ribosome) solvent-exposed residues interact with the long loop-connecting domains 3 and 4 (data not shown). This linker between the domains contains two other amino acid substitutions detected in our study, Q349R and S352I, that interact directly with the mutations located on the surface of the ␤-sheet of domain 2 but only when mRF1 adopts the conformation observed in the complex with the ribosome. As a consequence, the V163D, V226F, R228G, A247P, Q349R, and S352I substitutions may influence the movement of domains 1 and 3 with respect to the functional unit formed by domains 2 and 4 and alter mRF1 interaction with the ribosome. In addition, change in physicochemical properties of the domain 2 surface caused by these mutations may also affect its direct binding to the ribosome (ribosomal protein S12 and 16 S rRNA).

mrf1 Mutant Cells Maintain Wild Type mtDNA and Mitochondrial Ribosomes-
The non-respiratory phenotype of mrf1-13 and mrf1-780 point to an essential role of Gly-164 and Arg-231 for mRF1 activity. These two mutants do not grow on medium containing glycerol or other non-fermentable carbon sources if wild type mtDNA is present (9). In contrast, other mrf1 alleles do not cause respiratory defects.
Because mRF1 is required for the maintenance of wild type mitochondrial genome, the loss of mtDNA could explain the respiratory deficiency of mrf1-13 and mrf1-780 cells. To determine mtDNA maintenance, mrf1-13 and mrf1-780 were crossed to a wild type [rho 0 ] strain. The resulting diploids grew on non-fermentable carbon sources, indicating that the mutants are recessive and that the cells retained mtDNA (Fig. 2A).
The fraction of accumulated [rho Ϫ ] mutations, determined as described under "Experimental Procedures," is 22% in mrf1-13 and 51% in mrf1-780, whereas the control strain and the remaining mrf1 alleles show 1-4% [rho Ϫ ]. It is obvious that stronger defects in translation termination lead to increased accumulation of [rho Ϫ ], although ϳ50% of cells with mtDNA retained is enough to ensure expression of mitochondrial genes in a cell population.
Unstable mitochondrial ribosomes in cells with a non-functional release factor could be an alternative reason of the respiratory deficiency of the mrf1-13 and mrf1-780 mutants. The steady-state levels of selected mitochondrial ribosomal proteins were examined to rule out the effect of ribosome degradation. It is known that Mrp13, Mrp20, and Mrp49 are relatively unstable in the absence of ribosome assembly (30,31) and free Mrp7 is relatively stable (32). As presented in Fig. 2B, the levels of all of the mitochondrial ribosomal proteins tested, Mrp7, Mrp13, Mrp20, and Mrp49 are not affected in mrf1-13 and mrf1-780 cells (Fig. 2B). This shows that despite the presence of a nonfunctional release factor, the mitochondrial ribosome is assembled and not degraded in the mrf1-13 and mrf1-780 cells.

Inhibition of Mitochondrial Translation Mediated by Mutated Release
Factor mRF1-We compared the rate of synthesis of mitochondrial gene products in wild type (MB143-3D), mrf1-13 (MB143-3D/13), and mrf1-780 (MB143-3D/780) cells by pulse labeling with [ 35 S]methionine in the presence of cycloheximide to inhibit cytoplasmic protein synthesis. Proteins were then analyzed by SDS-PAGE and autoradiography (Fig.  3A). The rate of label incorporation by the mrf1-13 and mrf1-780 cells was significantly lower than that in the control strain (Fig. 3B). Moreover, the synthesis of Cox subunits was decreased. Labeling of Cox2 was not visible, and labeling of Cox1 was reduced in the mutants relative to wild type. It was difficult to estimate Cox3 labeling because this band overlapped Atp6, but subsequent study showed that the level of Cox3 was decreased also (see below). Thus, mrf1-13 and mrf1-780 cells have a decreased level and, additionally, an altered pattern of mitochondrial protein synthesis.
The cytochrome spectrum of mutant mitochondria was determined for frozen cells (Fig. 3C). The absorption bands corresponding to cytochrome b and to cytochromes aa 3 (which reflects the amount of cytochrome c oxidase) were not visible in the mrf1-13 and mrf1-780 mutants. In contrast, the band of cytochrome cϩc 1 was unaffected, presumably because no mitochondrially encoded component was involved.
Western blot analysis of mutant cell extracts revealed considerably reduced levels of the mitochondrially encoded proteins Cox2, Cox3, and Atp6 (Fig. 3D). Reduction of Cox2 and Cox3 was by ϳ90%, and only traces of these proteins were Crude mitochondria (50 g of protein) prepared from MB143-3D (wt), MB143-3D/13 (mrf1-13), and MB143-3D/780 (mrf1-780) cells were analyzed by immunoblotting with antibodies directed against Mrp7, Mrp13, Mrp20, and Mrp49. The same blot was probed with antibodies specific for the nuclearly encoded mitochondrial malate dehydrogenase (Mdh1), serving as a loading control. wt, wild type. detectable in mrf1-13 and mrf1-780 cells at steady state. The defect of Cox2 and Cox3 synthesis was not the result of an mRNA synthesis or stability defect, because the steady-state levels of COX2 and COX3 mRNAs were similar in the mrf1-13, mrf1-780, and wild type strains (Fig. 3E).
To further monitor translation of the mitochondrial COX Trichloroacetic acid-precipitated radioactivity, corresponding to 1 A 600 of cell culture, was determined at the indicated time points. C, low temperature absorption spectra of cytochromes. The absorption peaks at 546, 558, and 602 nm correspond to cytochromes c, b, and aa 3 , respectively. The peak between 550 and 600 nm corresponds to porphyrines. D, steady-state levels of selected mitochondrially encoded proteins. Crude mitochondria (50 g of protein) were analyzed by immunoblotting with antibodies directed against Cox2 and Cox3 and subunit 6 of the F 0 F 1 -ATPase (Atp6). The same blot was probed with antibodies specific for nuclear-encoded mitochondrial malate dehydrogenase (Mdh1), serving as a loading control. To determine the relative amount of the various proteins in mrf1-13 and mrf1-780 mutants, respectively, the amount of each protein present in the wild type (wt) strain was set to 1. The numbers give the ratio mrf1-13/wt and mrf1-780/wt, respectively. E, steady-state levels of selected mitochondrial mRNAs. Northern blot analysis was carried out on mitochondrial RNAs. Specific probes were described under "Experimental Procedures." 16 S rRNA hybridization served as a loading control. wt, wild type. locus, we took advantage of the synthetic mitochondrially located ARG8 m gene kindly provided by T. Fox and successfully used in his laboratory as a reporter for mitochondrial gene expression (33). Arg8 protein is normally encoded in the nucleus, synthesized in the cytoplasm, and imported into the mitochondrial matrix where it participates in arginine biosynthesis. The synthetic ARG8 m when inserted to mtDNA encodes the same enzyme within the mitochondrial matrix and is able to complement chromosomal arg8-⌬ deletion. To check whether mutant release factor inhibits translation of the reporter gene, we performed a cytoduction experiment. We introduced mitochondria containing ARG8 m in the COX3 coding region to the recipient mrf1-780 arg8-⌬ cells devoid of mtDNA and containing a plasmid with MRF1 ( Fig. 4 and see "Experimental Procedures"). The resulting cytoductants were replica-plated on a medium lacking arginine. In the presence of the MRF1 gene supplied by the plasmid, the cytoductant was Arg ϩ , indicating efficient translation of the reporter cox3::ARG8 m gene. The effect of mrf1-780 was observed when the complementing pRS314-MRF1 plasmid was lost on a non-selective medium. mrf1-780 inhibited translation of cox3::ARG8 m , which resulted in weak growth on a medium lacking arginine. The results of this experiment (Fig. 4) verify that the mrf1-780 mutation inhibits translation of mitochondrial mRNAs.

DISCUSSION
Translation seems to be crucial for regulation of mitochondrial gene expression (34), but the role of the termination step in this regulation has not been studied so far. We identified ten alleles of the S. cerevisiae chromosomal gene encoding mitochondrial release factor mRF1 that change the fidelity of stop codon recognition. Only two alleles show a non-respiratory phenotype in the presence of wild type mtDNA, and this phenotype correlates with inhibition of mitochondrial translation.
It is known that mitochondrial translation in yeast is controlled at the initiation and elongation steps, although the mechanisms of this control are poorly understood. The scanning model does not function in start codon selection in yeast mitochondrial mRNA, and initiation might be analogous to that of internal ribosome entry in eukaryotic cytoplasmic translation (35). The major distinguishing feature is that mitochondrial initiation requires the action of membrane-bound translational activator proteins that recognize targets in the 5Ј-untranslated leaders and seem to mediate individual mRNA-mitoribosome interactions (34,36). Yeast mitoribosomes are also sensitive to secondary mRNA structures within coding sequences. Translation of COX2 is inhibited by ribosome pausing at negative regulatory sequences identified in the coding region that are antagonized by a positive sequence specifying the pre-Cox2 leader peptide (37,38).
The data presented here show that mitochondrial translation can also be inhibited at the termination step. Such inhibition was observed in the presence of specific mutations, mrf1-13 and mrf1-780, in the nuclear gene encoding the mitochondrial release factor, mRF1. The overall rate of translation on mitochondrial ribosomes was reduced in mrf1-13 and mrf1-780 cells, but the degree of inhibition varied among individual mitochondrially encoded polypeptides, being the most pronounced for COX genes. Decreased mitochondrial translation of COX1 and COX2 was previously reported for two other mrf1 mutants (11). Similar gene-specific effects on mitochondrial gene expression were reported previously for mutations in other general components of mitochondrial translation system (10, 39 -41). Additionally, we showed that the steady-state levels of Cox2 and Cox3 were decreased in mrf1-13 and mrf1-780 over 20-fold, whereas the levels of the respective mRNAs were not affected. A connection of this gene-specific effect with co-translational cytochrome oxidase assembly could be excluded because translation of the ARG8 m reporter gene inserted into the COX3 locus was also decreased. Additionally, to translational regulation, some post translational effects were also observed. Although COB translation occurred efficiently in mrf1-13 cells, cytochrome b was not observed in the spectrum. Moreover, whereas mrf1-13 and mrf1-780 showed Gly Ϫ phenotype with wild type mtDNA, the same mutations acted as suppressors of premature stop codons in mtDNA and caused leaky growth on YPGly with mit Ϫ mutant mtDNA (9). Although the detailed genetic analysis reported here fully confirms this phenotype, we do not understand the molecular basis of observed effect. Similarly, the pet309 allele isolated as a suppressor of a cox3 initiation codon mutation caused the Gly Ϫ phenotype with the wild type mtDNA but leaky growth with mutant mtDNA. 2 The molecular mechanism of translation inhibition in the 2 L. S. Folley, D. F. Steele, and T. D. Fox, personal communication.
FIG. 4. The mrf1-780 mutation inhibits translation of the cox3::ARG8 m reporter gene and Arg ؉ growth. As outlined in the scheme, strain MJ14-1D arg8-⌬ mrf1-780 (see Table I) was made [rho 0 ] and transformed with a plasmid encoding wild type MRF1 gene (pRS314-MRF1). The resulting strain was used in a cytoduction experiment as a recipient for the mitochondria of the DFS189 strain containing the cox3::ARG8 m reporter gene. Subsequently, the cytoductants were grown on a non-selective medium, allowing the loss of the plasmid coding for mRF1. To test translation of the cox3::ARG8 m reporter gene, the Arg ϩ growth competence of the cytoductants was assessed. MJ14-1D [cox3::ARG8 m ] with and without pRS314-MRF1 plasmid were replicated to SC-Trp and SC-Arg plates and incubated for 3 days at 30°C. presence of mutated release factor is unknown. The level of mRF1 protein is not decreased in the mrf1-13 and mrf1-780 mutants (data not shown); therefore, simple titration of the factor could be excluded. Instead, it is likely that mutations change the interaction of mRF1 with the mitoribosome as it was shown previously for two other mrf1 mutants (12). One could expect a slowing down of the codon recognition step in the mrf1-13 and mrf1-780 mutants because of amino acid substitutions close to the PST motif that is suggested to act as a peptide anticodon of mRF1. That would results in the stalling of the mitoribosome on the stop codon, delay of polypeptide hydrolysis, and, consequently, inhibition of translation.
Newly synthesized nascent polypeptide chains emerge from the ribosome through the exit tunnel. As revealed by recent cryoelectromagnetic studies, the structure of the exit tunnel in mitoribosome is distinctly different from the tunnels in cytoplasmic ribosomes. The tunnel has a wide opening allowing access of proteins that are not defined so far (42). It is probable that the nascent mitochondrial polypeptide chains interact with the release factor and this interaction stabilizes an intermediate termination complex. One could assume that particular amino acid substitutions in mutated mRF1 increase the stability of these interactions leading to ribosomal stalling. Moreover, the degree of inhibition could be differently modulated by the C-terminal sequences and that is a possible explanation why mrf1-13 and mrf1-780 differently affected translation of the various mitochondrially synthesized polypeptides. Interestingly, a mechanism of translation termination inhibition, mediated by direct interaction of the nascent polypeptide chain with the eRF1 release factor, has been described for cytomegalovirus (43).