Highly Diverged Homologs of Saccharomyces cerevisiae Mitochondrial mRNA-Specific Translational Activators Have Orthologous Functions in Other Budding Yeasts

Corresponding author: Thomas D. Fox, Department of Molecular Biology and Genetics, Biotechnology Bldg., Cornell University, Ithaca, NY 14853-2703., tdf1{at}cornell.edu (E-mail)

Communicating editor: F. WINSTON

	ABSTRACT

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Translation of mitochondrially coded mRNAs in Saccharomyces cerevisiae depends on membrane-bound mRNA-specific activator proteins, whose targets lie in the mRNA 5'-untranslated leaders (5'-UTLs). In at least some cases, the activators function to localize translation of hydrophobic proteins on the inner membrane and are rate limiting for gene expression. We searched unsuccessfully in divergent budding yeasts for orthologs of the COX2- and COX3-specific translational activator genes, PET111, PET54, PET122, and PET494, by direct complementation. However, by screening for complementation of mutations in genes adjacent to the PET genes in S. cerevisiae, we obtained chromosomal segments containing highly diverged homologs of PET111 and PET122 from Saccharomyces kluyveri and of PET111 from Kluyveromyces lactis. All three of these genes failed to function in S. cerevisiae. We also found that the 5'-UTLs of the COX2 and COX3 mRNAs of S. kluyveri and K. lactis have little similarity to each other or to those of S. cerevisiae. To determine whether the PET111 and PET122 homologs carry out orthologous functions, we deleted them from the S. kluyveri genome and deleted PET111 from the K. lactis genome. The pet111 mutations in both species prevented COX2 translation, and the S. kluyveri pet122 mutation prevented COX3 translation. Thus, while the sequences of these translational activator proteins and their 5'-UTL targets are highly diverged, their mRNA-specific functions are orthologous.

TRANSLATION of mitochondrially coded mRNAs in Saccharomyces cerevisiae is a surprisingly complex process. Most, if not all, of the seven major mitochondrially coded mRNAs (DIECKMANN and STAPLES 1994 ) are translated under the direction of mRNA-specific translational activator proteins specified by nuclear genes (FOX 1996 ). The translational activator proteins interact functionally with targets in the mRNA 5'-untranslated leaders (5'-UTLs) and, in at least one case, also with the mitochondrial ribosomal small subunit (FOX 1996 ).

The activator proteins specific for translation of four of the mRNAs encoding integral membrane proteins, COX1, COX2, COX3, and COB, are themselves bound to the mitochondrial inner membrane (MICHAELIS et al. 1991 ; MCMULLIN and FOX 1993 ; MANTHEY et al. 1998 ; N. S. GREEN-WILLMS and T. D. FOX, unpublished data), suggesting that they could localize translation to assembly sites of respiratory complexes (COSTANZO and FOX 1990 ; FOX 1996 ). Indeed, the 5'-untranslated regions of the COX2 and COX3 mRNAs contain information necessary for proper targeting of the proteins they encode (SANCHIRICO et al. 1998 ). COX2 and COX3 mRNA-specific translational activation is also a rate-limiting step in gene expression (STEELE et al. 1996 ; N. S. GREEN-WILLMS and T. D. FOX, unpublished data). Thus, the S. cerevisiae translational activator proteins appear to have dual roles in regulating mitochondrial gene expression and in localizing translation.

Why translation of the mitochondrially coded mRNAs specifying cytochrome c oxidase subunits Cox1p, Cox2p, and Cox3p should be dependent on distinct activators remains an open question. One possible rationalization is that mRNA specificity allows for relative topological distinctions between the sites where these mitochondrially coded proteins are synthesized, which could promote efficient assembly of cytochrome c oxidase complexes in the inner membrane. However, this effect would be relatively subtle since Cox1p, Cox2p, and Cox3p can be assembled into cytochrome c oxidase after translation of experimentally derived chimeric mRNAs bearing 5'-UTLs derived from certain other mitochondrial genes, under the direction of their respective activators (MULLER et al. 1984 ; COSTANZO and FOX 1986 , COSTANZO and FOX 1988 ; POUTRE and FOX 1987 ; RODEL and FOX 1987 ; MULERO and FOX 1993B ; MANTHEY and MCEWEN 1995 ). These experiments shed no light on possible quantitative differences in efficiency of cytochrome c oxidase assembly, owing to the genetic instability of the mitochondrially heteroplasmic strains employed. However, they demonstrate that the one-to-one correspondence between translational activators and the mitochondrial genes they govern can be experimentally altered without destroying function. Thus, if there were no adaptive value in maintaining these correspondences, then they could diverge during evolution.

In this study we have sought to use phylogenetic comparisons to ask whether the correspondences between translational activators and mitochondrial mRNAs have been conserved and to shed light on the function of the activator proteins themselves. We have focused on the COX2-specific activator Pet111p (STRICK and FOX 1987 ) and the subunits of the COX3-specific activator complex, Pet54p, Pet122p, and Pet494p (BROWN et al. 1994 ), none of which had known homologs at the start of this study. We began a generally unsuccessful search for orthologous genes in the divergent budding yeasts S. servazzii, S. kluyveri, Kluyveromyces lactis, and the dimorphic yeast Candida albicans, using cross-hybridization and cross-complementation. These approaches worked only for a sister species of S. cerevisiae, S. bayanus. However, we were able to isolate S. kluyveri genes homologous to PET111 and PET122, as well as a K. lactis homolog of PET111, by screening for complementation of highly conserved genes that are adjacent to PET111 and PET122 in S. cerevisiae. We found that Pet111p and Pet122p are among the most rapidly diverging proteins known in these species. They are nevertheless orthologous, since the mutations we constructed in the S. kluyveri and K. lactis homologs generated phenotypes in those species that are similar to those of the corresponding S. cerevisiae mutants. Thus, while the sequences of these translational activator proteins are highly diverged, their one-to-one correspondences with mitochondrial mRNAs have been conserved. These findings suggest that the specificity of translational activation plays an important role in fungal mitochondrial biogenesis.

	MATERIALS AND METHODS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Yeast strains, media, and genetic methods:
Strains used in this study are listed in Table 1. All S. cerevisiae strains were isogenic or congenic to the wild-type strain D273-10B (ATCC #25657), except strains JM43-GD7, NGB108, PTH43, and PTH352. Media and genetic methods were as described (SHERMAN et al. 1986 ). Respiratory growth was assessed on YPEG medium (3% ethanol, 3% glycerol, 1% yeast extract, 2% bacto-peptone, 2% agar). Saccharomyces strains were transformed either by treatment with lithium acetate and polyethylene glycol (ITO et al. 1983 ) or by using the Yeast EZ Transformation Kit (Zymo Research, Orange, CA). K. lactis was transformed as described (CHEN and CLARK-WALKER 1996 ).

Plasmid manipulations, nucleotide sequencing, and computer analysis:
Plasmids were constructed and transformed into Escherichia coli DH5F'IQ using standard techniques (SAMBROOK et al. 1989 ). Oligonucleotide synthesis and nucleotide sequencing were performed by DNA Services, Cornell University. Nucleotide sequence data were analyzed using Lasergene Biocomputing Software (DNAStar, Madison, WI) and Sequencher (Gene Codes, Ann Arbor, MI). The Basic Local Alignment Search Tool (BLAST; ALTSCHUL et al. 1990 ) program was accessed through the National Center for Biotechnology Information or the Saccharomyces Genome Database to search for nucleotide and protein sequence similarities. RNA structures were predicted using the mfold version 3.0 program accessed at http://mfold2.wustl.edu/~mfold/rna/form1-2.3.cgi (ZUKER 1994 ). The C. albicans genomic sequence determined by the Stanford DNA Sequencing and Technology Center was searched at http://www-sequence.stanford.edu/group/candida. The S. pombe genomic sequence was searched at http://www.sanger.ac.uk/Projects/S_pombe/blast_server.shtml. The C. elegans genomic sequence was searched at http://www.sanger.ac.uk/Projects/C_elegans/blast_server.shtml.

Genomic libraries:
Genomic DNA from S. bayanus (NRRL #Y-12624), S. kluyveri (NRRL #Y-12651), or S. servazzii (NRRL #Y-12661) was prepared using either the QIAGEN Genomic-tip 500 kit (QIAGEN Inc., Valencia, CA) or the procedure of PHILIPPSEN et al. 1991 . DNA was partially digested with Sau3AI and the partial digestion products were separated by size in a 10–40% sucrose gradient as described by ROSE and BROACH 1991 . Fragments of 15–20 kb were pooled and ligated to BamHI-cleaved YEp352 (HILL et al. 1986 ) and the libraries were amplified by standard methods (ROSE and BROACH 1991 ). The S. bayanus library was composed of ~15,000 independent E. coli transformants, 45% with genomic DNA inserts; the S. servazzii library, of ~10,000 independent transformants, 70% with inserts; and the S. kluyveri library, of ~40,000 independent transformants, 74% with inserts. A K. lactis (strain CBS2359) genomic library of 4- to 10-kb BamHI fragments cloned into YepLac195 (MULDER et al. 1994 ) was obtained from L. Grivell.

Isolation of complementing plasmids from libraries:
In all cases, Ura⁺ clones were selected on minimal glucose medium after transformation of host strains (Table 1) with library or defined plasmid DNAs. Respiring clones were then identified by replica plating to YPEG medium. PET111 homologs were obtained either by complementation of pet111 in strain NB39-5D or by complementation of cox7 in strain JM43-GD7. We obtained only a 3' fragment of S. servazzii PET111: the cox7-complementing clone carried DNA coding only the C-terminal 238 residues of the protein, and we failed to isolate a larger clone with the entire gene. PET122 homologs were obtained either by complementation of pet122 mutation in strain PTH43 or by complementation of oxa1 in strain MCC318. PET494 homologs were sought by complementation of pet494 in strain NGB108. S. bayanus PET54 was isolated as a 3.0-kb BamHI fragment of S. bayanus genomic DNA that hybridized at high stringency to an S. cerevisiae PET54 probe. Its ability to complement was demonstrated by transformation of strain PTH352.

Analysis of the S. kluyveri LYS9 region:
PET494 and LYS9 are ~2.4 cM (6.6 kb) apart on S. cerevisiae chromosome XIV (MULLER and FOX 1984 ). We isolated the S. kluyveri LYS9 region in an attempt to clone a PET494 ortholog. Six plasmids were obtained from the S. kluyveri library that complemented S. cerevisiae lys9 strain TF112 (Table 1). Sequencing the ends of each insert allowed orientation of the six fragments based on overlapping end sequences. Sequencing of the LYS9-containing region on one clone revealed that S. kluyveri LYS9 is adjacent to a gene similar to S. cerevisiae MSO1 (GenBank accession no. AF170311), the same arrangement as in S. cerevisiae. However, flanking MSO1 on the other side is a homolog (GenBank accession nos. AF170309 and AF170310) of S. cerevisiae YKL215c, located on S. cerevisiae chromosome XI, instead of an S. kluyveri PET494 homolog. Flanking S. kluyveri LYS9 on the other side is a homolog (GenBank accession no. AF170312) of S. cerevisiae YCR095c, located on S. cerevisiae chromosome III. We did not obtain a contiguous sequence of the entire region.

Cloning the mitochondrial COX2 and COX3 genes from other yeasts:
Probes corresponding to the S. cerevisiae COX2 and COX3 coding sequences were produced by PCR, ³²P-labeled using standard techniques, and used to probe Southern blots of restricted genomic DNA from S. kluyveri (strain GRY1175) or K. lactis (COX3 only; strain 2105-1D). Hybridizations were done at low stringency (55° in aqueous hybridization solution containing 6x SSC, 0.9 M NaCl, and 0.09 M sodium citrate). DNA fragments within the size range of cross-hybridizing bands were isolated and cloned to create minilibraries that were screened by colony hybridization to the COX2 and COX3 probes. S. kluyveri COX2 was cloned in part on a 1.1-kb DraI fragment (Table 2) that lacked the 5' end and 5'-UTL coding region. Sequence of the remaining coding sequence and upstream region was obtained by J. Piskur by direct sequencing of purified S. kluyveri mitochondrial DNA. S. kluyveri COX3 was cloned as overlapping 1.0-kb SspI and 4.0-kb DraI fragments (Table 2). K. lactis COX3 was cloned on a 1.5-kb MspI fragment. All genomic clones were confirmed to be devoid of rearrangements, either by Southern blot hybridization to genomic DNA or by production of fragments of the expected size using PCR on genomic DNA.

Construction of null alleles in S. kluyveri PET111 and PET122 and in K. lactis PET111:
S. kluyveri PET111 was subcloned on a 5.8-kb EcoRI-XhoI fragment in pBluescriptKS(-) (Stratagene, La Jolla, CA) to create plasmid pNB143. pNB143 was cleaved with HindIII to remove a 2.1-kb fragment of the PET111 gene containing the C-terminal 689 codons. A fragment with HindIII ends carrying the URA3 gene, generated by PCR, was ligated into the HindIII-cleaved pNB143 backbone to generate pNB145. The pNB145 EcoRI-XhoI insert was gel-purified and used to transform the ura3 mutant S. kluyveri strain GRY1175 (Table 1) to uracil prototrophy. PCR reactions were performed using genomic DNA of a transformant as a template, with primers corresponding to the PET111 region and URA3, to verify that the gene replacement occurred as expected, leaving only the N-terminal 108 codons of PET111.

S. kluyveri PET122 was subcloned on a 2.4-kb BglII-ClaI fragment in pBluescriptM13(-) (Stratagene) to create plasmid pMC367. pMC367 was digested with EcoRV to remove 726 bp of the PET122 coding sequence carrying 242 codons and religated with a BamHI linker; then a hisG::URA3::hisG cassette (ALANI et al. 1987 ) was ligated into the resulting BamHI site to create plasmid pMC369. pMC369 was digested with SpeI and SalI to cleave the pet122::URA3 insert from the plasmid backbone and was used to transform S. kluyveri GRY1175 (Table 1) to uracil prototrophy. PCR reactions were performed using genomic DNA of transformants as a template, with primers corresponding to the PET122 region and URA3, to verify that the gene replacement occurred as expected leaving only the N-terminal 12 codons and the C-terminal 59 codons of PET122.

A 1.5-kb EcoRI fragment carrying the 5' half of K. lactis PET111 was subcloned from a cox7-complementing plasmid isolated from the K. lactis genomic library into pTZ19U (Bio-Rad, Richmond, CA) to generate the plasmid pKLN20. To disrupt PET111 in pKLN20, a 1.4-kb fragment containing a kanamycin resistance cassette was obtained from plasmid pUG6 (GULDENER et al. 1996 ) by digestion with XhoI and BglII, made blunt-ended using T4 DNA polymerase, and ligated to a 3.26-kb fragment generated by cleaving pKLN20 with ApaI and EcoRV. This disruption removed the 5' 847 bp of PET111 carrying the N-terminal 282 codons. A 1.8-kb EcoRI fragment from the resulting plasmid, pKLN22B, was used to transform K. lactis. An atp2.1 mutant strain (CK56-7A; see Table 1) was used as a recipient for transformation because atp mutant strains recover better after transformation than wild type (G. D. CLARK-WALKER, unpublished results). Transformants resistant to G418 (200 µg/ml in YPD medium) were checked for stability of G418 resistance, and correct integration of the fragment at PET111 was verified by Southern analysis. To isolate a strain carrying the pet111 disruption and a wild-type ATP2 gene, the pet111 disruptant was crossed to a wild-type strain, the diploid was sporulated, and 15 asci were dissected. All spores were viable, and all tetrads showed 2:2 segregation of G418 resistance, marking the pet111 disruption. All G418-resistant spores were respiratory deficient. Resistance to ethidium bromide (16 µg/ml in YPD medium), marking the atp2.1 mutation, also segregated 2:2 and was unlinked to pet111::kan^r. A strain carrying the pet111 mutation but not the atp2.1 mutation, CW64-1C (Table 1), was chosen for further study.

Isolation of mitochondria and immunological methods:
Mitochondria were prepared from S. cerevisiae and S. kluyveri cells by differential centrifugation as described (DAUM et al. 1982 ). Mitochondria were isolated from K. lactis as described for S. cerevisiae (GLICK and PON 1995 ), except that liquid cultures were grown in YPD, and crude mitochondria were purified on Nycodenz step gradients with 5% intervals. Wild-type K. lactis mitochondria floated at the 20%/25% interface of a 5% to 25% gradient, while K. lactis pet111 mutant mitochondria floated at the 15%/20% interface of a 5% to 35% gradient.

Antibody against Cox2p was the mouse monoclonal CCO6 (a gift of T. L. Mason; PINKHAM et al. 1994 ). Anti-Cox3p was the mouse monoclonal DA5 (Molecular Probes, Eugene, OR). Anti-Cox1p was the rabbit polyclonal SS7-T, obtained from G. Schatz. SDS-polyacrylamide gel electrophoresis and Western blotting were performed using standard techniques (HARLOW and LANE 1988 ). Antigen-antibody complexes were visualized using horseradish peroxidase-conjugated goat anti-mouse or anti-rabbit IgG (Gibco BRL, Bethesda, MD) secondary antibody and the enhanced chemiluminescence system (Amersham Life Science, Arlington Heights, IL). Immunoprecipitations were performed by standard techniques (HARLOW and LANE 1988 ) except that goat anti-mouse IgG was added to the immunoprecipitation mixture before the addition of protein A-Sepharose.

In vivo labeling of mitochondrial translation products:
In vivo ³⁵S-labeling in the presence of cycloheximide was performed as described previously (FOX et al. 1991 ), except that S. kluyveri labelings were performed on exponential-phase rather than stationary-phase cells. SDS gel electrophoresis of labeled mitochondrial translation products was performed as described (GREEN-WILLMS et al. 1998 ).

RNA gel blot hybridization analysis:
Total RNA was isolated from yeast strains grown to exponential phase in galactose-containing complete medium, using the hot-phenol extraction method (AUSUBEL et al. 1993 ). Equal amounts (10 µg) were subjected to electrophoresis, blotted to a filter, and probed with radioactively labeled probes produced by random-primed labeling (Boehringer Mannheim, Indianapolis). Blots were probed sequentially with labeled PCR products corresponding to the S. kluyveri COX2 gene, the S. kluyveri COX3 gene, and the S. cerevisiae mitochondrial 15S rRNA gene. mRNA abundance was quantitated using a Storm 840 Phosphoimager (Molecular Dynamics, Sunnyvale, CA) and the ImageQuant software package.

	RESULTS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Isolation of genes from divergent budding yeasts, homologous to S. cerevisiae genes for mRNA-specific translational activators:
Complementation of S. cerevisiae mutations by functional homologs only from closely related species: The phylogenetic relationships of the various yeasts used in this study are diagrammed in Fig 1. S. bayanus is very closely related to S. cerevisiae, based on analysis of ribosomal RNA sequences (KURTZMAN and ROBNETT 1991 , KURTZMAN and ROBNETT 1998 ; JAMES et al. 1997 ). They are nearly identical phenotypically (BARNETT et al. 1990 ) and have the same number of chromosomes, although their electrophoretic karyotypes are distinguishable (NAUMOV et al. 1992 ). Many genes are arranged in the same relative order and are similar enough to cross-hybridize (RYU et al. 1996 ). The divergence of S. servazzii and S. kluyveri from S. cerevisiae, as well as from each other, is much higher and comparable to that of K. lactis (KURTZMAN and ROBNETT 1991 , KURTZMAN and ROBNETT 1998 ; JAMES et al. 1997 ). Each has distinct physiological characteristics and fewer chromosomes than S. cerevisiae (VAUGHAN-MARTINI et al. 1993 ).

View larger version (5K): [in this window] [in a new window] [Download PPT slide]   Figure 1. Phylogenetic relationships of yeasts used in this study and the distantly related fission yeast S. pombe. Relative distances were calculated using the neighbor-joining method (SAITOU and NEI 1987 ) to compare sequences of the D1/D2 domains of 26S rRNA (KURTZMAN and ROBNETT 1998 ).

To seek complementing functional homologs, we constructed libraries of S. bayanus, S. servazzii, and S. kluyveri genomic DNA in a high-copy plasmid vector carrying the URA3 gene (MATERIALS AND METHODS). Appropriate S. cerevisiae strains (MATERIALS AND METHODS; Table 1) were transformed to Ura⁺ with the libraries and transformants were screened for growth on nonfermentable carbon sources (Pet⁺ phenotype). In cases where Pet⁺ transformants were obtained, linkage to the URA3 marker was tested in plasmid loss assays. Complementing plasmids were isolated from the yeast transformants, propagated in E. coli, and reintroduced into the same host strains to confirm complementation.

The S. bayanus library yielded complementing clones at a frequency of about 1 per 10,000 Ura⁺ transformants when introduced into the S. cerevisiae pet111, pet122, and pet494 mutants. S. bayanus PET54 was isolated independently by cross-hybridization (MATERIALS AND METHODS) and complemented when introduced into an S. cerevisiae pet54 mutant. For all four genes, the S. bayanus homologs complemented as strongly as the corresponding wild-type S. cerevisiae genes.

The S. servazzii library yielded two clones that exhibited weak complementation, among 32,000 Ura⁺ transformants of the pet122 mutant. However, no Pet⁺ colonies were found in 700,000 Ura⁺ transformants of the pet111 mutant strain, in more than 200,000 transformants of the pet54 mutant strain, or in more than 130,000 transformants of the pet494 mutant strain.

The S. kluyveri library failed to yield any complementing clones after screening of 170,000 Ura⁺ transformants of the pet111 mutant, 45,000 transformants of the pet54 mutant, 131,000 transformants of the pet122 mutant, and 51,000 transformants of the pet494 mutant. This library did yield one leu2 complementing clone out of 9,500 transformants, and four oxa1 complementing clones (see below) out of 16,000 transformants, suggesting that the failure to complement translational activator mutations was not due to poor quality of the library. As shown below, direct tests with the S. kluyveri orthologs of PET111 and PET122, isolated by a different approach, confirmed that they do not function in S. cerevisiae. These complementation results are summarized in Table 3.

Isolation of sequence homologs by complementation of mutations in neighboring S. cerevisiae genes: The S. servazzii DNA fragment with weak pet122-complementing activity contained a sequence homolog of PET122 adjacent to a sequence homolog of the OXA1 gene. This arrangement is similar to that in S. cerevisiae, where PET122 is 215 bp away from OXA1. OXA1 encodes a protein that is required for the assembly of cytochrome c oxidase and assembly or stability of ATP synthase and is functionally conserved in humans (BONNEFOY et al. 1994A , BONNEFOY et al. 1994B ; ALTAMURA et al. 1996 ). This observation suggested that translational activator homologs might be isolated indirectly through cloning of genes from other yeasts by complementation of neighboring S. cerevisiae mutations, if synteny were preserved during evolutionary divergence (KUWABARA and SHAH 1994 ; AGNAN et al. 1997 ). The only other S. cerevisiae specific translational activator gene that is closely linked to a highly conserved gene is PET111, which is situated 502 bp from the gene encoding subunit VII of cytochrome c oxidase, COX7 (CALDER and MCEWEN 1991 ). A Cox7p ortholog is present in the mammalian enzyme (CAPALDI 1990 ), suggesting that the gene is likely to be present in other yeasts.

Null mutants in cox7 and oxa1 have a tight Pet^- phenotype (CALDER and MCEWEN 1991 ; BONNEFOY et al. 1994A ). We therefore transformed cox7 and oxa1 mutant strains (Table 1) with genomic libraries from S. kluyveri and K. lactis and screened as described above for Pet⁺ transformants whose phenotype depended on plasmid-encoded genes. In addition, we transformed the cox7 mutant with the S. servazzii library in an attempt to isolate a PET111 homolog.

We obtained plasmids complementing cox7 from the S. servazzii, S. kluyveri, and K. lactis (MULDER et al. 1994 ) genomic libraries, and sequencing revealed the presence of a PET111 homolog in the genomic fragments from each of these species (Table 2). The S. servazzii clone contained only the 3' end of the PET111 gene, but the others carried complete coding sequences. We also obtained plasmids complementing oxa1 from the S. kluyveri library, and they contained a complete PET122 homolog (Table 2). Direct testing of complementation, by transformation of the corresponding S. cerevisiae mutant strains with S. kluyveri PET111 and PET122 and K. lactis PET111, demonstrated that none of these genes functions in S. cerevisiae (Table 3). We were unable to find any oxa1 complementing plasmids from the K. lactis library among more than 1.3 x 10⁶ transformants. Since the library appears to be good (it yielded four COX7 clones among 25,000 transformants), and since the Oxa1p ortholog from humans functions in S. cerevisiae (BONNEFOY et al. 1994B ), the most likely explanation for this result is that the K. lactis OXA1 gene is not expressed in S. cerevisiae.

There are no characterized S. cerevisiae genes immediately adjacent to PET54 or PET494 suitable for cloning by complementation. However, we attempted to clone PET494 from the S. kluyveri library by complementation of the linked gene LYS9 (MULLER and FOX 1984 ), which is located 6.6 kb from PET494 in S. cerevisiae. We isolated six independent S. kluyveri genomic fragments that complemented the lys9 strain TF112 (Table 1). Unfortunately, partial sequencing of these fragments revealed that PET494 and LYS9 are not syntenic in S. kluyveri. Instead, the S. kluyveri LYS9 chromosomal region has sequences similar to at least three different S. cerevisiae chromosomes (MATERIALS AND METHODS).

Sequence comparisons among specific translational activators from budding yeasts: Comparison of the proteins coded by orthologous genes of the sister species S. cerevisiae and S. bayanus reveals that they are all similar in sequence: 73% identity for Pet111p, 72% for Pet54p and Pet122p, and 76% for Pet494p. Sequence identity is evenly dispersed over the lengths of these protein pairs, with one notable exception. Residues 198–259 of S. cerevisiae Pet54p probably comprise an RNA recognition motif (RRM; reviewed in BURD and DREYFUSS 1994 ). This region of Pet54p has 27% identity and 50% similarity to the identified RRM (residues 143–216) of the C. elegans fox-1 protein, a known RNA-binding protein involved in sex determination (HODGKIN et al. 1994 ). Although the match between Pet54p and the RRM consensus is weak enough that the motif is not identifiable by comparison with databases such as PROSITE (BAIROCH et al. 1997 ) or BLOCKS (HENIKOFF and HENIKOFF 1994 ), it was identified in a search using a profile generated from more than 300 known RRMs (M. E. CUSICK, personal communication). In the predicted RRM region, S. bayanus Pet54p is 90% identical to S. cerevisiae Pet54p, while the overall identity between the two proteins is 72%. This is not surprising given the ability of S. bayanus Pet54p to interact with the S. cerevisiae COX3 mRNA 5'-UTL.

Comparisons of the complete Pet111p-homologous sequences from S. kluyveri and K. lactis reveal that they are highly diverged from that of S. cerevisiae and from each other: the S. cerevisiae and S. kluyveri proteins are 32% identical while the S. cerevisiae and K. lactis proteins are only 20% identical (Fig 2; Table 4). While we obtained the sequence of only 238 C-terminal residues of S. servazzii Pet111p, these residues exhibited a similar degree of sequence divergence (Table 4). Similarity between all the sequences is evenly distributed over their entire lengths and there are no highly conserved regions (Fig 2).

View larger version (96K): [in this window] [in a new window] [Download PPT slide]   Figure 2. Alignment of Pet111p orthologs from S. cerevisiae (S.c.), S. kluyveri (S.k.), and K. lactis (K.l.). Sequences were aligned using the Jotun-Hein method with the MegAlign program (MATERIALS AND METHODS). Identical amino acids are shaded. Dashes represent gaps.

Comparisons of the Pet122p-homologous sequences from S. servazzii and S. kluyveri also revealed a high degree of divergence. The very weakly complementing S. servazzii PET122 encodes a protein that is 27% identical to S. cerevisiae Pet122p and has a C-terminal extension of 31 residues relative to S. cerevisiae Pet122p. S. kluyveri Pet122p, which failed to complement, is 33% identical to S. cerevisiae Pet122p and has a C-terminal extension of 42 residues relative to it. Sequence identity is distributed evenly over the whole lengths of Pet122p sequences except for the C-terminal extensions, which do not resemble each other (not shown). The longest stretch of amino acid identity shared by all four Pet122p sequences is eight amino acids, corresponding to residues 145–152 of S. cerevisiae Pet122p.

Mitochondrially coded RNA targets of S. kluyveri and K. lactis translational activators: The COX2- and COX3-mRNA- specific translational activators interact functionally with targets in the 5'-UTLs of the mitochondrially coded mRNAs (COSTANZO and FOX 1988 , COSTANZO and FOX 1993 ; MULERO and FOX 1993A , MULERO and FOX 1993B ; BROWN et al. 1994 ; WIESENBERGER et al. 1995 ). In the case of the 54-nucleotide (nt) COX2 5'-UTL of S. cerevisiae, a central 32-nt region containing a stem-loop structure is necessary and partially sufficient for Pet111p activation of translation (DUNSTAN et al. 1997 ). This region contains the sequence UCUAA, which has been found upstream of COX2 in a number of budding yeasts, including K. lactis (HARDY and CLARK-WALKER 1990 ; CLARK-WALKER and WEILLER 1994 ). UCUAA comprises part of the stem structure and is necessary for translation in S. cerevisiae (DUNSTAN et al. 1997 ). The nature of the target(s) in the 613-nt COX3 5'-UTL of S. cerevisiae is less well understood. However, a 151-nt region in its upstream half is necessary and, with modifications, partially sufficient for translational activation (WIESENBERGER et al. 1995 ).

To allow further comparisons, we cloned and sequenced COX2 and COX3 from S. kluyveri and COX3 from K. lactis (MATERIALS AND METHODS and Table 2). These data revealed that Cox2p and Cox3p are far more highly conserved than the translational activators: S. kluyveri and K. lactis Cox2p share 89 and 86% amino acid identity to S. cerevisiae Cox2p, respectively, and 91% identity to each other. S. kluyveri and K. lactis Cox3p have 85 and 82% identity to S. cerevisiae Cox3p, respectively, and 87% identity to each other.

Comparison of the COX2 and COX3 mRNA 5'-UTLs is complicated by the fact that the 5' ends of the S. kluyveri and K. lactis mRNAs have not been experimentally determined. However, S. cerevisiae, K. lactis, and several other budding yeasts share the same mitochondrial promoter consensus sequence (TATAAGTAA) (OSINGA et al. 1982 ; CLARK-WALKER et al. 1985 ; RAGNINI and FRONTALI 1994 ; BISWAS 1998 ), and thus S. kluyveri is likely to as well. Indeed, upstream of the S. kluyveri COX2 gene there is a perfect match to the promoter consensus at positions -95 to -87, predicting an 88-nt 5'-UTL. The K. lactis COX2 promoter appears to direct cotranscription of tRNA^Val and the COX2 mRNA from 330 nt upstream of the COX2 initiation codon (HARDY and CLARK-WALKER 1990 ). In S. cerevisiae, tRNA^Val is cotranscribed with the COX3 mRNA, and 3'-processing of the tRNA generates the mRNA 5'- end (WIESENBERGER et al. 1995 ). Similar events in K. lactis would produce a 226-nt COX2 mRNA 5'-UTL.

In the S. cerevisiae COX2 mRNA 5'-UTL, the downstream side of the 5-bp stem contains the first four bases of the conserved sequence UCUAA and ends at -20 relative to the initiation codon (DUNSTAN et al. 1997 ; Fig 3). The S. kluyveri COX2 5'-UTL also contains a putative stem-loop structure with a 4-bp stem whose downstream side contains the first three bases of the UCUAA sequence and ends at position -21 (Fig 3). A potential stem structure containing the conserved sequence UCUAA is also present in the K. lactis COX2 5'-UTL (Fig 3). The 4-bp stem could form in a position similar to those of S. cerevisiae and S. kluyveri, with the 3' end of the stem at position -30. However, in this case, the UCUAA sequence would be in the upstream half of the stem. Other than the UCUAA element, there is no significant primary sequence conservation between the COX2 5'-UTLs of S. cerevisiae, S. kluyveri, and K. lactis.

View larger version (9K): [in this window] [in a new window] [Download PPT slide]   Figure 3. A conserved sequence and secondary structure in the 5'-untranslated leaders of COX2 mRNAs from S. cerevisiae, S. kluyveri, and K. lactis. Lines represent 5'-UTLs, while boxes represent the 5' ends of the COX2 coding sequences. Distances between the coding sequence and the predicted hairpin loop (19 nt in S. cerevisiae, 20 nt in S. kluyveri, 29 nt in K. lactis) are drawn to scale. Thick black line indicates the location of UCUAA (HARDY and CLARK-WALKER 1990 ) in a predicted stem loop (DUNSTAN et al. 1997 ) in each 5'-UTL. Asterisk at the 5' end of S. cerevisiae and S. kluyveri 5'-UTLs indicates that they are known or predicted primary transcripts; the 5' end of the K. lactis COX2 mRNA is predicted to arise from a processing event.

The S. kluyveri COX3 gene is apparently transcribed from a perfect match to the consensus mitochondrial promoter sequence at -331 to -323. A tRNA^Val gene is located at positions -289 to -217. If the tRNA^Val and COX3 are cotranscribed and processed as in S. cerevisiae, then the COX3 5'-UTL would be 216 nt in length. Upstream of K. lactis COX3 there is a consensus promoter sequence at -147 to -139, suggesting that its COX3 5'-UTL would be 140 nt in length. We could find no significant similarities between the COX3 5'-UTL sequences of S. cerevisiae, S. kluyveri, and K. lactis as detectable by analysis with the BLAST or MegAlign programs (MATERIALS AND METHODS).

Mutations in homologous genes of S. kluyveri and K. lactis demonstrate orthologous mRNA-specific translational activator functions:
Null mutations in S. kluyveri PET111 and PET122 and in K. lactis PET111 cause nonrespiratory phenotypes: The highly diverged PET111 and PET122 homologs isolated from S. kluyveri and K. lactis failed to complement the corresponding mutations in S. cerevisiae, raising the question of whether they were truly orthologous in function. To answer this question, we constructed null mutations in these genetically tractable yeast species by deleting substantial portions of each coding region from their chromosomal DNA and replacing them with either the S. cerevisiae URA3 gene or a kanamycin resistance cassette (MATERIALS AND METHODS). These null mutations in S. kluyveri PET111, K. lactis PET111, or S. kluyveri PET122 all prevented respiratory growth on nonfermentable carbon sources. In each case, transformation with the corresponding S. cerevisiae gene failed to restore respiratory growth of the null mutant (not shown).

S. kluyveri pet111 and pet122 mutants and a K. lactis pet111 mutant are specifically deficient in cytochrome c oxidase: To test whether the S. kluyveri pet111 and pet122 mutants affected cytochrome c oxidase we compared their whole-cell cytochrome absorption spectra to that of an isogenic wild-type S. kluyveri strain, as well as to the corresponding S. cerevisiae strains (Fig 4). The cytochrome aa₃ peak, corresponding to cytochrome c oxidase, was missing from the absorption spectrum of each mutant strain, while the other cytochrome peaks were unaffected. We have obtained a similar result for a K. lactis pet111 null mutant (N. BONNEFOY and G. D. CLARK-WALKER, unpublished results). This indicates that the respiratory defects of the S. kluyveri pet111 and pet122 mutant strains and of the K. lactis pet111 mutant are due to a specific cytochrome c oxidase deficiency, as is the case in S. cerevisiae.

View larger version (20K): [in this window] [in a new window] [Download PPT slide]   Figure 4. Whole-cell absorption spectra of S. cerevisiae and S. kluyveri wild-type and pet111 or pet122 mutants. Low temperature absorption spectra of galactose-grown cells were recorded as described by CLAISSE et al. 1970 . The absorption maximum expected for the alpha band of cytochromes aa3 at 602 nm is indicated. The absorbance scale is indicated at the top right of each panel. The strains used were the following: S. cerevisiae: wild type, DL1; pet122, GW226; pet111, NB39-5D. S. kluyveri: wild type, GRY1175; pet122, MCC328; pet111, NB180.

pet111 and pet122 null mutations in S. kluyveri and K. lactis block translation of the COX2 and COX3 mRNAs, respectively: The fact that pet111 and pet122 null mutations blocked respiration in S. kluyveri and K. lactis and specifically affected cytochrome c oxidase in S. kluyveri suggested that they were orthologous to their S. cerevisiae counterparts. To test this possibility directly, we examined expression of the mitochondrial COX2 and COX3 genes in the mutant strains.

Accumulation of Cox2p and Cox3p in mitochondria from the S. kluyveri mutant strains was assayed by Western blotting using monoclonal antibodies against the S. cerevisiae proteins (Fig 5A). Cox2p and Cox3p from wild-type S. kluyveri mitochondria had approximately the same SDS-gel mobility as the corresponding S. cerevisiae proteins and cross-reacted well with the antibodies. The S. kluyveri pet111 null mutant strain completely lacked Cox2p and had significantly reduced levels of Cox3p, while the S. kluyveri pet122 mutant strain completely lacked Cox3p and had significantly reduced levels of Cox2p (Fig 5A). Thus the pet111 and pet122 null mutations in S. kluyveri specifically blocked accumulation of Cox2p and Cox3p, respectively, as they do in S. cerevisiae. The reduced accumulation of Cox3p in the pet111 mutant and of Cox2p in the pet122 mutant is probably due to degradation of the unassembled subunits. S. cerevisiae Cox2p and Cox3p are known to be unstable when cytochrome c oxidase assembly is blocked (PEARCE and SHERMAN 1995 ; GLERUM and TZAGOLOFF 1997 ; LEMAIRE et al. 1998 ).

View larger version (44K): [in this window] [in a new window] [Download PPT slide]   Figure 5. Accumulation of mitochondrial translation products in S. kluyveri and K. lactis mutant strains. (A) A total of 20 µg per lane of mitochondrial protein from the indicated S. cerevisiae and S. kluyveri strains (MATERIALS AND METHODS) were subjected to electrophoresis on a 16% SDS-polyacrylamide gel and Western blotted. The blot was probed with mouse monoclonal antibodies against S. cerevisiae Cox2p and Cox3p (MATERIALS AND METHODS). Lane 1, wild-type S. cerevisiae (strain PTY11); lane 2, wild-type S. kluyveri (strain GRY1175); lane 3, pet111 mutant S. kluyveri (strain NB180); lane 4, pet122 mutant S. kluyveri (strain MCC328). (B) Left, 30 µg per lane of mitochondrial protein from the indicated K. lactis strains (MATERIALS AND METHODS) were subjected to electrophoresis on a 12% SDS-polyacrylamide gel and blotted. The blot was probed with rabbit polyclonal antisera against S. cerevisiae Cox1p (MATERIALS AND METHODS). Right, 10 µg per lane of the indicated mitochondrial proteins were treated as above and probed with mouse monoclonal antibody against S. cerevisiae Cox2p. Lane 1, wild-type K. lactis (strain KB101); lane 2, pet111 mutant K. lactis (strain CW64-1C).

The phenotype of the K. lactis pet111 mutant was examined by Western blotting of mitochondrial proteins with antibodies against S. cerevisiae Cox1p and Cox2p (Fig 5B). (The available antibody against S. cerevisiae Cox3p failed to cross-react with K. lactis Cox3p.) K. lactis Cox1p and Cox2p comigrated with the homologous S. cerevisiae proteins (data not shown). The K. lactis pet111 mutant mitochondria had no detectable Cox2p, but contained Cox1p, albeit at a reduced level relative to wild type. Thus, the K. lactis pet111 mutation appears to specifically block accumulation of Cox2p. The reduced Cox1p level in the pet111 mutant strain is presumably due to instability of the unassembled protein.

To test whether the S. kluyveri pet111 and pet122 mutations affected synthesis of Cox2p and Cox3p, we examined mitochondrial translation products labeled in vivo in the presence of cycloheximide by SDS-gel electrophoresis and autoradiography (Fig 6). Unfortunately, the pattern of mitochondrial translation products in wild-type S. kluyveri (Fig 6B) was neither as reproducible nor as clear as that in S. cerevisiae (Fig 6A). To positively identify bands corresponding to S. kluyveri Cox2p and Cox3p, we immunoprecipitated each protein from in vivo-labeled mitochondria using monoclonal anti-Cox2p and anti-Cox3p antibodies before gel electrophoresis (Fig 6B). This analysis showed that synthesis of Cox2p was greatly reduced in the S. kluyveri pet111 mutant while synthesis of Cox3p was nearly normal. Conversely, no synthesis of Cox3p was detectable in the pet122 mutant but Cox2p was synthesized at wild-type levels (Fig 6B). These data suggested that S. kluyveri Pet111p and Pet122p are required for translation of Cox2p and Cox3p, respectively. This analysis could not be carried out in K. lactis since it is resistant to cycloheximide (DEHOUX et al. 1993 ).

View larger version (48K): [in this window] [in a new window] [Download PPT slide]   Figure 6. Effects of S. kluyveri pet111 and pet122 mutations on mitochondrial protein synthesis. Mitochondrial translation products were 35S-labeled in vivo in the presence of cycloheximide. Portions of each sample were solubilized in SDS and immunoprecipitated; immunoprecipitates and total labeled proteins were subjected to electrophoresis on SDS-16% polyacrylamide glycerol-containing gels and autoradiographed (see MATERIALS AND METHODS). (A) In vivo-labeled total mitochondrial translation products from S. cerevisiae wild type (wt; strain DL1), pet111 mutant (strain NB39-5D), and pet122 mutant (strain TWM10-41). (B) In vivo-labeled total S. kluyveri mitochondrial translation products (total) or immunoprecipitated S. kluyveri mitochondrial translation products (IP). Immunoprecipitations were performed with no serum (––), with mouse monoclonal antibody against S. cerevisiae Cox2p (Cox2p), or with mouse monoclonal antibody against S. cerevisiae Cox3p (Cox3p). S. kluyveri strains used were the following: GRY1175, wild type; NB180, pet111 mutant; and MCC328, pet122 mutant.

Finally, we compared steady-state levels of the COX2 and COX3 mRNAs in wild-type S. kluyveri to the pet111 and pet122 mutants by Northern hybridization (Fig 7). The mRNA levels were normalized to levels of the mitochondrial 15S ribosomal RNA. In the pet111 mutant, both the COX2 and COX3 mRNA levels were reduced to ~50% of wild type. In the pet122 mutant, the COX2 mRNA level was slightly reduced, while the COX3 mRNA level was reduced to ~7% of wild type. These results are consistent with a role for S. kluyveri Pet111p and Pet122p in translational activation. They apparently also affect mRNA stability. In S. cerevisiae, pet111 mutations reduce the levels of COX2 mRNA to 10–30% of wild-type levels (POUTRE and FOX 1987 ), while pet122 mutations reduce the level of the COX3 mRNA to ~50% of wild type (WIESENBERGER and FOX 1997 ).

View larger version (41K): [in this window] [in a new window] [Download PPT slide]   Figure 7. Abundance of the mitochondrial COX2 and COX3 mRNAs in S. kluyveri translational activator mutant strains. Total RNA from S. kluyveri wild type (wt, GRY1175), pet111 mutant (NB180), and pet122 mutant (MCC328) was subjected to electrophoresis, blotted to a membrane, and probed with radioactively labeled S. kluyveri COX2 or COX3 genes or with the S. cerevisiae mitochondrial 15S rRNA gene (see MATERIALS AND METHODS). Table at bottom shows the level of each mRNA, normalized to the level of 15S rRNA, in the mutant strains.

	DISCUSSION

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Mitochondrial gene expression systems have diverged to an extraordinary degree during the evolution of eukaryotes (GRAY et al. 1999 ). In this study, we have begun to explore the extent to which the mRNA-specific translational activation system uncovered in S. cerevisiae may be conserved in other organisms. We found that among budding yeasts, both the nuclearly coded activator proteins and their mitochondrially coded mRNA 5'-UTL targets have diverged extensively in sequence. However, the correspondences between activator protein orthologs and the mitochondrial translation products they control have been conserved.

Among S. bayanus and S. kluyveri, an assessment of divergence from S. cerevisiae is hampered by the fact that only a small set of gene sequences is available (<20 complete sequences at present), and it is biased in favor of those that function in S. cerevisiae. Nevertheless, the orthologs from S. bayanus and S. kluyveri are all more diverged from the S. cerevisiae translational activators than any other known protein from those species. (No protein-coding sequences from the nuclear genome of S. servazzii are currently available, except those reported in this article.)

A set of K. lactis random partial sequences representing 296 genes has recently been generated and compared to S. cerevisiae (OZIER-KALOGEROPOULOS et al. 1998 ). Overall, the sequences in the random set have a mean amino acid identity of 63.6% to their S. cerevisiae counterparts (OZIER-KALOGEROPOULOS et al. 1998 ), considerably higher than the 20% identity observed between K. lactis and S. cerevisiae Pet111p. Eight of the genes identified by Ozier-Kalogeropoulos et al. encode proteins with similarity to S. cerevisiae proteins known to be involved in mitochondrial biogenesis, and the amino acid identity of these sequence fragments to the S. cerevisiae proteins ranges from 77 to 42%. Interestingly, the least conserved of these genes encodes a potential Pet54p ortholog, with 42% identity over 57 residues to S. cerevisiae Pet54p (OZIER-KALOGEROPOULOS et al. 1998 ). Thus, mRNA-specific mitochondrial translational activators appear to be among the most rapidly diverging proteins in budding yeasts, despite their conserved functions.

In some cases, rapid protein divergence driven by positive Darwinian selection has been detectable through an elevated ratio of missense to silent nucleotide substitutions (WHITFIELD et al. 1993 ; LEE et al. 1995 ; SUTTON and WILKINSON 1997 ). Much more often, however, rapid divergence appears to reflect the absence of positive or negative selection. In Drosophila species, a large number of proteins exhibit very rapid neutral divergence (SCHMID and TAUTZ 1997 ). Our analysis of substitutions in the genes reported here is consistent with rapid neutral evolution of the translational activator proteins. In this context it is interesting to note that even Pet111p and Pet122p amino acids that can mutate to suppress mutations in their target mRNA 5'-UTLs or in other proteins in the S. cerevisiae genetic system (COSTANZO and FOX 1993 ; MULERO and FOX 1993A ; BROWN et al. 1994 ) are not highly conserved in the other species.

The mitochondrially coded targets of the mRNA-specific translational activators in the COX2 and COX3 mRNA 5'-UTLs also appear to have evolved rapidly. This contrasts sharply with the highly conserved mitochondrially coded protein sequences, which are known to evolve more slowly in yeasts than the nuclearly encoded cytochrome c (CLARK-WALKER 1991 ). Even the lengths (known or predicted) of the 5'-UTLs are quite different from each other: the COX2 mRNA 5'-UTLs in S. cerevisiae, S. kluyveri, and K. lactis are 54, 88, and 226 nt, respectively, while the COX3 mRNA 5'-UTLs in the same species are 613, 216, and 140 nt. The only significant conserved sequence element among any of the 5'-UTLs is the UCUAA pentanucleotide previously noted in the COX2 5'-UTLs of S. cerevisiae, K. lactis, Hansenula saturnus, and Torulopsis (Candida) glabrata (HARDY and CLARK-WALKER 1990 ); we found the same element in S. kluyveri. However, while it is clear that other sequences in the COX2 5'-UTL are important functionally in S. cerevisiae (DUNSTAN et al. 1997 ), these are not recognizably conserved in the other yeasts.

The correspondences of translational activator proteins to their target mRNAs are conserved in the species studied here. The phenotypes of S. kluyveri pet111 and pet122 null mutants and of a K. lactis pet111 null mutant show that these genes are truly orthologous to their S. cerevisiae counterparts. The weak functional complementation of the S. cerevisiae pet122 mutation by S. servazzii PET122 indicates conservation of orthologs in that species as well. Thus, the proteins and their RNA targets are coevolving.

Experimentally, the activator dependence of mitochondrial mRNAs can be altered without eliminating translation by in vivo expression of chimeric mRNAs containing 5'-UTLs and coding sequences derived from different genes (MULLER et al. 1984 ; COSTANZO and FOX 1986 , COSTANZO and FOX 1988 ; POUTRE and FOX 1987 ; RODEL and FOX 1987 ; MULERO and FOX 1993B ; MANTHEY and MCEWEN 1995 ). The conservation of the mRNA specificity during the evolution of budding yeasts—despite this fact and despite the sequence divergence of the system's components—suggests that the mRNA specificity has adaptive value in promoting efficient synthesis of respiratory complexes. The untranslated regions of the COX2 and COX3 mRNAs play a role in correctly localizing translation of Cox2p and Cox3p (SANCHIRICO et al. 1998 ). Thus, it seems plausible that the membrane-bound mRNA-specific translational activators aid in assembly of the respiratory chain by topologically ordering the synthesis of key components on the inner membrane surface.

mRNA-specific translational activation may be a general feature of fungal mitochondrial gene expression. The nuclearly coded Neurospora crassa gene cya-5 is required at a post-transcriptional step in mitochondrial expression of Cox1p (NARGANG et al. 1978 ; COFFIN et al. 1997 ). Interestingly, its protein product has significant sequence similarity to Pet309p, a membrane-bound protein specifically required for COX1 mRNA translation in S. cerevisiae (MANTHEY and MCEWEN 1995 ; MANTHEY et al. 1998 ), strongly suggesting that the CYA-5 protein and Pet309p are orthologous (COFFIN et al. 1997 ). Possible PET309 orthologs are also present in the currently available genomic sequences of C. albicans and Schizosaccharomyces pombe, two fungi for which nearly complete sequence data are available. Interestingly, the soluble maize chloroplast translational activator protein CRP1 exhibits sequence similarity to Pet309p (FISK et al. 1999 ). There are no significant matches to Pet111p, Pet54p, Pet122p, or Pet494p in the available C. albicans and S. pombe sequences (see MATERIALS AND METHODS). However, given the high divergence we find for these proteins among more closely related species, it seems possible that orthologs are present in those species but are difficult to detect by sequence similarity alone.

There is little information on how translation initiation occurs in animal systems. Animal mitochondrial mRNAs typically lack 5'-UTLs (ATTARDI and SCHATZ 1988 ). Thus, if translational activation occurs in animal mitochondria, there must be significant differences in mechanism compared to fungi. There are no predicted sequences significantly similar to yeast translational activators in the C. elegans genome (MATERIALS AND METHODS; CONSORTIUM 1998 ), except for similarity between Pet54p and the fox-1 gene product (HODGKIN et al. 1994 ), which is confined to the RRM domain. Thus, mitochondrial translational activators could represent a possible target for antifungal drugs.

Estimates of the conservation of gene order between K. lactis and S. cerevisiae range from 50% (OZIER-KALOGEROPOULOS et al. 1998 ) to ~74% (KEOGH et al. 1998 ). While there are not enough data available from S. servazzii or S. kluyveri to make accurate predictions, our results suggest that the approach of cloning conserved genes syntenic to highly diverged genes of interest (KUWABARA and SHAH 1994 ; AGNAN et al. 1997 ) will often be fruitful for genes of these yeasts.

	FOOTNOTES

¹ Present address: Proteome, Inc., Beverly, MA 01915.
² Present address: Centre de Génétique Moléculaire, Laboratoire propre du CNRS associé à l'Université Pierre et Marie Curie, 91198 Gif-sur-Yvette Cedex, France.

	ACKNOWLEDGMENTS

We thank J. Piskur for performing direct sequencing of S. kluyveri mitochondrial DNA and D. Pillai for isolation of the K. lactis COX3 clone. We also thank M. E. Cusick for advice on RRM domains, K. J. Schmid for help with analysis of synonymous and nonsynonymous substitutions, X. J. Chen for helpful discussions concerning K. lactis disruptions, L.-J. Ouyang for skilled technical assistance, and G. Weiller for help with analysis of phylogenetic relationships. Finally, we thank T. Mason for the anti-Cox2p antibody, G. Schatz for the anti-Cox1p antibody, L. Grivell for the K. lactis genomic library, and C. P. Kurtzman, J. E. McEwen, J. Sloan, and J. Strathern for strains. This work was supported by the U.S. National Institutes of Health grant (GM-29362) to T.D.F. and by a Human Frontier Science Program Organization Fellowship (LT22/96) to N.B.

Manuscript received September 13, 1999; Accepted for publication November 1, 1999.

	LITERATURE CITED

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

AGNAN, J., C. KORCH, and C. SELITRENNIKOFF, 1997  Cloning heterologous genes: problems and approaches. Fungal Genet. Biol. 21:292-301[Medline].

ALANI, E., L. CAO, and N. KLECKNER, 1987  A method for gene disruption that allows repeated use of URA3 selection in the construction of multiply disrupted yeast strains. Genetics 116:541-545[Abstract/Free Full Text].

ALTAMURA, N., N. CAPITANIO, N. BONNEFOY, S. PAPA, and G. DUJARDIN, 1996  The Saccharomyces cerevisiae OXA1 gene is required for the correct assembly of cytochrome c oxidase and oligomycin-sensitive ATP synthase. FEBS Lett. 382:111-115[Medline].

ALTSCHUL, S. F., W. GISH, W. MILLER, E. W. MYERS, and D. J. LIPMAN, 1990  Basic local alignment search tool. J. Mol. Biol. 215:403-410[Medline].

ATTARDI, G. and G. SCHATZ, 1988  Biogenesis of mitochondria. Annu. Rev. Cell Biol. 4:289-333.

AUSUBEL, F. M., R. BRENT, R. E. KINGSTON, D. D. MOORE, J. G. SEIDMAN et al., 1993 Current Protocols in Molecular Biology. John Wiley & Sons, New York.

BAIROCH, A., P. BUCHER, and K. HOFMANN, 1997  The PROSITE database, its status in 1997. Nucleic Acids Res. 25:217-221[Abstract/Free Full Text].

BARNETT, J., R. PAYNE and D. YARROW, 1990 Yeasts: Characteristics and Identification. Cambridge University Press, Cambridge, United Kingdom.

BISWAS, T. K., 1998  Usage of non-canonical promoter sequence by the yeast mitochondrial RNA polymerase. Gene 212:305-314[Medline].

BONNEFOY, N., F. CHALVET, P. HAMEL, P. P. SLONIMSKI, and G. DUJARDIN, 1994a  OXA1, a Saccharomyces cerevisiae nuclear gene whose sequence is conserved from prokaryotes to eukaryotes controls cytochrome oxidase biogenesis. J. Mol. Biol. 239:201-212[Medline].

BONNEFOY, N., M. KERMORGANT, O. GROUDINSKY, M. MINET, and P. P. SLONIMSKI et al., 1994b  Cloning of a human gene involved in cytochrome oxidase assembly by functional complementation of an oxa1^- mutation in Saccharomyces cerevisiae.. Proc. Natl. Acad. Sci. USA 91:11978-11982[Abstract/Free Full Text].

BROWN, N. G., 1994 Interactions among PET54, PET122 and PET494, three nuclearly encoded translational activators of the mitochondrial COX3 gene in Saccharomyces cerevisiae and the search for other translational components that interact with them. Ph.D. Thesis, Cornell University, Ithaca, NY.

BROWN, N. G., M. C. COSTANZO, and T. D. FOX, 1994  Interactions among three proteins that specifically activate translation of the mitochondrial COX3 mRNA in Saccharomyces cerevisiae.. Mol. Cell. Biol. 14:1045-1053[Abstract/Free Full Text].

BURD, C. G. and G. DREYFUSS, 1994  Conserved structures and diversity of functions of RNA-binding proteins. Science 265:615-621[Abstract/Free Full Text].

CALDER, K. and J. MCEWEN, 1991  Deletion of the COX7 gene in Saccharomyces cerevisiae reveals a role for cytochrome c oxidase subunit VII in assembly of remaining subunits. Mol. Microbiol. 5:1769-1777[Medline].

CAPALDI, R. A., 1990  Structure and function of cytochrome c oxidase. Annu. Rev. Biochem. 59:569-596[Medline].

CHEN, X. J. and G. D. CLARK-WALKER, 1996  The mitochondrial genome integrity gene, MGI1, of Kluyveromyces lactis encodes the beta-subunit of F1-ATPase. Genetics 144:1445-1454[Abstract].

CLAISSE, M. L., G. A. PERE-AUBERT, L. P. CLAVILIER, and P. P. SLONIMSKI, 1970  Méthode d'estimation de la concentratio