Advertisement

Accumulation of mitochondrially synthesized Saccharomyces cerevisiae Cox2p and Cox3p depends on targeting information in untranslated portions of their mRNAs

Marie E. Sanchirico, Thomas D. Fox, Thomas L. Mason

Author Affiliations

  1. Marie E. Sanchirico1,
  2. Thomas D. Fox2 and
  3. Thomas L. Mason*,1
  1. 1 Department of Biochemistry and Molecular Biology and The Graduate Program in Molecular and Cellular Biology, University of Massachusetts, Amherst, MA, 01003‐4505, USA
  2. 2 Section of Genetics and Development, Cornell University, Ithaca, NY, 14853‐2703, USA
  1. *Corresponding author. E-mail: tmason{at}biochem.umass.edu

Abstract

The essential products of the yeast mitochondrial translation system are seven hydrophobic membrane proteins and Var1p, a hydrophilic protein in the small ribosomal subunit. Translation of the membrane proteins depends on nuclearly encoded, mRNA‐specific translational activators that recognize the 5′‐untranslated leaders of their target mRNAs. These translational activators are themselves membrane associated and could therefore tether translation to the inner membrane. In this study, we tested whether chimeric mRNAs with the untranslated sequences normally present on the mRNA encoding soluble Var1p, can direct functional expression of coding sequences specifying the integral membrane proteins Cox2p and Cox3p. DNA sequences specifying these chimeric mRNAs were inserted into mtDNA at the VAR1 locus and expressed in strains containing a nuclearly localized plasmid that supplies a functional form of Var1p, imported from the cytoplasm. Although cells expressing these chimeric mRNAs actively synthesized both membrane proteins, they were severely deficient in cytochrome c oxidase activity and in the accumulation of Cox2p and Cox3p, respectively. These data strongly support the physiological importance of interactions between membrane‐bound mRNA‐specific translational activators and the native 5′‐untranslated leaders of the COX2 and COX3 mRNAs for localizing productive synthesis of Cox2p and Cox3p to the inner membrane.

Introduction

The targeting of membrane proteins synthesized in the well‐studied bacterial and eukaryotic cytoplasmic systems has typically been found to depend on signals contained within the proteins themselves or their precursors (Na et al., 1992; Corsi and Schekman, 1996; Rapoport et al., 1996; Schatz and Dobberstein, 1996; Wickner and Leonard, 1996; Neupert, 1997; Saavedra et al., 1997). Within mitochondria, the signals and mechanisms that target membrane proteins encoded by mtDNA have not been studied directly, owing to the lack of in vitro protein synthesis systems derived from the organelles.

Genetic analysis in Saccharomyces cerevisiae has demonstrated that translation of several mitochondrially coded mRNAs depends on membrane‐bound mRNA‐specific translational activators that recognize the 5′‐untranslated leaders (UTLs) of their target mRNAs (Fox, 1996). These activator proteins, encoded by nuclear genes, appear to mediate the productive interaction between the mRNAs and mitochondrial ribosomes at the surface of the inner membrane. They are required for the translation of both normal mitochondrial mRNAs (Fox, 1996), and chimeric mRNAs specifying the soluble mitochondrial reporter‐protein Arg8mp (Steele et al., 1996; N.S.Green‐Willms and T.D.Fox, unpublished data). Taken together, these studies have suggested that membrane insertion of mitochondrially coded proteins such as cytochrome c oxidase subunits II and III (Cox2p and Cox3p), as well as cytochrome b, could depend on this system to tether mitochondrial translational initiation complexes to the membrane (Costanzo and Fox, 1990; Maleszka et al., 1991; Michaelis et al., 1991; McMullin and Fox, 1993; Fox, 1996). Such a system could prevent mislocalized translation of proteins destined for the inner membrane. However, unlike the analogous SRP/SRP‐receptor system of eukaryotic cytoplasms (Walter and Lingappa, 1986; Ng and Walter, 1994), this model has the interesting and novel implication that information contained in the yeast mitochondrial mRNA 5′‐UTLs should be important for membrane localization of the proteins encoded by those mRNAs.

There are eight major translation products specified by yeast mtDNA, seven of which are hydrophobic subunits of energy‐transducing enzyme complexes located in the inner mitochondrial membrane (Tzagoloff and Myers, 1986). However, one major yeast mitochondrial gene product, termed Var1p, is a hydrophilic ribosomal protein in the mitochondrial small ribosomal subunit (Groot et al., 1979; Terpstra and Butow, 1979; Terpstra et al., 1979; Hudspeth et al., 1982). Whereas nothing is presently known about what, if any, specific factors may be required for activation of VAR1 mRNA translation, it is clear from the exceptional nature of the polypeptide and its ultimate destination, that its synthesis should not depend upon a mechanism devoted to membrane protein insertion. Thus, the 5′‐UTL of the VAR1 mRNA should not contain information for the localized translation of membrane proteins.

In this study, we have taken advantage of our ability to relocate synthetic yeast mitochondrial and nuclear genes (Sanchirico et al., 1995; Steele et al., 1996) to ask, in vivo, whether any targeting information for mitochondrially coded membrane proteins is contained in the untranslated portions of their mRNAs. By inserting the COX2 and COX3 coding sequences into chimeric mRNAs containing the 5′‐ and 3′‐untranslated regions (UTRs) of the VAR1 mRNA, we tested whether the VAR1 untranslated sequences, normally present on a ribosomal protein mRNA, can direct functional expression of coding sequences specifying the integral membrane proteins Cox2p and Cox3p. DNA sequences specifying the chimeric mRNAs were inserted into mtDNA and expressed in strains containing a nuclearly localized plasmid that supplies, in trans, a functional form of Var1p imported from the cytoplasm (Sanchirico et al., 1995). Our results demonstrate that although both membrane proteins were translated from the chimeric mRNAs, their incorporation into active cytochrome c oxidase complexes was severely defective. These data strongly support the physiological importance of interactions between membrane‐bound, mRNA‐specific translational activators and the native 5′‐UTLs of the COX2 and COX3 mRNAs for localizing productive synthesis of Cox2p and Cox3p to the inner membrane.

Results

Mitochondrial transformation and integration of the var1::ARG8m, var1::COX2 and var1::COX3 chimeric genes into the mitochondrial genome

The var1::ARG8m, var1::COX2 and var1::COX3 genes were designed to place the functional expression of the synthetic mitochondrial reporter gene ARG8m (Steele et al., 1996), or COX2 or COX3, respectively, under VAR1 transcriptional and translational control (Figure 1; Materials and methods). It is important to note that the var1::COX2 and var1::COX3 chimeric genes specify wild‐type Cox2p and Cox3p, respectively. As a first step towards integrating these constructs into mtDNA, synthetic rho strains, each carrying the var1::ARG8m, var1::COX2 or the var1::COX3 chimeric genes, were obtained by mitochondrial transformation of the rho0 strain DFS160 with plasmids pMES7, pMES18 or pMES21, respectively (Materials and methods). Mitochondrial transformants carrying the var1::ARG8m gene were identified by mating to the arg8 rho+ strain DFS188 containing the VAR1u expression vector, pEVA1, and selecting for Arg+ diploids. Only synthetic rho mitochondrial transformants carrying the var1::ARG8m sequence are capable of forming Arg+ diploids in this mating assay. Mitochondrial transformants carrying either var1::COX2 or var1::COX3 were identified by mating to the cox2 mutant strain TF192 or the cox3 mutant strain LSF5, respectively. Only synthetic rho mitochondrial transformants carrying COX2 or COX3 sequences are capable of forming respiratory competent (Pet+) diploids in these mating assays. Stable synthetic rho strains were isolated through repeated rounds of subcloning and testing in the mating assays.

Figure 1.

Schematic representation of the mitochondrial DNA from strains expressing the coding sequences of ARG8m, COX2 or COX3 from VAR1. The coding sequences are indicated as follows: VAR1 (black box), COX2 (dark grey box), COX3 (light grey box), ARG8m (open box). The mature VAR1 transcript begins at bp −162 (black circle) (Smooker et al., 1988) and ends at the conserved dodecamer sequence (small open box). The var1::ARG8m was constructed such that the first 17 codons from VAR1, followed by three codons generated by insertion of an AccI site, are fused in‐frame at the start codon for the ARG8m structural gene. The var1::COX2 and var1::COX3 5′‐end gene fusions occur exactly at the VAR1 ATG codon, and the fusion junctions at the 3′ ends of the COX2 and COX3 ORFs immediately downstream of the normal COX2 and COX3 stop codons are out of frame with respect to the 166‐bp segment of the VAR1 ORF (see Methods and materials). The arrows in the var1::ARG8m, var1::COX2 and var1::COX3 chimeric genes indicate the positions of the termination codons used in the respective ORFs. Thus, the Cox2p and Cox3p proteins expressed from the chimeric genes are identical to wild‐type. The construction of the cox3::ARG8m gene fusion was described previously (Steele et al., 1996). The cox2::ARG8m construct is a precise fusion of the ARG8m initiation codon to the corresponding position of COX2: it contains no COX2‐coding sequence (H.M. Dunstan and T.D.Fox, unpublished data).

Next, we isolated a rho+ diploid strain in which the VAR1 coding sequence was replaced by ARG8m. To do this, the synthetic rho transformant MSY362 (Table I) carrying the var1::ARG8m construct was crossed to the rho+ strain DFS188 (Table I), which contained the VAR1u expression plasmid pEVA1 in its nucleus, such that mitochondrial translation could be maintained in the absence of the mitochondrial VAR1 gene. As expected, this cross yielded a respiratory competent Arg+ diploid, MSY573 (Table I), indicating that the ARG8m gene was expressed functionally from the var1::ARG8m chimeric locus and that VAR1u covered the function of the deleted VAR1 coding sequence. Both respiratory growth and Arg+ prototrophy were lost in mitotic segregants of MSY573 that lacked the nuclear VAR1u expression plasmid pEVA1, confirming that the mitochondrial VAR1 gene had been inactivated. Southern analysis confirmed the replacement of VAR1 by var1::ARG8m in the mtDNA of this diploid (data not shown).

View this table:
Table 1. Strains used in this study

To study expression of COX2 from the VAR1 locus, we generated a diploid in which VAR1 was replaced by var1::COX2, and COX2 was replaced by cox2::ARG8m. This diploid, MSY575 (Table I), was made by mating the synthetic rho transformant MSY493 (Table I) carrying the var1::COX2 construct, with the rho+, cox2::ARG8m strain HMD22 (Table I), carrying the VAR1u expression plasmid. It was capable of slow respiratory growth, indicating that the var1::COX2 construct was partially functional and Arg+. Replacement of VAR1 by var1::COX2 was confirmed by the dependence of both Arg+ prototrophy and slow respiratory growth in the presence of the VAR1u expression plasmid, and by Southern analysis of mtDNA (not shown).

A similar diploid, MSY577 (Table I), in which VAR1 was replaced by var1::COX3, and COX3 was replaced by cox3::ARG8m, was generated by mating the synthetic rho transformant MSY523 (Table I) with the rho+, cox3::ARG8m strain DFS189 (Table I), carrying the VAR1u expression plasmid. This diploid also exhibited slow respiratory growth and Arg+ prototrophy, both of which were VAR1u‐dependent. Southern analysis confirmed the replacement of VAR1 by var1::COX3 (not shown).

Synthesis of Cox2p and Cox3p

As a prerequisite to asking whether Cox2p or Cox3p could be stably incorporated into the inner membrane when translated from chimeric mRNAs bearing the VAR1 5′‐ and 3′‐UTRs, we first determined whether these chimeric mRNAs were efficiently translated in vivo. To do this, we compared the pulse‐labeling of mitochondrial gene products in the var1::ARG8m strain (MSY573), containing wild‐type COX2 and COX3 loci, with the var1::COX2 strain (MSY575) and the var1::COX3 strain (MSY577). Cells were grown to mid‐exponential phase in complete medium containing galactose (YPGal), and mitochondrial translation products were pulse‐labeled for 5 and 10 min in vivo with [35S]methionine, following the addition of cycloheximide to prevent cytoplasmic protein synthesis (Materials and methods). The labeled mitochondrial translation products were analyzed by SDS–PAGE of samples containing equal amounts (80 μg) of mitochondrial protein, followed by phosphoimager scanning of the dried gel (Figure 2). This experiment demonstrated that the var1::COX2 chimeric mRNA was translated at roughly the same rate as the wild‐type COX2 mRNA. Translation of the var1::COX3 mRNA was also robust, but appeared to be lower than that of the wild‐type COX3 mRNA by ∼3‐fold, as determined by phosphoimager quantitation of labeling after 5 min. However, there was little increase in Cox3p labeling after 5 min, suggesting that the protein translated from the var1::COX3 mRNA could be highly unstable. In any event, the untranslated regions of the VAR1 mRNA are clearly capable of directing translation of sequences coding the membrane proteins Cox2p and Cox3p.

Figure 2.

Cox2p and Cox3p are labeled efficiently in vivo during short pulses when translated from var1::COX2 and var1::COX3 chimeric mRNAs, respectively. Strains MSY573 (var1::ARG8m), MSY575 (var1::COX2) and MSY577 (var1::COX3) were grown to mid‐exponential phase in YPGal medium at 30°C. Mitochondrial translation products were pulse‐labeled in vivo with [35S]methionine in the presence of cycloheximide. Cells were labeled during pulses of 5 and 10 min, as indicated. Mitochondrial proteins (80 μg per lane) were resolved by SDS–PAGE in a 12.5% gel, and the dried gel was exposed to a phosphoimager screen. The mitochondrially encoded polypeptides are indicated on the right of the computer image from the phosphoimager scan.

The synthesis of mature Cox2p requires N‐terminal cleavage of a larger precursor by the IMP protease complex (reviewed in He and Fox, 1997). This processing step requires translocation of the N‐terminus of pre‐Cox2p across the inner membrane. It is interesting, therefore, that radiolabeled Cox2p has the same electrophoretic mobility in each of the three strains examined in Figure 2, indicating that Cox2p expressed in the var1::COX2 strain (MSY575) is apparently processed correctly. This result suggests that translation of pre‐Cox2p from the var1::COX2 chimeric mRNA does not prevent transport of its N‐terminus through the inner membrane.

Another interesting aspect of the results shown in Figure 2 is that newly synthesized Arg8mp is only observed in the var1::COX3 strain. All of the strains examined in Figure 2 are phenotypically Arg+ and therefore express the ARG8m reporter gene, and although longer exposures show the presence of labeled Arg8mp in all three strains, there is clearly differential labeling of Arg8mp translated from the three different chimeric mRNA species. At present, we do not understand the basis for these differences, but it may be relevant that the N‐terminal residues of the altered pre‐Arg8p primary translation product are not identical (Figure 1; Methods and materials). Whereas matrix processing of these pre‐proteins should yield the same mature product, it could be that either synthesis of the primary translation products or the kinetics of their processing and stability are different.

Respiratory phenotypes of var1::COX2 and var1::COX3 strains

Next, we examined the ability of strains expressing Cox2p or Cox3p from coding sequences located at the VAR1 locus to grow on nonfermentable carbon sources. When streaked on medium containing only the carbon sources glycerol and ethanol, the strains bearing either var1::COX2 or var1::COX3 exhibited a slow‐growth phenotype relative to the control strain bearing var1::ARG8m and wild‐type COX2 and COX3 genes (Figure 3). Whereas these slow‐growth phenotypes indicated that functional expression of the var1::COX2 and var1::COX3 chimeric genes was defective relative to wild‐type, they were clearly distinct from the completely nonrespiratory null phenotypes of cox2::ARG8m and cox3::ARG8m control strains. As expected, the slow growth on nonfermentable medium was accompanied by reduced rates of cellular respiration. The polarographically measured rates of cyanide‐sensitive oxygen consumption were ∼85% lower in the var1::COX2 and var1::COX3 strains compared with the var1::ARG8m control strain (Table II).

Figure 3.

Respiratory growth is reduced in strains expressing var1::COX2 and var1::COX3 chimeric mRNAs. Isonuclear diploid strains (Table I) MSY573 [rho+ var1::ARG8m], MSY510 [rho+ cox2::ARG8m], MSY575 [rho+ cox2::ARG8m var1::COX2], MSY509 [rho+ cox3::ARG8m] and MSY577 [rho+ cox3::ARG8m var1::COX3] were streaked on YPGE agar, and incubated for 4 days at 30°C. Also streaked, were strains carrying the indicated chimeric genes heteroplasmically with wild‐type mtDNA, in nuclear backgrounds that demanded expression of the chimeric genes (Costanzo and Fox, 1988; Mulero and Fox, 1993b): JJM194×JJM102 [rho+, rho cox3‐COX2], JJM195×PTH44 [rho+, rho cox2‐COX3] and MCC60R2‐16×MCC62 [rho+, rho cox3‐COB].

View this table:
Table 2. Mutant respiratory activities

Dramatically reduced respiratory growth is not a general consequence of expressing cytochrome c oxidase subunits from chimeric mRNAs. Strains that express these membrane proteins from chimeric genes that specify cox2::COX3 or cox3::COX2 mRNAs, carried in an unstable heteroplasmic state together with wild‐type mtDNA (Mulero and Fox, 1993b), exhibited vigorous respiratory growth, as did a similar strain expressing a cox3::COB mRNA (Costanzo and Fox, 1988) (Figure 3). In each of these cases, the chimeric mRNAs bear the 5′‐UTLs known to interact with membrane‐bound, mRNA‐specific translational activators, in contrast to the var1::COX2 and var1::COX3 mRNAs.

Reduced Cox2p and Cox3p levels, and cytochrome c oxidase activity in mutant mitochondria

To determine the effects of var1::COX2 and var1::COX3 expression on cytochrome c oxidase activity and the accumulation of Cox2p and Cox3p, we performed enzyme activity assays and immunoblot analysis with mitochondria purified from MSY573, MSY575 and MSY577 cells grown in complete medium containing galactose. The specific activities of cytochrome c oxidase in mitochondria isolated from the var1::COX2 (MSY575) and var1::COX3 (MSY577) mutants were, respectively, 12.6 and 7.9% of the activity in the mitochondria from the control strain MSY573 (Table II). Moreover, the quantitative immunoblot analysis shown in Figure 4 indicated that the levels of Cox2p and Cox3p in MSY575 and MSY577 mitochondria are, respectively, ∼5% of the amounts detected in the mitochondria from MSY573. Thus, the respiration‐deficient phenotype of these strains appears to be caused by decreased levels of Cox2p or Cox3p when they are expressed from chimeric mRNAs bearing the untranslated regions of the VAR1 mRNA. Since, as shown above, translation of the var1::COX2 mRNA was essentially normal whereas translation of the var1::COX3 mRNA was reduced only 3‐fold, we conclude that decreased protein stability, not reduced translation of the chimeric mRNAs, is the major factor responsible for the low steady‐state levels of Cox2p and Cox3p in the mutants.

Figure 4.

The steady‐state levels of Cox2p and Cox3p are severely reduced after translation of var1::COX2 and var1::COX3 chimeric mRNAs, respectively. Strains MSY573 (var1::ARG8m), MSY575 (var1::COX2) and MSY577 (var1::COX3) were grown to mid‐logarithmic phase in YPGal at 30°C. Mitochondria were isolated and the indicated amounts of total mitochondrial protein were separated on a SDS–PAGE gel (12.5%), transferred to nitrocellulose, and incubated with monoclonal antibodies to either Cox2p (upper panel) or Cox3p (lower panel). The immune complexes were decorated with [125I]anti‐mouse immunoglobulin and the blots were exposed to a phosphoimager screen. The figure shows the computer‐generated image of the phosphoimager scan. The relative signal intensities per microgram protein were: Cox2p; MSY573, 17.2 ± 1.9, MSY577, 0.7 ± 0.2 and Cox3p; MSY573, 15.7 ± 1.7, MSY577, 0.8 ± 0.3.

Discussion

The key function of mitochondrial genetic systems is to promote respiration in eukaryotic cells by providing a few hydrophobic protein subunits of energy‐transducing complexes in the inner membrane (Attardi and Schatz, 1988). This fact suggests that mitochondrial genetic systems are specialized for the synthesis of integral membrane polypeptides. If so, these organellar systems should possess features that localize translation of proteins on the surface of the inner membrane.

One interesting feature of the S.cerevisiae mitochondrial genetic system is that translation of at least five of the seven mitochondrial mRNAs encoding membrane proteins depends on nuclearly encoded mRNA‐specific translational activator proteins (Fox, 1996). For example, translation of the COX2 mRNA is specifically activated by the Pet111p nuclear gene product (Poutre and Fox, 1987; Mulero and Fox, 1993b), whereas translation of COX3 mRNA is specifically activated by a complex containing the nuclear gene products of Pet54p, Pet122p and Pet494p (Müller et al., 1984; Costanzo and Fox, 1986; Costanzo et al., 1986; Fox et al., 1988a; Brown et al., 1994). In both cases, translation of the mRNAs is limited by the levels of their respective activators (Steele et al., 1996; N.S.Green‐Willms and T.D.Fox, unpublished data). These translational activator proteins interact functionally with sites in the 5′‐UTLs of their target mRNAs (Costanzo and Fox, 1988, 1993; Mulero and Fox, 1993a,b; Wiesenberger et al., 1995; Dunstan et al., 1997). In the case of the COX3 mRNA‐specific activator Pet122p, a functional interaction with mitochondrial ribosomes has also been demonstrated (Haffter et al., 1990, 1991; McMullin et al., 1990). Pet122p is an integral inner mitochondrial membrane protein (McMullin and Fox, 1993; C.A.Butler and T.D.Fox, unpublished data), as is the COX2 mRNA‐specific activator Pet111p (N.S.Green‐Willms and T.D.Fox, unpublished data). Thus, in addition to their regulatory role, these activator proteins could tether translation initiation complexes involving the mRNAs encoding Cox2p and Cox3p to the mitochondrial inner membrane, by virtue of their interactions with the 5′‐untranslated mRNA leaders. This model predicts that targeting information for mitochondrially synthesized proteins would reside, at least in part, in the leaders of the mRNAs that encode them.

The ribosomal protein Var1p is the only major yeast mitochondrial translation product that is not a hydrophobic membrane protein, and it is widely assumed, although not actually demonstrated, that Var1p is translated on free ribosomes in the matrix. Thus, the 5′‐UTL of the VAR1 mRNA is unlikely to interact with a membrane‐bound, mRNA‐specific, translational activator. Based on this premise, we asked whether the untranslated portions of the VAR1 mRNA could promote the functional expression in vivo of COX2 and COX3 coding sequences in chimeric mRNAs. If membrane‐bound, mRNA‐specific translational activators are physiologically important for efficient, productive insertion of Cox2p and Cox3p in the inner membrane, then we expected that phenotypic expression of the var1::COX2 and var1::COX3 chimeric genes would exhibit post‐translational defects.

Pulse‐labeling of cells containing chimeric genes indicated that the rate of Cox2p synthesis directed by the var1::COX2 mRNA was equivalent to that directed by the wild‐type COX2 mRNA. Thus, the VAR1 5′‐UTL and 3′‐UTR were fully competent in translation of this membrane protein. In sharp contrast to the normal rate of Cox2p synthesis in cells containing the var1::COX2 chimeric mRNA, we found that those cells had severely reduced rates of respiration and growth on nonfermentable carbon sources. Furthermore, cytochrome c oxidase activity and the steady‐state level of Cox2p were also roughly an order of magnitude lower than wild‐type. Taken together, these data can be most easily explained by a model (Figure 5) in which the var1::COX2 chimeric mRNA is translated at a location, perhaps in the matrix, that makes it difficult for Cox2p to assemble efficiently into cytochrome c oxidase complexes in the inner membrane. The unassembled Cox2p must be degraded rapidly, a phenomenon that has been observed previously (Dowhan et al., 1985). Apparently, translation of the Cox2p coding sequence under the direction of untranslated regions from an mRNA that normally encodes a soluble protein, causes mislocalized translation leading to a physiological defect. We conclude that untranslated sequences of the wild‐type COX2 mRNA are important for targeting Cox2p for correct insertion in the inner membrane, presumably through the interaction of Pet111p with the COX2 mRNA 5′‐UTL.

Figure 5.

Model for expression of the var1::COX2 and var1::COX3 chimeric mRNAs in yeast mitochondria. The UTRs of the VAR1 mRNA (thick wavy lines) direct mislocalized translational initiation of either the COX2 or COX3 coding sequences, possibly activated by an unknown protein on free ribosomes in the matrix. Only a small proportion of the Cox2p or Cox3p translated in this way is assembled into cytochrome c oxidase in the membrane (dotted arrow), whereas the bulk of these proteins are degraded. Translation of the wild‐type COX2 mRNA is activated at the surface of the inner membrane (IM) by interaction of the activator protein Pet111p (111) with the COX2 mRNA 5′‐UTL (thin wavy line). Membrane‐bound translation of the wild‐type COX3 mRNA is similarly activated by the Pet54p–Pet122p–Pet494p complex (54, 122, 494).

Although our results show that productive synthesis of Cox2p cannot be supported by the 5′‐ and 3′‐UTRs of the VAR1 mRNA, the apparently correct processing of newly synthesized pre‐Cox2p in the var1::COX2 strain (Figure 2) suggests that translation from the chimeric mRNA does not prevent translocation through the inner membrane. If the assumption is correct that translation of the chimeric mRNAs with VAR1 leaders initiates on free ribosomes, then the hydrophobic, membrane‐spanning domains of nascent Cox2p are apparently sufficient to cause insertion into the phospholipid bilayer of the inner membrane. This insertion, while allowing processing of the pre‐Cox2p leader peptide, could either be aberrant or occur in the wrong place, thus preventing efficient assembly.

The apparent rate of Cox3p synthesis directed by the var1::COX3 chimeric mRNA was ∼60–70% lower than that directed by the wild‐type COX3 mRNA, but was nevertheless robust. The reason for the reduced pulse‐labeling of Cox3p in the var1::COX3 strain remains unclear, but it may reflect instability of the protein product. In any event, the steady‐state level of Cox3p was reduced far more, relative to wild‐type, than the rate of Cox3p synthesis. Thus, translation of the var1::COX3 mRNA also appears to yield an unstable product due to mislocalized translation (Figure 5), suggesting that membrane targeting of Cox3p is normally strongly dependent on the interaction of the Pet54p–Pet122p–Pet494p complex with the COX3 mRNA 5′‐UTL.

These results demonstrate that the UTRs of the COX2 and COX3 mRNAs contain information necessary for normal membrane localization and/or assembly of their protein products. However, it is important to emphasize that translation of a mitochondrial mRNA under the control of membrane‐bound activators is not sufficient to cause an otherwise soluble protein to insert into the inner membrane. The synthetic mitochondrial coding sequence ARG8m specifies the same soluble matrix protein as the wild‐type nuclear gene ARG8 (Steele et al., 1996). When this sequence was translated from a chimeric mRNA with the UTRs of COX3 mRNA, the resulting Arg8mp was soluble (Steele et al., 1996). When a derivative of this sequence, which lacks the first 21 codons specifying the matrix targeting signal, was translated from a mitochondrial chimeric mRNA with the COX2 mRNA UTRs, most of the product was soluble and none of it was inserted into the membrane (He and Fox, 1997). Distinct signals for the translocation of Cox2p N‐ and C‐termini through the inner membrane are indeed present within the Cox2p precursor protein (He and Fox, 1997).

Our results suggest that the VAR1 mRNA 5′‐UTL is recognized either by a soluble translational activator, or directly by free mitochondrial ribosomes. Genetic studies of VAR1 mRNA translation have previously been hampered by the fact that Var1p is required globally for mitochondrial translation. However, it is now possible to screens for mutations that affect the expression of the var1::ARG8m reporter gene in a strain expressing VAR1u from the nucleus. Such studies should reveal VAR1 mRNA‐specific translational activators if they exist.

Genetic evidence for mRNA‐specific translational activation mediated through 5′‐UTLs has also been obtained for chloroplasts in both Chlamydomonas and maize (Gillham et al., 1994; Rochaix, 1996; Cohen and Mayfield, 1997). In addition, several RNA‐binding proteins which could play a role in this process have been detected biochemically, and one of them is associated with a chloroplast membrane fraction (Zerges and Rochaix, 1998). Thus, the chloroplast genetic system may employ a mechanism for targeting membrane protein translation similar to that of S.cerevisiae mitochondria.

Protein localization mediated by signals in mRNAs is certainly not restricted to eukaryotic organellar systems. mRNA localization appears to be important in targeting cytoplasmically translated proteins in eukaryotic cells (Deshler et al., 1997; Lithgow et al., 1997; Chicurel et al., 1998). Furthermore, it was recently reported that the Type III secretion machinery of the Gram‐negative bacterial pathogen Yersinia enterocolitica recognizes signals embedded in the RNA sequence of the first 15 codons of the mRNAs for the secreted Yersinia outer membrane proteins YopE and YopN (Anderson and Schneewind, 1997). These signals, present in the translated portion of the mRNAs, are sufficient to signal secretion of downstream polypeptide and are also necessary for efficient translation of the mRNAs. Whereas there is no direct evidence for a homologous relationship between bacterial Type III secretion systems and organellar translational activation systems, the similarities between them are intriguing.

Materials and methods

Yeast strains, growth conditions, DNA manipulations and the VAR1u plasmid

The yeast strains used in this study are listed in Table I. Complete glucose (YPD), galactose (YPGal) and glycerol–ethanol (YPGE) were prepared using 2% glucose, 2% galactose or 2% glycerol plus 2% ethanol, respectively (Sherman et al., 1986). Minimal medium (0.67% yeast nitrogen base without amino acids) was supplemented with specific amino acids, uracil and adenine as required, and a carbon sources as specified for rich medium. Standard genetic methods (Sherman et al., 1986; Fox et al., 1991) were used. Transformation of yeast cells with nuclear plasmids was performed using the lithium acetate method (Ito et al., 1983) or a modification (Elble, 1992). The VAR1u yeast expression vector, pEVA1, was constructed by ligation of a 3.1‐kb SalI fragment from pAM2 (Sanchirico et al., 1995) into SalI digested TVS30A. pEVA1 carries the VAR1u open reading frame (ORF) fused to the COX4 mitochondrial targeting sequence under the transcriptional control of the ADH1 promoter (UASADH1), 2m replication origin, LEU2 and ADE3. Standard techniques were used for all DNA manipulations and Escherichia coli transformations (Sambrook et al., 1989). Restriction enzymes and Vent DNA polymerase were used as recommended by the supplier (New England Biolabs, Beverly, MA). Plasmid DNA was isolated using Qiagen columns, and DNA fragments were isolated from 0.8% agarose gels using the Quiax II extraction kit (Qiagen, Chatsworth, CA). PCR amplification was performed using standard PCR conditions in either a PTC‐100 (MJ Research, Watertown, MA) or Progene thermal cycler (Techne, Princeton, NJ). DNA was sequenced with the Sequenase, version 2.0 kit (United States Biomedical Corp., Cleveland, OH). The nucleotide sequence of the synthetic VAR1uORF is available in the EMBL Nucleotide Sequence Database (accession No. AJ010480).

Construction of var1::ARG8m, var1::COX2 and var1::COX3 chimeric genes

Purified mtDNA was used as a template for PCR amplification of sequences from the VAR1 locus. For the var1::ARG8m construct, the downstream VAR1 sequence containing the last 166 bp of the ORF including the stop codon was generated by PCR using primers that added a BamHI site to the 5′ end and a PstI site at the 3′ end of PCR amplification product. The BamHI and PstI sites were used to insert the fragment into BamHI–PstI‐digested pUC19, creating pMES5. The sequence from 8 bp upstream of the tRNASer to +60 of the VAR1 ORF was generated by PCR with primers that added an EcoRI site to the upstream 5′ end and AccI and BamHI sites to the downstream 3′ end of the fragment. This DNA fragment was digested with EcoRI and BamHI and ligated into EcoRI–BamHI‐digested pMES5 to create pMES6. The 1.3 kb AccI–BamHI fragment from pDS24 containing the sequence for the ARG8m gene (Steele et al., 1996) was then ligated to AccI–BamHI‐digested pMES6 to create the var1::ARG8m gene fusion plasmid, pMES7. This construct encodes for a fusion protein that contains the first 20 amino acids from Var1p followed by an in‐frame fusion of the precursor Arg8mp.

To construct the var1::COX2 locus, a DNA fragment extending from −434 to −3 with respect to the ATG of VAR1 ORF and a second DNA fragment containing the last 166 bp of the VAR1 ORF including the VAR1 stop codon were generated by PCR amplification using two sets of primers. The primers were designed so that 18 bp at the 3′ end of the upstream fragment would overlap 18 bp at the 5′ end of the downstream fragment. The region of overlap contained NcoI and XbaI restriction sites. These two PCR‐amplified DNA fragments were used as templates in a second PCR amplification using the two primers that hybridize to the 5′ end of the −434 to −3 fragment and to the 3′ end of the VAR1 ORF fragment. The sequence overlap between the two fragments made it possible to obtain a 661 bp DNA fragment containing the mtDNA region upstream of the VAR1 ORF, the NcoI and XbaI restriction sites, and the last 166‐bp of the VAR1 ORF. This 661 bp fragment was amplified so that EcoRI and PstI restriction sites introduced at the ends of the fragment could be used to ligate the DNA into EcoRI–PstI‐digested pUC19 to create pMES17. The complete COX2 ORF from the ATG through the stop codon was amplified by PCR from the plasmid pJM2 (Mulero and Fox, 1994) using primers that introduced an AflIII site at the ATG and an XbaI site directly downstream of the stop codon. The amplified COX2 DNA fragment was digested with AflIII and XbaI and ligated into NcoI–XbaI‐digested pMES17 to create the var1::COX2 gene fusion plasmid pMES18. The 5′‐end fusion between the VAR1 UTL and the COX2 ORF was exactly at the ATG with the ATG context altered from TAATG to CCATG.

To construct the var1::COX3 locus, the sequence from 8‐bp upstream of the tRNASer to +3 of the VAR1 ATG was generated by PCR with primers that added an EcoRI site to the upstream 5′ end and the sequence for codons 2–5 of COX3 and BglII and XbaI sites to the downstream 3′ end of the fragment. This DNA fragment was digested with EcoRI and XbaI and ligated into EcoRI–XbaI‐digested pUC19 to create pMES19. The sequence containing the last 166 bp of the VAR1 ORF including the stop codon was generated by PCR using primers that added BglII and XbaI sites to the 5′ end and a PstI site at the 3′ end of PCR amplification product. The BglII and PstI sites were used to insert the fragment into BglII–PstI‐digested pMES19 creating pMES20. pMES20 contains the sequence from 8 bp upstream of tRNASer through the VAR1 ATG and followed downstream by in‐frame codons 2–5 from COX3, the BglII and XbaI sites, and the last 166 bp of the VAR1 ORF. The sequence from bp 22 in the COX3 ORF downstream to the stop codon was PCR‐amplified from pLSF600 (Folley and Fox, 1991) using primers that added BglII and XbaI sites to the upstream and downstream ends, respectively. This COX3 DNA fragment was digested with BglII and XbaI and ligated into BglII–XbaI‐digested pMES20 to create the var1::COX3 gene fusion plasmid pMES21. The BglII site in COX3 introduced from the PCR primer sequence replaced bp 16–21 in the COX3 ORF and changed the codon at position seven from AGT(Ser) to TCT(Ser).

It is important to note that translation of Cox2p and Cox3p from the chimeric mRNAs begins at the normal intiation codons and terminates at the normal termination codons present in the COX2 and COX3 coding sequences, respectively. Moreover, the 166 bp segment of the VAR1‐coding sequence in the chimeric genes is out of frame with respect to the upstream COX coding sequences.

Mitochondrial transformation and gene replacement

High‐velocity microprojectile bombardment cotransformation was performed as described (Fox et al., 1988b) with a helium‐charged Biolistics PDS‐1000 (Dupont, Boston, MA). The recipient strain was DFS160 rho0 and the nuclear selectable plasmid was YEp351 (Hill et al., 1986). For var1::ARG8m, pMES7 was cotransformed and Leu+ nuclear transformants were replica mated to a lawn of DFS188 containing the VAR1u plasmid pEVA1. The mated cells were then replica plated to SGE lacking Arg to identify mitochondrial transformants carrying pMES7 and to select for diploid mitochondrial recombinants in which var1::ARG8m was integrated at the VAR1 locus in mtDNA. The stable synthetic rho strains were purified through five to ten rounds of subcloning. The replacement of VAR1 by the var1::ARG8m construct was confirmed by Southern analysis, and as expected, mitochondrial translation in the var1::ARG8m strain was dependent on the presence of the VAR1u expression plasmid pEVA1.

For var1::COX2 and var1::COX3, DFS160 rho0 was cotransformed as above with pMES18 or pMES21, respectively. Leu+ nuclear transformants were replica mated to lawns of testers strains TF192 or LSF5, and the mated cells were then replica plated to YPGE to detect marker rescue by synthetic rho mitochondrial transformants carrying either pMES18 or pMES21, respectively. Stable synthetic rho strains were purified through three to five rounds of subcloning. To integrate the var1::COX2 and the var1::COX3 constructs into rho+ mtDNA, the synthetic rho strains MSY493 and MSY523 were mated, respectively, to the respiratory deficient strains HMD22 [arg8 (rho+ cox2::ARG8m)] and DFS189 [arg8 (rho+ cox3::ARG8m)], each of which carried the VAR1u expression plasmid pEVA1. Diploids with either the var1::COX2 or the var1::COX3 allele integrated at the VAR1 locus were identified by their ability to grow, albeit slowly, on YPGE. Each was confirmed by Southern analysis and dependence of mitochondrial translation on the presence of the VAR1u expression plasmid pEVA1.

Isolation of mitochondria, mtDNA and immunological analysis

Yeast cells were grown to mid‐exponential phase in YPGal medium and mitochondria were isolated as described by Daum et al. (1982). Mitochondrial DNA was isolated by resuspending purified mitochondria in lysis buffer (50 mM Tris–HCl pH 7.5, 300 mM NaOAc pH 5.2, 10 mM EDTA, 2% SDS), extracting three times with phenol, twice with phenol–chloroform, and once with chloroform. Mitochondrial DNA was precipitated from the aqueous phase with ethanol. Immunoblots were performed using the culture supernatants from the CCO6 and CCO20 hybridoma cell lines that secrete antibodies to Cox2p and Cox3p, respectively. Immunoblot analysis was performed as described previously (Fearon and Mason, 1988). 125I‐labeled goat anti‐mouse IgG (Amersham, Arlington Heights, IL) was used to decorate immune complexes.

Labeling of mitochondrial translation products

In vivo pulse‐labeling of mitochondrial translation products with [35S]methionine Trans label (ICN Biomedicals, Irvine, CA) was essentially as described previously (Fox et al., 1991), except that log‐phase cells were incubated in the presence of 1 mCi Translabel for 5 or 10 min at 30°C in the presence of 500 μg per ml cycloheximide. Labeling was stopped by rapid chilling of the cells by dilution into 12 ml of crushed ice and water and all subsequent steps were performed on ice. Mitochondria were isolated from labeled cells after breakage by vortexing with glass beads as described by Douglas et al. (1979). The radiolabeled proteins were analyzed by SDS–PAGE in a 12.5% gel and analysis of the dried gel using a Storm 840 PhosphorImager (Molecular Dynamics, Sunnyvale, CA). Quantitation of the radioactivity was performed using the ImageQuant v1.1 program with local average background correction (Molecular Dynamics, Sunnyvale, CA).

Assays of whole‐cell respiration and cytochrome c oxidase

Cyanide‐sensitive respiration of whole cells was measured polarographically with a Clark electrode essentially as described by Rouslin and Schatz (1969). Cells were grown to mid‐exponential phase in YPGal liquid medium at 30°C. Cells in ∼5 ml culture medium were harvested by centrifugation, resuspended in 500–750 ml of respiration buffer (40 mM KPO4 pH 7.4, 0.2% glucose, 1% ethanol) and stored on ice. Measurements were made with 50–200 ml of the cell suspension injected into the respiration chamber containing 1.5 ml of air‐saturated respiration buffer at 30°C. Cyanide‐sensitive oxygen uptake was measured in the presence of 1.0 mM KCN. The cell concentration was estimated by optical density at 600 nm. The rate of cyanide‐sensitive respiration was expressed as the change in the percent O2 saturation per minute per OD600. Protein concentrations were determined according to Lowry et al. (1951) with bovine serum albumin (BSA) as a standard. Cytochrome c oxidase activity was determined spectrophotometrically as described previously (Mason et al., 1973).

Acknowledgements

We thank H.M.Dunstan for constructing the cox2::ARG8m chimeric gene. This work was supported by a National Institutes of Health grant (GM29362) to T.D.F. and a National Science Foundation grant (MCB‐9419340) to T.L.M.

References