Members of the Oxa1/YidC/Alb3 family of protein translocases are essential for assembly of energy-transducing membrane complexes. In Saccharomyces cerevisiae, Oxa1 and its paralog, Cox18, are required for assembly of Cox2, a mitochondrially encoded subunit of cytochrome c oxidase. Oxa1 is known to be required for cotranslational export of the Cox2 N-terminal domain across the inner mitochondrial membrane, while Cox18 is known to be required for post-translational export of the Cox2 C-tail domain. We find that overexpression of Oxa1 does not compensate for the absence of Cox18 at the level of respiratory growth. However, it does promote some translocation of the Cox2 C-tail domain across the inner membrane and causes increased accumulation of Cox2, which remains unassembled. This result suggests that Cox18 not only translocates the C-tail, but also must deliver it in a distinct state competent for cytochrome oxidase assembly. We identified respiring mutants from a cox18Δ strain overexpressing OXA1, whose respiratory growth requires overexpression of OXA1. The recessive nuclear mutations allow some assembly of Cox2 into cytochrome c oxidase. After failing to identify these mutations by methods based on transformation, we successfully located them to MGR1 and MGR3 by comparative hybridization to whole-genome tiling arrays and microarray-assisted bulk segregant analysis followed by linkage mapping. While Mgr1 and Mgr3 are known to associate with the Yme1 mitochondrial inner membrane i-AAA protease and to participate in membrane protein degradation, their absence does not appear to stabilize Cox2 under these conditions. Instead, Yme1 probably chaperones the folding and/or assembly of Oxa1-exported Cox2 in the absence of Mrg1 or Mgr3, since respiratory growth and cytochrome c oxidase assembly in a cox18 mgr3 double-mutant strain overexpressing OXA1 is YME1 dependent.
CYTOCHROME c oxidase is the last enzyme in the pathway of aerobic respiration. Its catalytic core consists of the three largest subunits, Cox1, Cox2, and Cox3, which are encoded in mitochondrial DNA (mtDNA) in fungi and animals, and surrounded by nuclear gene products. The synthesis of these subunits and the assembly of active cytochrome oxidase is a highly complex process that requires the action of at least 30 nuclear genes in Saccharomyces cerevisiae (reviewed in Barrientos et al. 2002; Herrmann and Funes 2005; Cobine et al. 2006; Fontanesi et al. 2008). For example, functional expression of the mitochondrial COX2 gene specifically requires, at least, Pet111 to activate mRNA translation; Oxa1 for translocation of the N-terminal domain through the inner membrane; Cox20 to chaperone the processing of the Cox2 leader peptide by the inner membrane protease (Imp1, Imp2, and Som1); Cox18, Mss2, and Pnt1 to translocate the Cox2 C-terminal domain; and Sco1 and Cox17 to insert copper into the CuA site in the C-terminal domain. These functions generate a mature protein with two transmembrane helices in the inner membrane and N- and C-tail domains in the intermembrane space (IMS) that is assembled into the complex in steps involving additional factors.
Oxa1 is the founding member of the Oxa1/YidC/Alb3 family of integral membrane proteins that facilitate the insertion of respiratory and energy-transducing complexes into bacterial, mitochondrial, and thylakoid membranes through protein translocase and membrane insertase activities (reviewed in Bonnefoy et al. 2009). Mitochondria of fungi, animals, and plants contain both Oxa1 proteins and paralogously related Cox18 (also known as Oxa2) proteins (Funes et al. 2004; Gaisne and Bonnefoy 2006). These proteins, and bacterial YidC proteins, share similar core topologies with five transmembrane domains. Oxa1 has a large C-terminal domain facing the matrix that interacts with mitochondrial ribosomes (Jia et al. 2003; Szyrach et al. 2003). Bacterial YidC and mitochondrial Cox18 proteins lack this domain.
In S. cerevisiae, Oxa1 and Cox18 have distinct functions in the biogenesis of Cox2. Oxa1 is required for translocation of the Cox2 N-tail domain (He and Fox 1997) via a cotranslational mechanism. It is also required for translocation of the C-tail domain, although it is unclear whether this requirement involves direct participation of Oxa1 or reflects a requirement of prior N-tail topogenesis for C-tail export (Bonnefoy et al. 2009). Yeast Oxa1 also participates in the assembly of the ATP synthase (Altamura et al. 1996; Hell et al. 2001; Jia et al. 2007). Yeast Cox18 is not required for N-tail export, but in conjunction with the peripheral inner membrane protein Mss2 and the integral membrane protein Pnt1, Cox18 is required for the export of the Cox2 C-tail post-translationally and has no other known substrate (He and Fox 1999; Broadley et al. 2001; Saracco and Fox 2002; Fiumera et al. 2007). These observations show that Oxa1 alone is not capable of translocating the Cox2 C-tail, whether or not it directly participates in that reaction.
S. cerevisiae OXA1 fails to complement cox18 mutations when overexpressed in otherwise wild-type cells (Saracco and Fox 2002). Similarly, COX18 fails to complement oxa1 mutations (L. E. Elliott, H. L. Fiumera and T. D. Fox, unpublished results), confirming that these proteins have distinct functions. While the precise roles of human Oxa1 and Cox18 in human cells have not been established (Stiburek et al. 2007), expression in yeast of cDNAs encoding these human proteins does partially complement the corresponding yeast mutations (Bonnefoy et al. 1994b; Gaisne and Bonnefoy 2006). Furthermore, expression of mitochondrially targeted Escherichia coli YidC in yeast partially complements cox18 mutations but not oxa1 mutations (Preuss et al. 2005). Addition of the yeast Oxa1 C-terminal ribosome-binding domain to YidC allows it to partially complement oxa1 mutations but not cox18 mutations.
If Cox2 is correctly inserted into the inner membrane but prevented from assembling into cytochrome oxidase, it is degraded by a pathway largely dependent on Yme1 (Nakai et al. 1995; Pearce and Sherman 1995; Weber et al. 1996). Yme1 is a member of a conserved family of ATP-dependent AAA proteases (reviewed in Koppen and Langer 2007), whose human ortholog functions in yeast (Shah et al. 2000). Yme1 comprises the i-AAA protease, an integral inner membrane protein whose AAA and proteolytic domains are exposed in the IMS (Leonhard et al. 1996) where they interact with exported domains of Cox2 (Graef et al. 2007). When export of the Cox2 C-tail domain is prevented by an mss2 deletion, Cox2 is instead largely degraded by the m-AAA protease (Broadley et al. 2001), an enzyme homologous to Yme1 with catalytic domains in the matrix (Leonhard et al. 1996). The AAA domain of Yme1 exhibits the chaperone-like property of binding to unfolded substrates in isolated mitochondria and in vitro, an interaction that precedes degradation (Leonhard et al. 1999; Graef et al. 2007). However, Yme1 has not previously been shown to participate as a chaperone in the productive folding of mitochondrial proteins in vivo.
In this study we have examined the phenotype of a nonrespiring cox18Δ deletion strain overproducing Oxa1 from multiple plasmid-borne copies of the wild-type OXA1 gene. Surprisingly, we found that overproduced Oxa1 does support limited export of the Cox2 C-tail domain, but cytochrome oxidase is not assembled. Thus, in wild-type cells Cox18 appears not only to translocate the Cox2 C-tail, but also to do so in a fashion that promotes its proper folding and/or assembly. Respiring mutants selected from this strain inactivate either of two recently discovered adaptor subunits of the i-AAA protease. Respiratory growth of these strains remains dependent upon Yme1 activity, suggesting that under these conditions Yme1 can function as a chaperone in the assembly of Cox2 into cytochrome oxidase.
MATERIALS AND METHODS
Yeast strains and genetic analysis of suppressor mutations:
Standard genetic manipulations for nuclear genes of S. cerevisiae were performed as previously described (Guthrie and Fink 1991; Ausubel et al. 2007). Complete media containing dextrose, ethanol and glycerol, or raffinose were prepared as described. Minimal media (CSM based) were purchased from Bio101 Systems. Transformation of plasmids and PCR products was accomplished with the EZ-transformation kit (Zymo Research). S. cerevisiae strains used in this study are listed in Table 1. All strains are congenic to D273-10B (ATCC 24657), except CUY564, which is congenic to S288C.
Three spontaneously occurring, weakly respiring pseudorevertants were isolated from SCS59 [pNB189] (pseudorevertant strain SCS207) and CAB115 (pseudorevertants CAB115/1-2 and CAB115/2-8) on YPEG medium after incubation at 30° for 5–10 days. To determine the number of mutations in each pseudorevertant strain, pseudorevertants were mated with wild-type strains JKR101 (SCS207) or CAB9 (CAB115/1-2 and CAB115/2-8), diploids were sporulated, and cox18Δ meiotic progeny were analyzed for respiratory growth. In addition, COX18 meiotic progeny carrying the nuclear recessive suppressor mutation from SCS207 were mated with pseudorevertant strain CAB115/1-2. The resulting diploids were sporulated and cox18Δ meiotic progeny were analyzed to determine that the complementing nuclear recessive loci in pseudorevertants SCS207 and CAB115/1-2 were unlinked. The mgr1Δ∷KanMX4 and mgr3Δ∷KanMX4 cassettes were amplified using PCR from Open Biosystems strains (Winzeler et al. 1999) and used to create strains HFC1 and HLF636, respectively, by transformation of a D273-10B congenic recipient.
Direct detection of mutations and genetic mapping by bulk segregant analysis using microarrays:
To detect the suppressor mutations, genomic DNAs prepared from strains CAB115, CAB115/1-2, SCS59, and SCS207 were hybridized to Affymetrix GeneChip S. cerevisiae Tiling 1.0R arrays, and the data for each mutant were compared to those of the parent strain as previously described (Gresham et al. 2006). Bulk segregant analysis was performed as previously described (Brauer et al. 2006) from pools of genomic DNA with and without a suppressor mutation, prepared from cox18Δ meiotic progeny of a cross between SCS207 and CUY564.
Analysis of mitochondrial proteins:
Isolation and purification of mitochondria were carried out as previously described (Glick and Pon 1995). Mitochondrial protein concentrations were determined by the Lowry method using the DC protein assay kit (Bio-Rad). Protease treatments of solubilized mitochondrial and protease protections of mitoplasts were carried out as previously described (Broadley et al. 2001; Fiumera et al. 2007). Proteinase K was inactivated through incubation with 5% trichloroacetic acid and heat, as previously described (Glick 1995). For immunoblotting, proteins were separated on 12% SDS–PAGE, transferred to Immobilon-P polyvinylidene difluoride membrane (Millipore), and incubated with monoclonal anti-Cox2 CCO6, polyclonal anti-cytochrome b2, or polyclonal anti-citrate synthase antibodies. Secondary anti-mouse or anti-rabbit antibodies were detected using the ECL+ kit (Amersham Pharmacia).
In the absence of Cox18, overproduced Oxa1 increases Cox2 levels and promotes Cox2 C-tail export, but not cytochrome oxidase assembly:
Mutants lacking Cox18 synthesize Cox2 in the matrix and export the Cox2 N-tail domain, but fail to export the C-tail domain (Souza et al. 2000; Saracco and Fox 2002; Fiumera et al. 2007). As a result, cytochrome oxidase is not assembled and the cells exhibit a Pet− nonrespiratory growth phenotype. Overexpression of OXA1 from a multicopy plasmid does not compensate for the absence of Cox18 at the level of respiratory growth (Saracco and Fox 2002) (Figure 1A).
Since Cox2 is not assembled in cox18 mutants, its steady-state level is greatly reduced relative to wild type. To determine whether overproduced Oxa1 affected Cox2 in the absence of Cox18, we first tested the effect of excess Oxa1 on the level of Cox2 in a cox18Δ mutant by immunoblotting whole-cell extracts with a Cox2-specific antibody. Interestingly, while overexpression of OXA1 in a cox18Δ mutant does not support respiratory growth, it does lead to increased accumulation of Cox2 at steady state in these cells (Figure 2). Thus, overproduced Oxa1 interacts, directly or indirectly, with unassembled Cox2 to stabilize it.
We wanted to investigate whether overproduced Oxa1 could affect translocation of the Cox2 C-tail across the inner membrane in the absence of Cox18 by determining protease sensitivity in mitoplasts. For this purpose we previously used epitope tags placed at the C terminus of Cox2; these tags are accessible to proteases when exported, whether or not Cox2 is assembled into cytochrome oxidase (Saracco and Fox 2002; Fiumera et al. 2007). However, we found that, in the mutant strains described below, the addition of either HA or FLAG epitopes to Cox2 affected respiratory growth, rendering this approach unusable here (data not shown). Instead, we evaluated Cox2 C-tail translocation by examining protease sensitivity in mitoplasts using a monoclonal antibody (CCO6) that recognizes an endogenous epitope in the Cox2 C-tail domain (He and Fox 1997; H. L. Fiumera and T. D. Fox, unpublished results). In assembled cytochrome oxidase from wild-type cells, the Cox2 C-tail domain is protected from exogenous protease in mitoplasts, despite being exported to the IMS (Broadley et al. 2001) (Figure 3). In mitoplasts from cox18 mutants, the unassembled Cox2 C-tail is resistant to protease since it remains in the matrix and is protected by the inner membrane (Saracco and Fox 2002). However, since the Cox2 N-tail is exported but not assembled in cox18 mutants, the N-tail is digested by exogenous protease, resulting in a shorter protein fragment bearing the C-tail domain (Figure 3). In contrast, mitoplasts from a cox18 mutant overexpressing Oxa1 failed to fully protect the Cox2 C-tail domain from protease, indicating that the C-tail was relatively efficiently exported to the IMS. Despite this C-tail export, Cox2 is apparently not assembled into active cytochrome oxidase since these cells fail to grow on nonfermentable carbon sources (Figure 1A).
After assembly into cytochrome c oxidase, Cox2 from wild-type mitochondria is highly resistant to degradation by proteases added to solubilized mitochondrial extracts (Eytan and Schatz 1975; Broadley et al. 2001). We used this test to confirm the presumed assembly states of Cox2 in these strains by evaluating protease sensitivity after solubilization of purified mitochondria with the nondenaturing detergent octylglucopyranoside (Figure 4). As expected, Cox2 from wild-type mitochondria was stable in the presence of 50 μg/ml proteinase K, whereas Cox2 from a cox18Δ mutant was sensitive to protease concentrations >1 μg/ml. Overexpression of Oxa1 in the cox18 mutant had little, if any, effect on the protease resistance of Cox2, confirming that, despite the export of both N- and C-terminal domains under these conditions, Cox2 was not assembled into cytochrome c oxidase complexes.
Mutations inactivating the nuclear genes MGR1 and MGR3 allow respiratory growth in a cox18Δ mutant overexpressing OXA1:
While overexpression of OXA1 does not affect respiratory growth of cox18 mutants, we could select weakly respiring spontaneous pseudorevertants of cox18Δ mutants containing the OXA1, LEU2, 2μ plasmid pNB189 on YPEG medium after 5–10 days of incubation at 30°. Three such pseudorevertants were characterized in this study and shown to contain mgr1 or mgr3 mutations as described below. Respiratory growth of these pseudorevertants depended upon elevated levels of Oxa1p, since loss of the OXA1 plasmid produced a tight Pet− phenotype in both mgr1 (not shown) and mgr3 pseudorevertants (Figure 1A). Pseudorevertant strains whose OXA1 plasmids had been lost were retransformed with the OXA1 plasmid pNB189, resulting in restored respiratory growth (data not shown). Thus, the pseudorevertant phenotype did not depend upon plasmid-linked mutations.
The three pseudorevertants produced nonrespiring diploids when mated to cox18Δ rho0 tester strains, indicating the presence of nuclear recessive mutations. The mutations segregated as single Mendelian loci in crosses of the pseudorevertants to wild type (materials and methods), and 50% of the cox18Δ [OXA1] meiotic progeny of these crosses were weakly respiring ([OXA1] indicates the presence of a high-copy plasmid bearing OXA1). Two of the pseudorevertants, SCS207 and CAB115/2-8, formed weakly respiring diploids when mated to each other, indicating failure of the recessive mutations to complement. Tetrads from this respiring diploid all contained four Pet+ spores, indicating that the noncomplementing suppressors were tightly linked and thus in the same gene. The nuclear recessive mutation in the third pseudorevertant strain, CAB115/1-2, complemented the other two and segregated independently, indicating an unlinked locus.
We first attempted to identify genes bearing these nuclear recessive mutations by transformation with a library of mini-transposon mutagenized genomic DNA fragments (Burns et al. 1994). However, we were unable to recover pseudorevertants whose respiratory phenotype was linked to the transposon marker. We therefore employed two recently developed microarray-based mapping strategies (Brauer et al. 2006; Gresham et al. 2006) to find the mutations.
To identify base-pair changes that could correspond to the mutations in the pseudorevertants, we employed comparative hybridization of genomic DNAs to DNA tiling microarrays (Gresham et al. 2006; materials and methods). The highest-scoring predicted difference between CAB115/1-2 and its parent strain CAB115 was in a previously identified (Dunn et al. 2006) gene, MGR1. Sequence analysis of MGR1 revealed a G-to-T transversion at nucleotide 778, resulting in a premature stop at codon 261 (of 417), termed mgr1-1. DNA sequence analysis of bulk meiotic segregants, pooled on the basis of their respiratory phenotype, confirmed that this substitution was genetically linked to the mutation. Furthermore, an mgr1 deletion mutation constructed in our cox18Δ [OXA1] strain background caused a Pet+ phenotype (Figure 1A) similar to that of mgr1-1 (not shown), indicating that the stop codon causes loss of function.
We were unable to identify a clear candidate mutation by comparative hybridization of DNA from pseudorevertant SCS207 and its parent, SCS59. Instead, we employed microarray-assisted bulk segregant analysis (Brauer et al. 2006) to detect genetic linkage between the mutation and polymorphic differences between SCS207 (D273-10B genetic background) and a wild-type strain related to S288c (materials and methods). This analysis indicated linkage to polymorphisms on chromosome XIII in the region around base pair 560,000 (not shown). We confirmed that the mutation was in this region by meiotic mapping, which located the mutation to a position 9.7 cM from YMR102C and 21.4 cM from IMP1 (not shown). This location corresponded closely to that of ORF YMR115W, centered around base pair 498,500 of chromosome XIII. This ORF was first characterized as encoding a mitochondrial protein (Sickmann et al. 2003) and was later named MGR3 (Dunn et al. 2008). DNA sequence analysis of MGR3 amplified from pseudorevertant SCS207 revealed a deletion of a single A within a monotonous run of 11 A's (nucleotides 217–227), leading to a frameshift early in the 501-codon MGR3 ORF, termed mgr3-1. The nature of this mutation probably accounts for the failure of direct detection by hybridization to tiling microarrays. We cannot account for the small but significant discrepancy between the actual position and that predicted by microarray-assisted bulk segregant analysis. As expected, an mgr3 deletion mutation also caused a Pet+ phenotype in the cox18Δ [OXA1] background (Figure 1A). Furthermore, the mutation in pseudorevertant CAB115/2-8, which failed to complement mgr3-1 and was genetically linked to it, also affected the MGR3 ORF: insertion of a single T into a monotonous run of six T's (nucleotides 913–918), causing a frameshift. This pseudorevertant strain exhibited a weaker respiratory phenotype (not shown), suggesting an incomplete loss of function possibly due to ribosomal frameshifting.
The Mgr1 and Mgr3 proteins have recently been shown to associate with each other and with Yme1 (Dunn et al. 2006, 2008), the inner mitochondrial membrane AAA ATP-dependent zinc protease (Nakai et al. 1995; Pearce and Sherman 1995; Weber et al. 1996) facing the intermembrane space (Leonhard et al. 1996). These newly discovered subunits of the i-AAA complex bind to known Yme1 substrates and are required for normal degradation rates in vivo, suggesting that they recruit substrate proteins to Yme1 for degradation (Dunn et al. 2008).
Respiratory growth of a cox18Δ mgr3Δ [OXA1] strain is Yme1 dependent:
Mutant strains lacking either Mgr1 or Mgr3, or both, exhibit normal respiratory growth (Dunn et al. 2008). Similarly, yme1 mutants exhibit robust respiratory growth at 30°, although they are Pet− at 37° (Thorsness and Fox 1993). The genetic analysis described above demonstrated that the absence of Mgr1 and/or Mgr3 proteins allows overproduced Oxa1 to partially compensate for the lack of Cox18 in the assembly of cytochrome c oxidase. Since mgr1 and mgr3 mutations appear to decrease proteolysis by Yme1 (Dunn et al. 2006, 2008), we asked whether a yme1 deletion might also allow respiratory growth in the cox18Δ [OXA1] background. Interestingly, the yme1Δ cox18Δ [OXA1] strain failed to grow on nonfermentable carbon sources (Figure 1B). Furthermore, an mgr3Δ yme1Δ cox18Δ [OXA1] strain was also nonrespiratory. Thus, the respiratory growth of mgr3Δ cox18Δ [OXA1] strains is not simply due to loss of Yme1 activity, and indeed depends upon functional Yme1.
Yme1 contains a highly conserved Walker A motif within the AAA ATPase domain and a conserved metallopeptidase catalytic site. Missense mutations that catalytically inactivate either of these active sites have been previously observed to cause the same phenotypes as the yme1 deletion (Leonhard et al. 1996, 1999; Weber et al. 1996). We nevertheless tested whether both of these Yme1 catalytic domains were required for the respiratory growth of a yme1Δ cox18Δ mgr3Δ strain containing extra copies of OXA1 and transformed with additional plasmids bearing defined yme1 missense alleles. Neither yme1-K327R nor yme1-E541A (Weber et al. 1996) supported respiratory growth, regardless of the incubation time, indicating that both the ATP-binding and zinc-dependent protease domains, respectively, are required for this activity (Figure 1B). Transformation with a corresponding plasmid carrying wild-type YME1 was able to restore respiration, albeit not to the level supported by the chromosomal YME1 in the control strain, perhaps because competition between the two related plasmids in these transformants reduced the Oxa1 levels.
To better understand the basis for the Yme1-dependent respiratory growth of the cox18Δ mgr3Δ [OXA1] pseudorevertant, we examined the status of Cox2 in cells of that genotype. Interestingly, the steady-state level of Cox2 was only modestly elevated in these cells relative to the nonrespiring cox18Δ [OXA1] parent strain (Figure 2). Thus, the absence of Mgr3 did not appear to greatly reduce degradation of Cox2 under these conditions, despite the fact that Mgr3 appears to function in delivering substrates to the Yme1 protease (Dunn et al. 2008). Furthermore, deletion of YME1 actually reduced the steady-state level of Cox2 relative to the cox18Δ [OXA1] parent strain to roughly the level seen in the simple cox18Δ mutant (Figure 2). Thus, unassembled Cox2 molecules whose C-tails are exported by excess Oxa1 appear to be stabilized by Yme1, suggesting that, under these conditions, Yme1 provides a chaperone-like function that, in the absence of Mgr3, can promote limited assembly of active cytochrome oxidase. In this connection, it is interesting to note that plasmid-borne extra copies of YME1 did not promote respiratory growth of a cox18Δ [OXA1] strains (not shown).
To confirm that limited amounts of Cox2 were assembled in the cox18Δ mgr3Δ [OXA1] pseudorevertant, we examined its sensitivity to protease digestion in mitoplasts (Figure 3) and in detergent solubilized mitochondria (Figure 4). In both cases, full-length Cox2 from the pseudorevertant exhibited increased protease stability relative to the cox18Δ [OXA1] parent, which is consistent with its assembly into functional cytochrome c oxidase complexes.
Functional expression of protein-coding genes in yeast mtDNA requires a surprisingly complex suite of accessory functions, many of which are specific for individual mitochondrially encoded proteins. In this study, we have explored the functions of two paralogous, nuclearly encoded mitochondrial inner membrane translocases, Oxa1 and Cox18, in the topogenesis and subsequent assembly of the mitochondrially encoded Cox2 subunit of cytochrome c oxidase. Our results have revealed a new role for Cox18 and latent activities of Oxa1 and the i-AAA ATP-dependent protease Yme1. Orthologs of all three of these proteins are encoded in the human genome.
Previous studies have shown that Cox18 is specifically required in otherwise wild-type cells for translocation of the mitochondrially synthesized Cox2 C-tail through the inner membrane to the IMS (Saracco and Fox 2002). This fact alone would account for the cytochrome c oxidase deficiency of cox18 mutants and the low steady-state level of Cox2 (Souza et al. 2000). However, we now find that elevated levels of Oxa1 in cells lacking Cox18 lead to increased accumulation of Cox2, although not to wild-type levels, and significant translocation of the Cox2 C-tail through the inner membrane to the IMS. Despite this overproduced-Oxa1-dependent C-tail export, there is no assembly of active cytochrome c oxidase. Thus, the Oxa1-translocated Cox2 C-tail is not available for assembly in otherwise wild-type cells despite the fact that it exhibits an N-out, C-out topology. We conclude that, under normal conditions, Cox18 not only translocates the Cox2 C-tail to the IMS, but also is required to deliver in a state, as yet unknown, that is competent for downstream steps leading to active cytochrome c oxidase.
Cox18 interacts functionally and physically with Mss2 (Saracco and Fox 2002), a peripheral inner membrane protein on the matrix side that is also required for C-tail export (Broadley et al. 2001). Both Cox18 and Mss2 interact physically with full-length amino terminally processed, newly synthesized Cox2, and C-tail export depends upon the amino acid sequence and/or structural features of the 40 C-terminal Cox2 residues (Fiumera et al. 2007). These facts indicate that Cox2 C-tail translocation is post-translational (Fiumera et al. 2007), in contrast to cotranslational Oxa1-dependent N-tail export (Szyrach et al. 2003). One may conjecture that the normal C-tail export apparatus delivers this mitochondrially synthesized domain to the IMS in a specific, partially folded conformation, while, in the absence of Cox18, overproduced Oxa1 does not.
Why does excess Oxa1 promote limited Cox2 C-tail export in the absence of Cox18, while normal levels of Oxa1 do not? We do not have a clear answer to this question. Oxa1 has a large C-terminal matrix localized domain that interacts with mitochondrial ribosomes to promote cotranslational protein insertion (Jia et al. 2003; Szyrach et al. 2003). Perhaps overproduction increases the number of Oxa1 molecules to the point that many are unable to bind to ribosomes and are thus available to act post-translationally on Cox2 C-tail domains left on the inner surface of the membrane in the absence of Cox18. Our data do not shed light on the question of whether Oxa1 participates directly in Cox2 C-tail export in wild-type cells or indirectly through a requirement for prior N-tail export. It is noteworthy, however, that Oxa1 can be crosslinked to newly synthesized full-length mitochondrial translation products, including Cox2, in wild-type mitochondria (Hell et al. 1998).
The presence of the C-terminal ribosome-binding domain on Oxa1, and its absence from Cox18, appears to play a role in the functional distinction between these paralogous proteins. The homologous YidC translocase of E. coli resembles Cox18 in lacking this domain (reviewed in Bonnefoy et al. 2009), and expression in yeast of mitochondrially targeted YidC partially complements a cox18 deletion, but not an oxa1 deletion (Preuss et al. 2005). Addition of the yeast Oxa1 C-terminal ribosome-binding domain to mitochondrially targeted YidC allows it to partially complement oxa1, but not cox18, deletions at the level of respiratory growth (Preuss et al. 2005). Whether this modified YidC-Oxa1 chimeric protein causes translocation of the Cox2 C-tail domain in the cox18Δ mutant, but fails to allow cytochrome oxidase assembly, was not determined. In any event, it is also clear that the presence or absence of the Oxa1 C-terminal domain is not the only feature underlying the distinct functions of Oxa1 and Cox18 since expression of truncated forms of Oxa1 lacking its C-terminal ribosome-binding domain (Lemaire et al. 2004), at either normal or high levels, does not suppress the respiratory defect caused by a cox18 deletion (H. L. Fiumera, L. E. Elliott and T. D. Fox, unpublished results).
Spontaneous recessive nuclear mutations in two genes allowed cox18Δ strains overexpressing Oxa1 to assemble low levels of active cytochrome c oxidase and exhibit weak respiratory growth on nonfermentable carbon sources. We mapped these mutations to MGR1 and MGR3 by comparative hybridization of genomic DNAs to tiling microarrays (Gresham et al. 2006) and by microarray-assisted bulk segregant analysis (Brauer et al. 2006) after failing to identify them by transposon-tagging mutagenesis.
Mgr1 and Mgr3 have recently been identified as physically and functionally associated with the i-AAA protease Yme1, which is bound to the inner membrane with catalytic domains facing the IMS (Dunn et al. 2006, 2008). Mgr1 and Mgr3 interact with each other, with substrates of the Yme1 protease, and together bind to Yme1 itself. Both Mgr1 and Mgr3 are required for normal binding of proteolytic substrates to Yme1 and for normal rates of substrate degradation. Thus, Mgr1 and Mgr3 appear to play the role of adaptor subunits in the recognition and delivery of substrates to be degraded by Yme1 (Dunn et al. 2008). For reasons that are not clear (see below), yme1 mutations are synthetically lethal in combination with the loss of mtDNA (rho− or rho0) (Thorsness et al. 1993), as are mgr1 and mgr3 mutations in certain strain backgrounds (Dunn et al. 2006, 2008). However, respiratory growth of mgr1 and mgr3 mutants is normal, while that of yme1 mutants is defective only at elevated temperature, and steady-state levels of Cox2 are normal in all three mutants (Dunn et al. 2008). Thus, these functions are not needed for assembly of cytochrome c oxidase or other respiratory complexes in otherwise wild-type strains.
The fact that Mgr1 and Mgr3 are necessary for normal proteolysis by Yme1 initially suggested that the respiratory growth of cox18Δ mgr1 and cox18Δ mgr3 strains overproducing Oxa1 could be due to decreased degradation of Cox2 by Yme1. An effect like this has been previously reported, in which a yme1Δ mutant can reverse the oxa1Δ-induced defect in ATP synthase activity by stabilizing the integral membrane components of ATP synthase, although yme1Δ has no effect on the cytochrome c oxidase deficiency (Lemaire et al. 2000).
However, our data clearly indicate that decreased proteolysis by Yme1 is not the mechanism operating here. First, the steady-state level of Cox2 in these strains was not greatly increased by the mgr3 deletion. Second, a yme1 deletion itself did not allow respiratory growth of a cox18Δ strain overexpressing Oxa1 and actually decreased the steady-state level of Cox2 relative to the corresponding YME1 control. Thus, in this situation Yme1 partially protects Cox2 from degradation by other, unknown, proteases. Finally, both yme1 deletion and catalytically inactivating yme1 missense mutations prevented the respiratory growth of a cox18Δ mgr3Δ strain overexpressing Oxa1. Thus, cytochrome c oxidase assembly and respiratory growth are dependent on active Yme1 under these conditions.
For reasons that are not clear, yme1 mutations are synthetically lethal in combination with the loss of mtDNA (rho− or rho0), as are mgr1 and mgr3 mutations in certain strain backgrounds. rho0 cells lacking Yme1 are virtually inviable (Thorsness et al. 1993), and a similar phenotype is observed in some yeast strain backgrounds for rho0 cells lacking Mgr1 and Mgr3 (Dunn et al. 2006, 2008). This inviability is apparently due to the inability of these nonrespiratory mitochondria to maintain an inner-membrane potential by F1-ATPase-dependent hydrolysis of fermentatively derived ATP in the matrix, although the mechanism underlying this defect remains unclear (Kominsky et al. 2002). Since both yme1Δ and mgr3Δ cause similar membrane potential related growth phenotypes in rho0 cells, it is unlikely that the distinct respiratory growth phenotypes caused by these deletions in the cox18 [OXA1] strains reported here are due to differential effects on membrane potential.
We propose that, in the absence of Cox18 and Mgr1 or Mgr3, Yme1 acts as a chaperone to promote the assembly of Cox2 C-tail domains that were aberrantly exported by excess Oxa1. In this view, the absence of the substrate adaptor subunits could allow latent chaperone activity of Yme1 to affect the structure of the Cox2 C-tail domain neighboring it on the outer surface of the inner membrane. While evidence indicating chaperone activity of Yme1 in vivo has not been previously reported, the purified AAA domain of Yme1 does exhibit chaperone-like function in vitro, preferentially binding to denatured dihydrofolate reductase and promoting its renaturation (Leonhard et al. 1999).
The homologous, matrix-facing, m-AAA protease complex was previously thought to have chaperone activity in the assembly of respiratory complexes (Paul and Tzagoloff 1995; Arlt et al. 1996). However, the phenotypes caused by mutations affecting the m-AAA protease are now known to be due to a defect in the processing of an imported mitochondrial ribosomal protein (Nolden et al. 2005). Thus, the apparent in vivo chaperone activity of Yme1 in the absence of Mgr1 or Mgr3 is a novel observation for AAA proteases.
We thank C. M. Tucker, H. F. Chen, and C. G. DeSevo for excellent technical assistance; N. Bonnefoy, P. E. Thorsness, and J. E. Pike for plasmids used in this study; M. J. Brauer and D. M. Ruderfer for scripts; and C. E. Dunn and R. E. Jensen for communicating results prior to publication. M.J.D. was supported in part by National Institutes of Health (NIH) grant P50 GM071508 to the Lewis Sigler Institute at Princeton University. This study was supported by NIH grant GM29362 to T.D.F.
↵1 Present address: Department of Biological Sciences, Binghamton University, Binghamton, NY 13902.
↵2 Present address: Department of Genome Sciences, University of Washington, School of Medicine, Seattle, WA 98195.
↵3 Present address: Monsanto Company, Chesterfield, MO 63017.
Communicating editor: J. Engebrecht
- Received January 27, 2009.
- Accepted March 20, 2009.
- Copyright © 2009 by the Genetics Society of America