Mitochondrial DNA Instability Mutants of the Bifunctional Protein Ilv5p Have Altered Organization in Mitochondria and Are Targeted for Degradation by Hsp78 and the Pim1p Protease*

Ilv5p is a bifunctional mitochondrial protein in Saccharomyces cerevisiae required for branched-chain amino acid biosynthesis and for the stability of wild-type ( (cid:1) (cid:2) ) mitochondrial DNA (mtDNA). Mutant forms of Ilv5p defective in mtDNA stability (a (cid:2) D (cid:3) ) are present as 5–10 punctate structures in mitochondria, whereas mutants lacking enzymatic function (a (cid:3) D (cid:2) ) show a reticular distribution, as does wild-type Ilv5p. a (cid:2) D (cid:3) ilv5 mutations are recessive, and the mutant protein is redistributed to a reticular form when co-expressed with wild-type Ilv5p. Ilv5p proteins that are punctate in vivo are also less soluble in detergent extracts of isolated mitochondria, suggesting that the punctate foci in a (cid:2) D (cid:3) Ilv5p mutants are aggregates of the protein. a (cid:2) D (cid:3) Ilv5p proteins are selectively degraded in cells lacking a functional mitochondrial genome, but only in cells grown under derepressing conditions. The targeted degradation of a (cid:2) D (cid:3) Ilv5p, which occurs even when co-ex-pressed with wild-type Ilv5p, is

Ilv5p is a bifunctional mitochondrial protein in Saccharomyces cerevisiae required for branched-chain amino acid biosynthesis and for the stability of wildtype ( ؉ ) mitochondrial DNA (mtDNA). Mutant forms of Ilv5p defective in mtDNA stability (a ؉ D ؊ ) are present as 5-10 punctate structures in mitochondria, whereas mutants lacking enzymatic function (a ؊ D ؉ ) show a reticular distribution, as does wild-type Ilv5p. a ؉ D ؊ ilv5 mutations are recessive, and the mutant protein is redistributed to a reticular form when co-expressed with wild-type Ilv5p. Ilv5p proteins that are punctate in vivo are also less soluble in detergent extracts of isolated mitochondria, suggesting that the punctate foci in a ؉ D ؊ Ilv5p mutants are aggregates of the protein. a ؉ D ؊ Ilv5p proteins are selectively degraded in cells lacking a functional mitochondrial genome, but only in cells grown under derepressing conditions. The targeted degradation of a ؉ D ؊ Ilv5p, which occurs even when co-expressed with wild-type Ilv5p, is mediated by the glucoserepressible chaperone, Hsp78, and by the ATPdependent Pim1p protease, whose activity may be modulated by ؉ mtDNA.
The stability and inheritance of mitochondrial DNA (mtDNA) 1 in Saccharomyces cerevisiae depends on a surprisingly large number of proteins, some of which would not have been anticipated to be involved in mtDNA transactions (1). One example is Ilv5p, a mitochondrial NADPH-requiring acetohydroxyacid reductoisomerase that catalyzes parallel steps in the biosynthesis of branched-chain amino acids (2,3). Cells that lack Ilv5p are blocked in branched-chain amino acid biosynthesis and therefore require isoleucine, leucine, and valine for growth. An unanticipated function for Ilv5p in mtDNA transactions was revealed in studies of suppression of a mtDNA instability phenotype in cells with a deletion of the ABF2 gene (4). Abf2p is an abundant high mobility group-box protein that binds non-specifically to mtDNA (5) and is likely to be involved in mtDNA packaging (6 -8). mtDNA is rapidly lost from abf2⌬ cells grown under conditions that are not selective for respiration, i.e. on fermentable carbon sources (8). Overexpression of Ilv5p by as little as 2-to 3-fold is sufficient to suppress the mtDNA instability phenotype of abf2⌬ cells (4). Moreover, wild-type ( ϩ ) mtDNA is unstable in ilv5⌬ cells, leading to the production of Ϫ petites (which contain amplified fragments of the ϩ genome). Finally, Ilv5p has also been shown to be involved in the redistribution of mtDNA nucleoids, which occurs in response to amino acid starvation (9). These effects do not depend on the operation of the branched-chain amino acid biosynthetic pathway, because deletion of another essential gene in that pathway, ILV2, has no effect on mtDNA stability (4). These studies led to the conclusion that Ilv5p is a bifunctional protein required for the synthesis of branched-chain amino acids and for mtDNA transactions.
Further evidence for the bifunctional nature of Ilv5p has come from a detailed mutational analysis of the protein (10). Recessive point mutants of Ilv5p were identified that resulted in either the loss in Ilv5p of branched-chain amino acid biosynthetic function (a Ϫ D ϩ ) or its mtDNA stability function (a ϩ D Ϫ ). Like ilv5⌬ cells, ϩ mtDNA was unstable in a ϩ D Ϫ mutants and gave rise to Ϫ petites. In some of the a ϩ D Ϫ mutants, Ϫ petites were produced at rates comparable to that of ilv5⌬ cells, whereas other mutants of this class displayed either a weaker or a stronger (hypermorphic) mtDNA instability phenotype. The affected residues of the two classes of point mutants of Ilv5p cluster to distinct and different regions of the 3-dimensional structure of the spinach ortholog of Ilv5p. That protein has 31% sequence identity with yeast Ilv5p and is the only structure currently available for an acetohydroxyacid reductoisomerase (11). Mutations of the a Ϫ D ϩ class map to a region buried within the core of the protein at or close to residues known to bind the substrate and cofactors, NADPH and Mg 2ϩ . By contrast, the majority of mutations of the a ϩ D Ϫ class map within or adjacent to two ␣-helices, also present in Ilv5p of yeast, that are on the surface of the spinach protein. These data suggested that the a ϩ D Ϫ mutations could affect intermolecular interactions of the yeast Ilv5p. The mtDNA redistribution function of Ilv5p (9) was also found to be defective in a ϩ D Ϫ mutants (10), suggesting that the stability and organizational state of mtDNA nucleoids are functionally linked. Finally, none of these mtDNA transaction defects is observed in the a Ϫ D ϩ mutants, further confirming that branched-chain amino acid biosynthesis per se is not connected to these mtDNA activities.
We show here that the a ϩ D Ϫ mutant proteins are present as a small number of aggregated, punctate structures within mitochondria and that the extent of aggregation is related to the severity of the mtDNA instability phenotype. In contrast, wildtype Ilv5p and the a Ϫ D ϩ mutants have a distinctly reticular distribution within mitochondria. Surprisingly, we find that the a ϩ D Ϫ mutant proteins are unstable in petite cells, but not in ϩ cells, and that the instability is glucose-repressible. This instability is independent of the organizational state of the a ϩ D Ϫ mutant protein and can be attributed to the Pim1p protease, whose activity we have found to be regulated by the glucose-repressible chaperone Hsp78.
Plasmid Construction-Construction of plasmids containing a ϩ D Ϫ mutants I267F and W327R and a Ϫ D ϩ mutant D255E was described previously (10). The GFP-tagged fusion of wild-type ILV5 was constructed as follows. Using the primers 5Ј-ACCTCTAGTGGATCCGTA-GATGTAATC-3Ј and 5Ј-TTATTTTCCTCGAGATTGGTTTTCTGGTC-3Ј, a 2379-bp fragment containing the ILV5 promoter and coding sequence was amplified from pRS-ILV5 (4) and digested with BamHI and XhoI. The coding sequence of bGFP was then excised from pGEM7zf(ϩ)bGFP, containing bGFP (13), cloned into the XhoI/KpnI sites in pGEM7zf(ϩ) (Promega), using XhoI and HindIII. The ILV5 fragment was fused to bGFP by excising the BamHI/HindIII fragment from pBS-ILVcϩ3Ј (10) and ligating the BamHI/XhoI ILV5 fragment, along with the XhoI/HindIII bGFP fragment, into the remaining vector to produce pBS-ILV5GFP. The BamHI/BlpI fragment from this vector was then used to replace the same fragment in pRS-ILV5 to produce pRS-ILV5GFP. The mutant ILV5 proteins were GFP-tagged in a similar fashion.
Wild-type ILV5 and the W327R ilv5 mutant were tagged with His 6 by mutagenizing both pRS-ILV5 (4) and pRSW327R (10), using the Muta-Gene Phagemid In Vitro Mutagenesis kit (version 2, Bio-Rad), so as to incorporate the His 6 sequence before the stop codon.
Mitochondrial Purification, Lysis, and Affinity Chromatography-Mitochondria were purified as described previously (15), except that the sucrose gradient floatation step was omitted. Mitochondria were divided into 0.5 mg or 1 mg aliquots, resuspended in 100 l of lysis buffer (50 mM NaCl, 50 mM imidizole (Sigma)/HCl, 5 mM ⑀-amino-n-caproic acid (Sigma), 0.5 mM phenylmethylsulfonyl fluoride, 1 g/ml aprotinin (U. S. Biological), 1 g/ml leupeptin (Sigma)), incubated on ice for 5 min, and lysed by adding 2 g of n-dodecylmaltoside (Roche) per g of mitochondria and incubating on ice for 10 min. Insoluble material was then isolated by centrifugation of the lysed mitochondria for 20 min at 20,817 ϫ g. The pellet was then resuspended in an equal volume of 1ϫ SDS-PAGE loading buffer (0.125 M Tris/HCl, 1% SDS, 4% glycerol, 100 mM dithiothreitol, 0.002% bromphenol blue, pH 6.8). For Ni 2ϩ chromatographic purification of His 6 -tagged Ilv5p, purified mitochondria were resuspended to between 1 and 2 mg/ml in column buffer (50 mM NaH 2 PO 4 , 50 mM KCl, 10 mM imidizole, 0.5 mM phenylmethylsulfonyl fluoride, 1 g/ml aprotinin, 1 g/ml leupeptin, 1ϫ Complete™ EDTAfree protease inhibitor mixture (Roche)). Resuspended mitochondria were then broken by sonication on ice for four periods of 15 s at 7% output, using a Branson Sonifier 450 (Danbury, CT). Broken mitochondria were centrifuged for 30 min at 20,400 ϫ g. The pellet was then resupended in an equal volume of column buffer, and the supernatant was added to 1 ml of nickel-nitrilotriacetic acid agarose (Qiagen) which had been washed in column buffer and gently mixed in a Poly-Prep chromatography column (Bio-Rad) for 1 h at 4°C. The column was then washed with 15 ml of wash buffer (the same as column buffer with the imidizole concentration increased to 20 mM), and the bound protein was eluted with three sequential additions of 0.5 ml of elution buffer (the same as column buffer with the imidizole concentration increased to 250 mM and the addition of 20% glycerol).
Western Blot Analysis-Trichloroacetic acid precipitates of total yeast cell proteins were prepared from OD 600 0.6 -1.0 cultures as described previously (16). For SDS-PAGE, equal volumes of extract were solubilized in 1ϫ SDS-PAGE loading buffer (see above); samples were loaded onto 10% SDS-PAGE gels and separated using the Ready Gel system (Bio-Rad). Proteins were transferred to nitrocellulose membranes (Schleicher & Schuell) using a Hoefer SemiPhor semidry transfer apparatus (Amersham Biosciences). Immunodetection of proteins was carried out using primary rabbit anti-Ilv5p (15) and mouse anti-Porin (Molecular Probes). Anti-rabbit or anti-mouse immunoglobulin G-coupled horseradish peroxidase (Bio-Rad) were used as secondary antibodies and were visualized using the ECL system (Amersham Biosciences).
Fluorescence Microscopy-Log-phase cells were observed live using a Leica microscope (model DMRXE) equipped with an HBO 100 W/2 mercury arc lamp, and a 100ϫ Plan-Apochromat objective lens. Differential interference contrast and green fluorescent protein were visualized using filter sets described previously (13). Mitotracker was visualized using 515-560 nm excitation and Ͼ590 nm long pass emission filters. Images were captured with a color-chilled three charge-coupled device camera system (model C5810; Hamamatsu Photonics) and processed using Adobe Photoshop (Adobe Systems, Inc.). Cells were stained with Mitotracker Red CM-H 2 XRos (Molecular Probes), by adding Mitotracker to log-phase growing cells to a final concentration of 0.5 M and incubating for 20 min at 30°C, and pelleting the cells and washing three times in distilled water, before microscopy.
Petite Assay-To measure petite formation in the various strains, mid-log-phase cultures of strains grown in YNBGlyϩcas medium were transferred to YNBDϩcas medium and grown for at least thirty generations. At various time points, cells were plated onto YNBDϩcas plates and grown for 3 days at 30°C. ϩ and petite colonies were distinguished by 2,3,5-triphenyltetrazolium chloride (TTC) overlay (17), using 0.2% TTC (Sigma) with 0.8% agarose. Between fifty and three hundred colonies were counted for each time point.

a ϩ D Ϫ ilv5 Mutants Have Variable mtDNA Instability Phenotypes and Form Punctate Structures within Mitochondria-To
obtain additional insight into the bifunctional nature of Ilv5p, we analyzed three Ilv5p mutants: two a ϩ D Ϫ mutants, I267F and W327R, defective in ϩ mtDNA stability, and one a Ϫ D ϩ mutant, D255E, which has no effect on mtDNA stability but is defective in branched-chain amino acid biosynthesis. Centromeric vectors containing these mutant alleles, wild-type ILV5, or no insert were transformed into ilv5⌬ cells. These transformants were pre-grown on glycerol medium (YNBGlyϩcas) to maintain ϩ mtDNA and then transferred to glucose medium (YNBDϩcas) and grown for at least thirty generations. At various time points, samples of each culture were plated onto YNBDϩcas medium to determine the percentage of ϩ cells in the population. In accordance with our previous results (10), ϩ mtDNA was unstable in ilv5⌬ cells containing the empty vector, whereas it was completely stable in cells containing wildtype ILV5, or the a Ϫ D ϩ mutant D255E allele (Fig. 1). Although ϩ mtDNA was unstable in both of the a ϩ D Ϫ mutants, the degree of mtDNA instability was significantly different between them and ilv5⌬ cells containing the control plasmid. Specifically, a ϩ D Ϫ mutant I267F has a weaker mtDNA-instability phenotype than ilv5⌬ cells, whereas a ϩ D Ϫ mutant W327R has a much stronger mtDNA-instability phenotype than ilv5⌬ cells (Fig. 1). We refer to these a ϩ D Ϫ mutant alleles as weak and strong, respectively.
To determine how these mutant forms of Ilv5p are distributed in mitochondria, we constructed C-terminal GFP fusion genes of wild-type Ilv5p and each of the mutant proteins and expressed these under control of the ILV5 promoter from centromeric plasmids transformed into ϩ cells of strain 14CWW⌬ilv5u Ϫ . The wild-type yeast GFP fusion protein complemented all of the phenotypes of ilv5⌬ cells, whereas the mutant derivatives reproduced the phenotypes of their respective untagged forms (data not shown). Although all of the Ilv5p-GFP fusions localized to mitochondria when visualized by fluorescence microscopy, there was a dramatic difference in the distribution of these proteins, which correlated with their effects on mtDNA stability: both the wild-type and the a Ϫ D ϩ mutant fusion protein, D255E, had a largely reticular morphology typical of a mitochondrial matrix protein ( Fig. 2A). By contrast, both the strong (W327R) and the weak (I267F) a ϩ D Ϫ GFP-tagged alleles were mainly localized to just a few foci in mitochondria (on average between 5 and 10 when all sections throughout the cells were examined). Closer inspection of the distribution of these GFP-tagged proteins in mitochondria also indicated that some of the I267F a ϩ D Ϫ Ilv5p-GFP was also reticular. The punctate distribution of the a ϩ D Ϫ Ilv5p-GFP mutant proteins is not due to an alteration in mitochondrial morphology, because strains expressing these proteins exhibit normal mitochondrial structures as revealed by staining with MitoTracker (Fig. 2B). This reticular structure persists in wildtype and a ϩ D Ϫ mutant Ϫ and o petite cells (data not shown).
The above findings raise the possibility that the mtDNA instability phenotype of the ilv5 a ϩ D Ϫ mutants is related to their propensity to form punctate structures in mitochondria. Previous genetic studies established that all of the ilv5 a ϩ D Ϫ mutants are recessive (10). This includes not only the mtDNA instability phenotype, but also the temperature-sensitive growth phenotype of ilv5 a ϩ D Ϫ cells grown in medium containing a non-fermentable carbon source. Thus it was of interest to determine whether the punctate morphology of the a ϩ D Ϫ mutant Ilv5p-GFP persists when co-expressed with wild-type Ilv5p. To address this, we introduced a centromeric plasmid containing the a ϩ D Ϫ mutant, W327R Ilv5p-GFP, into ILV5 14CWW wild-type cells expressing untagged Ilv5p and compared the mutant Ilv5p-GFP morphology to that in strain 14CWW⌬ilv5u Ϫ transformed with the same plasmid. The results of this experiment (Fig. 2C) show that there was a dramatic redistribution of W327R Ilv5p-GFP from tight, punctate structures when expressed alone to a largely reticular distribution when co-expressed with untagged, wild-type Ilv5p. This redistribution of a ϩ D Ϫ mutant Ilv5p-GFP was also seen when a strain containing an integrated copy of W327R Ilv5p-GFP was transformed with pRS-ILV5 encoding wild-type Ilv5p (data not shown). Taken together, these data suggest that the unusual punctate distribution of mutant a ϩ D Ϫ Ilv5p is related to the defects in mtDNA transactions.
Although wild-type Ilv5p shows a largely reticular distribution within mitochondria, Ilv5p has been recovered as a component of the mtDNA nucleoid (15), suggesting some functional partitioning of the protein. Because there are fewer a ϩ D Ϫ Ilv5p-GFP punctate structures than ϩ mtDNA nucleoids, it has been difficult to assess with certainty whether these punctate structures co-localize with ϩ mtDNA. However, we can ask whether the punctate morphology of a ϩ D Ϫ Ilv5p-GFP depends on the presence of mtDNA. To this end, we transformed the centromeric plasmid containing W327R Ilv5p-GFP into a °d erivative of strain 14CWW⌬ilv5u Ϫ and examined the distri- Ilv5 mutants have different mtDNA-instability phenotypes. 14CWW⌬ilv5u Ϫ cells transformed with either wild-type ILV5 on a centromeric plasmid, empty vector (ilv5⌬), or the indicated ilv5 a Ϫ D ϩ or a ϩ D Ϫ mutant allele were pre-grown in YNBGlyϩcas then shifted to YNBDϩcas medium. Aliquots were removed at the time points indicated and scored by TTC overlay (17) for the fraction of petites in the population.
FIG. 2. a ϩ D Ϫ mutant Ilv5p-GFP fusion proteins localize as punctate structures within mitochondria. A, Ilv5⌬ cells were transformed with plasmids containing GFP fusions of wild-type ILV5, a ϩ D Ϫ mutants W327R and I267F, and a Ϫ D ϩ mutant D255E. Representative cells are shown, using differential interference contrast and fluorescence microscopy to visualize GFP. B, Ilv5p distribution was visualized in strain 14CWW⌬ilv5u Ϫ expressing either wild-type Ilv5p-GFP, or a ϩ D Ϫ W327R Ilv5p-GFP, and mitochondrial morphology was visualized by staining with MitoTracker (MT). C, a plasmid containing a ϩ D Ϫ ilv5 mutant W327R Ilv5p-GFP was transformed into either the ILV5 wild-type strain 14CWW or into strain 14CWW⌬ilv5u Ϫ . D, Ϫ (HS40) or o cells of strain 14CWW⌬ilv5u Ϫ were transformed with a plasmid containing a ϩ D Ϫ ilv5 mutant W327R Ilv5p-GFP. All strains were grown to mid-log phase in YNBGlyϩcas medium. Bar ϭ 3 m.
bution of the protein in cells. The same plasmid was also transformed into the strain 14WW⌬ilv5u Ϫ HS40 containing the Ϫ mtDNA HS40 (which consists of a 760-bp tandemly repeated fragment of the ϩ genome). The same punctate foci of the a ϩ D Ϫ mutant Ilv5p-GFP were seen in these °and Ϫ cells as in the ϩ parent strain (Fig. 2D), indicating that the punctate distribution of a ϩ D Ϫ protein occurs independently of mtDNA.
a ϩ D Ϫ Ilv5p Solubility Is Decreased in Mitochondria-To gain additional insight to the intriguing differences between the a ϩ D Ϫ mutant proteins and the a Ϫ D ϩ mutant and wild-type proteins, we determined the relative efficiency of solubilization of these proteins by extracting mitochondria isolated from cells expressing each of these proteins with the non-ionic detergent n-dodecylmaltoside (see "Experimental Procedures"). The mitochondrial extracts were separated into supernatant and pellet fractions, which were then analyzed by Western blotting to determine the distribution of Ilv5p and the outer membrane marker protein, porin (Fig. 3A). Whereas porin, wild-type Ilv5p, and a Ϫ D ϩ Ilv5p were nearly completely solubilized by n-dodecylmaltoside, only about 40% of the strong (W327R) and 70% of the weak (I267F) a ϩ D Ϫ Ilv5p mutants were solubilized by this treatment. These data correlate with the respective morphological distribution of the wild-type and mutant Ilv5p-GFP fusion proteins ( Fig. 2A) and suggest that the punctate structures formed by Ilv5p in the a ϩ D Ϫ mutants are aggregated species with reduced detergent solubility compared with wild-type and a Ϫ D ϩ mutant Ilv5p.
Because a ϩ D Ϫ mutant Ilv5p is redistributed in mitochondria from a punctate to a largely reticular morphology when coexpressed with wild-type Ilv5p, we could ask whether the reduced detergent solubility of a ϩ D Ϫ mutant Ilv5p was a consequence of its organizational state within mitochondria, or some intrinsic property of the mutant protein, independent of how it is organized. When mitochondria from glycerol-grown diploid cells co-expressing wild-type Ilv5p-GFP and untagged a ϩ D Ϫ W327R mutant protein were extracted with n-dodecylmaltoside, much more of the mutant protein was solubilized by the detergent than when both untagged and GFP-tagged mutant proteins were co-expressed (Fig. 3B). As expected, both the GFP-tagged and untagged wild-type Ilv5p were nearly completely solublized by n-dodecylmaltoside. These data suggest that the solubility of the mutant protein in n-dodecylmaltoside is related to its organizational state in mitochondria.
Interaction between Wild-type and a ϩ D Ϫ Mutant Ilv5p-One explanation for the redistribution of a ϩ D Ϫ mutant Ilv5p when co-expressed with wild-type Ilv5p is that the wild-type protein interacts with the a ϩ D Ϫ mutant protein, preventing its aggregation. To test for an interaction between these proteins, we constructed C-terminal His 6 -tagged derivatives of wild-type ILV5 and W327R a ϩ D Ϫ mutant ilv5 and integrated these at the ILV5 locus to produce strains ILV5his and W327Rhis, respectively. Strain ILV5his was then transformed with a plasmid expressing wild-type Ilv5p-GFP, whereas strain W327Rhis was transformed with plasmids expressing either wild-type Ilv5p-GFP or a ϩ D Ϫ mutant W327R Ilv5p-GFP. As a control, the wild-type strain 14CWW was transformed with the plasmid expressing wild-type Ilv5p-GFP. Mitochondria from these strains were isolated and broken by sonication, and the soluble extracts were chromatographed on nickel-nitrilotriacetic acid agarose (see "Experimental Procedures"). Bound and unbound material was then analyzed by Western blotting with anti-Ilv5p antiserum, which detects both the His 6 -and GFP-tagged forms. The results of these experiments (Fig. 4, A-C) show that, in all of the combinations of wild-type and mutant Ilv5p examined, affinity purification of either wild-type or mutant His 6 derivatives resulted in co-elution of the corresponding wildtype or mutant GFP-tagged form, whereas no material was found in the bound fraction in the control lacking a His 6 -tagged derivative (Fig. 4D). These data show that different forms of Ilv5p can interact and suggest that the association between a ϩ D Ϫ mutant and wild-type Ilv5p prevents aggregation of the mutant Ilv5p. FIG. 3. a ϩ D Ϫ ilv5 mutant proteins have decreased detergent solubility, which can be reversed by co-expression of wild-type Ilv5p. A, mitochondria were isolated from cells expressing wild-type Ilv5p (ILV5), a Ϫ D ϩ mutant (D255E), strong (W327R), or weak (I267F) a ϩ D Ϫ mutant derivatives, solubilized by treatment with n-dodecylmaltoside, separated into supernatant (S) and pellet (P) fractions, and the distribution of Ilv5p was analyzed by Western blotting. All cells were grown in YPG medium. To confirm that mitochondria had been efficiently solubilized, the distribution of the outer membrane protein porin was also analyzed. B, mitochondria isolated from diploid strains (303V5/14V5G, 303W327R/14V5G, and 303W327R/14W327RG) expressing either wild-type Ilv5p and wild-type Ilv5p-GFP (ILV5/ILV5-GFP), a ϩ D Ϫ W327R mutant Ilv5p, and wild-type Ilv5p-GFP (a ϩ D Ϫ / ILV5-GFP), or a ϩ D Ϫ W327R mutant Ilv5p and a ϩ D Ϫ W327R mutant Ilv5p-GFP (a ϩ D Ϫ /a ϩ D Ϫ -GFP) grown in YPG medium were solubilized with n-dodecylmaltoside, separated into supernatant (S) and pellet (P) fractions, and analyzed as in A.

FIG. 4. Interaction between wild-type Ilv5p and a ϩ D Ϫ mutant
Ilv5p. Plasmids containing wild-type ILV5-GFP or a ϩ D Ϫ mutant W327R-GFP were transformed into strain ILV5his and W327Rhis, expressing His 6 -tagged isoforms of the wild-type or a ϩ D Ϫ mutant W327R Ilv5p, respectively. As a control, the plasmid containing ILV5-GFP was also transformed into strain 14CWW expressing untagged, wild-type Ilv5p as a control. Soluble mitochondrial extract (S) from these strains, grown in YNBGlyϩcas medium, was then chromatographed on Ni 2ϩ resin. The bound (B) material was eluted by addition of 250 mM imidazole, and the bound and unbound (UB) fractions were analyzed for the presence of Ilv5p by Western blotting. A, strain ILV5his co-expressing wild-type Ilv5p-GFP. B, strain W327Rhis co-expressing wild-type Ilv5p-GFP. C, strain W327Rhis co-expressing a ϩ D Ϫ mutant W327R Ilv5p-GFP. D, strain 14CWW co-expressing wild-type Ilv5p-GFP. a ϩ D Ϫ Mutant Ilv5p Is Unstable in Petite Cells, and the Instability Is Glucose-repressible-Although microscopy experiments indicated that the punctate morphology of the strong a ϩ D Ϫ mutant protein is independent of mtDNA (Fig. 2D), we noticed that there was a significant decrease in the abundance of a ϩ D Ϫ mutant W327R Ilv5p in °petite cells grown in raffinose medium. To investigate this further, we compared the abundance of Ilv5p in wild-type, the a Ϫ D ϩ mutant, and both the strong and weak a ϩ D Ϫ Ilv5p mutants in ϩ , Ϫ , and °p etite cells grown in media containing different carbon sources. Because ϩ mtDNA is unstable in a ϩ D Ϫ cells grown on fermentable carbon sources, we carried out our initial analysis of the stability of the various forms of Ilv5p in ϩ cells grown in glycerol medium. Under these conditions, wild-type and mutant forms of Ilv5p were present at comparable abundance (Fig.  5A). However, in either a Ϫ or a °petite strain grown on raffinose, a non-repressing, fermentable carbon source, both of the a ϩ D Ϫ mutant proteins were unstable, whereas neither wild-type Ilv5p nor the a Ϫ D ϩ mutant Ilv5p were affected (Fig.  5, B and D). By contrast, when these petite strains were grown in glucose medium, the a ϩ D Ϫ mutant proteins were as stable as the wild-type or a Ϫ D ϩ mutant protein (Fig. 5, C and E). These data suggest that one or more factors accounting for the instability of the a ϩ D Ϫ mutant proteins in petites is glucose-repressible. The extent of the instability of Ilv5p in the strong and weak a ϩ D Ϫ mutants (Fig. 5, B and D) also correlated with the severity of the mtDNA instability phenotype associated with these mutations (Fig. 1). Fig. 5 (B and D) show that the level of Ilv5p in the strong mutant, W327R, was severalfold lower than that of the weak mutant, I267F. Northern blot analysis revealed that there was no difference in the mRNA abundance of wild-type and mutant ilv5 transcripts in cells grown in the different media noted above (data not shown), indicating that the cause of the instability of the a ϩ D Ϫ mutant proteins is post-transcriptional.
We wanted to compare the stability of the wild-type and mutant forms of Ilv5p in ϩ and petite cells grown on identical carbon sources and to determine whether the organizational state of a ϩ D Ϫ mutant Ilv5p proteins is an underlying factor in the instability of these mutant proteins. To do this we used a ϩ diploid strain (W327R/ILV5-GFP), in which the W327R a ϩ D Ϫ mutant allele was complemented by a wild-type copy of ILV5-GFP. Given that a ϩ D Ϫ ilv5 mutations are recessive, ϩ mtDNA would be stable in those cells grown in raffinose or glucose medium. As a control, we used the diploid strain ILV5/ILV5-GFP containing untagged and GFP-tagged wild-type ILV5 alleles. In W327R/ILV5-GFP °cells, the mutant Ilv5p was unstable when those cells were grown on raffinose medium, exactly as we had observed in cells expressing only the mutant Ilv5p (Fig. 6A). As expected, the mutant Ilv5p was stable in W327R/ILV5-GFP °cells grown on dextrose (data not shown). By contrast, in W327R/ILV5-GFP ϩ cells grown on raffinose medium, the mutant Ilv5p was stable (Fig. 6B). These findings exclude the possibility that the instability of the a ϩ D Ϫ mutant Ilv5p in petites is simply a consequence of growth on raffinose medium. Importantly, these data also suggest that the instability of the a ϩ D Ϫ mutant Ilv5p is not related to the in vivo distribution of the protein, but rather, that the instability is an intrinsic property of the a ϩ D Ϫ proteins. Finally, we have also observed that a ϩ D Ϫ mutant Ilv5p is stable in ϩ W327R/ILV5-GFP cells grown on raffinose medium in the presence of the respiratory chain inhibitor antimycin A or the ATP synthase inhibitor oligomycin (data not shown). We conclude that the instability of the a ϩ D Ϫ mutant Ilv5p in petites is a consequence of the absence of a functional mitochondrial genome rather than simply the loss of oxidative phosphorylation capacity. a ϩ D Ϫ Mutant Ilv5p Is Degraded by the Pim1p Protease, and the Activity Is Regulated by the Glucose-repressible Chaperone Hsp78 -The results presented thus far indicate that the a ϩ D Ϫ mutant Ilv5p is unstable in derepressed cells lacking a functional mitochondrial genome. Examination of microarray data from cells undergoing the diauxic shift (18) did not reveal any known mitochondrial protease in which expression was significantly repressed by glucose. However, the expression of Hsp78, a mitochondrial chaperone that has significant sequence similarity to Escherichia coli ClpB, as well as to ClpA and ClpX-polypeptides that form a complex with the proteolytic subunit ClpP to yield an oligomeric ATP-dependent protease in E. coli (19,20)-was increased by 6-fold on depletion of glucose from the medium. This observation concurs with an earlier report showing that expression of HSP78 was repressed 3-to 5-fold in cells grown on glucose medium (21). Although there is no known protein with significant sequence similarity to E. coli ClpP in yeast, we nevertheless investigated the possibility that Hsp78 might function in concert with other pro-FIG. 5. a ϩ D Ϫ mutant Ilv5p is selectively unstable in petite cells grown under derepressing conditions. A, whole-cell extracts of ϩ strains expressing wild-type (WT) Ilv5p, strong W327R (S), and weak I267F (W) a ϩ D Ϫ Ilv5p mutants, or the a Ϫ D ϩ mutant, D255E, were grown to mid-log phase in YPG medium and analyzed by Western blotting for the abundance of the different forms of Ilv5p, using porin as a loading control. Ϫ (HS40) derivatives of the same strains used in A were grown in either raffinose medium (B) or glucose medium (C), and the level of Ilv5p determined was determined as in A. o derivatives of the same strains used in A were grown in either raffinose medium (D) or glucose medium (E) and analyzed as in A.
FIG. 6. The instability of a ϩ D Ϫ mutant Ilv5p does not depend on its aberrant organization within mitochondria. A, o derivatives of diploid strains 14V5/14V5G and 14W327R/14V5G expressing either wild-type Ilv5p and wild-type Ilv5p-GFP (ILV5/ILV5-GFP) or a ϩ D Ϫ W327R mutant Ilv5p and wild-type Ilv5p-GFP (a ϩ D Ϫ /ILV5-GFP) were grown to mid-log phase in raffinose medium. The relative amount of Ilv5p in cell extracts from each strain was then analyzed by Western blotting, using porin as a loading control. B, ϩ parent strains 14V5/14V5G (ILV5/ILV5-GFP) and 14W327R/14V5G (a ϩ D Ϫ /ILV5-GFP) were grown and analyzed as in A.
teins in the turnover of a ϩ D Ϫ mutant Ilv5p.
We confirmed that in the strains used in this study, the level of the HSP78 transcript was glucose-repressible (data not shown). To determine whether Hsp78 was involved in the instability of a ϩ D Ϫ mutant Ilv5p, we first deleted HSP78 in wild-type and a ϩ D Ϫ mutant strains and determined the level of Ilv5p in o isolates of these strains grown in raffinose medium, conditions in which a ϩ D Ϫ mutant Ilv5p is otherwise unstable. The level of wild-type Ilv5p was unaffected in hsp78⌬ o cells, whereas a ϩ D Ϫ mutant Ilv5p was still unstable in that genetic background (Fig. 7A). However, we observed a slight but reproducible increase in the level of a ϩ D Ϫ mutant Ilv5p in hsp78⌬ cells compared with the wild-type HSP78 control cells. Although these data rule out the possibility that Hsp78 is the sole factor accounting for the instability of a ϩ D Ϫ mutant Ilv5p in derepressed petite cells, we nevertheless explored the notion that Hsp78 functions together with some protease activity directed to a ϩ D Ϫ mutant Ilv5p, particularly in light of the findings that the regulatory Clp subunits in E. coli (20), as well as the ortholog of Hsp78, ClpB, are not essential for protease activity.
We next determined whether constitutive expression of Hsp78 in glucose medium destabilizes a ϩ D Ϫ mutant Ilv5p. To this end, we constructed the centromeric plasmid p416TEFHSP78, which contains HSP78 under the control of the translation elongation factor promoter, which is not glucose-repressed. This plasmid, or a control plasmid with no insert (pRS416), was transformed into wild-type (14CWW o ) and a ϩ D Ϫ mutant (W327R o ) o strains. These strains were then grown in glucose medium, where the mutant protein is normally stable, and the level of Ilv5p was determined by Western blotting. As expected, a ϩ D Ϫ mutant Ilv5p was stable in the strain containing the control plasmid (Fig. 7B). In striking contrast, the level of the mutant protein was now significantly lower when HSP78 expression was under the control of the constitutive TEF promoter. These data suggest that Hsp78 may function in concert with one or more proteins to effect the turnover of a ϩ D Ϫ mutant Ilv5p.
A major protease of S. cerevisiae mitochondrial matrix proteins is the Pim1p protease, which is an ortholog of the Lon protease in E. coli (22). Besides its protease function, Pim1p is also required for the maintenance of ϩ mtDNA (23). To determine whether Pim1p is involved in the degradation of a ϩ D Ϫ mutant Ilv5p, we deleted the PIM1 gene in wild-type and a ϩ D Ϫ mutant W327R °strains and compared the abundance of the wild-type and mutant Ilv5p by Western blotting of extracts from cells grown in raffinose medium. This experiment (Fig.  7C) shows that deletion of PIM1 completely reverses the instability of the a ϩ D Ϫ mutant Ilv5p in °cells grown in raffinose medium, indicating that Pim1p is responsible for the proteolytic degradation of a ϩ D Ϫ mutant Ilv5p. Because PIM1 expression is not glucose-repressible (data not shown), these data further suggest that Hsp78 cooperates with Pim1p for protease activity.

DISCUSSION
In this study, we show that a ϩ D Ϫ Ilv5p mutant proteins, which give rise to a ϩ mtDNA instability phenotype, have a dramatically altered localization within mitochondria compared with wild-type and a Ϫ D ϩ proteins. We also show that the a ϩ D Ϫ Ilv5p mutants are selectively degraded in Ϫ and o petite cells grown on a non-repressing, fermentable carbon source. This selective degradation is determined by the ClpB ortholog, Hsp78, a glucose-repressible mitochondrial chaperone, and the Pim1p protease, an ortholog of the E. coli Lon protease. Our results suggest that Pim1p works in concert with Hsp78 and implicate a role for wild-type mtDNA in modulating Pim1p protease activity.
A Relation between a ϩ D Ϫ Mutant Ilv5p Organization and mtDNA Stability-The a ϩ D Ϫ Ilv5p-GFP's are distributed in mitochondria as a few (5-10) punctate structures, in striking contrast to the largely reticular distribution of wild-type and a Ϫ D ϩ Ilv5p-GFPs. A number of observations support the conclusion that the ϩ mtDNA instability phenotype associated with the a ϩ D Ϫ Ilv5p mutants is a direct consequence of the unusual organization of these proteins within mitochondria. First, the punctate structures are not observed for wild-type or a Ϫ D ϩ mutant forms of Ilv5p in which ϩ mtDNA is stable. Second, the weak a ϩ D Ϫ mutant Ilv5p-GFP has a more reticular distribution underlying its punctate foci than does the strong a ϩ D Ϫ mutant. Finally, co-expression of wild-type and a ϩ D Ϫ mutant proteins, which results in a suppression of the ϩ mtDNA instability phenotype and temperature sensitivity of a ϩ D Ϫ cells grown on a non-fermentable carbon source (10), also results in a redistribution of the mutant protein from a punctate to a largely reticular structure. These observations correlate well with the difference in biochemical fractionation between the a ϩ D Ϫ class of Ilv5p mutant proteins and that of wild-type and the a Ϫ D ϩ Ilv5p mutant: both of the a ϩ D Ϫ mutant proteins have a marked decrease in detergent solubility compared with wild-type and a Ϫ D ϩ mutant Ilv5p. The extent of the insolubility of Ilv5p in the strong and weak mutants correlates with the observed differences in their punctate distribution in mitochondria and their mtDNA instability phenotypes. These effects can be explained in part by the propensity of the a ϩ D Ϫ mutant proteins to self-associate, probably into aggregates, which may be disrupted by interaction with wild-type Ilv5p. Finally, in cells expressing the a ϩ D Ϫ mutant proteins, mitochondrial morphology appears normal, making it unlikely that the mtDNA instability phenotype associated with these mutant proteins is caused by the known mtDNA instability observed in cells with gross defects in mitochondrial morphology (24).
What is the mechanism by which the aberrant organization of a ϩ D Ϫ mutant proteins leads to the instability of ϩ mtDNA? It is unlikely that the presence of a ϩ D Ϫ mutant Ilv5p in mito- FIG. 7. Hsp78 activates the degradation of a ϩ D Ϫ mutant Ilv5p by the Pim1p protease. A, o derivatives of wild-type strain 14CWW (WT), a ϩ D Ϫ mutant W327R, or isogenic strains carrying a null-mutation in the HSP78 gene (hsp78⌬), were grown to mid-log phase in raffinose medium and the relative amount of Ilv5p in cell extracts from each strain was then analyzed by Western blotting using porin as a loading control. B, a plasmid containing HSP78 whose expression is under the constitutive TEF promoter (pTEF-HSP78) or a control, empty plasmid (pRS416), were transformed into o derivatives of wild-type strain 14CWW (WT), or a ϩ D Ϫ mutant W327R and grown to mid-log phase in glucose medium. The levels of Ilv5p were determined as in A. C, o derivatives of wild-type strain 14CWW (WT), a ϩ D Ϫ mutant W327R, or isogenic strains carrying a null-mutation of the PIM1 gene (pim1⌬) were grown to mid-log phase in raffinose medium and analyzed as in A. chondria leads to a complete block in the segregation of nucleoids because a ϩ D Ϫ mutants grown in a fermentable carbon source produce stable Ϫ petites (10). We showed previously that the affected residues in a ϩ D Ϫ ilv5 mutants lie on the surface of the protein and proposed that this might affect intermolecular interactions of the mutant proteins (10). In preliminary experiments, we have identified several proteins present in mtDNA nucleoids that are also associated with wild-type and a ϩ D Ϫ mutant forms of Ilv5p. Moreover, the relative abundance of some of these nucleoid proteins in the Ilv5p complexes differs depending on whether the complex was isolated from wild-type or a ϩ D Ϫ mutant strains. Experiments are currently in progress to characterize further the composition and properties of those complexes. These altered interactions of the a ϩ D Ϫ mutant forms of Ilv5p with other nucleoid proteins combined with its tendency to aggregate could reflect differences in mtDNA packing that might, for example, affect intramolecular mtDNA recombination events leading to the formation of Ϫ mtDNAs. In this connection, we know from previous studies that when overexpressed, Ilv5p effectively suppresses the loss of ϩ mtDNA in cells lacking the mtDNA packaging protein, Abf2p (4), thus implicating a role for Ilv5p in the organization of mtDNA.
Factors Affecting the Stability of a ϩ D Ϫ Mutant Ilv5p-We found unexpectedly that a ϩ D Ϫ mutant Ilv5p forms are unstable in petite cells grown under derepressing conditions. The instability of a ϩ D Ϫ mutant Ilv5p is not just a consequence of respiratory deficiency because the mutant protein was stable in ϩ cells grown in the presence of antimycin A or oligomycin. Interestingly, unlike the mtDNA-instability phenotype, the instability of a ϩ D Ϫ mutant Ilv5p appeared to be unrelated to the punctate organization of the protein, because it was equally unstable when dispersed in mitochondria by co-expression with wild-type Ilv5p.
The degradation of a ϩ D Ϫ mutant Ilv5p in petite cells is activated by Hsp78, a member of the Clp/Hsp100 family of heat shock proteins (21). We suspected that Hsp78 might be involved in the degradation of a ϩ D Ϫ mutant Ilv5p because of its high degree of similarity to ClpA, a regulatory ATP-dependant subunit of the E. coli Clp protease, and because the expression of Hsp78 is glucose-repressible (18,21). Before this study, the only known in vivo mitochondrial functions for Hsp78 were its requirement for mitochondrial thermotolerance under extreme heat stress (25) and the observation that it could substitute for a defective mt-Hsp70 in an ssc1-3 mutant (26). Although the a ϩ D Ϫ Ilv5p mutant protein is only marginally stabilized in petite cells with an hsp78⌬ mutation, constitutive expression of HSP78 in glucose-repressed petite cells results in the clear destabilization of the mutant Ilv5p. These data provide direct evidence for a role of Hps78 in the degradation of a ϩ D Ϫ Ilv5p and suggest that under derepressing conditions, one or more other chaperones can substitute for Hsp78. Cooperation of Hsp78 with other heat-induced chaperones was previously suggested in its role in reactivation of the mitochondrial protein synthesis apparatus after heat stress (25). Hsp78 has also been shown to be capable of substituting for Hsp104 in mediating cellular thermotolerance when misexpressed in the cytosol (25), suggesting that, like Hsp104, Hsp78 may act to disassemble insoluble protein complexes (27). However, Hsp78 is not acting in this way for a ϩ D Ϫ mutant Ilv5p, because Hsp78 is still necessary for degradation of the mutant protein when it is largely disaggregated when co-expressed with wild-type Ilv5p. A likely scenario is that the chaperone activity of Hsp78 alters the conformation of a ϩ D Ϫ mutant Ilv5p, so that it becomes a target for Pim1p protease activity.
Deletion of PIM1 completely eliminated the instability of a ϩ D Ϫ mutant Ilv5p in petite cells grown in raffinose medium. This finding establishes that Pim1p can degrade an abnormal isoform of an endogenous protein in vivo. Pim1p has been shown to be involved in the turnover of some unassembled subunits of mitochondrial complexes such as the F1 ATPase and mitochondrial ribosomal proteins (23,28), and in vitro studies have shown that Pim1p degrades chimeric and foreign misfolded proteins (29,30). With a ϩ D Ϫ mutant Ilv5p, Pim1p is likely to be recognizing a subtle alteration in the mutant protein rather than an aggregated population, because it is still degraded when largely disaggregated by the wild-type protein.
Indeed, it is unlikely that the a ϩ D Ϫ mutant Ilv5p is grossly misfolded, because it retains its enzymatic activity in branched chain amino acid biosynthesis (10) and interacts with wild-type Ilv5p when co-expressed. Aside from the turnover of unassembled or aberrant proteins, the proteolytic activity of Pim1p is required for the stability of ϩ mtDNA (pim1⌬ cells produce Ϫ petites (22,23,31) and for the expression of the mitochondrial COX1 and COB genes (32,33). In most yeast strains, these genes contain introns, some of which encode maturases that function in the splicing of the intron that encodes them. The requirement for Pim1p in the expression of these genes occurs at multiple levels, including maturase-dependent splicing, RNA stability, and translation (32). Although it is not clear how Pim1p functions in these diverse activities, or in maintaining the integrity of ϩ mtDNA, these findings suggest that there is a significant degree of control over the substrate specificity and activity of the Pim1p protease. Our findings add new dimensions to this control that include the participation of Hsp78 and ϩ mtDNA in regulating Pim1p activity. Additional studies will be required to determine whether these effects occur by direct modulation of Pim1p protease activity or by substrate presentation. Finally, it is intriguing that, like the Lon protease in E. coli (34,35), the human ortholog of PIM1 has been demonstrated to bind DNA, specifically to single-stranded sequences in mtDNA (36). These observations have led to speculation that Lon proteases might use DNA binding to control mtDNA metabolism by degrading regulatory proteins at sites adjacent to promoters (33). It is reasonable to imagine therefore that an interaction between mtDNA and Pim1p could modulate its activity and substrate specificity.