Saccharomyces cerevisiae Multidrug Resistance Gene Expression Inversely Correlates with the Status of the F0Component of the Mitochondrial ATPase*

Loss of the mitochondrial genome (ρ0 cell) or elimination of the mitochondrial inner membrane protein Oxa1p causes a dramatic increase in expression of the ATP binding cassette transporter-encoding gene PDR5 in the yeast Saccharomyces cerevisiae. This increase in gene expression occurs via activation of the function of the Cys6-Zn(II)2 cluster transcription factor Pdr3p, which in turn autoregulates expression of its structural gene. Surprisingly, the acquisition of PDR5-dependent multidrug resistance occurs at a very high frequency, consistent with the appearance of ρ− cells in a fermentatively growing culture (∼2%). The degree of activation of Pdr3p target genes was found to vary considerably and to be influenced by the presence of the homologous protein, Pdr1p. Mutagenesis and overexpression studies provided evidence that the control of Pdr3p expression was the major control point of this transcription factor by mitochondrial retrograde signaling. Because both ρ0 and oxa1mutant cells have multiple defects including loss of normal respiratory chain function and oxidative phosphorylation, a series of mutant strains with more selective defects in mitochondrial function was employed to identify the molecular signal that triggersPDR5 transcriptional activation. Only mutations that influenced the functional status of the F0 subunit of the mitochondrial ATPase were found to lead to activation ofPDR5 expression.

porter proteins is a major problem for successful chemotherapeutic treatment of cancer in human patients (for a recent review, see Ref. 6).
Fungi also express a range of different ABC transporter proteins, several of which are capable of effluxing antifungal drugs (7,8). In the yeast Saccharomyces cerevisiae, the multidrug resistance phenotype is referred to as pleiotropic drug resistance or Pdr (9). One of the best studied proteins contributing to pleiotropic drug resistance is the PDR5 gene product (10 -12). This ABC transporter protein is localized to the plasma membrane (13) where, like its mammalian counterparts, it is believed to act as a drug efflux pump of broad specificity (14,15). PDR5 is able to mediate resistance to an impressive range of different compounds (16) including the translational elongation inhibitor cycloheximide.
Control of PDR5 expression is a critical determinant of drug resistance in S. cerevisiae (9). Substitution mutant forms of the Pdr1p or Pdr3p transcription factors have been identified that lead to overexpression of PDR5 with associated increases in multidrug resistance (17,18). Pdr1p and Pdr3p are Cys 6 -Zn(II) 2 transcription factors (19,20) that are both capable of recognizing and binding to a DNA element designated the Pdr1p/Pdr3p response element (PDRE) (21). The PDRE is present in the promoter of all genes responsive to Pdr1p and/or Pdr3p and is found in three copies in the PDR5 promoter (22). Importantly, the PDR3 gene contains two PDREs and is under autoregulation (23).
Previously, we have found that loss of the mitochondrial genome ( 0 cells) or elimination of the mitochondrial inner membrane protein Oxa1p (24) led to a pronounced up-regulation of Pdr3p function and a large increase in PDR5 expression (25). Because both 0 and oxa1 mutant cells are defective in electron transport and oxidative phosphorylation, little could be concluded about the nature of the defect triggering Pdr3p activation and PDR5 induction. Here we provide evidence that loss of normal functional levels of the F 0 complex of the mitochondrial ATPase is a key event monitored by Pdr3p. Activation of Pdr3p by mitochondrial defects has been found to be an extremely high frequency event, occurring in several percent of cells grown in glucose. Use of a conditional allele of OXA1 indicates that the induction of PDR5 gene expression is an event closely linked to loss of Oxa1p function. Finally, the autoregulatory circuit of PDR3 is critical for normal mitochondrial to nuclear (retrograde (26)) signaling to activate expression of PDR5.

MATERIALS AND METHODS
Yeast Strains and Media-All yeast strains used in this study were derived from SEY6210, W303, or BY4741, and their genotypes are listed in Table I. Yeast cells were grown in YPD (2% yeast extract, 1% peptone, and 2% dextrose) or synthetic complete (SC) medium (27) at 30°C with shaking. Derivatives of strains lacking their mitochondrial genomes ( 0 ) were obtained through growing yeast cells on YPD media containing 25 g/ml ethidium bromide to saturation for 2 cycles. A gradient spot test assay (28) was carried out to determine the relative drug resistance of each yeast strain on YPD plates containing cycloheximide or 4-nitroquinoline-N-oxide. Respiratory deficient (petite) cells were identified by their inability to grow on non-fermentable YPGE (2% yeast extract, 1% peptone, 3% glycerol, and 3% ethanol) plates. A standard lithium acetate method was used for yeast transformations as described (29).
Plasmids-PDR3, PDR5, TRP5, YOR1, SNQ2, and YRR1 promoter fusions to the lacZ gene in the low copy number plasmid pSEYC102 were described before (21, 25, 30 -32). The 2ϫHA-tagged PDR3-containing plasmid pFL38PP3HA was obtained courtesy of Karl Kuchler. The plasmid pXTZ132 was obtained through SacII digestion of pFL38PP3HA followed by religation by T4 ligase. SacII cleaves within the center of both PDREs in the PDR3 promoter. This results in pXTZ132 lacking the sequence between the two PDREs of the PDR3 promoter, whereas the religation reforms a new full-length single PDRE. To mutate both PDREs of the PDR3 promoter, primers PDR3dmF (ttA-gCT-GCC-TCC-TCT-GCC-GCT-CGA-GCT-TTa-aGC) and PDR3dmR (ttA-AAG-CTC-GAG-CGG-CAG-AGG-AGG-CAG-cTa-a-GC) were designed. Note that the lowercase nucleotides are different from the wild-type sequence. These two primers were annealed, phosphorylated with T4 polynucleotide kinase, and cloned into the SacII site of pXTZ132. The resulting plasmid was named pXTZ133, and the PDRE mutation was confirmed by automated sequencing. This construct has the exact same sequence as the wild-type PDR3 promoter except for the mutations in the PDREs as shown in Fig. 4. The SalI and SphI fragment of pTH233 (low copy number plasmid pRS315 (33) containing the wild-type PDR3 gene) was replaced with the analogous fragment from pFL38PP3HA or pXTZ133 resulting in pXTZ134 and pXTZ136, respectively. The PGK1 promoter was PCR-amplified from genomic DNA using primers PGKxbaI-733F (CAT-CTA-GAT-AAT-AGG-CAT-TTG-C) and PGKNcoIR (AGA-CCA-TGG-TTT-TAT-ATT-TGT-TG) and cloned into the Topo2.1 vector (Invitrogen). The resulting plasmid pXTZ138 was digested with XbaI and NcoI and cloned into pXTZ134 cleaved with the same enzymes. This construct contains a PGK1 promoter-driven 2ϫHA-tagged PDR3 and was named pXTZ138.
␤-Galactosidase Assay-The enzymatic activities of ␤-galactosidase expressed from lacZ genes fused to PDR5, TRP5, YOR1, and SNQ2 promoters were determined by using o-nitrophenyl-␤-D-galactopyranoside as substrate. Yeast cells were grown on SC media with appropriate supplements to mid-log phase. The ␤-galactosidase activity of appropriate transformants was determined as described (35). ␤-Galactosidase activities of cells containing PDR3-lacZ or YRR1-lacZ were assayed by a chemiluminescent ␤-galactosidase assay method as described previously (25). All assays were done at least twice from two independent transformations, and enzyme activities were reported as an average Ϯ S.D.
Western Blot Analysis-Yeast cells were grown in 10 ml of YPD or SC media to an A 600 ϭ 0.6 -1.0. Cells were harvested by centrifugation and subjected to glass bead breakage by a Tomy shaker for 20 min at 4°C. Whole cell extracts were obtained by centrifugation at 2000 rpm for 3 min at 4°C. Protein concentrations were determined by the Bradford method (36). 100 g of proteins of each sample were incubated with an equal volume of gel-loading buffer containing 15% SDS and 8 M urea for 30 min at 37°C. Protein samples were then electrophoresed through a 15% SDS-PAGE and transferred to a nitrocellulose membrane. After blocking with 2.5% nonfat dry milk in phosphate-buffered saline, the membrane was probed with polyclonal anti-Atp4p or anti-Cox2p at a dilution of 1:1000. Horseradish peroxidase-conjugated anti-rabbit secondary antibody and an ECL kit (Pierce) were used to detect the immunoreactive proteins. The same nitrocellulose membranes were then blocked overnight with 2.5% nonfat dry milk in phosphate-buffered saline and reprobed with monoclonal anti-Vph1p followed by incubation with horseradish peroxidase-conjugated anti-mouse secondary antibody, and the immunoreactive band was detected by the ECL kit.

Cycloheximide-resistant Cells
Occur at a High Frequency in S. cerevisiae-It has been previously shown that mitochondrial status is a critical determinant of drug resistance in S. cerevisiae. Our previous work demonstrated that loss of mitochondrial DNA ( 0 cell) or deletion of the mitochondrial inner membrane protein-encoding gene OXA1 caused a large elevation in the cycloheximide resistance phenotype of the cells (25). The increased cycloheximide tolerance was due to an elevation in the function of the zinc cluster-containing transcription factor Pdr3p. Pdr3p in turn stimulated expression of the PDR5 gene, encoding an ATP binding cassette transporter protein that is believed to efflux cycloheximide and other drugs from the cell (see Ref. 37 for a recent review of Pdr5p function). Because both Mat␣ leu2 ura3 his3 met15 cyc1-⌬1::kanMX4 Research Genetics 0 and oxa1 cells lack the ability to grow on non-fermentable carbon sources like glycerol (petite phenotype) and the frequency of petite mutants occurring in S. cerevisiae can be 5% or higher of the culture (38), we wanted to determine how frequently cycloheximide hyper-resistant colonies arose spontaneously.
To accomplish this, an isogenic series of strains (wild type, ⌬pdr1, ⌬pdr3, and ⌬pdr1 ⌬pdr3) were grown in rich medium to mid-log phase. About 1500 cells of each strain were plated directly to YPD plates containing 0.25 g/ml cycloheximide, and the colonies formed were counted (Fig. 1). Wild-type or ⌬pdr1 cells were found to give rise to cycloheximide-resistant cells at a frequency of about 2%, a number comparable with the frequency of petite mutants in the strains (data not shown). However, deletion of PDR3 in either the wild-type or ⌬pdr1 background totally abolished the presence of any cycloheximide-resistant cells. These data are consistent with the idea that these high frequency spontaneous mutants are drug resistant via activation of a Pdr3p-dependent pathway.
To determine whether these cycloheximide resistant cells were petite and assess their level of PDR5 expression, seven clones were picked at random for further analysis. First, each colony was streaked on YPGE plates. None of these strains was able to grow when glycerol and ethanol were present as the sole carbon source. Second, the level of PDR5-dependent ␤-galactosidase activity was determined by assay of the integrated PDR5-lacZ fusion present in each strain. All of these cycloheximide-resistant colonies also exhibited a large increase in PDR5 expression. These data strongly argue that these cycloheximide resistant colonies that appear at remarkably high frequency arise due to mitochondrial defects that trigger an increase in PDR5 expression. The high frequency of appearance of these drug-resistant colonies is most consistent with defects caused by mitochondrial genome lesions; it is possible that some may be caused by nuclear mutations. Irrespective of the genomic origin of these mutant strains, there is a tight linkage between the petite phenotype and elevation in drug resistance.
Pdr3p Differentially Regulates Target Gene Expression in 0 Cells-Pdr3p has been shown by our work and others to be a key regulator of cycloheximide resistance via PDR5 gene activation (21,39). However, Pdr3p is capable of stimulating expression of a number of different target genes and resistance phenotypes, consistent with the role of this transcriptional regulator as a multidrug resistance determinant (40). Mitochondrial activation of Pdr3p leads to a 10 -13-fold increase in expression of the PDR5 and PDR3 genes, respectively ( Fig. 2 and Ref. 25). The relative expression of other Pdr3p target genes, YOR1 (41), SNQ2 (42), and YRR1 (32) was compared in 0 and ϩ genetic backgrounds by introducing lacZ gene fusions and measuring ␤-galactosidase activity. Compared with the Ն10-fold activation of PDR3 and PDR5, the other three Pdr3p-responsive genes increased in expression by no more than 2-fold in 0 cells (Fig. 2). PDR5, SNQ2, and YOR1 all encode ABC transporter genes that are directly involved in drug resistance (10 -12, 30, 43, 44), whereas YRR1 and PDR3 encode zinc cluster-containing transcriptional regulators (20,45). Whereas the range of response to loss of the mitochondrial genome ranges from 10-to 2-fold for PDR5 and SNQ2, respectively, this cannot be correlated with the number of binding sites for Pdr3p (PDRE). Both of these loci contain 3 PDREs but have a 5-fold difference in their response to loss of the mitochondrial genome. YOR1 and YRR1 both contain a single PDRE and were induced by 2-fold in response to the 0 signal.

FIG. 1. Analysis of spontaneous cycloheximide hyper-resistant colonies.
A, an isogenic series of strains containing the indicated PDR1 and PDR3 alleles were grown to an A 600 of ϳ1, and 1500 cells of each strain were plated on YPD media containing 0.25 g/ml cycloheximide. The plates were incubated overnight at 30°C, and the number of colonies appearing were counted. B, seven spontaneous cycloheximideresistant colonies were selected from the wild-type (WT, SEY6210) strain and streaked on YPD and YPGE to determine their ability to grow on a nonfermentable carbon source. C, each of the colonies indicated above contain an integrated copy of a PDR5-lacZ reporter gene and were grown to mid-log phase in SC medium. PDR5-dependent ␤-galactosidase activity was determined as described (35). These experiments were carried out in a strain expressing both Pdr1p and Pdr3p, which poses a potential complication since both of these factors are able to bind to the PDRE (22). We have previously demonstrated that only Pdr3p is required for transduction of the mitochondrial signal to the PDR5 promoter (25). To examine the behavior of these Pdr3p target genes in response to mitochondrial signaling in the absence of Pdr1p, we introduced these same reporter plasmids into 0 and ϩ cells lacking the PDR1 structural gene. The level of ␤-galactosidase produced from each fusion gene was then determined (Fig. 3).
Removal of the PDR1 locus from ϩ cells led to a slight reduction in expression of all three reporter plasmids. Strikingly, loss of Pdr1p from 0 cells caused a clear increase in the ability of both YOR1 and SNQ2 to respond to the 0 signal. Similarly, the maximal level of PDR5 expression increased by 2-fold when a ⌬pdr1 allele was introduced into the 0 background. This analysis indicates that the presence of Pdr1p dampens the mitochondrial signal through Pdr3p.
Because assay of SNQ2 expression indicated that this locus was induced by loss of the mitochondrial genome, the response of the 4-nitroquinoline-N-oxide resistance phenotype mediated by Snq2p was also tested (Fig. 3). Cycloheximide resistance was assayed in parallel as a control for the genetic backgrounds tested. Both 4-nitroquinoline-N-oxide and cycloheximide tolerance was markedly elevated in 0 cells, and a further small increase was seen when the PDR1 gene was removed from the 0 background. These data are consistent with the results of the expression analysis of PDR5 and SNQ2. We were unable to test the effect of these mutations on YOR1 since the resistance phenotype mediated by this gene (oligomycin) requires cells to grow on a non-fermentable carbon source. These analyses suggest that the differential response of Pdr3p target genes to mitochondrial retrograde signaling may be a consequence of differential interactions between Pdr1p and Pdr3p at PDREcontaining promoters.
PDREs of the PDR3 Promoter Are Required for PDR5 Expression Up-regulation in 0 Cells-The key role of Pdr3p in the retrograde signaling pathway from the mitochondria to the nucleus makes understanding the regulation of PDR3 of paramount importance. We have previously provided evidence that Pdr3p was post-translationally regulated and that autoregulation of PDR3 was involved (25). Autoregulation of PDR3 expression was demonstrated by the loss of the activation of PDR3 gene transcription in 0 cells when Pdr3p was not present. Here we confirm that the PDREs in the PDR3 gene promoter are required for 0 activation of PDR3 expression and B, a strain containing ⌬pdr1 and ⌬pdr3 alleles along with an integrated PDR5-lacZ fusion gene (⌬pdr1,⌬pdr3) was transformed with the low copy number plasmids listed on the left. A 0 variant (⌬pdr1,⌬pdr3, 0 ) of this strain was generated by ethidium bromide treatment and then transformed with the same plasmids. The vector plasmid was pRS315, into which all PDR3 derivatives were cloned. The 2ϫHA-PDR3 and mPDREs-2ϫHA-PDR3 contain the wild-type or the PDRE-less PDR3 promoter controlling expression of the 2ϫHA-tagged form of Pdr3p. The PGK-2ϫHA-PDR3 construct has the PGK1 promoter-controlling expression of the epitope-tagged form of Pdr3p. These plasmids were transformed into the ϩ and 0 derivatives of the ⌬pdr1 ⌬pdr3 strain and assayed for PDR5-lacZ-encoded ␤-galactosidase as above. C, PB4 (⌬pdr1 ⌬pdr3) or PB4 0 (⌬pdr1 ⌬pdr3 0 ) strains were transformed with the PGKcontaining plasmid expressing the 2ϫHA-tagged form of Pdr3p. PB4 transformed with the empty pRS315 vector (control) was included as a control for nonspecific background binding of the anti-HA antibody (Berkeley Antibody Co.). Selected transformants were grown to mid-log phase, whole cell protein extracts were prepared, and equal amounts (100 g) were electrophoresed through SDS-PAGE. After transfer to a nitrocellulose filter, the blot was probed with anti-HA antibody followed by detection of the bound antibody using a chemiluminescent method (Pierce). The position of HA-Pdr3p is indicated, whereas the lower band represents a nonspecific reaction with the antibody. The two PDREs in the PDR3 promoter were removed by site-directed mutagenesis and placed upstream of a HA-tagged form of wild-type Pdr3p. This PDR3 promoter lacking the two PDREs (mPDREs) or a wild-type version was placed upstream of 2ϫHA-Pdr3p in low copy number plasmids. The presence of the HA tags did not interfere with Pdr3p function in this assay. 2 These two different PDR3 genes were introduced into cells lacking PDR1 and PDR3 that were either ϩ or 0 . The levels of PDR5-lacZ supported by the PDR3 clones were then determined (Fig. 4).
Introduction of the HA-tagged form of Pdr3p under control of the wild-type PDR3 promoter produced normal PDR5 expression levels in the presence of either ϩ or 0 mitochondria. Importantly, loss of the two PDREs eliminated the response of PDR5 to the 0 signal and dropped the level of PDR5 expression in ϩ cells to 10% of normal. This analysis is most consistent with induction of PDR3 mediated directly by binding of Pdr3p to the PDREs.
Along with lowering the expression of Pdr3p, the relatively weak PDR3 promoter was replaced with the extremely active PGK1 transcriptional control region. Introduction of this PGK-PDR3 fusion gene into ϩ cells led to a dramatic increase in expression of PDR5, which could be modestly elevated in response to the 0 mitochondrial signal (Fig. 4). The levels of PGK-driven Pdr3p were not significantly different in the ϩ or 0 cells as determined by Western blot analysis. These data confirm the importance of expression control of Pdr3p in the ability of this factor to respond to specific defects in the mitochondria.
F 0 Is a Key Determinant in Regulating PDR5 Expression-The data above and our previous work (25) demonstrate that loss of the mitochondrial genome is a potent activator of PDR5 expression and drug resistance in S. cerevisiae. Interpretation of this result is complicated by the fact that loss of the organellar genome causes many defects in the mitochondria including compromising electron transport chain, mitochondrial ATPase production, and a host of other metabolic problems (38). Earlier, we found that loss of the mitochondrial inner membrane Oxa1p, which is required for biogenesis of the mitochondrial ATPase and cytochrome c oxidase (46), also caused an increase in PDR5 expression and associated drug resistance (25). Because oxa1 and 0 mutant strains share the common defects of loss of both mitochondrial ATPase and cytochrome c oxidase, we employed other mutations individually defective in these protein complexes to assess the contribution of the individual defects to the activation of PDR5 expression.
Strains lacking either the COX4 gene encoding a subunit of the cytochrome c oxidase (47) or ATP10 lacking a folding/ assembly factor for the F 0 complex of the mitochondrial ATPase (48) were transformed with low copy number plasmids carrying lacZ fusions to PDR5 or TRP5. Appropriate transformants were grown to mid-log phase and then assayed for ␤-galactosidase activity as well as for cycloheximide resistance (Fig. 5).
Loss of the ATP10 gene led to a strong elevation in PDR5 expression, whereas elimination of COX4 failed to induce PDR5 transcription. Expression of the tryptophan biosynthetic enzyme-encoding TRP5 gene was essentially unaffected in these strains. Additionally, ⌬atp10 cells exhibited a marked increase in cycloheximide tolerance compared with wild-type cells, whereas a ⌬cox4 mutant was not significantly different from an isogenic COX4 cell. These experiments suggest that reduction in the functional levels of the F 0 complex of the mitochondrial ATPase triggers PDR5 overexpression, whereas elimination of electron transport chain function via deletion of COX4 does not.
To test in more detail the idea that F 0 is the critical determinant modulating Pdr3p control of PDR5 expression, an isogenic series of disruption mutations in various components of both the electron transport chain and the mitochondrial ATPase were obtained. These yeast strains include ⌬atp1, ⌬atp11, ⌬ndi1, ⌬sdh1, ⌬cyt1, ⌬coq4, and ⌬cyc1. ATP1 encodes the ␣ subunit of the mitochondrial F 1 complex (49), whereas ATP11 is required for the assembly of the F 1 complex (50). NDI1 encodes the NADH dehydrogenase of electron transport chain complex I (51). Sdh1p and Cyt1p are essential components of electron transport chain complex II succinate dehydrogenase (52) and complex III ubiquinol-cytochrome c reductase (53), respectively. CYC1 encodes iso-1 cytochrome c, which functions in accepting electrons from cytochrome c reductase and transferring them to the downstream complex IV (54). COQ4 is involved in biosynthesis of coenzyme Q that serves as an electron transfer point from complexes I and II to complex III (55). Each strain was grown to mid-log phase and compared for the ability to tolerate a gradient of cycloheximide relative to wild-type and 0 cells (Fig. 6).
None of the mutant strains tested was able to significantly increase cycloheximide tolerance. Mutants lacking either the F 1 component Atp1p or the F 1 assembly factor Atp11p were unable to enhance the ability to grow on the cycloheximide gradient, quite unlike the ⌬atp10 strain (Fig. 5). Mutant strains lacking electron chain components Ndi1p, Cyt1p, 2 X. Zhang and W. S. Moye-Rowley, unpublished data.
FIG. 5. Loss of F 0 but not cytochrome c oxidase production leads to PDR5 gene activation. A, an isogenic series of strains lacking proper F 0 assembly (⌬atp10), cytochrome c oxidase production (⌬cox4), or the wild type was transformed with lacZ fusions to the tryptophan biosynthetic enzyme-encoding gene TRP5 or PDR5. TRP5 does not respond to mitochondrial regulatory signals controlling PDR5 expression and serves as a control for global gene expression (25). Transformants were grown and assayed for ␤-galactosidase activity as described above. B, cycloheximide resistance of ⌬atp10 and ⌬cox4 strains was compared with that of wild-type using the gradient plate assay as described above.
Sdh1p, coenzyme Q, or Cyc1p were not found to increase cycloheximide tolerance. A strain lacking subunit b of the F 0 complex (Atp4p (56)) exhibited an increase in both cycloheximide resistance and PDR5 expression, 3 again supporting the notion that loss of normal F 0 levels trigger elevation of PDR5 transcription. This survey of a range of loci leading to different mitochondrial defects is consistent with our hypothesis that loss of the F 0 complex from the mitochondrial ATPase is a key determinant in modulating the functional status of Pdr3p.
Deletion of YME1 Restores Normal Cycloheximide Sensitivity of ⌬oxa1 Cells-As mentioned above, loss of the Oxa1p assembly factor leads to an inability to produce normal levels of either the mitochondrial ATPase or cytochrome c oxidase (46). Recently, it has been found that elimination of the matrix protease Yme1p in an oxa1 mutant background restores ATPase levels but not cytochrome c oxidase production (57). If our model that F 0 levels are the key feature controlling Pdr3p activity is correct, then introduction of a yme1 mutation into a ⌬oxa1 strain would be expected to reduce the elevated cycloheximide resistance that arises in the absence of Oxa1p.
To test this prediction, a set of isogenic strains was constructed that lacked either YME1 or OXA1 individually or simultaneously. The cycloheximide resistance phenotypes of these strains were then tested (Fig. 7). A strain carrying the ⌬yme1 allele produced a cycloheximide resistance profile that was very similar to that of the wild-type. As seen before, loss of OXA1 strongly increased cycloheximide resistance. Importantly, the ⌬yme1 ⌬oxa1 strain no longer exhibited an increase in cycloheximide tolerance. This experiment indicates that restoring levels of the mitochondrial ATPase is sufficient to block Pdr3p-mediated induction of PDR5 expression.
To confirm that the previously described selective stabilization of the ATPase (57) was reproduced in our strains, Western blot analysis was carried out. Whole cell protein extracts were prepared from wild-type, ⌬yme1, ⌬oxa1, and ⌬yme1 ⌬oxa1 strains and electrophoresed on SDS-PAGE. After transferring the polypeptides to a nitrocellulose membrane, levels of the F 0 component Atp4p, the cytochrome c oxidase constituent Cox2p, and the vacuolar membrane protein Vph1p were assayed by blotting with appropriate antisera. Consistent with previous data, Atp4p and Cox2p steady-state levels were dramatically reduced in a ⌬yme1 strain, but only Atp4p levels were restored in a strain lacking both Oxa1p and Yme1p.
A Temperature-sensitive Form of Oxa1p Leads to Rapid Induction of PDR5 Expression at the Restrictive Temperature-Although the deletion mutants used in the analysis above support our hypothesis of the critical nature of F 0 levels in controlling Pdr3p activity, this type of a mutation represents a chronic defect in mitochondrial function. The long term lack of mitochondrial function may lead to a slow adjustment by the cell that in turn eventually elevates expression of PDR5. To explore the issue of temporal linkage between a defect in the mitochondria and activation of PDR5 expression, a strain con-3 E. Freire, unpublished data.

FIG. 7.
Restoration of F 0 production in an oxa1 mutant background returns cycloheximide resistance to wild-type levels. A, the relative cycloheximide resistance of the indicated strains was compared using the gradient plate assay. B, the strains from A were grown to mid-log phase, whole cell protein extracts were prepared, and equal amounts (100 g) were analyzed by Western blotting as described above. The blot was probed with anit-Atp4p (F 0 component), anti-Cox2p (cytochrome c oxidase subunit), or anti-Vph1p (vacuolar ATPase subunit; serves as a loading control).
FIG. 6. Electron transport and F 1 ATPase mutants fail to activate cycloheximide resistance. The indicated mutant strains were analyzed for their cycloheximide resistance using the gradient plate assay. See "Results" for the defect of each mutant. taining a temperature-sensitive form of Oxa1p (oxa1 ts ) was employed (58). This strain produces sufficient Oxa1p at the permissive temperature (30°C) to allow the cell to grow on a nonfermentable carbon source, but this ability is lost at the restrictive temperature (37°C). A PDR5-lacZ fusion gene was introduced into the wild-type, oxa1 ts , and ⌬oxa1 strains on a low copy number plasmid, and transformants were assayed for ␤-galactosidase activity after overnight growth at 30 and 37°C (Fig. 8).
Consistent with the ability of the oxa1 ts cell to grow on nonfermentable carbon sources at 30°C, no difference in PDR5 expression was seen between wild-type or the temperaturesensitive mutant. However, at the restrictive temperature, a clear elevation of PDR5-lacZ expression was seen in the oxa1 ts cell but not the wild-type strain. This observation confirms that the oxa1 ts cell lacks sufficient Oxa1p activity at the restrictive temperature, and this deficiency elicits Pdr3p activation of PDR5 expression.
A perplexing feature of the degree of PDR5 activation in the oxa1 ts strain was that only a 2-fold activation was seen compared with the nearly 7-fold activation at 30°C for the ⌬oxa1 strain. To determine whether this reduced activity was due to a feature of the oxa1 ts allele or the elevated temperature used to inactivate the temperature-sensitive Oxa1p, the level of PDR5 expression in a ⌬oxa1 strain was also examined at 37°C. Whether in the presence of the oxa1 ts or the ⌬oxa1 allele, essentially the same level of PDR5 expression was observed. Two important findings about the oxa1 ts -dependent induction of PDR5 can be derived from these results. First, 37°C treatment depresses the ability to fully induce PDR5 expression in the absence of Oxa1p. Second, the elimination of Oxa1p activity produced by the oxa1 ts at 37°C is as complete as the ⌬oxa1 strain as both lead to activation of PDR5. These findings indicated that the oxa1 ts strain was suitable to use as a tool to investigate the time-dependent activation of PDR5 caused by rapid depletion of functional Oxa1p.
To examine the rate of induction of PDR5 in response to loss of Oxa1p function, a time course assay was carried out. Wildtype or oxa1 ts cells carrying the PDR5-lacZ fusion gene were grown to mid-log phase at 30°C and divided into two equal aliquots. One culture of each was returned to 30°C, whereas the second was incubated at 37°C. Samples were withdrawn at 1-h intervals after resumption of growth and assayed for levels of ␤-galactosidase (Fig. 8).
Rapid induction of PDR5 expression was seen in the oxa1 ts strain after transferring this mutant to the restrictive temperature. Induction could be seen as early as 1 h after the temperature shift, and PDR5-lacZ expression increased until a new steady-state level was reached by 3 h. Neither the oxa1 ts strain grown at the permissive temperature or the wild-type strain grown at either temperature showed any significant induction. We interpret these data as arguing for the close coupling between loss of Oxa1p function and activation of PDR5 expression by Pdr3p. DISCUSSION A striking result from these experiments is the finding that the frequency of multidrug-resistant cells approaches that of the appearance of petite cells in a culture of S. cerevisiae. This indicates that in cells growing in glucose (or another fermentable carbon source), several percent of the population will have activated PDR5 expression to extremely high levels. This finding complicates the interpretation of PDR5 expression in glucose-grown cultures, since this gene will be expressed at two different levels depending on mitochondrial status. Because we have compared fully induced PDR5 in a pure 0 culture to partially induced PDR5 in a culture with a small fraction of induced PDR5, the changes in gene expression reported here actually underestimate the magnitude of PDR5 activation by appropriate mitochondrial defects.
Along with our identification of mitochondrial induction of PDR5 in S. cerevisiae, work in the human pathogenic fungus Candida glabrata has detected a similar high frequency acquisition of azole resistance that is linked to loss of normal mitochondrial function eliciting overproduction of ABC transporterencoding genes (59). Intriguingly, the promoter region of at least one of the C. glabrata ABC transporter-encoding loci contains a close match to a PDRE (60), suggesting that a Pdr3p-like transcription factor might also be involved in regulation in this yeast. The presence of a similar regulatory system in another yeast argues for the importance of the mitochondrial control of ABC transporter genes. The precise rationale for this regulatory circuit remains under investigation, but because loss of normal mitochondrial function is so common, perhaps PDR5 expression is activated to efflux some toxic product from the dysfunctional mitochondria from the cell.
The experiments reported here are consistent with the level of expression of the F 0 subunit of the mitochondrial ATPase under surveillance by Pdr3p. As levels of F 0 drop, the activity of Pdr3p is enhanced. Earlier work demonstrated that either 0 or oxa1 cells strongly induced Pdr3p activity and PDR5 expression (25). A goal of this work was to narrow the large number of different parameters influenced by these mutations and determine which contributed to modulation of Pdr3p function. Because both 0 or oxa1 mutants exhibit defects in the respiratory chain as well as oxidative phosphorylation, a panel of different mutants specifically defective in one of these proc- FIG. 8. Rapid loss of Oxa1p function triggers activation of PDR5 expression. A, a series of strains containing different OXA1 alleles was transformed with the PDR5-lacZ fusion plasmid. These strains contained either a normal form (wild type), a temperaturesensitive form (oxa1 ts ), or a deletion allele of OXA1. Transformants were grown overnight at 30 or 37°C and analyzed for PDR5-dependent lacZ activity. B, the wild-type and oxa1 ts strains from A were grown to early log phase at 30°C and then shifted to 37°C or allowed to remain at 30°C. Aliquots of each culture were withdrawn at the indicated hourly time points after the shift, and ␤-galactosidase activities were measured. esses was employed to identify the specific problem leading to PDR5 induction. Multiple different blocks inhibiting electron transport uniformly failed to elevate drug resistance. Additionally, defects in the F 1 complex of the mitochondrial ATPase also did not influence Pdr5p function as measured by cycloheximide tolerance. However, loss of a F 0 component (⌬atp4: data not shown) or two different mutants preventing normal F 0 assembly (⌬atp10 and oxa1) consistently enhanced both PDR5 expression and cycloheximide resistance.
A concern for the use of mutant strains that lack the F 0 sector of the mitochondrial ATPase is their propensity to convert to a Ϫ phenotype. This has been described to occur at a high frequency in ⌬atp4 strains (56) and is a significant concern as we have already seen that 0 cells have elevated PDR5 expression. Although we cannot conclusively rule this possibility out for the ⌬atp4 mutant strains, the ATPase defects of both the ⌬atp10 and oxa1 mutants are suppressible by mutations in multiple different nuclear (57,61) or mitochondrial (62) genes. This finding demands the presence of the mitochondrial genome since three of the F 0 components are encoded by this genome (reviewed in Ref. 63). Additionally, the rapid activation of PDR5 expression after shift of the oxa1 ts strain to the nonpermissive temperature is most consistent with loss of F 0 function, leading to stimulation of Pdr3p activity.
Although the loss of F 0 triggers activation of Pdr3p, the level of Pdr3p is critical for the normal retrograde signaling to occur. The fact that high level overproduction of Pdr3p leads to dramatic overproduction of PDR5 in a ϩ strain suggests that control of Pdr3p expression is the major regulatory component determining activity of this protein. This view is supported by the finding that a strain containing a PGK-PDR3 fusion gene can only be 2-fold-induced by retrograde signaling in 0 cells compared with ϩ cells. We suggest from this result that the post-translational regulation of Pdr3p may only be in the range of 2-fold, and it is amplification of this signal by autoregulation that leads to the large observed increase in PDR5 expression. This autoregulatory circuit could serve to make PDR3 expression highly sensitive to the activity state of Pdr3p since a small change in activity (2-fold) could lead to a large increase in expression of this gene (13-fold (25)).
These data provide strong evidence in support of the idea that the levels of the F 0 subunit determine in large part the activity of Pdr3p. Given that the F 0 complex consists primarily of integral membrane proteins (recently discussed in Ref. 64), it seems likely that some of the response of Pdr3p to changes in the status of this complex will require other participants to transduce the F 0 defect to Pdr3p. Localization studies demonstrate that Pdr3p is found exclusively in the nucleus (65), whereas the F 0 complex is present in the mitochondrial inner membrane, indicating the likely existence of some means of communication between these spatially separated factors. Consistent with the presence of other members of a retrograde signaling pathway connecting F 0 status with Pdr3p, we have carried out genetic analysis and identified several soluble proteins that are required for elevated PDR5 expression in 0 cells. 2 Although these data suggest that levels of the F 0 subunit are predictive of Pdr3p activity, the precise nature of the signal controlling this transcription factor remains to be seen. Because one of the important functions of the mitochondrial ATPase is to consume the proton gradient generated by electron transport (66), a possible explanation is an effect on the essential membrane potential that must be maintained across the mitochondrial membrane (67). Other evidence in the literature argues against this since direct measurements of membrane potential in strains lacking Atp2p or the mitochondrial genome ( 0 ) show a similar reduction in membrane potential in these two strains (68). This comparable effect does not correlate with the differential activation of PDR5 expression seen in these two backgrounds. Alternatively, it is possible that strains with an improperly assembled or absent F 0 component have structural defects in the mitochondrial inner membrane, allowing metabolites to leak into the cytoplasm. Strains lacking F 1 components like Atp1p or Atp2p have been suggested to remain relatively tightly sealed (69). This potential membrane compromise in F 0 mutants may ultimately result in Pdr3p activation. The molecular basis underlying mitochondrial signaling to Pdr3p is currently under investigation.