Multiple signals from dysfunctional mitochondria activate the pleiotropic drug resistance pathway in Saccharomyces cerevisiae.

Multiple or pleiotropic drug resistance most often occurs in Saccharomyces cerevisiae due to substitution mutations within the Cys(6)-Zn(II) transcription factors Pdr1p and Pdr3p. These dominant transcriptional regulatory proteins cause elevated drug resistance and overexpression of the ATP-binding cassette transporter-encoding gene, PDR5. We have carried out a genetic screen to identify negative regulators of PDR5 expression and found that loss of the mitochondrial genome (rho(o) cells) causes up-regulation of Pdr3p but not Pdr1p function. Additionally, loss of the mitochondrial inner membrane protein Oxa1p generates a signal that results in increased Pdr3p activity. Both of these mitochondrial defects lead to increased expression of the PDR3 structural gene. Importantly, the signaling pathway used to enhance Pdr3p function in rho(o) cells is not the same as in oxa1 cells. Loss of previously described nuclear-mitochondrial signaling genes like RTG1 reduce the level of PDR5 expression and drug resistance seen in rho(o) cells but has no effect on oxa1-induced phenotypes. These data uncover a new regulatory pathway connecting expression of multidrug resistance genes with mitochondrial function.

The appearance of multidrug-resistant cells in human tumors is often associated with overproduction of ATP-binding cassette (ABC) 1 transporter proteins, like Mdr1 or Mrp (1,2). These plasma membrane-localized molecules act as ATP-dependent drug efflux pumps and eliminate chemotherapeutic agents from cells, allowing resistance to these cytotoxic drugs (3). In the yeast Saccharomyces cerevisiae, a similar multidrug resistant phenotype also typically involves overproduction of ABC transporter proteins and is referred to as pleiotropic drug resistance (Pdr) (reviewed in Ref. 4). Examples of the relevant ABC transporter genes that are overproduced in multiply drug resistant S. cerevisiae include PDR5 and YOR1, loci that confer resistance to cycloheximide and oligomycin, respectively (5)(6)(7)(8).
Loss of normal transcriptional control of these S. cerevisiae ABC transporter protein-encoding genes is frequently associated with single amino acid substitutions within related Cys 6 -Zn(II) transcription factors designated PDR1 (9) and PDR3 (10). Previous work has shown that these substitution mutant forms of Pdr1p (11) and Pdr3p (10) behave as dominant, hyperactive transcriptional regulatory proteins and elicit marked overproduction of target genes like PDR5 and YOR1 (8,12). Both Pdr1p and Pdr3p bind to a sequence element referred to as the Pdr1p/Pdr3p response element (PDRE) that is found in the promoter region of all genes responsive to these transcription factors (13). Importantly, loss of PDR1 or PDR3 does not produce the same phenotypic effect on cells. ⌬pdr1 cells are extremely sensitive to cycloheximide or oligomycin while ⌬pdr3 cells are relatively normal in terms of tolerance to these drugs (10,14). However, loss of both PDR1 and PDR3 causes a pronounced drug sensitivity, much greater than loss of either gene alone.
More recent work has demonstrated that Pdr1p but not Pdr3p is positively regulated by a Hsp70-related protein, Pdr13p (15). This finding indicated that, while Pdr1p and Pdr3p are 36% identical across their lengths, these proteins are regulated by different mechanisms. Since several different substitution mutations could be isolated that produced hyperactive forms of Pdr1p or Pdr3p, we set out to examine the possibility that these mutant proteins were able to escape negative regulation. To test this hypothesis, we utilized transposon mutagenesis to identify loci that served as negative regulators of PDR5 gene expression through reduction of Pdr1p or Pdr3p function. From this screen, three different genes were identified that acted genetically as negative regulators of PDR5 expression and resulting cycloheximide resistance: FZO1, encoding a mitochondrial GTPase required for normal organelle inheritance (16); OXA1, an inner mitochondrial membrane protein involved in biogenesis of cytochrome c oxidase (17); and TIM17, a gene encoding a component of machinery required to import proteins across the inner membrane (18). Similar induction of PDR5 transcription was found when cells were induced to lose the mitochondrial genome ( o ). These data provide important new insight into the coordination of PDR gene expression with the functional status of the mitochondria and suggest that activation of ABC transporter gene expression may be required for cell viability upon compromise of mitochondrial function.

MATERIALS AND METHODS
Yeast Strains and Media-The genotypes of the yeast strains used in this study are listed in Table I. Yeast transformations were performed using the lithium acetate procedure (19) or a high-efficiency technique (20). Standard YPD (1% yeast extract, 2% peptone, 2% dextrose) and synthetic complete medium (21) were used for growth of cells and drug spot test assay (22). ␤-Galactosidase activity determinations using o-nitrophenyl-␤-D-galactopyranoside hydrolysis was measured as described previously (23). Luminescent ␤-galactosidase assays were performed following the manufacturer's specifications (CLONTECH). 0 Derivatives of strains were generated using 25 g/ml ethidium bromide as described in Ref. 24 and loss of mitochondrial DNA was assessed by 4,6-diamidino-2-phenylindole staining and visualizing by fluorescence microscopy. Gradient plate assays were carried out as described (8).
Selection for Cycloheximide-resistant Strains with Linked Transposon Insertions-A wild-type strain (SEY6210) containing an integrated PDR5-lacZ gene was transformed with a transposon-mutagenized S. cerevisiae genomic library (25) using a high-efficiency protocol. Approximately 5000 transformants were plated on SC (21) plates containing 0.25 g/ml cycloheximide. 14 transformants were recovered that exhibited elevated cycloheximide resistance. These transformants were then assayed for the level of expression of their integrated PDR5-lacZ reporter gene. Colonies with both elevated ␤-galactosidase expression and cycloheximide resistance were crossed to a wild-type strain and the resulting diploids subjected to standard genetic analysis. 5 colonies were found in which both the PDR5 overproduction and cycloheximide hyper-resistance phenotypes were linked to the transposon. The locations of transposon insertions in the genome were determined using plasmid rescue as described (25).
Plasmids-Low-copy number plasmids containing lacZ gene fusions with the TRP5 (tryptophan synthase), PDR5 (plasma membrane ABC transporter), CTT1 (cytosolic catalase), CUP1 (copper metallothionein), and HSP12 (small heat shock protein) (15) or GSH1 (␥-glutamylcysteine synthetase) (26) and TRX2 (thioredoxin) (27) genes were described before. The PDR5-lacZ gene containing mutations in all three PDREs in the promoter was produced earlier, along with the plasmids carrying PDRE-CYC1-lacZ and mutant PDRE-CYC1-lacZ (13). PDR1 and PDR3 genes were carried on the low-copy number vector pRS316 (28). The FZO1 gene, carried on the centromeric pRS416 vector, was obtained from Janet Shaw (16). The ATP2 gene (29) was carried on the low-copy vector YCplac22 and was provided by David Bedwell (University of Alabama, Birmingham, AL). The PDR3-lacZ gene (pTH38) was carried on a low-copy number vector and contains 652 bp of 5Ј-flanking sequence from PDR3. The PDR1-lacZ gene (pTH223) contains 600 nucleotides upstream of translation start and was amplified from genomic DNA with the 5Ј primer GAGCCTGAATTCCCGGTTGCTTGGACTTTT and the 3Ј primer GAGCCTGGATCCATCTTCCAGTTTCTTGGA. The PCR product was cut with BamHI and EcoRI and cloned in-frame with lacZ in the pSEYC102 plasmid (30). 600 bp of the PDR3 promoter was PCR amplified as a SacI-SacI fragment using the 5Ј primer GGATTC-GAGCTCACTTGACAGGGCTTCCAA and the 3Ј primer GGATC-CGAGCTCTGCGGTCACGCAATAAGA and cloned as a SacI fragment upstream of the PDR1 gene with a SacI site just 5Ј of translation start. This plasmid (pTH221a) carried the PDR1 gene under control of the PDR3 promoter. The plasmid pTH223 carrying the PDR3 gene under PDR1 promoter control was constructed by PCR amplifying 600 nucleotides of the PDR1 promoter with the 5Ј primer GTCGGCGTCGAC-CCGGTTGCTTGGACTTTT and the 3Ј primer GTCGG CCCATGGTC-CAGTTTCTTGGATTCT. This PCR product was cut with SalI/NcoI and cloned into a plasmid (pDK4) carrying the PDR3 gene with a NcoI site introduced at the start codon.
Gene Disruptions-fzo1-⌬1::kanMX4 and rtg1-⌬1::kanMX4 gene disruption strains were obtained from Research Genetics. PCR primers specific for sequences 500 nucleotides upstream and downstream of the kanMX4 gene replacement were used to PCR amplify the locus. The product was used to disrupt the corresponding genes in SEY6210 by homologous recombination and selection on G418-containing plates (31). The rtg2-⌬1::his5ϩ gene disruption was generated following PCRbased protocol (32). Briefly, PCR primers were designed with 50 bp from either the 5Ј or 3Ј end of the RTG2 gene and with 16 bp specific for amplification of the his5ϩ gene. The rtg2-⌬1::his5ϩ allele was generated using the 5Ј primer GTGTCCTTTACTAAGGATTGTTTTGAACG-AAAAGTGTAGGCGTGCCACAACGGATCCCCGGGTTAATTAA and the 3Ј primer TATATAAGGATTTCGTATTTATTGTTCAAGTATTTAA-AGACTAGATGTCTGAATTCGAGCTCGTTTAAAC to PCR amplify the S. pombe his5ϩ gene carried on the pFA6:HIS5 plasmid (32), transforming cells with the PCR product and selecting for strains that grew on plates without histidine. Correct integration in all strains constructed was determined by Southern blotting and PCR amplification techniques.

Genetic Screen for Negative Regulators of Pleiotropic Drug
Resistance-Previous searches for mutants with an enhanced pleiotropic drug resistance phenotype have identified alterations in the genes encoding the zinc finger-containing transcription factors Pdr1p and Pdr3p (see Ref. 33 for a review). Characterization of these gain-of-function alleles has shown that these lesions are single amino acid replacements that result in a constitutive elevation in activity of Pdr1p or Pdr3p. One hypothesis, consistent with these data, is that the function of Pdr1p and Pdr3p is subject to some form of negative regulation that is relieved in the mutant alleles. To identify these putative negative regulators and avoid isolation of more single amino acid replacements in the transcription factors, we used transposon mutagenesis to select cells that exhibited both elevated pleiotropic drug resistance and enhanced expression of a Pdr1p/Pdr3p target gene, the ATP-binding cassette transporter-encoding gene, PDR5 (5-7).
This screen was carried out by transforming a wild-type strain containing an integrated PDR5-lacZ fusion gene with a yeast genomic library containing random transposon insertions (25). Approximately 5000 transformants were grown on selective media containing 0.25 g/ml cycloheximide. Cycloheximide was included in the plates as an indicator of PDR5 expression, the key target gene for Pdr1p/Pdr3p-mediated cycloheximide resistance (12,14). Cycloheximide hyper-resistant colonies were isolated and assayed for PDR5-lacZ-dependent ␤-galactosidase activity to ensure that PDR5 was also overproduced. The 14 transformants that exhibited both enhanced cycloheximide This study resistance and overproduction of PDR5 were crossed to an isogenic wild-type strain and subjected to standard genetic analysis to ensure that the increased cycloheximide resistance and PDR5-lacZ expression were linked to a single transposon insertion. Of the 14 recovered strains, 5 isolates were found that exhibited 100% linkage of the transposon to both phenotypes. The site of transposon insertion was determined by plasmid rescue as described (25). The five different transposon insertions were found in three different nuclear genes: FZO1, OXA1, and TIM17 (Fig. 1). Three different insertions were found in FZO1, with single isolates found in OXA1 and TIM17. The insertions in FZO1 and OXA1 were all within the coding sequences of these loci while the insertion at TIM17 was found 11 bp downstream of the translation stop for this gene.
Each of the genes affected by these transposon insertions is involved in mitochondrial function in S. cerevisiae. Fzo1p spans both mitochondrial membranes, contains a cytoplasmic GTPase domain and is required for normal mitochondrial fusion and maintenance of the mitochondrial genome (16). Oxa1p spans the inner mitochondrial membrane (34) and is required for proper assembly of cytochrome c oxidase (complex IV) and the ATP synthase (complex V) (35). Tim17p is a component of the preprotein import machinery of the mitochondrial inner membrane (36). Since TIM17 is an essential locus (18), the transposon insertion likely depresses but does not eliminate production of Tim17p. All these mutant strains failed to grow on medium with glycerol/ethanol as the carbon sources, consistent with their predicted defect in mitochondrial function.
One concern underlying the isolation of these transposon insertion alleles is the possible production of truncated gene products that could possibly give rise to the observed phenotypes. To ensure that the effects on drug resistance and PDR5 expression seen in these transposon mutants were due to loss of gene function, we constructed two deletion derivatives in which either the entire coding sequence of FZO1 or a large segment of the OXA1 coding sequence was removed and replaced with kanMX4 or S. cerevisiae HIS3, respectively. The essential role of TIM17 precluded a similar type of analysis for this locus and the transposon-generated allele was assayed for comparison to the gene deletion alleles of FZO1 and OXA1.
These deletion mutants were then tested for drug resistance using medium containing a gradient of increasing concentration of cycloheximide (Fig. 2).
Loss of the normal FZO1 or OXA1 coding sequences caused a strong elevation of cycloheximide tolerance. Direct comparisons of the insertion mutations present in the original transposon insertion isolates with the deletion derivatives indicated that all alleles described here of either FZO1 or OXA1 behave identically (data not shown). Throughout this study, these different alleles were used interchangeably to facilitate strain construction.
PDR5-dependent ␤-galactosidase was also assayed to quantitate the effects of these mutants on expression of PDR5 (Fig.  2). Expression of a gene that is not a member of the Pdr1p/ Pdr3p regulon, a gene involved in tryptophan biosynthesis (TRP5 (37)), was also assayed as a control for the specificity of the effect of loss of FZO1, OXA1, or TIM17.
Wild-type cells carrying a PDR5-lacZ fusion gene produce approximately 72 units/OD of ␤-galactosidase activity. Expression of PDR5 increased 4-fold to 308 units/OD upon loss of OXA1. The ⌬fzo1 and tim17 mutations both led to over a 10-fold increase in PDR5 expression generating 834 and 1116 units/ OD, respectively. PDR5-lacZ expression in the ⌬fzo1 and tim17 mutant backgrounds is comparable to the highest observed PDR5 expression generated by gain-of-function alleles of Pdr1p and Pdr3p (11,38). Expression of the TRP5-lacZ gene fusion showed little response to alterations in mitochondrial function, varying from 22 units/OD in the wild-type and oxa1 strains to approximately 13 units/OD of ␤-galactosidase activity in the ⌬fzo1 and tim17 backgrounds.
Loss of FZO1, OXA1, or TIM17 Does Not Lead to Wide Scale Induction of Stress-responsive Gene Expression-While expression of the TRP5 gene is not significantly influenced in this collection of mitochondrial mutant strains, it is possible that compromising the mitochondrial function might elicit stress responses in the cell that would not be expected to alter TRP5 expression. As previous work has suggested that PDR5 may be a stress-responsive gene (39), we examined the possibility that the mutants we isolated that led to PDR5 activation also induced expression of other stress-related genes. We analyzed a variety of stress-responsive genes and found that none exhibited the same level of induction as seen at the PDR5 promoter.
The stress genes tested included the cytosolic catalase gene CTT1 (40) and a small heat shock protein-encoding locus, HSP12 (41). Both of these genes are regulated by the stressresponsive transcription factors Msn2p and Msn4p (42,43). The yeast copperthionein-encoding gene, CUP1 (44), was also assayed for its response to these mitochondrial defects. CUP1 is responsive to activation of the yeast heat shock transcription factor (45,46). Finally, the oxidative stress-regulated genes GSH1 and TRX2 were examined in these different genetic backgrounds. Both GSH1 and TRX2 are regulated by the basic region-leucine zipper-containing transcription factor Yap1p and are induced upon oxidative stress in cells (26,47,48). TRX2 expression is also controlled by the stress-responsive transcription factor Skn7p (49). Together, this collection of genes represents targets for most of the major stress-responsive transcription factors in the cell.
CUP1-lacZ expression increased 2-fold in an oxa1 background and 300% in the ⌬fzo1 and tim17 backgrounds (Fig. 2). HSP12-lacZ expression increased 2-fold in the ⌬fzo1 and tim17 strains but was unchanged in the presence of the oxa1 strain. CTT1-lacZ, GSH1-lacZ, and TRX2-lacZ expression was unaffected by any of the mutant backgrounds. The 2-3-fold increase in CUP1 and HSP12 expression was consistent with the generation of a mild stress response upon loss of the normal function of either FZO1 or TIM17. However, in these same genetic backgrounds, PDR5 expression was increased by 10-fold. We interpret these data to support the idea that the elevation in PDR5 expression represents one of the normal avenues used to respond to loss of mitochondrial function rather than a global activation of stress-responsive genes.
Loss of the Mitochondrial Genome Is Sufficient to Induce PDR5 Expression-An important feature of cells lacking the FZO1 gene is the concomitant loss of the mitochondrial genome (16). This absence of mitochondrial DNA is referred to as the o state (50) and importantly cannot be complemented through introduction of the wild-type FZO1 gene (16). To determine if the observed activation of PDR5 expression and drug resistance was due to loss of FZO1 or the cells becoming o , we introduced a low-copy plasmid carrying FZO1 back into the ⌬fzo1 strain. Although these transformants contained wildtype FZO1 gene function, the high level PDR5 expression and drug resistance phenotype was not affected (data not shown). This suggested that loss of the mitochondrial genome was sufficient to trigger elevation of PDR5 expression. To evaluate this possibility, o derivatives of two different wild-type strains were generated corresponding to our standard wild-type strain (SEY6210) or a second commonly used wild-type strain (W303).
Loss of the mitochondrial DNA in either wild-type strain produced a marked increase in cycloheximide tolerance (Fig. 3). Similarly, PDR5 expression was increased by nearly 6-fold in the o derivative of SEY6210 compared with the ϩ form. This analysis strongly suggests that loss of the mitochondrial genome results in a signal that acts to up-regulate PDR5 expression and corresponding drug resistance.
The finding that a chemically-induced o strain also exhibited high level PDR5 expression and cycloheximide resistance suggested that any mutant strain lacking normal mitochondrial function would potentially display these same resistance phenotypes. To address this issue, a mutant strain that lacked the ␤-subunit of the F1 mitochondrial ATPase was employed. This mutant strain is unable to grow on nonfermentable carbon sources due to an elimination of oxidative phosphorylation. The ⌬atp2 strain was transformed with a PDR5-lacZ reporter plasmid and with a low-copy number vector alone or carrying wild-type ATP2. Expression of PDR5 and the cycloheximide resistance phenotype of these transformants were then assayed (Fig. 4).
Loss of the ATP2 gene did not increase cycloheximide resistance and caused a modest 2-fold enhancement of PDR5 expression. Assay of TRP5-lacZ expression in the same genetic backgrounds demonstrated that TRP5-dependent ␤-galactosidase activity decreased from 85 units/OD in the ATP2 cells to 47 units/OD upon loss of this locus. While the ⌬atp2 strain is petite and lacks the ability to carry out oxidative phosphorylation, there is a minimal effect on PDR5 gene expression and no detectable effect on cycloheximide tolerance. Note that the oxa1 strain is also petite but exhibits a larger increase in PDR5 expression as well as cycloheximide resistance. These data FIG. 2. Drug resistance and expression phenotypes of strains lacking normal mitochondrial function. A, wild-type (SEY6210) and isogenic derivatives carrying the indicated gene disruptions were grown to mid-log phase and assayed for cycloheximide resistance on YPD medium containing a gradient of drug. The concentration of cycloheximide increases from left to right as denoted by the bar of increasing width. The highest concentration of drug is 0.5 g/ml cycloheximide. 1000 cells of each type were placed along the gradient and the plate allowed to develop at 30°C. B, wild-type cells or isogenic mutant strains were transformed with low-copy number plasmids containing the indicated lacZ gene fusions. Transformants were grown in minimal medium to mid-log phase and assayed for plasmid-dependent ␤-galactosidase activity as described (23). Activities are expressed as units/A 600 of cells. support the idea that not all petite mutants will lead to the same PDR5-dependent phenotypes.
Intact PDREs Are Necessary and Sufficient for Response to Mitochondrial Signals-Previous analyses of the PDR5 promoter have indicated the central importance of three binding sites for the transcriptional regulatory proteins Pdr1p and Pdr3p in normal expression of this gene (13). These binding sites are referred to as PDREs and are found in the promoters of all genes regulated by Pdr1p or Pdr3p (13). To address the role the three PDREs in the PDR5 promoter in the transcriptional response to these mutants lacking normal mitochondrial function, the expression of a wild-type PDR5-lacZ fusion gene was compared with that of a mutant variant lacking the three PDREs. The wild-type and PDRE-less PDR5-lacZ fusions were introduced into the various mutant backgrounds and PDR5-dependent ␤-galactosidase activity measured (Table II).
Removal of the PDREs from the PDR5 promoter (mPDR5) abolished expression of the PDR5-lacZ fusion and eliminated any response to the mutant backgrounds. One concern with this result arose due to the potential presence of other transcriptional regulatory elements in the PDR5 promoter. Although this experiment demonstrates that the PDREs in the PDR5 5Ј noncoding region are required for the up-regulation by mitochondrial mutants, it does not indicate that this effect is mediated through these specific binding sites.
To address this issue, a different reporter plasmid was used. This plasmid contained a single PDRE from the PDR5 pro-moter in place of the normal upstream activation sequences of a CYC1-lacZ gene fusion (PDRE-CYC1-lacZ) or an identical plasmid containing a mutant form of the PDRE to which neither Pdr1p nor Pdr3p could bind in vitro (mPDRE-CYC1-lacZ). These two plasmids were introduced into wild-type and ⌬fzo1 strains, followed by measurement of CYC1-directed ␤-galactosidase levels (Table III).
The PDRE-CYC1-lacZ fusion plasmid produced 4.5-fold more ␤-galactosidase activity in the ⌬fzo1 strain than when in a FZO1 background. Introduction of the mutant PDRE into this same plasmid eliminated the ⌬fzo1 induction of lacZ expression. These data strongly support the involvement of the PDREs in PDR5 as recipients of the response to mitochondrial dysfunction in these mutant backgrounds.
Mitochondrial Signals Require Pdr3p but Not Pdr1p to Activate PDR5 Expression-While the data above implicate the PDR5 PDREs as important participants in the response of this gene to mitochondrial defects, Pdr1p, Pdr3p, or both of these factors could be mediating the observed activation of gene expression. To determine the relative contribution of these zinc finger-containing transcription factors to PDR5 induction, various combinations of ⌬pdr1 and ⌬pdr3 disruption mutations FIG. 3. Loss of the mitochondrial genome activates PDR5 expression and increases cycloheximide resistance. A, wild-type cells were treated with ethidium bromide to produce a derivative lacking mitochondrial DNA (24). A selected o cell was transformed with low-copy number plasmids carrying a TRP5-or PDR5-lacZ fusion gene. Transformants were grown to mid-log phase and assayed for ␤-galactosidase activity as described above. B, an ethidium bromide-treated o version of W303 was prepared. Our standard wild-type strain and its o derivative (SEY6210 and SEY6210 o ) along with W303 and its o derivative were grown to an A 600 of 1 and assayed for cycloheximide resistance using a gradient plate.

FIG. 4.
A mutant defective in oxidative phosphorylation has no effect on PDR5 gene function. A, a mutant strain lacking the gene encoding the ␤ subunit of the mitochondrial ATPase was transformed with pRS314 (⌬atp2) or with a plasmid containing the wild-type ATP2 gene (YCplac-ATP2: ATP2). Along with these plasmids, the TRP5-or PDR5-lacZ fusion plasmids were also introduced. Appropriate transformants were grown to mid-log phase and assayed for ␤-galactosidase activity. B, the relative cycloheximide resistance of cells lacking (⌬atp2) or containing (ATP2) the ATP2 gene was assessed by spot test assay using a gradient plate as described above. were constructed in ⌬fzo1 or oxa1 genetic backgrounds. All strains were assayed for their cycloheximide resistance phenotype. The ⌬fzo1 strains were also transformed with either PDR5-or TRP5-lacZ fusion plasmids and appropriate transformants then assayed for ␤-galactosidase activity. Loss of the PDR3 gene completely ablated the induction of both PDR5 expression and cycloheximide resistance seen in the ⌬fzo1 strain (Fig. 5 and Table IV). In contrast, removal of the PDR1 gene had no significant effect on either phenotype. Loss of both genes further reduced expression of PDR5 and drug resistance. Expression of the TRP5-lacZ fusion was unaffected by these different strain backgrounds. Similarly, the increased cycloheximide resistance produced in the oxa1 background was eliminated upon disruption of PDR3, but was unaffected by loss of PDR1. These data indicate that the elevation of PDR5 expression and accompanying drug resistance requires the presence of PDR3 but not PDR1.
Components of the Retrograde Signaling Pathway Are Important for Propagating the Signal from ⌬fzo1/rho 0 but Not from oxa1-The up-regulation of Pdr3p function in response to mitochondrial defects strongly resembles a previously described regulatory network termed retrograde regulation (51). Retrograde regulation refers to the nuclear response to reduction of mitochondrial function and involves activation of the expression of several different genes. Three genes have been identified as critical for this transcriptional up-regulation: RTG1 and RTG3 encoding two related basic helix-loop-helix transcription factors and RTG2, a gene encoding a potential ATPase (51,52). Loss of any of these genes prevents o cells from increasing the expression of the citrate synthase-encoding locus, CIT2 (53). To determine if the retrograde signaling pathway and Pdr3p interacted, disruption mutations in RTG1 and RTG2 were constructed and assayed for drug resistance and PDR5 expression phenotypes.
Elimination of RTG1 or RTG2 from our wild-type background strain caused a modest decrease in both cycloheximide tolerance and PDR5-lacZ expression levels (Fig. 6). Similarly, introduction of the ⌬rtg1 or ⌬rtg2 mutation into a oxa1 background had no effect on either PDR5-dependent phenotype. Interestingly, TRP5-lacZ expression consistently increased, possibly due to the amino acid imbalance generated by the rtg alleles (51). In contrast, removal of either RTG locus from ⌬fzo1 cells caused a 50% reduction in PDR5-lacZ expression and a similar decline in cycloheximide resistance. These findings indicate that oxa1 activation of PDR5 expression is independent of the function of Rtg1p and Rtg2p, whereas a significant component of the ⌬fzo1 signal requires an intact retrograde signaling pathway.
PDR3 Requires Autoregulation and RTG1 for Induction in 0 Cells-Although Pdr1p and Pdr3p share 36% sequence identity across their length, a striking difference between the two structural genes encoding these proteins is the presence of two PDREs in the PDR3 promoter which are not found upstream of the PDR1 gene (54). Previous studies have provided evidence that PDR3 is subject to autoregulation that requires the pres-ence of these PDREs (54). To determine if activation of Pdr3p function upon loss of normal mitochondrial function involved an increase in PDR3 expression, we constructed a PDR3-lacZ fusion gene and introduced this reporter construct into several different strains to evaluate PDR3 gene expression (Fig. 7).
Expression of PDR3 was very low in wild-type cells but was dramatically up-regulated in o cells. Introduction of a ⌬rtg1 allele into o cells decreased PDR3-lacZ expression to less than 40% the level seen in RTG1 o cells. Finally, the presence of the wild-type PDR3 structural gene was required for the induction of PDR3-lacZ expression in the o genetic background, strongly suggesting that autoregulation is the cause of the observed elevation of PDR3 expression.
These data indicate that the observed increase in Pdr3p response to loss of FZO1 Wild-type or ⌬fzo1 cells were transformed with the indicated CYC1-lacZ fusion plasmids containing a single copy of a wild-type or mutant PDRE from the PDR5 promoter. ␤-Galactosidase activities were determined as described in Table II. 5. Activation of PDR5 gene function by mitochondrial defects requires the presence of the PDR3 gene. A, an isogenic series of strains containing the indicated alleles of FZO1, PDR1, and PDR3 were assayed for cycloheximide resistance using a gradient plate. B, an isogenic series of strains varying at the OXA1 locus, PDR1, or PDR3 loci were tested for relative resistance using a cycloheximide gradient plate. The finding that PDR3 expression is elevated in response to mitochondrial defects suggests that the promoter of this gene serves as the link between mitochondrial status and PDR5 expression. To explore this possibility, we constructed two different chimeric genes by exchanging the promoters of PDR1 and PDR3. In this fashion, we generated a PDR3 structural gene that responded to the transcriptional signals of PDR1 (PDR1:PDR3), as well as the reciprocal construct (PDR3:PDR1). The goal of this experiment was to evaluate if the observed elevation of drug resistance seen in o cells was linked to the presence of the PDR3 promoter only or if the PDR3 gene product was the target for response to mitochondrial defects. The chimeric genes were introduced into o or ϩ cells lacking both PDR1 and PDR3 as well as o ⌬pdr1,pdr3 cells carrying the ⌬rtg1 allele. Appropriate transformants were then assayed for their relative cycloheximide resistance (Fig. 8).

␤-Galactosidase activity (units/A 600 )
Only the presence of the wild-type PDR3 gene or the PDR1: PDR3 chimeric gene supported the elevation in cycloheximide resistance seen in o cells. Neither the PDR1 gene nor the PDR3 promoter driving expression of the PDR1 structural gene (PDR3:PDR1) conferred increased cycloheximide resistance in the presence of a o lesion. These data provide important support for the view that the response to mitochondrial deficiency comes directly through the PDR3 gene product, which is both necessary and sufficient to enhance cycloheximide resistance.
Unlike the clear dependence on the presence of the PDR3 coding sequence seen for response to the o state of the cell, loss of the RTG1 gene appeared to decrease drug resistance in all genetic backgrounds. We interpret this result to indicate that loss of Rtg1p reduces the overall fitness of cells in addition to the specific effect on PDR3 expression. DISCUSSION One of the challenges in the analysis of multidrug resistance protein is to understand the nature of the physiological role of FIG. 6. Retrograde signaling pathway genes affect the response of PDR5 to mitochondrial defects. Strains containing the indicated alleles of RTG1, RTG2, OXA1, or FZO1 were tested for their relative cycloheximide resistance by gradient plate assay (left-hand figure). Additionally, these same strains were transformed with either the TRP5-or PDR5-lacZ fusion genes. Transformants were grown to mid-log phase and assayed for ␤-galactosidase activity (right hand  table). containing the indicated alleles of RTG1 or PDR3 were transformed with a low-copy number plasmid carrying a PDR3-lacZ or PDR1-lacZ fusion gene. Transformants were grown in SC medium and assayed for PDR3-dependent lacZ expression using a chemiluminescent assay (CLONTECH). A, the structure of the PDR3-lacZ fusion plasmid is shown at the top of the figure. The two PDREs have been previously shown to be involved in autoregulation of the gene (54). B, levels of ␤-galactosidase activity produced by either fusion gene in the various genetic backgrounds are listed. the genes encoding these factors. The data reported here indicate that a physiological activator of Pdr3p and its target genes is the status of the mitochondria. Although the precise nature of the Pdr3p-inducing signal cannot be gleaned from this work, two common defects in all the mutants we have examined is the absence of a functional electron transport chain and lack of a normal F 0 complex of the mitochondrial ATPase. Mutants lacking Oxa1p fail to normally assemble cytochrome c oxidase (55) while o cells (or ⌬fzo1 strains) lack mitochondrial genes that encode subunits of cytochrome c oxidase (50). Similarly, oxa1 and o cells do not properly produce the F 0 complex of the mitochondrial ATPase, although the soluble F 1 component is apparently normal (56). It is possible that the critical component that is under surveillance by Pdr3p are these integral membrane protein complexes. Alternatively, it may be the lack of normal electron transport chain function and/or F 0 activity that causes activation of Pdr3p. Since we recovered only one oxa1 allele, we believe that the transposon screen is not saturated for mutations that could lead to activation of Pdr3p with subsequent elevation of cycloheximide resistance. Assessing drug resistance and PDR5 expression phenotypes of other mutations lacking wild-type electron transport chain function should shed light on the specific contribution of these proteins to Pdr3p regulation.
Irrespective of the specific signal being perceived by Pdr3p, the receptor of this signal is Pdr3p itself rather than another factor that up-regulates PDR3 expression. Our expectation is that the center region of Pdr3p is the ultimate target site for the mitochondrially-generated signal by analogy with other Cys 6 -Zn(II) zinc cluster proteins like Leu3p (57,58) and Gal4p (59). Both of these proteins are positively regulated through changes in function that are controlled by modulation of the center domain of these transcriptional regulators (60 -62). Computer alignments of a large number of Cys 6 -Zn(II) zinc cluster proteins suggest the presence of a conserved structural motif in this family of transcription factors (63). Interestingly, mutations that lead to high level activity of either Pdr1p or Pdr3p map to the center regions of these proteins (11,38), supporting the idea that these domains of Pdr1p and Pdr3p will be critical in controlling their function.
The identification of Pdr3p as another member of the retrograde signaling response in S. cerevisiae provides important insight into the complex nature of the physiological processes that are induced by mitochondrial dysfunction. Analyses of RTG genes have shown that loss of these factors produces important phenotypes even in ϩ cells. For example, elimination of the RTG1 gene results in a glutamate auxotrophic strain (51). Loss of PDR3 does not have a similar effect as ⌬pdr3 cells can grow in the absence of glutamate supplementation. 2 Likewise, loss of RTG1 has a relatively modest effect on drug resistance. Clearly, different loci are responsive to RTG-and PDR-encoded transcription factors.
While RTG and PDR genes appear to control distinct sets of 2 T. C. Hallstrom and W. Scott Moye-Rowley, unpublished data.
FIG. 8. The response of PDR3 to the absence of the mitochondrial genome maps to the coding sequence of the gene. Cells that lacked both the PDR1 and PDR3 loci as well as o and o , ⌬rtg1 derivatives were transformed with the indicated plasmids expressing various forms of PDR1 or PDR3. All plasmids are based on the low-copy number vector pRS316 (28) and express wild-type PDR1 (PDR1), PDR3 (PDR3), or PDR3 coding sequence under control of the PDR1 promoter (PDR1:PDR3) or the PDR1 coding sequence under control of the PDR3 promoter (PDR3:PDR1). Transformants were grown under selective conditions and assayed for cycloheximide resistance using a gradient plate.
genes, there is strong evidence for interaction between these two pathways that both respond to mitochondrial defects. Cycloheximide resistance, PDR3 and PDR5 expression is depressed in cells that lack either RTG1 or RTG2, indicating that these genes play a role in PDR gene function. The mutants we have isolated suggest that two different signal pathways feed into modulation of Pdr3p. These pathways can be discriminated by the involvement of the RTG genes in the ultimate activity of Pdr3p. Loss of the mitochondrial genome ( o , fzo1 mutants) requires RTG gene function for normal up-regulation of Pdr3p but oxa1 mutants do not. A diagram of the genetic interactions we have observed is shown in Fig. 9. While the interaction of Pdr3p with RTG genes is clear, more work is required to clarify at what level RTG genes impact on PDR gene function. Inspection of the PDR3 promoter for Rtg1p/ Rtg3p-binding sites suggested that these factors are not likely to directly control expression of this locus (data not shown). Irrespective of the exact mechanisms at work here, PDR3 and the RTG genes appear to act together to elicit response to mitochondrial defects.
Finally, these studies provide valuable new insight into the physiological connections regulating PDR gene function, especially control of Pdr3p. Previously, we have observed distinct colony morphologies that appear on cycloheximide-containing medium upon loss of PDR1 or PDR3 alone (14). A small number of highly cycloheximide-resistant cells (Ͻ5%) were found in wild-type or ⌬pdr1 cells but not in cells lacking the PDR3 gene. This small number of cells correlates well with the estimated occurrence of o cells in a population of growing S. cerevisiae (50). An attractive explanation for this behavior is that in this small number of o cells, Pdr3p is activated and PDR5 expression increases to very high levels. Understanding the nature of the Pdr3p-inducing signal elicited upon loss of normal mitochondrial activity and why activation of PDR gene expression is an important response to this physiological problem represent the next set of experimental goals. FIG. 9. Model for interaction of RTG and PDR loci. A scheme outlining the genetic interactions between the RTG and PDR genes described in this work is shown. Loss of the FZO1 locus produces a o cell which has previously been shown to activate function of RTG genes (51). Since oxa1 induction of PDR5 expression is insensitive to loss of RTG genes, we suggest that oxa1 and o cells share a common pathway of Pdr3p activation with o cells having an additional RTG-dependent component.