Zinc cluster protein Rdr1p is a transcriptional repressor of the PDR5 gene encoding a multidrug transporter.

The yeast PDR5 gene encodes an efflux pump that confers multidrug resistance. Expression of PDR5 is positively regulated by the transcription factors Pdr1p and Pdr3p that recognize the same pleiotropic drug resistance elements (PDREs) in the PDR5 promoter. Pdr1p and Pdr3p belong to the Gal4p family of zinc cluster proteins. The function of RDR1 (YOR380W), which also encodes a member of this family, is unknown. To identify target genes for Rdr1p, we have performed whole-genome analysis of gene expression with DNA microarrays. Our results show that Rdr1p is a transcriptional repressor of five genes, including PDR5. A Deltardr1 strain has increased resistance to cycloheximide, as expected from the overexpression of PDR5. In addition, the activity of a PDR5-lacZ reporter is increased in a Deltardr1 strain. All (but one) genes affected by removal of Rdr1p contain PDREs in their promoters. We tested if the effect of Rdr1p is mediated through PDREs by inserting this DNA element in front of a minimal promoter. Activity of this reporter was increased in a Deltardr1 strain. Moreover, mutations known to reduce binding of Pdr1/Pdr3p abolished the induction observed in the Deltardr1 strain. Thus, we have identified a transcriptional repressor involved in the control of multidrug resistance.

Multidrug resistance is a widespread phenomenon that allows organisms ranging from bacteria to humans to defend themselves against a variety of toxic compounds. Drug resistance is mediated through membrane-bound transporters that act as drug efflux pumps. This process has been extensively studied in the yeast Saccharomyces cerevisiae. Two major classes of multidrug transporters have been identified: the major facilitator superfamily (MFS) 1 (1) and the ABC (ATP binding cassette) family of transporters (2)(3)(4). ABC transporters act in an ATP-dependent manner, and their overexpression leads to increased drug resistance. Members of the family of ABC transporters include Pdr5p, Snq2p, and Yor1p (2).
The transcriptional activators Pdr1p and Pdr3p control the expression of many ABC transporters (3,4). These activators belong to a family of transcriptional regulators called zinc cluster or binuclear cluster proteins (5)(6)(7). Members of this family contain six highly conserved cysteines that coordinate binding to two zinc atoms to allow proper folding of the DNA binding domain. The cysteine-rich region (or zinc finger) is usually followed by a short linker sequence that bridges the zinc finger to a dimerization domain (6). This domain is involved in recognition of specific DNA sequences through interaction of the zinc finger with DNA (for references see Ref. 8). Pdr1p and Pdr3p both recognize CGG triplets oriented in opposite directions (CCGCGG) to form an everted repeat (9). This motif is found in the target genes of Pdr1p and Pdr3p (10 -17). For example, three binding sites called PDREs (pleiotropic drug response elements) for Pdr1p and Pdr3p have been mapped in the PDR5 promoter (10,12). Thus, Pdr1p and Pdr3p recognize the same DNA sequences in target promoters for activation of transcription. In addition, the PDR3 promoter also contains two PDREs and is positively regulated by its own gene product and by Pdr1p (18).
Deletion of PDR1 or PDR3 results in increased sensitivity to drugs while a double knockout is hypersensitive to drugs. Moreover, increased drug resistance is primarily caused by mutations in PDR1 and PDR3. These mutants are hyperactive transcription factors (19,20) resulting in increased expression of target genes (such as PDR5, SNQ2, YOR1) and, hence, increased drug resistance. Whole-genome analysis of gene expression with hyperactive forms of Pdr1p and Pdr3p (or a chimeric Pdr1p activator) has allowed the identification of additional targets for these transcriptional activators (21,22). These targets comprise genes encoding MFS and permeases (e.g. YOR049C) as well as genes involved in lipid and cell wall metabolism.
The yeast genome contains 55 genes encoding putative zinc cluster proteins (for a complete list see Akache et al. (7) and Ref. 23). Many of these genes are uncharacterized and their function unknown. To better understand the role of these genes, we have performed a phenotypic analysis of 33 members of the family of zinc cluster proteins (7). For example, deletion of YOR380W results in inability to grow on non-fermentable carbon sources and hypersensitivity to calcofluor white, a compound that has high affinity to chitin, a cell wall component (7). However, the function and the target genes of YOR380W are unknown. To identify target genes of YOR380W, we have performed whole-genome analysis of gene expression with DNA microarrays. Our results show that a very limited number of genes have their expression altered by deletion of YOR380W. Strikingly, all the affected genes are up-regulated in the deletion strain, and they predominantly encode membrane proteins, including ABC transporters such as Pdr5p. Moreover, deletion of YOR380W results in increased resistance to cycloheximide. These data suggest that YOR380W is a transcrip-tional repressor; hence, it was named RDR1 for repressor of drug resistance.

MATERIALS AND METHODS
Strains and Media-S. cerevisiae strains used in this study are derived from FY73 (24) and are listed below in Table I. Deletion of the PDR5 open reading frame (ORF) was performed using the PCR method of Baudin et al. (25) using oligos with 45 nucleotides of homology to the target gene at their 5Ј-end and 14 nucleotides complementary to the KanMX (G418 R ) selection marker. Plasmid pFA6 (26) was used as a template for PCR with the oligonucleotides CTTTTAAGTTTTCGTAT-CCGCTCGTTCGAAAGACTTTAGACAAAACAGCTGAAGCTTCG and GTCCATCTTGGTAAGTTTCTTTTCTTAACCAAATTCAAAATTCTAG-CATAGGCCACTAGTGGATCTG. Strains KH51 and KH52 were obtained by deleting the PDR5 ORF in strains FY73 and FZP, respectively. Deletions were verified by Southern blot analysis using a probe located upstream of the ATG of the PDR5 ORF. Yeast extract/peptone/dextrose medium (YPD) and synthetic dextrose medium (SD) were prepared according to Ref. 37.
Microarray Analysis-Yeast cells (strains FY73 and FZP, Table I) were grown in rich medium (YPD (27)), to an A 600 of 0.8 -1.0, and total RNA was isolated by the hot phenol procedure (28). RNA was further purified with Qiagen columns according to the manufacturer's protocol except that RNA was eluted for 15 min. cDNA labeling and hybridization were performed exactly as described previously (29,30). Custommade yeast whole-genome microarrays (Ͼ6200 yeast ORFs) were obtained from the Microarray Center at the Ontario Cancer Institute (Toronto, Canada). Scanning and quantification were performed exactly as described (31). Briefly, chip A was hybridized with a mixture of wild-type cDNA labeled with Cy3 (ϭWT-Cy3) and Cy5 (ϭWT-Cy5); chip B, mixture of WT-Cy3 and ⌬rdr1-Cy5; chip C, mixture of ⌬rdr1-Cy3 and WT-Cy5; chip D, mixture of ⌬rdr1-Cy3 and ⌬rdr1-Cy5. The ratio of ⌬rdr1/WT of each ORF obtained from chips B and C were normalized with the corresponding ratio of the same ORF from chips A and D. Because each ORF is duplicated on the same chip, four ratios obtained from chips B and C were normalized individually with the four ratios obtained from chips A and D; i.e. 16 values were obtained for each ORF. The results presented in Table II are an average of two independent experiments performed with independent RNA preparations.
Southern and Northern Blot Analysis-Genomic DNA was isolated according to Ref. 32. Southern and Northern blot analyses were done according to standard procedures (33). Hybridizations were performed at 42°C in 50% formamide, 1 M NaCl, 2.8ϫ Denhardt's solution, 0.5% SDS, and 10% dextran sulfate.
Probes for Northern Blot Analysis-The PDR16 probe was obtained by PCR using genomic DNA with the oligos CGGAATTCATATTCCTT-GGTTAGCATGG and CGGAATTCGGTACTGCTTTCCGATTTTA to amplify nucleotides ϩ731 to ϩ1050 (relative to the ATG) of the PDR16 gene. The PDR5 probe (containing ϩ111 to ϩ447 bp relative to the ATG) was obtained by digesting plasmid pPCR-PDR5 (34) with EcoRI and the fragment was gel-purified. PCR products or fragments were randomly labeled (33) for Northern blot analysis. The actin probe has been described elsewhere (30).
LacZ Reporters-A PDR5 reporter was constructed by amplifying its promoter using genomic DNA isolated from strain YPH499 (35) and the oligos ATCGACTCGAGGTCTTCCTCTTTGATTCCA and CGGGATC-CCATTTTTGTCTAAAGTCTTTCG. The PCR product was cut with XhoI and BamHI and subcloned into plasmid pSLF⌬178K, a high copy plasmid with a URA3 selection marker (36) cut with BamHI and XhoI to give PDR5-lacZ-2. The reporter contains about 1000 bp of sequences upstream of the ATG. The ATG of the promoter is used to initiate translation of the lacZ gene. A reporter for chromosomal integration was obtained by deleting 2-m sequences by cutting with HindIII and re-ligating the backbone to give PDR5-lacZ-IP. The resulting plasmid was linearized by cutting at the unique ApaI site located in the URA3 marker. DNA was transformed into FY73 or FZP (Table I) and colonies selected on plates lacking uracil. Proper integration of the reporter was verified by Southern blot analysis. A low copy version of the reporter was obtained by amplifying autonomously replicating sequence and centromere sequence with oligos ATCAGACAAGCTTTTCCCCGAAAA-GTGCCAC and ATGACTTAAGCTTAAAGGGCCTCGTGATACG using plasmid pRS314 (35) as a template. The PCR product was cut with HindIII and subcloned into the HindIII site of PDR5-lacZ-IP. The same strategy was used to generate reporters for the SNQ2 gene (using oligos ATCGACTCGAGTGGAAAAGAACGGAGACATA and CGGGATCCCA-TTGAATTCTCTTTACGTA). The SNQ2 reporter contains ϳ700 bp of promoter sequences upstream of the ATG codon. Promoter sequences were verified by DNA sequencing.
LacZ reporters containing a Pdr1p/Pdr3p binding site (or mutants) upstream of a minimal CYC1 promoter were constructed by inserting double-stranded oligos into the XhoI site of pSLF⌬178K (36). Oligos correspond to site number 3 of the PDR5 promoter (12) and span sequences Ϫ372 to Ϫ337 bp relative to the ATG. Reporter PDRE3-CYC1 was constructed using the oligo TCGAAAAAGAGAAATGTCTCCGCG-GAACTCTTCTACGCCG and its complement TCGACGGCGTAGAAG-AGTTCCGCGGAGACATTTCTCTTTT. PDRE3A-CYC1 with TCGAAA-AAGAGAAATGTCTCTGCGGAACTCTTCTACGCCG and TCGACGG-CGTAGAAGAGTTCCGCAGAGACATTTCTCTTTT and PDRE3B-C-YC1 with TCGAAAAAGAGAAATGTCTCCGCAGAACTCTTCTACG-CCG and TCGACGGCGTAGAAGAGTTCTGCGGAGACATTTCTCT-TTT (mutations are underlined). PDREs are in the same orientation and were verified by DNA sequencing.
␤-Galactosidase Assays-Strains FY73 and FZP were transformed with reporters. Colonies were grown overnight in YPD and diluted about 500-fold in minimal medium (synthetic dextrose (37)) supplemented with appropriate amino acids and adenine. ␤-galactosidase assays were performed with permeabilized cells (38). Values are the average of at least 2 independent experiments performed at least in duplicate. ␤-galactosidase assays for integrated reporters were performed as described above except that YPD was used instead of synthetic dextrose.

RESULTS
To identify genes regulated by Rdr1p, RNA was isolated from wild-type strain FY73 (24) and a strain carrying a deletion of the putative DNA binding domain of RDR1 (7). We then performed a whole-genome analysis of gene expression using DNA microarrays. Data were obtained on about 6000 genes and are an average of two independent experiments performed with duplicate genomes. First, we did not observe any gene whose expression was decreased more than 1.5-fold in the deletion strain (data not shown). Thus, these data strongly suggest that Rdr1p is not a transcriptional activator under the conditions tested (rich medium). On the contrary, this study revealed that five genes were overexpressed more than 2-fold in the deletion strain (Table II). Interestingly, four of these genes encode membrane-associated transporters. The strongest induction (6.3fold) was observed with PDR5. Two other genes (PDR15 and PDR16) also involved in drug resistance were up-regulated in the absence of the zinc cluster protein encoded by RDR1. Expression of ORF YOR049C was up-regulated by a factor of 4 (Table II). The function of this (putative) ORF is unknown. However, the sequence of the ORF predicts a protein of 354 amino acids with a high probability of possessing seven transmembrane domains (data not shown). Finally, PHO84, a member of the MFS family and encoding a high affinity inorganic phosphate/Hϩ symporter was up-regulated in absence of Rdr1p. Thus, our results suggest that Rdr1p is a transcriptional repressor.
To confirm the microarray results, we performed Northern blot analysis of RNA isolated from wild-type and ⌬rdr1 strains (Fig. 1). PDR5 mRNA level was increased about five times in the deletion strain in close agreement with the microarray analysis. Similarly, deletion of the RDR1 gene resulted in a modest increase of PDR16 mRNA (about 2-fold), again in agreement with the results generated with microarrays. Equal loading and transfer of RNA isolated for wild-type and knockout strains was shown by the similar signals obtained with an actin Rdr1p Is a Repressor of PDR5 Expression probe (Fig. 1, bottom). Thus, the microarray results for PDR5 and PDR16 were confirmed by an independent method.
Overexpression of ABC transporters has been shown to increase resistance to various drugs (4). For example, overexpression of the ABC transporter encoded by PDR5 results in greater resistance to the translation inhibitor cycloheximide (39 -41). Thus, we tested if removal of Rdr1p results in altered sensitivity to cycloheximide as predicted from the up-regulation of PDR5 in a strain deleted of RDR1. Cells were grown overnight in rich medium, serially diluted, and spotted on plates containing cycloheximide or not (Fig. 2). Increased resistance to cycloheximide was observed in a strain carrying a deletion of the RDR1 gene as compared with the parental strain FY73 (Fig. 2,  lower panel). Thus, we have identified an additional phenotype for cells lacking Rdr1p. Because deletion of RDR1 results in increased expression of genes (other than PDR5) that confer drug resistance (Table II), we tested whether the observed phenotype is mediated through PDR5 or not. We constructed two different strains: one that carries a deletion of the PDR5 gene and another that is deleted of both PDR5 and RDR1. As expected (39 -41), deletion of PDR5 results in hypersensitivity to cycloheximide (Fig. 2, bottom). However, deletion of RDR1 in a ⌬pdr5 background does not result in increased resistance unlike what is observed in a wild-type background. These results strongly suggest that increased resistance to cycloheximide in a ⌬rdr1 strain is mainly mediated by PDR5.
To determine if changes in PDR5 mRNA levels are due to altered promoter activity, we constructed a reporter containing the PDR5 promoter driving lacZ expression. The reporter was stably integrated into a wild-type strain and a strain carrying a deletion of the RDR1 gene. Cells were grown in rich medium (YPD) and assayed for ␤-galactosidase activity (Fig. 3). The activity was increased by about 10-fold in cells lacking Rdr1p (Fig. 3, lanes 1 and 2). As a control, we used a gene related to PDR5, SNQ2, which also encodes an ABC transporter. According to our microarray analysis, steady-state levels for SNQ2 RNA were unchanged in cells lacking Rdr1p (data not shown). A SNQ2 reporter stably integrated in the yeast genome showed similar activity in wild-type and ⌬rdr1 strains (Fig. 3, lanes 3 and 4). Similar results were obtained with low copy episomal reporters introduced into wild-type and ⌬rdr1 strains (Fig. 3, lanes 5-8).
The activity of the PDR5 reporter was increased about 5-fold in a ⌬rdr1 strain whereas a SNQ2 reporter showed a slight reduction in activity. In conclusion, these results strongly suggest that TABLE II Microarray analysis: genes whose expression is altered by removal of Rdr1p Genes whose mRNA levels are affected (more than 2-fold) by deletion of the RDR1 gene are listed. "Expression" refers to the ratio of a given mRNA level in the ⌬rdr1 strain as compared to the wild-type strain FY73.  Rdr1p negatively regulates the PDR5 promoter.
The PDR5 promoter has been shown to contain three sites that are recognized by the transcriptional activators Pdr1p and Pdr3p (12). The sites contain CGG triplets oriented in opposite direction (CCGCGG) forming an everted repeat (9). All genes (except PHO84) identified with the microarray analysis (Table  II) contain DNA sequences that match the consensus Pdr1p/ Pdr3p binding site in their promoters (Table III). Thus, Rdr1p may exert its repressive effect by acting on Pdr1p/Pdr3p recognition sites. To test this possibility, we inserted oligonucleotides corresponding to a Pdr1p/Pdr3p binding site (site number three in Ref. 12) in front of a minimal CYC1 promoter driving lacZ transcription. Insertion of a PDRE in front of a minimal CYC1 promoter greatly increased promoter activity when compared with the minimal promoter (Fig. 4, lanes 1 and 3). Strikingly, activity of the PDRE3-CYC1 reporter was increased about 8-fold when assayed in a ⌬rdr1 strain (Fig. 4, lanes 3 and  4) whereas the activity of a reporter lacking the PDRE remained at basal levels (Fig. 4, lanes 1 and 2). These results strongly suggest that the effect of Rdr1p on PDR5 transcription is mediated by the Pdr1p/Pdr3p binding site. We then assayed the activity of reporters that contain mutations in the PDRE that reduce binding of Pdr1p and Pdr3p. The mutations are located in either of the CGG triplets that are crucial for binding of Pdr3p (9). As expected, the activity of the mutant was reduced when tested in a wild-type background (Fig. 4, lanes 5  and 7). Mutations in either CGG triplet abolished the induction observed in a ⌬rdr1 strain (Fig. 4, lanes 5-8). Thus, this mutant analysis further suggests that the negative effect of Rdr1p on PDR5 expression is mediated by PDREs. Rdr1p and Pdr1p/ Pdr3p appear to function on highly related DNA sequences. DISCUSSION This study focused on a member of the Gal4p family of transcriptional regulators, RDR1 (YOR380W). Our previous work has shown that deletion of RDR1 results in absence of growth on non-fermentable carbon sources and sensitivity to calcofluor white, a phenotype associated with cell wall defects (42,43). To gain insights into the role of RDR1, we have performed whole-genome analysis of gene expression with RNA isolated from wild-type and cells lacking Rdr1p. No genes had their expression decreased by a factor of 1.5-fold or more in the ⌬rdr1 strain suggesting that Rdr1p is not a transcriptional activator. On the other hand, five genes had RNA levels increased by at least 2-fold in the deletion strain (Table II). Most of these genes encode membrane proteins.
Microarray data were confirmed by Northern blot analysis for PDR5 and PDR16 (Fig. 1). Furthermore, deletion of RDR1 results in increased resistance to cycloheximide (Fig. 2), a phenotype expected from the increased expression of the ABC transporter Pdr5p (39 -41). Even though removal of Rdr1p results in increased expression of multidrug resistance genes other than PDR5, genetic analysis showed that the major target of Rdr1p is PDR5 with regard to cycloheximide resistance (Fig. 2). The effect of Rdr1p is likely to be at the transcriptional level, because a PDR5-lacZ reporter mimicked the activity observed for the endogenous gene. For example, increased activity of the PDR5 promoter was observed with both integrated and episomal lacZ reporters when assayed in a strain deleted of RDR1 (Fig. 3). This effect was specific, because no increase in reporter activity was observed with a SNQ2 reporter when tested in a ⌬rdr1 background.
Consensus target DNA sequences for Pdr1p/Pdr3 are found in all (but one) promoters of genes identified in our analysis (Table III). Therefore, we focused on this DNA element. A reporter consisting of a minimal CYC1 promoter under the control of a single PDRE shows increased activity in absence of Rdr1p (Fig. 4). Thus, the DNA sequences involved in activation by Pdr1p/Pdr3p and repression by Rdr1p are the same or overlap. Moreover, PDRE mutants with decreased binding to Pdr3p (9) (and presumably Pdr1p) do not show increased activity in the absence of Rdr1p. In summary, our studies strongly suggest that Rdr1p acts through Pdr1p/Pdr3p binding sites. However, many genes (11,(13)(14)(15)(16)(17) (like SNQ2) that have been shown to contain PDREs and to be regulated by Pdr1p/ Pdr3p are not affected by removal of Rdr1p. Thus, the effect of Rdr1p appears to be mediated by a subset of the target genes of FIG. 3. Activity of a PDR5-lacZ reporter is increased in a ⌬rdr1 strain. ␤-Galactosidase activity was measured in wild-type (FY73, black bars) and ⌬rdr1 strains (FZP, hatched bars) containing reporters PDR5 or SNQ2 integrated at the URA3 locus as described under "Material and Methods." Alternatively, the same strains were transformed with the low copy reporters ("episomal") PDR5-lacZ or SNQ2-lacZ and assayed for ␤-galactosidase activity as described under "Material and Methods." ␤-Galactosidase units and standard deviations are given on the histogram.

Rdr1p Is a Repressor of PDR5 Expression
Pdr1p/Pdr3p. We have shown that the zinc cluster proteins Leu3p and Uga3p both recognize CGG triplets oriented in opposite directions (an everted repeat) and spaced by 4 bp (CCGN 4 CGG) (44). However, target genes of Leu3p and Uga3p are completely different. Discrimination of these highly related DNA sequences is achieved by nucleotides located between the CGG triplets (44). Thus, nucleotides outside the PDRE core sequence (CCGCGG) may allow targeting of Rdr1p to a subset of the Pdr1p/Pdr3p-responsive genes.
What is the mechanism of action of Rdr1p? It is possible that Rdr1p binds directly to PDREs found in promoters of target genes resulting in decreased activity. For example, Rdr1p could compete with Pdr1p and Pdr3p for binding to PDREs. A related model would involve the formation of heterodimers between Rdr1p and Pdr1p or Pdr3p. Such heterodimers would be impaired for activation of the PDR5 gene. In support of this model, zinc cluster proteins Oaf1 and Pip2p, involved in activation of peroxisomal genes, have been shown to bind to DNA as heterodimers (45). Ada3p, a subunit of the SAGA complex involved in chromatin remodeling, represses activity of Pdr1p (46). Thus, the action of Rdr1p may be mediated by Ada3p. Another possibility is that the effect of Rdr1p is indirect. Rdr1p may control the level of a transcription factor involved in activation of PDR5 gene expression. A number of hyperactive alleles of PDR1 and PDR3 have been isolated (19,20). One may speculate that these mutations relieve the inhibitory effect of Rdr1p. However, activation of a PDR5 reporter by some Pdr1p mutants is increased more than 80-fold as compared with wildtype Pdr1p (19) whereas the maximal effect achieved by removing Rdr1p is only 10-fold. Thus, inactivation of Rdr1p is unlikely to account in full for the hyperactivity of the Pdr1p/Pdr3p mutants.
Even though our microarray analysis points to a very limited number of affected genes, removal of Rdr1p results in an interesting phenotype, an increased resistance to cycloheximide. Previous studies (3, 4) have identified Pdr1p and Pdr3p as positive regulators of PDR5, and our laboratory has recently discovered an additional transcriptional activator of PDR5. 2 This study shows that PDR5 is also negatively regulated by Rdr1p. Thus, the regulation of the PDR5 gene expression appears to be more complex than initially anticipated. Further studies will be required to better understand the mechanism of action of RDR1, but the results presented here raise the possibility that similar transcriptional repressors would be present in pathogenic yeasts such as Candida albicans. FIG. 4. A Pdr1p/Pdr3p binding site mediates repression by Rdr1p. ␤-Galactosidase activity was measured in wild-type (FY73, black bars) and ⌬rdr1 strains (FZP, hatched bars) transformed with lacZ reporters driven by a minimal CYC1 reporter or a CYC1 containing a wild-type Pdr1p/Pdr3p binding site (PDRE3) or mutant sites (PDRE3A and PDRE3B) inserted upstream of the CYC1 promoter. The core sequence of the PDREs is shown in the lower panel with the mutations underlined. Reporters were assayed for ␤-galactosidase activity as described under "Material and Methods." ␤-Galactosidase units and standard deviations are given on the histogram.