The Tup1-Cyc8 Protein Complex Can Shift from a Transcriptional Co-repressor to a Transcriptional Co-activator*

Cyc8(Ssn6)-Tup1, a general co-repressor complex, is recruited to promoter DNA via interactions with DNA-binding regulatory proteins and inhibits the transcription of many different yeast genes. Previous studies have established that repression function of the complex is performed by one subunit of the complex, the Tup1 protein, and requires specific components of the RNA polymerase II holoenzyme such as Sin4 and Rgr1. In this study we test the transcriptional activity of the Cyc8 subunit using a lexAoperator-containing reporter. We show that a LexA-Cyc8 hybrid stimulates transcription when expressed in atup1Δ, a sin4Δ, or a rgr1Δstrain, suggesting that transcriptional activation is an intrinsic property of the Cyc8-Tup1 co-repressor. In support of this notion we demonstrate that Cyc8-Tup1 has a dual function on CIT2, a gene encoding a citrate synthase that is expressed upon mitochondrial dysfunction. First, we show that Cyc8-Tup1 is tethered toCIT2 promoter by interacting with the activation domain of Rtg3, a bHLH/L-Zip DNA-binding transactivator of CIT2. Next we demonstrate that Cyc8-Tup1 activates CIT2 transcription in response to mitochondrial dysfunction, and this stimulatory effect is mediated by Cyc8. In contrast, basal (noninduced) expression of this gene is inhibited by Tup1. These findings establish a positive role for the Cyc8-Tup1 complex in transcription and support a model by which specific metabolic signals may convert the Cyc8-Tup1 transcriptional co-repressor to a co-activator of certain promoters.

Cyc8(Ssn6)-Tup1, a general co-repressor complex, is recruited to promoter DNA via interactions with DNAbinding regulatory proteins and inhibits the transcription of many different yeast genes. Previous studies have established that repression function of the complex is performed by one subunit of the complex, the Tup1 protein, and requires specific components of the RNA polymerase II holoenzyme such as Sin4 and Rgr1. In this study we test the transcriptional activity of the Cyc8 subunit using a lexA operator-containing reporter. We show that a LexA-Cyc8 hybrid stimulates transcription when expressed in a tup1⌬, a sin4⌬, or a rgr1⌬ strain, suggesting that transcriptional activation is an intrinsic property of the Cyc8-Tup1 co-repressor. In support of this notion we demonstrate that Cyc8-Tup1 has a dual function on CIT2, a gene encoding a citrate synthase that is expressed upon mitochondrial dysfunction. First, we show that Cyc8-Tup1 is tethered to CIT2 promoter by interacting with the activation domain of Rtg3, a bHLH/L-Zip DNA-binding transactivator of CIT2. Next we demonstrate that Cyc8-Tup1 activates CIT2 transcription in response to mitochondrial dysfunction, and this stimulatory effect is mediated by Cyc8. In contrast, basal (noninduced) expression of this gene is inhibited by Tup1. These findings establish a positive role for the Cyc8-Tup1 complex in transcription and support a model by which specific metabolic signals may convert the Cyc8-Tup1 transcriptional co-repressor to a co-activator of certain promoters.
An important class of pleiotropic transcriptional regulators includes intermediary proteins such as co-activators and corepressors. These protein factors are tethered to specific promoters mainly by contacting DNA-binding factors and regulate transcription either by interacting with components of the Pol II holoenzyme or by modifying chromatin structure, or both (1,2). The human co-activator CBP/p300 (3), the yeast SAGA complex (4), and the human nuclear receptor co-repressors SMRT and N-CoR (5) are among the best characterized examples of this growing protein family. Interestingly, some of these factors have dual function on specific promoters; for example CBP/p300, which mediates activation of interferon ␤ gene expression in response to virus induction is also responsible for post-induction turn off (35).
In the yeast Saccharomyces cerevisiae, two physically asso-ciated proteins, Cyc8(Ssn6) and Tup1, inhibit the transcription of many diversely regulated genes when their expression is not required (6 -8). It is well established that Cyc8-Tup1 acts as a co-repressor complex that does not bind DNA directly but is recruited to different promoters via interactions with specific DNA-binding regulatory proteins. The repression function of the complex is performed by a specific domain of Tup1 (8).
When the Tup1 repression domain is brought upstream of an active test promoter through the DNA binding domain of LexA, it inhibits transcription independently of Cyc8. Moreover, this domain is required for repression of natural genes such as glucose, oxygen, and cell-type regulated genes. It has been postulated that multiple mechanisms are responsible for Tup1 repression. Tup1 interacts with histones H3 and H4 and may position nucleosomes over the transcription start point, suggesting that Tup1 might repress transcription by modifying chromatin structure (9,10). However, evidence from other studies argue that Tup1 inhibits the function of the basic transcription machinery; Tup1 repression was reconstituted in an in vitro transcription system in the absence of chromatin (11), and mutations in specific components of the RNA polymerase II holoenzyme complex, such as the mediator proteins Sin4, Rgr1, Srb10, and Srb11, weaken the Tup1 repression activity (12)(13)(14)(15). Previous studies suggested that Cyc8 does not directly inhibit transcription but contacts specific DNA-binding regulatory proteins (8,18). The N-terminal region of Cyc8 consists of 10 tandem repeats of a sequence motif termed tetratrico peptide repeat (TPR) 1 (16). TPRs serve as protein-protein interaction domains, and more importantly in the case of Cyc8, TPRs exhibit distinct interaction specificity although they are similar in primary structure (17)(18)(19)(20). TPR1, TPR2, and TPR3 contact Tup1, while different combinations of TPR4 to TPR10 mediate recruitment of Cyc8-Tup1 to different promoters (18). Based on these observations it was proposed that the function of Cyc8 is to link Tup1 to distinct, structurally dissimilar, DNA-bound repressor proteins. Consistently with this linker function, derivatives of Cyc8 that contain only the TPR domain are sufficient for repression. On the other hand, the C-terminal domain which comprises more than half of the protein appears to be dispensable (16 -18).
Recent genetic data suggested that Cyc8-Tup1 might also play a positive role in transcriptional control. More specifically, activation of the CYC1 gene transcription by the Hap1 transactivator and maximal induction of SUC2 gene both require functional Cyc8 protein (26,27). In this report, we present direct evidence that Cyc8-Tup1 indeed plays diverse roles in transcriptional regulation. Expression of Cyc8-Tup1 repressible genes is stimulated in response to specific signals, and under these conditions, the fate of the Cyc8-Tup1 co-repressor has been unclear for most of the cases. Our data suggest that this complex can convert to a co-activator of CIT2, a gene encoding a peroxisomal isoform of citrate synthase that is expressed upon inducing conditions of mitochondrial dysfunction (21). We show that Cyc8-Tup1 is tethered to the CIT2 promoter by interacting with Rtg3, a DNA-binding transactivator of CIT2. Genetic analysis indicates that basal (uninduced) expression of CIT2 is inhibited by Tup1, but its transcriptional activation is mediated by the second component of the complex, the Cyc8 subunit and specifically requires the C-terminal domain of this protein.
The transcriptional activity of Cyc8 was further examined using a synthetic reporter promoter and a LexA-Cyc8 hybrid. Previous studies have shown that LexA-Cyc8 represses transcription by recruiting the Tup1 repressor (18). Here we show that LexA-Cyc8 can activate transcription when Tup1 is absent or when the Tup1 repression is substantially impaired, as it is in a sin4⌬ or a rgr1⌬ mutant strain. Taken together, these data suggest an inherent potential of the Cyc8-Tup1 co-repressor for transcriptional activation function and establish a dual role (positive and negative) of this complex in transcriptional control.

EXPERIMENTAL PROCEDURES
Yeast Strains, Media, and Growth Conditions-All strains are derivatives of FT5 strain (MAT␣ ura3-52 trp1-⌬63 leu2::pet56 his3-⌬200). The two-hybrid screening was performed in the strain L9FT5, which was constructed by replacing his3-⌬200 with the L9His3 allele. L9His3 promoter contains a single synthetic and perfectly symmetric LexAbinding site in place of the Gcn4 UAS (8). cyc8⌬ and tup1⌬ alleles have been described previously (8). The sin4⌬ allele was constructed by inserting HIS3 between the internal PvuII and PstI sites, while the rgr1⌬ allele was constructed by inserting HIS3 between the NdeI and NsiI restriction sites. sin4⌬ and rgr1⌬ strains were generated by onestep gene replacement using linear DNA fragments and were confirmed by Southern analysis.
Standard synthetic media were used; YPD-and YPR-rich media contained 2% glucose or 2% raffinose, respectively. CS minimal media were supplemented with 0.6% casamino acids and glucose or raffinose as the carbon source. Mitochondrial dysfunction was caused by prolonged treatment of cells (ϳ48 h) with 20 g/ml ethidium bromide in YPD medium. CIT2 induction was monitored in exponential cultures growing in YPR medium. Standard procedure was used for routine yeast transformations, while for high efficiency yeast transformation the TRAFO protocol was followed (22).
Plasmid Constructs-All Cyc8, Tup1, and LexA derivatives expressed from the YCp91 vector and have been described previously (18). Briefly, the centromeric vector YCp91 (TRP1 marked) contains the ADH1 promoter and 5Ј-untranslated sequence (including the ATG start codon), followed by the SV40 nuclear localization signal and the HA1 epitope from the influenza virus (flu epitope), a polylinker sequence, and the CYC8 termination region (8). The reporter plasmid JK103 (23) is a URA3 marked multicopy plasmid that expresses the LacZ reporter gene from a minimal promoter consisting of four overlapping LexAbinding sites upstream of the GAL1 TATA box. The Gal4 activation domain-genomic library (used for the two-hybrid screening) was constructed in the pACT2 vector. 2 pACT2, a 2, LEU2 marked vector contains the ADH1 promoter and transcription start site followed by sequences consisting of nuclear localization signal, the activation domain of Gal4, and polylinker sequence (in which random genomic fragments have been inserted) ending to the termination region of the ADH1 gene.
Two-hybrid Screening for Cyc8-Tup1 Interacting Proteins-LexA-Cyc8, cloned in the YCp91 expression vector, was used to transform L9FT5 along with the LacZ reporter plasmid JK103 (23). L9FT5 yeast transformants appear white on X-Gal indicator plates and are sensitive for growth in low concentrations of 3-aminotriazole (0.5-1.0 mM), a competitive inhibitor of His3 enzymatic activity. The pACT2 library was used to transform this strain, and cells were recovered by shaking in selective liquid medium (SC, containing 2% glucose and 2% galactose) for 10 h at 30°C. Five million independent transformants were scored for growth in minimal media containing 3-aminotriazole at a concentration of 20 mM. 200 colonies, scored as positives, were selected (normal growth in 20 mM 3-aminotriazole), but less than half of them (ϳ96) appeared blue on X-Gal plates. Positive transformants containing pACT2-derived plasmids were rescued in the Escherichia coli KC8 strain (constructed by K. Struhl) based on the ability of the yeast LEU2 gene (pACT2 is a LEU2 marked plasmid) to complement the respective E. coli auxotrophy. 39 of 96 plasmids analyzed reproducibly supported 3-aminotriazole resistance and high ␤-galactosidase activity after retransformation into L9FT5 strain. Sequencing analysis revealed seven different ORFs encoding proteins capable of two-hybrid interaction with Cyc8.
GST Interaction Assay-A BamHI-PvuII fragment containing Rtg3-68 was cloned in the T7 expression vector pRSETC and was used to direct coupled transcription translation (Promega T7 TNT). 35 S-Labeled Rtg3 protein (10 l) was incubated with approximately 2 g of agarose bead-immobilized GST-Cyc8 protein (18) in a volume of 100 l containing 20 mM Tris-acetate, pH 7.4, 10% glycerol, 0.2 mM EDTA, 1 mM dithiothreitol, 0.15 M potassium acetate, and 1ϫ complete protease inhibitors (Boehringer) for 2 h at 4°C. Following incubation the beads were extensively washed in the above buffer, eluted in SDS-PAGE gel loading buffer, and analyzed by SDS-PAGE.
EMSA-Whole cell protein extracts from wild type and cyc8⌬ strains were isolated as described previously (24), and protein concentrations were determined by the Bradford assay. A CIT2 promoter region (Ϫ530 to ϩ1) was amplified from L9FT5 genomic DNA by polymerase chain reaction, and a BssHII-AatII fragment of it (106 base pairs) containing the two Rtg1/Rtg3-binding sites, was end-labeled by standard methods using "Klenow" DNA polymerase and purified by Sephadex G-50 chromatography (Amersham Pharmacia Biotech). EMSA reactions were performed with 10 g of protein extract, 10,000 cpm of CIT2 probe, 4 g of poly(dI/dC) competitor DNA, 1 mM dithiothreitol, 10 mM Tris-HCl, 100 mM KCl, 2 mM MgCl 2 , and 5% glycerol in a final volume of 20 l. They were incubated at 4°C for 25 min, and samples were separated on a 5% polyacrylamide, 1ϫ TBE gel at 4°C, 250 V for 2.5 h and visualized by autoradiography.
RNA Analysis-Total cellular RNA was extracted from yeast cells grown in the appropriate medium, using the acid phenol method (25), and was fractionated in 1.4% agarose gels containing 5.5% formaldehyde. RNA was transferred to nylon membrane and hybridized with 32 P-labeled probes generated by nick translation. For CIT2 and TBP probes, polymerase chain reaction fragments containing the entire coding sequence (from ATG to termination codon) of the respective genes were used.
LacZ Assays-␤-Galactosidase assays were performed on yeast cultures grown in the appropriate media and harvested during early log phase (A 600 Ͻ 1.0). Cells were washed with 20 mM Tris (pH 7.5), 1.0 mM EDTA in order to disperse the clumpy cyc8 and tup1 cells. LacZ values normalized to a 600 represent the average of at least three independent transformants, and they are accurate to 20 -30%.

RESULTS
Transcriptional Activation by Cyc8 -Recent genetic evidence suggested that Cyc8-Tup1 might play a positive role in transcriptional regulation. Mutations in the CYC8 gene adversely affect both Hap1-mediated stimulation of CYC1 and maximal induction of SUC2 transcription (26,27). Based on these observations, we directly tested whether Cyc8 can stimulate transcription by analyzing the activity of a LexA-Cyc8 hybrid protein on a GAL1-LacZ synthetic reporter that contains a LexA operator upstream of the TATA element. Wild type and isogenic tup⌬1, sin4⌬, and rgr1⌬ strains were cotransformed with plasmids carrying genes that express LexA-Cyc8 and the GAL1-LacZ reporter, and transformants were assayed for ␤-galactosidase activity. As shown in Fig. 1, LexA-Cyc8 represses transcription from the reporter promoter when functional Tup1 is present (wild type strain, line 1). However, in the absence of Tup1 (tup⌬1), LexA-Cyc8 stimulates transcription by 7-fold (line 2). This result indicates that Tup1 not only actively represses transcription (8) but it also antagonizes the activation potential of Cyc8. This cannot be explained sim-ply by an intermolecular masking of Cyc8 by Tup1 because, similarly to the Cyc8 protein alone, the Cyc8-Tup1 protein complex can also activate transcription when the repression activity of Tup1 is impaired. In sin4⌬ or rgr1⌬ strains, in which the respective components of the holoenzyme mediator complex that are essential for the establishment of the Tup1 repression activity are missing, the LexA-Cyc8/Tup1 protein complex is converted to a transcriptional activator (line 3 and line 4). Noticeably, activation by LexA-Cyc8/Tup1 in these mutant strains is virtually higher (9-fold) than that in the tup1⌬ strain. These observations suggest that Cyc8 has the potential to act as a transcriptional co-activator and that this function is performed more efficiently in the context of the Cyc8-Tup1 protein complex.
The Activation Domain of Rtg3 Associates with the Cyc8-Tup1 Protein Complex-If Cyc8-Tup1 acts as a co-activator of certain natural promoters, then it would be expected to directly interact with promoter-specific DNA-binding activator protein(s). We explored this possibility by seeking such Cyc8-Tup1 interacting proteins (activator proteins), using a yeast twohybrid screen (see "Experimental Procedures"). We generated a strain that expresses HIS3 and LacZ reporter genes under the control of promoters containing a LexA-binding site. This strain was co-transformed with a LexA-Cyc8 expressing plasmid along with a library of random yeast genomic fragments fused to the Gal4 activation domain. Library plasmids supporting high levels of both HIS3 and LacZ reporter genes were recovered and sequenced. This selection scheme revealed 39 positive clones encoding regions from seven different proteins. Consistent with the co-repression function of Cyc8-Tup1, most of these were DNA-binding repressor proteins that inhibit transcription in a Cyc8-Tup1-dependent manner. 3 Interestingly though, two independently isolated clones encoded Rtg3, a well characterized transcriptional activator protein. Rtg3-68 (residues 305 to 486) and Rtg3-36 (residues 326 to 486) interact equally well with LexA-Cyc8 in a two-hybrid assay (not shown), and both contain sequences from the C-terminal region of Rtg3, known to be the activation domain of the protein (28, Fig. 2).
To further investigate the Cyc8-Rtg3 interaction, we used a GST affinity chromatography assay, in which 35 S-labeled Rtg3-68 protein was synthesized in vitro and was incubated with agarose bead-immobilized GST or GST-Cyc8 hybrid protein expressed in and purified from bacteria. As shown in Fig.  3, Rtg3 is specifically retained on the GST-Cyc8 column (lane 3) but not on the column containing GST alone (lane 2), strongly suggesting that Rtg3 directly associates with Cyc8 in the absence of any other yeast protein.
Rtg3 activates the transcription of CIT2, a gene encoding a peroxisomal isoform of citrate synthase, and probably the transcription of additional genes involved in peroxisome biogenesis (29). Thus, we subsequently explored the function of Cyc8 and Tup1 in the context of the natural CIT2 promoter.
Dual Function of Cyc8-Tup1 on CIT2 Transcription-CIT2-Transcription is induced by mitochondrial dysfunction, and this regulatory pathway, through which nuclear gene transcription responds to the functional state of mitochondria, is termed retrograde regulation (24,30). Both basal expression and retrograde response of CIT2 is mediated by the Rtg3 transactivator, which is always bound to the CIT2 UAS (29).
A typical retrograde response of CIT2 gene transcription is shown in Fig. 4A. CIT2 expression is much higher in wild type cells growing under conditions of mitochondrial dysfunction (lane 2), compared with the basal level of expression observed in normally growing cells (lane 1). However, basal expression and retrograde response of CIT2 transcription are dramatically reduced in a strain carrying a chromosomal deletion of CYC8 (lanes 3 and 4), suggesting that CIT2 is positively regulated by Cyc8. On the other hand, basal expression of CIT2 is increased in a tup1⌬ strain (lane 5) indicating that CIT2 transcription is yet another target of the Tup1 repression activity. In the tup1⌬ strain, retrograde response appears to be comparable, although slightly lower, than in wild type strain (lane 6). Thus, Tup1 inhibits basal expression of CIT2 but it might also be required along with Cyc8 for maximal CIT2 induction. In agreement with this notion, CIT2 expression is fully de-repressed in Tup1 repression defective strains, such as sin4⌬ and rgr1⌬ (lanes 7 and 8), in which Tup1 is expressed normally. In fact, the expression levels of CIT2 in these mutant strains are comparable with the levels observed under conditions of mitochondrial dysfunction. Taken together, these results strongly suggest that Cyc8-Tup1 has a dual function on the CIT2 promoter; it inhibits basal transcription, but moreover it acts as a coactivator that mediates retrograde response. It is noteworthy that Rtg3, which is the limiting factor for CIT2 transcription (28), is present at equal levels in wild type, cyc8⌬, and tup1⌬ cells growing either at normal or at inducing conditions (28, 29, and data not shown).   35 S-Labeled Rtg3 (residues 326 -486) interacting with a GST-Cyc8 hybrid protein or GST alone immobilized in Sepharose beads. Lane labeled input contains only 20% of the amount of the protein that was incubated with the beads.

Distinct TPR Motifs of Cyc8
Interact with Tup1 and Rtg3-Two-hybrid assays performed in cyc8⌬ and tup1⌬ strains indicated that Rtg3 specifically interacts with Cyc8 even in the absence of Tup1, while Tup1 interacts with Rtg3 only in the presence of Cyc8 (data not shown). The TPR domain of Cyc8 mediates protein-protein interactions and was proposed to link specific DNA-binding proteins to Tup1 (18). Thus, we examined whether Rtg3 interacts with specific TPR motifs of Cyc8 by testing various deletion derivatives of Cyc8 for Rtg3 interaction in a two-hybrid assay (Table I and Fig. 5). N175, that contains only three N-terminal TPRs (TPR1 to TPR3), does not activate transcription of the LacZ reporter demonstrating its failure to interact with Rtg3. In contrast, N300 that contains TPR1 to TPR7 strongly interacts with Rtg3 (it activates transcription over 90-fold), indicating that interaction with Rtg3 is mediated by specific TPR motifs, probably TPR4 to TPR7. However, the internally deleted Cyc8 derivative ⌬175-281 that lacks TPR4 to TPR7 but maintains TPR8 to TPR10 also interacts with Rtg3 as judged by its activity on the LacZ reporter, which is stimulated over 30-fold. Finally, derivatives such as C560, which comprise the C-terminal domain of Cyc8 but lack TPR sequences, are completely inactive. These data suggest that Rtg3 interacts with at least two independent combinations of TPR motifs, TPR4 -TPR7 and TPR8 -TPR10. It should be noted that none of these regions overlap with the Tup1 interaction domain, which consists of TPR1 to TPR3 (Ref. 18 and Fig. 5), and it explains how Rtg3 and Tup1 (which do not interact directly) can simultaneously associate with the TPR domain of Cyc8.
To test whether the TPR domain is sufficient to recruit Cyc8-Tup1 to the CIT2 promoter we performed a band shift experiment using whole yeast protein extracts and a probe encompassing two Rtg3/Rtg1-binding sites (Rtg3 binds DNA as a heterodimer with Rtg1, another bHLH/Zip protein, see "Discussion"). In agreement with previous data (29), a stable low mobility complex was detected in the presence of protein extracts derived from a wild type strain (Fig. 6, lane 2). The formation of this complex is dependent on the presence of Cyc8, because extracts from a cyc8⌬ strain do not give rise to shifted bands (lane 3). Moreover, ectopic expression of Cyc8 in the cyc8⌬ strain restores complex formation (lane 4), and more importantly, the complex is formed even by expressing only the TPR domain of Cyc8 (lane 5). These results strongly suggest that protein-protein interactions mediated by TPR motifs are sufficient to recruit Cyc8-Tup1 to the CIT2 promoter.
The Cyc8 C-terminal Domain Is Essential for Stimulation of CIT2 Transcription-The TPR domain of Cyc8 provides sufficient Cyc8 function for transcriptional repression by bringing Tup1 to specific DNA-binding proteins, while the C-terminal domain is dispensable. Because we showed that Cyc8-Tup1 activates CIT2, we examined whether recruitment of the complex by the TPR domain is sufficient for positive regulation of CIT2 transcription. For this purpose, derivatives of Cyc8 capable of interacting with both Rtg3 and Tup1, either containing or lacking C-terminal sequences, were expressed in a cyc8⌬ strain, and CIT2 mRNA levels were analyzed by RNA blotting. As shown in Fig. 4B and summarized in Fig. 5, ⌬175-281, which contains the entire C-terminal domain of the protein, supports wild type levels of CIT2 transcription (lanes 7 and 8).
In contrast, N300 and N597 that lack the C-terminal domain are inactive (lanes 3-6), despite their ability to interact with Rtg3 and to complement all previously described cyc8⌬ defects (18). Finally, a longer derivative, N816, that contains most of the C-terminal domain is only partially functional (lanes 1 and  2). These results indicate that Rtg3 interaction alone is not sufficient for transcriptional activation of CIT2; normal retrograde response of CIT2 expression requires the C-terminal domain of Cyc8 as well. To our knowledge, retrograde regulation is the only case where a specific function has been assigned to this domain of Cyc8, most likely reflecting the unique regulatory mode of Cyc8-Tup1 action on the CIT2 promoter.

DISCUSSION
In this study we provide direct evidence for a dual role of Cyc8-Tup1 in transcriptional control. We found that besides the well established repression activity, which is performed by Tup1, the Cyc8-Tup1 protein complex can also act as a transcriptional co-activator, and this function is predominantly mediated by the Cyc8 protein. When the Tup1 repression activity is impaired, as it is in a sin4 or a rgr1 mutant strain, Cyc8-Tup1 activates an artificial reporter gene, and in response to specific metabolic signals, activates the transcription of the natural CIT2 gene.
Transcription of CIT2 is controlled by Rtg3 and Rtg1, both members of the bHLH/Zip family of DNA-binding proteins. Recombinant Rtg3 and Rtg1 bind as a heterodimer at two sites within an upstream activation sequence of the CIT2 gene termed UAS r (31,32). Heterologous promoters bearing a UAS r respond to mitochondrial dysfunction in a Rtg1/Rtg3-dependent manner indicating that UAS r is sufficient to mediate CIT2 regulation (30). Notably, EMSAs using whole yeast extracts (instead of recombinant Rtg1 and Rtg3 proteins) suggested that additional yeast proteins, probably co-activators or co-repres- FIG. 4. Normal expression of the CIT2 gene requires functional Cyc8 and Tup1. A, total RNA was extracted from cultures of wild type (WT), cyc8, tup1, sin4, and rgr1 deletion strains exponentially grown in the indicated conditions (N, normal; MD, mitochondrial dysfunction). RNA was fractionated in formaldehyde agarose gel (1.5%), transferred to nylon membrane, and hybridized with probes specific for CIT2 and TBP (used to normalize RNA levels) as described under "Experimental Procedures." B, Cyc8 deletion derivatives (shown in Fig. 5), were expressed in a cyc8 deletion strain, and yeast transformants were cultured in normal (N) and mitochondrial dysfunction (MD) conditions. Total RNA was extracted, fractionated in a formaldehyde-agarose gel, transferred to nylon membrane, and probed with CIT2 and TPB specific probes. Several lines of evidence suggest that Cyc8-Tup1 is directly involved in the activation of CIT2 transcription and that this function is performed by the Cyc8 subunit. First, Cyc8-Tup1 specifically associates with the activation domain of Rtg3. This region of Rtg3 has been shown to be the major activation domain of the Rtg1/Rtg3 heterodimer because Rtg1, which does not possess independent transactivation properties, functions to recruit Rtg3 to its binding site (28). Thus, a possible role of the Rtg1/Rtg3 activation domain is to simply contact the Cyc8-Tup1 complex. Second, deletion of CYC8 or deletion of both CYC8 and TUP1 (data not shown) severely reduces the levels of CIT2 mRNA under inducing conditions of mitochondrial dysfunction. Under these conditions, CIT2 transcription is defective even in the presence of Cyc8 derivatives (N300 and N597) that fully complement all known cyc8⌬-specific phenotypes, including slow growth and temperature-sensitive lethality (18). These results suggest that lower CIT2 transcription is not an indirect physiological effect of the cyc8⌬ mutation. In agreement to this, deletion of TUP1, which causes similar pleiotropic defects as cyc8⌬, does not significantly affect the levels of CIT2 mRNA under inducing conditions. Third, when brought to a LexA operator-containing reporter, LexA-Cyc8 activates transcription in a tup1⌬ strain. Similarly, the LexA-Cyc8/Tup1 protein complex activates transcription in an isogenic sin4⌬ or a rgr1⌬ strain that lacks the respective factor essential for Tup1 repression. These data, together with previous observations (26,27), establish a positive role of Cyc8 in transcription and moreover they suggest a dual, positive and negative, function of the Cyc8-Tup1 protein complex on CIT2. Indeed Cyc8-Tup1 inhibits the basal (uninduced) expression of CIT2, and this function is performed by Tup1. CIT2 transcription is derepressed in cells carrying the tup1⌬ mutation while in cells that express Tup1, but lack Sin4 or Rgr1, CIT2 derepression occurs at even higher levels. This observation further suggests that Cyc8 activation function is better performed in the context of the Cyc8-Tup1 protein complex.
Our data indicate that CIT2 transcription requires the Cterminal domain of Cyc8, and in fact, this is the only case that a function has been assigned to this region. When bound upstream of a test promoter through a heterologous DNA-binding domain this C-terminal region of Cyc8 does not activate transcription (data not shown); therefore it does not function as a typical activation domain, but rather plays a regulatory role. Cyc8 is a phosphoprotein, and specific regions within this Cterminal domain, rich in serine and threonine residues, are potential phosphorylation sites (16). Similarly, Rtg3 contains a serine/threonine-rich region which might also play a regulatory role (28). Thus, it is conceivable that specific modifications of these protein domains, such as phosphorylation or dephosphorylation, may in fact modulate the transcriptional activity of Cyc8-Tup1. Some of these modifications, particularly in the C-terminal domain of Cyc8, are likely to be specific for the retrograde response of CIT2 as this domain is dispensable for the regulation of all other Cyc8-Tup1 repressible genes.
The signal(s) that mediate induction of peroxisomal genes upon mitochondrial dysfunction are presently unknown, and FIG. 6. The TPR domain is sufficient for recruitment of Cyc8-Tup1 to the CIT2 promoter. Total yeast protein was extracted from wild-type and cyc8⌬ strains or from a cyc8⌬ strain expressing either full-length Cyc8 protein or the N300 derivative that contains only TPRs (see Fig. 5). EMSAs were performed using a DNA fragment that contains the two Rtg1/Rtg3-binding sites of the CIT2 promoter.
although several possible models can be envisaged using the available data, the molecular mechanism by which Cyc8-Tup1 is converted from a co-repressor to a co-activator of CIT2 is not yet understood. One model predicts that, upon induction, Tup1 dissociates from the complex thus unmasking Cyc8 activation potential. However, EMSAs performed with protein extracts derived from either normal or mitochondria defective cells detect neither quantitative nor qualitative differences on UAS r DNA-protein complexes (21,29). 4 This observation also argues against the model according to which additional positive regulatory factors associate with Rtg3/1-bound Cyc8-Tup1 assembling an activator complex, although transient interactions with such factors cannot be excluded. Another, more plausible, model postulates that in response to specific signaling, Cyc8-Tup1 undergoes post-translational modifications which could reveal the intrinsic activation potential of Cyc8. Masking of Tup1 repression, although possible, cannot solely account for the activation function of the complex because Cyc8-mediated CIT2 induction is observed even in the absence of Tup1 (tup1⌬ strain, Fig. 4A). Regulatory mechanisms by which proteins undergo conformational changes and activate transcription have been previously reported, and in some cases, as that of the retinoic acid receptor, these mechanisms have been characterized extensively (34). It is conceivable that Cyc8 undergoes specific structural changes, i.e. by phosphorylation of the Cterminal domain, which is specifically required for stimulation of CIT2 transcription, and this could possibly be the key step through which the complex attains its activation potential. It must be emphasized that according to this model post-translational modifications of Cyc8 and probably of Tup1 have such an effect only when the complex is associated with the Rtg3/1 proteins. This hypothesis explains why Cyc8-Tup1 has a dual function only in CIT2 expression and why DNA-bound LexA-Cyc8/Tup1 is not converted to an activator of a synthetic reporter under inducing conditions of mitochondrial dysfunction (data not shown). Finally, it is known that Cyc8-Tup1 functionally interacts with the basic transcription machinery as well as with chromatin (15); thus it is conceivable that Cyc8 might activate CIT2 transcription by affecting either one or even both processes.