Synergy among Differentially Regulated Repressors of the Ribonucleotide Diphosphate Reductase Genes of Saccharomyces cerevisiae

Abstract

The Ssn6/Tup1 general repression complex represses transcription of a number of regulons through recruitment by regulon-specific DNA-binding repressors. Rox1 and Mot3 are Ssn6/Tup1-recruiting, DNA-binding proteins that repress the hypoxic genes, and Rfx1 is a Ssn6/Tup1-recruiting, a DNA-binding protein that represses the DNA damage-inducible genes. We previously reported that Rox1 and Mot3 functioned synergistically to repress a subset of the hypoxic genes and that this synergy resulted from an indirect interaction through Ssn6. We report here cross-regulation between Rox1 and Mot3 and Rfx1 in the regulation of the RNR genes encoding ribonucleotide diphosphate reductase. Using a set of strains containing single and multiple mutations in the repressor encoding genes and lacZ fusions to the RNR2 to -4 genes, we demonstrated that Rox1 repressed all three genes and that Mot3 repressed RNR3 and RNR4. Each repressor could act synergistically with the others, and synergy required closely spaced sites. Using artificial constructs containing two repressor sites, we confirmed that all three proteins could function synergistically but that two Rox1 sites or two Rfx1 sites could not. The significance of this synergy lies in the ability to repress gene transcription strongly under normal growth conditions, and yet allow robust induction under conditions that inactivate only one of the repressors. Since the interaction between the proteins is indirect, the evolution of dually regulated genes requires only the acquisition of closely spaced repressor sites.

Transcriptional repression in eukaryotes is an active process mediated through complexes that organize chromatin and/or interact with the basal transcriptional machinery (3, 24, 41, 45, 47). The evolutionarily conserved Ssn6/Tup1 general repression complex of Saccharomyces cerevisiae is one such example. Tup1 interacts with the N-terminal domain of histones H3 and H4 and, for at least some Ssn6/Tup1 repressed genes, positions a nucleosome over the TATA box to block access by the general transcriptional machinery (7, 12, 55). In addition, Tup1 recruits a histone deacetylase, presumably to lock this nucleosome in place (13, 50, 52). In the absence of a positioned nucleosome, the general repression complex still can repress transcription through a chromatin-independent mechanism (20, 32, 33, 38, 40). Tup1 has been demonstrated to interact with a number of subunits of the mediator complex which may play a role in this latter repression mechanism. For some genes, both mechanisms function, and repression can only be genetically disabled through mutations in both the histones and the mediator components (27, 45, 54).

Ssn6 and Tup1 are expressed constitutively and have no intrinsic DNA-binding activity (29, 45). The gene or regulon specificity of repression is controlled through specific DNA-binding proteins that recruit the general repression complex. Thus, for example, Rox1 represses the hypoxic genes (2, 35); Mig1 and -2, the glucose-repressed genes (48, 51); Rfx1 (Crt1), the DNA damage-repair genes (22); and α2-Mcm1, the a mating type genes (29, 31, 37, 46). In each case the transcription, activity, or cellular compartmentalization of these DNA-binding repressors is regulated to control when repression is effected (10, 22, 31, 34, 35, 36, 46).

Our studies have focused on the regulation of the hypoxic genes (reviewed in references 28, 56, and 57). These genes are repressed aerobically by Rox1, which binds to Ssn6 to recruit the general repression complex. ROX1 expression is dependent upon heme, the synthesis of which requires molecular oxygen. Hence under hypoxic conditions, heme levels are low, Rox1 is not made, and the Ssn6/Tup1 complex is not recruited to the hypoxic genes. Consequently, these genes are derepressed. Despite the large protein complex that localizes to the regulatory region of Rox1-repressed genes, including Ssn6/Tup1, positioned nucleosomes, deacetylases, and perhaps the general transcription machinery, repression is dependent on the number and quality of the Rox1-binding sites; a base pair change in a Rox1 site that caused a fewfold decrease in Rox1 binding in vitro resulted in a proportional decrease in repression in vivo (9). Recently, an additional DNA-binding protein, Mot3, was found to augment repression by Rox1. Strong hypoxic repression sites consist of a combination of Rox1 and Mot3 sites (27, 30, 35, 44). For example, the ANB1 gene is repressed 250-fold, 8-fold by operator B which consists of two Rox1 sites, and 76-fold by operator A, which contains two Rox1 sites flanking a Mot3 site (27). The HEM13 gene is repressed 50-fold by three Mot3 sites plus one Rox1 site (30). We have termed this repression combinatorial repression. Combinatorial repression can be additive or synergistic. Additive repression occurs when the two repressors act independently. While Rox1 can repress independent of Mot3, Mot3 can only repress independently through multiple sites, with each bound Mot3 molecule acting independently of the others. Synergy is defined as the two repressors combining to repress gene expression to a greater extent than the additive effects of the single proteins acting independently. Synergy results when Rox1 and Mot3 bind to closely spaced sites. This synergy does not result from an interaction between the two proteins but rather through each repressor's interaction with the Ssn6/Tup1 complex (30). Chromatin immunoprecipitation experiments demonstrated that Rox1 bound to DNA independently of Mot3 and enhanced Mot3 binding (30, 35). These findings and others led us to propose a model for the synergistic interaction. Rox1 interacts weakly with the Ssn6/Tup1 complex, and the role of Mot3 is to help recruit the general repression complex. On the other hand, Mot3 interacts weakly with DNA and requires Rox1 binding to DNA strengthen its interaction through Ssn6/Tup1 (30).

This model requires no hypoxic specificity on the part of Mot3 when it functions through synergistic interactions with Rox1. Indeed, although Mot3 is repressed under hypoxia (44), constitutive expression of Mot3 did not result in any repression of ANB1 during hypoxia (35). Since synergy requires no direct interaction between Rox1 and Mot3, and if Mot3 cannot repress alone through a single site, we posited that Mot3 could function in the same capacity at other Ssn6/Tup1 regulons. The Mot3 binding site, (A/C/T)AGG(C/T)A, is only 6 bp and degenerate (19), so that it has a 50% chance of appearing once every 340 bp in a random sequence. Consequently, we only scanned the regulatory regions of Ssn6/Tup1 repressed genes with well-characterized regulon-specific DNA binding sites, searching for Mot3 sites close to those sites. We found promising hits in the regulatory region of DNA damage-inducible RNR3 and -4 genes encoding two of the proteins of ribonucleotide diphosphate reductase (RNR) (15, 21, 49). To our surprise, we also found Rox1 binding sites in these two genes as well as in RNR2, encoding another RNR subunit (14, 23). In the present study, we demonstrate that these genes are regulated by oxygen through Rox1 and Mot3 and that both Mot3 and Rox1 can function synergistically with Rfx1. In hindsight, these results are not surprising; the formation of the active RNR enzyme requires molecular oxygen (4, 17), and, therefore, this enzyme falls in the class of those whose concentrations must increase to compensate for limiting oxygen concentrations under hypoxia. The results reported here indicate that synergy between repressors allows substantial derepression under conditions that relieve only one repressor' s activity, thus providing dual regulation without the need to evolve new regulatory patterns for the repressors.

MATERIALS AND METHODS

Strains and cell growth.

The yeast strains used in the present study are presented in Table 1. RZ53-6 derivatives were constructed by gene replacement (43). RZ53-6-related strains (the MZ series) were constructed by a mating of BY4741Δrfx1 with RZ53-6Δmot3 to obtain a mot3Δ rfx1Δ double deletion. This strain was backcrossed to RZ53-6 or RZ53-6Δrox1 to obtain the MZ series of strains listed in Table 1.

TABLE 1.

Yeast strains

Strain	Genotype	Source or reference
RZ53-6 derivatives		2
RZ53-6	MATα trp1 leu2 ura3 ade1	2
RZ53-6Δrox1	MATα trp1 leu2 ura3 ade1 rox1::LEU2	8
RZ53-6Δmot3	MATα trp1 leu2 ura3 ade1 mot3::kanMX	27
RZ56-6Δr1Δm3	MATα trp1 leu2 ura3 ade1 rox1::LEU2 mot3::kanMX	27
RZ53-6-related strains
MZ233-20	MATα trp1 leu2 ura3 ade1	This study
MZ233-2	MATα trp1 leu2 ura3 ade1 mot3::kanMX	This study
MZ233-13	MATα trp1 leu2 ura3 ade1 rfx1::kanMX	This study
MZ233-6	MATα trp1 leu2 ura3 ade1 mot3::kanMX rfx1::kanMX	This study
MZ255-6	MATα trp1 leu2 ura3 ade1 rox1::LEU2 rfx1::kanMX	This study
MZ269-8	MATα trp1 leu2 ura3 ade1 rox1::LEU2 mot3::kanMX rfx1::kanMX	This study
BY4741Δrfx1	MATaleu2 ura3 met15 rfx1::kanMX	Yeast Deletion Consortium

Yeast strains were maintained in YPD medium (26). For selection for plasmids in the experiments described here, cells were grown in SC medium (26) lacking the appropriate nutrient at 30°C with vigorous shaking. For anaerobic growth, the medium was supplemented with 0.2% Tween 80 and 20 μg of ergosterol/ml. Long-term anaerobic growth was carried out in flasks packed into an anaerobic jar with the BBL GasPak system (BD Diagnostics, Franklin Lakes, NJ), and the cultures were incubated at 30°C with shaking. For time course experiments, anaerobiosis was achieved by bubbling nitrogen through 10-ml cultures in 125-ml flasks hooked up in series to the nitrogen tank, and each time point was taken by removing the last culture in the series. Yeast cells were transformed as described previously (6).

Gradient plates used to assay cell sensitivity to 4-nitroquinoline-1-oxide (NQO) were made by first pouring yeast extract-peptone-dextrose (YPD) agar with 2.6 μM NQO into plates at an angle so that the agar solidified with a thin layer at one end and a thick layer at the other. Then, a second layer of YPD agar was poured over the first with the plates lying flat. The plates were used the same day. Cells to be tested were grown overnight to stationary phase and then diluted 50-fold and distributed into a 96-well microtiter dish. Drops were then stamped onto the plates, which were then incubated for 2 days at 30°C.

Plasmids.

The plasmid YCp(33)ROX1Z is a centromeric URA3 vector with the ROX1 upstream region from a HindIII site at −1290 to a PstI site at +4 fused to the lacZ coding sequence and the ANB1 3′ sequences (8). YCp(33)AZΔB is a centromeric URA3 vector containing the ANB1 gene sequences from positions −568 to +977 with an in-frame insertion of the lacZ coding sequence. The ANB1 regulatory region contains a deletion of operator B (positions −213 to −187) (9). YCp(33)AZΔBΔ5′ and YCp(33)AZΔBΔ3′ were derived from YCp(33)AZΔB and contain deletions of the 5′ or 3′ Rox1 binding sites, respectively (9). YCp(33)AZΔB(−10) and YCp(33)AZΔB(−5) are derivatives of YCp(33)AZΔB with a 10-bp or a 5-bp deletion between the Rox1 binding sites, respectively (9). YCp(22)MOT3 (30) and YCp(23)MOT3HA (35) are centromeric TRP1 vectors containing the MOT3 gene without and with the HA epitope, respectively. YCp(23)ROX1HA and YCp(22)ROX1 are centromeric TRP1 vectors containing the an hemagglutinin (HA)-tagged ROX1 and an untagged ROX1 allele, respectively (35).

All plasmids were constructed by using standard procedures (1). Enzymatic reactions were carried out under the conditions recommended by the vendor (New England Biolabs, Inc., Beverly, MA). Synthetic oligonucleotides were purchased from Sigma-Genosys (The Woodlands, TX). Genes are numbered with the A in the translational initiation codon at +1, and bases 5′-wards in negative numbers. The references to 5′ and 3′ ends of and directions along sequences refer to the coding strand.

YCp(33)RNR2Z, YCp(33)RNR3Z, YIp(211)RNR3Z, and YCp(33)RNR4Z.

The upstream regions of RNR2 (−957 to +18), RNR3 (−891 to +3), and RNR4 (−988 to +3) were PCR amplified from genomic DNA with primers that placed an EagI (RNR2) or a HindIII (RNR3 and RNR4) at the upstream end and a PstI site in the coding sequence. The fragments were digested with the enzymes described above and inserted into YCp(33)ROX1Z digested with EagI plus PstI (for RNR2) or HindIII plus PstI (for RNR3 and RNR4). In the resulting plasmids, the RNR regulatory sequences replaced those of ROX1.

The integrating plasmid YIp(211)RNR3Z was constructed by subcloning the RNR3Z HindIII-XmaI fragment from YCp(33)RNR3Z into the corresponding sites of YIplac211 (18). This plasmid contained a unique AvrII at −330 within the RNR3 regulatory region. The plasmid was digested with this enzyme and transformed into yeast to integrate the entire plasmid into the RNR3 locus.

YCp(33)AZ-M3R1.

The upstream region of ANB1 was amplified from positions −568 to −311 using YCp(33)AZΔB as a template, a 3′-wards upstream primer that hybridized to vector sequences, and a 5′-wards downstream primer that contained an XhoI site, followed by a Rox1 binding site, a Mot3 binding site, an EagI restriction site, and then the ANB1 sequences from −311 to −334. The resulting fragment was digested with XhoI plus HindIII and ligated into YCp(33)AZΔB digested with the same enzymes. The resulting construct contained a deletion of the ANB1 operator B and A (−275 to −313). OpA was replaced with one Rox1 and one Mot3 sites flanked by the restriction sites for EagI and XhoI. The inserted sequence is listed in Table 2.

TABLE 2.

Inserts with combinations of binding sites

Plasmid	Insert(s)a,b	Site(s)b
YCp(33)AZ-M3R1	cggccgtTGCCTGttttTCTATTGTTCTCgag	Mot3, Rox1
YCp(33)AZ-M3Rfs	cggccgtctTGCCTTcccatggcgTTGCCATGGCAACtcgag	Mot3, Rfx1
YCp(33)AZ-M3Rfw	cggccgtctTGCCTTcccatggcgTTGTTGTGGCAACtcgag	Mot3, Rfx1
YCp(33)AZ-R1Rf	cggccgCCCATTGTTCTCagatctgcgTTGTTGTGGCAACtcgag	Rox1, Rfx1
YCp(33)AZ-1RRf	cggccGAGAACAATGGGagatctgcgTTGTTGTGGCAACtcgag	Rox1, Rfx1
YCp(33)AZ-2Rf	cggccgTTGTTGTGGCAACtctagaaattTTGTTGTGGCAACtcgag	Rfx1

^aThe XhoI site on the right is at −270; the EagI site is on the left joined at −313.

^bUnderlined sequences represent the sites for the corresponding underlined repressors.

YCp(33)AZ-M3Rfw, YCp(33)AZ-M3Rfs, YIp(211)AZ-M3Rfs, YCp(33)AZ-R1Rf, YCp(33)AZ-1RRf, and YCp(33)AZ-2Rf.

The five plasmids YCp(33)AZ-M3Rfw, YCp(33)AZ-M3Rfs, YIp(211)AZ-M3Rfs, YCp(33)AZ-R1Rf, YCp(33)AZ-1RRf, and YCp(33)AZ-2Rf were constructed by digesting YCp(33)AZ-M3R1 with EagI plus XhoI and inserting synthetic double-stranded DNAs with the sequences listed in Table 2. When annealed, the synthetic DNAs had single-stranded ends complementary to those of the EagI- and XhoI-digested plasmid.

YIp(211)AZ-M3Rfs was constructed by subcloning the ANB1-lacZ HindIII-KpnI fragment from YCp(33)AZ-M3Rfs into the corresponding sites of YIplac211. This plasmid contained a unique BglII site within the ANB1 sequences 3′ to the lacZ coding sequence. The plasmid was digested with this enzyme and transformed into yeast to integrate the entire plasmid into the ANB1 locus.

β-Galactosidase and immunoblot assays.

β-Galactosidase assays were performed on pooled transformants (30). Transformants were selected on SC-uracil plates and, after several days of growth, the colonies on each plate were pooled and used to start overnight cultures. Cells were grown in SC-uracil medium to mid-exponential phase, and the enzyme assays were carried out as described previously (26). The data presented represent at least three independent assays with the standard deviations shown.

For immunoblots, cells were grown on selective media to mid-exponential phase, 1.5-ml samples were chilled, and the cells were harvested by centrifugation, resuspended in 40 μl of sodium dodecyl sulfate gel sample buffer, and boiled for 5 min. The samples were fractionated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis, and immunoblots were carried out as described previously (1). The blots were probed with monoclonal antibody against the HA epitope (Santa Cruz Biotechnology, Inc., Santa Cruz, CA). After the bands were visualized, the blot was stripped and reprobed with antisera against eIF5A (prepared by Alexander Kastaniotis). The ratio of HA to eIF5A was determined by using ImageQuant software (Amersham Biosciences, Piscataway, NJ).

RESULTS

Mot3 and Rox1 repress the RNR genes.

RNR2, -3, and -4 encode three of the four proteins of RNR (14, 15, 21, 23, 49). It has been reported that the expression of these genes is repressed by Rfx1 (Crt1) and Ssn6/Tup1 under normal growth conditions (16, 22). DNA damage induces the phosphorylation of Rfx1, which results in the loss of DNA-binding activity and the derepression of these genes (22). Each Rfx1-repressed gene contains at least two widely spaced binding sites for Rfx1, one strong and one weak (22). A search of these genes revealed Mot3 and Rox1 sites in the upstream region of all three, but their arrangements are quite different (Fig. 1). RNR2 contains two isolated Mot3 sites, far from each other and from the Rfx1 and Rox1 sites. Our studies of the HEM13 and ANB1 genes (27, 30) led us to predict that Mot3 would not influence the expression of this gene; Mot3 cannot repress from a single, isolated site. However, the two Rox1 sites should repress aerobic RNR2 expression, and the upstream site is quite close to the weak Rfx1 site. RNR4 contains two Mot3 sites far from each other, but one is very close to the weak Rfx1 site. Again, our experience with HEM13 and ANB1 (27, 30) led us to speculate that Mot3 might act synergistically with Rfx1 to repress RNR4 but would not be able to repress RNR4 in the absence of Rfx1. In addition, there is a Rox1 site close to the Mot3-Rfx1 cluster, suggesting an additional possibility for synergy. RNR2 and -4 encode the small subunit of ribonucleotide reductase, and the expression of both genes is required for growth under normal conditions (14, 23, 49). Therefore, these two genes cannot be completely repressed. However, RNR3 encodes an alternate large subunit protein that is not significantly expressed under normal growth conditions but is induced during DNA damage (15). This stronger repression of RNR3 is reflected in the number and location of repression sites. RNR3 contains two weak and one strong Rfx1 site. It contains six Mot3 sites. Three are clustered very close together, while a fourth is very close to a weak Rfx1 site. It also contains three Rox1 sites, two closely spaced to the strong Rfx1 site. We would predict in this case that Mot3 alone could repress RNR3 through the three clustered sites, but, in addition, Mot3 and Rfx1 might act synergistically through the Rfx1-Mot3 associated sites, and, if Rox1 and Rfx1 could act synergistically, they would do so in this gene.

FIG. 1.

Arrangements of repressor sites in the RNR genes. The upstream regions of the RNR2, RNR3, and RNR4 genes are presented. The sequences are numbered in negative integers from the first base preceding the ATG translational initiation codon, −1, 5′-wards to −800. The Rox1 sites are indicated as open boxes with a diagonal stripe, Mot3 sites are indicated as gray boxes, the strongly binding Rfx1 sites are designated as black boxes and the weakly binding Rfx1 sites are indicated as open boxes.

To test the effects of Mot3 and Rox1 on the repression of these RNR genes, lacZ fusions were constructed to the regulatory region of each gene, and plasmids carrying these fusions were transformed into a set of congenic strains with single deletions in each of the repressor genes, the three combinations of double deletions, and the triple deletion. The levels of repression were determined under repressing conditions (aerobic and non-DNA damage) by β-galactosidase activity assays, and the results are presented in Tables 3 to 5. The data clearly confirmed the predictions above; all three gene fusions were repressed by Rox1 as well as Rfx1, and that the RNR3 and -4 fusions were repressed by Mot3, too. To assess whether these repressors functioned additively or synergistically, we analyzed the data in the three possible pairwise combinations. For example, for the RNR3-lacZ data presented in Table 4, the possible interactions between Mot3 and Rfx1 were determined in the absence of Rox1 by comparing the data in the rox1Δ deletant (the combined Mot3-Rfx1 effect) to that in the rox1Δ rfx1Δ deletant (the Mot3 effect alone) and in the rox1Δ mot3Δ deletant (the Rfx1 effect alone) and in the triple deletant (complete derepression). The fold repression for this set of comparisons was calculated by dividing the β-galactosidase activity in the fully derepressed triple deletant by that in the rox1Δ cells, the rox1Δ mot3Δ deletant, or the rox1Δ rfx1Δ deletant. We then calculated the synergy factor, defined as the fold repression by Mot3 alone (enzyme activity in rox1Δ mot3Δ rfx1Δ/rox1Δ rfx1Δ cells) times the fold repression by Rfx1 alone (enzyme activity in rox1Δ mot3Δ rfx1Δ/rox1Δ mot3Δ cells) divided into the fold repression by both repressors (enzyme activity in rox1Δ mot3Δ rfx1Δ/rox1Δ cells). This comparison assumes that independent, or additive, repression should be measured using the product of the effects of Rfx1 and Mot3 alone rather than the sum, because we envision repression through a single repressor occurring whenever the repressor is bound. Therefore, the fraction of time a gene can be expressed is the fraction of time when neither repressor is bound, or the product of the fold repression by the two repressors alone. This is a conservative estimate of additive repression, and we feel more confident assigning synergy using this assumption. A synergy factor of 1.0 represents additive repression, whereas a value significantly higher than 1 indicates synergy. For the example above, the synergy factor was 4.7, indicating a synergistic interaction between Mot3 and Rfx1 at RNR3.

TABLE 3.

Repression of the RNR2 gene

Genotypea	Mean β-galactosidase activity ± SD	Fold Rfx1-Rox1 repression (synergy factor)b
Wild type	33 ± 5.1	45 (1.8)
mot3Δ	33 ± 5.1
rfx1Δ	638 ± 73	2.3
mot3Δ rfx1Δ	676 ± 24
rox1Δ	139 ± 22	11
rox1Δ rfx1Δ	1,490 ± 117
rox1Δ mot3Δ rfx1Δ	1,520 ± 44

^aThe strains used were MZ233-20 (wild type), MZ233-2 (mot3Δ), MZ233-13 (rfx1Δ), RZ53-6Δrox1 (rox1Δ), MZ233-6 (mot3Δ rfx1Δ), MZ255-6 (rox1Δ rfx1Δ), and MZ269-8 (rox1Δ mot3Δ rfx1Δ). The cells were transformed with YCp(33)RNR2Z and grown aerobically in SC-uracil.

^bThe fold repression was calculated by dividing the activity in the rox1Δ rfx1Δ cells by that of the wild type or designated mutant. The synergy factor was calculated by dividing the fold repression of wild-type cells by the product of the fold repression of the rfx1Δ and rox1Δ mutants.

TABLE 5.

Repression of the RNR4 gene

Genotypea	Mean β-galactosidase activity ± SD	Fold repression (synergy factor)b
		Rfx1-Mot3	Rfx1-Rox1	Rox1-Mot3
Wild type	27 ± 6.6
mot3Δ	50 ± 3.8		20 (5.7)
rfx1Δ	591 ± 120			1.7 (1.3)
rox1Δ	135 ± 8.9	7.5 (2.8)
mot3Δ rfx1Δ	794 ± 43		1.3	1.3
rox1Δ rfx1Δ	1,350 ± 43	0.7		0.7
rox1Δ mot3Δ	377 ± 16	2.7	2.7
rox1Δ mot3Δ rfx1Δ	1,010 ± 49

^aThe strains used were MZ233-20 (wild type), MZ233-2 (mot3Δ), MZ233-13 (rfx1Δ), RZ53-6Δrox1 (rox1Δ), RZ56-6Δr1Δm3 (rox1Δ mot3Δ), MZ233-6 (mot3Δ rfx1Δ), MZ255-6 (rox1Δ rfx1Δ), and MZ269-8 (rox1Δ mot3Δ rfx1Δ). The cells were transformed with YCp(33)RNR4Z and grown aerobically in SC-uracil.

^bThe fold repression was calculated as described in Table 4. For the calculations of the synergy factor with the rox1Δ rfx1Δ strain, the fold repression was assumed to be 1.0.

TABLE 4.

Repression of the RNR3 gene

Genotypea	Mean β-galactosidase activity ± SD	Fold repression (synergy factor)b
		Rfx1-Mot3	Rfx1-Rox1	Rox1-Mot3
Wild type	0.66± 0.2
mot3Δ	2.7 ± 0.7		436 (44)
rfx1Δ	22 ± 5.7			54 (1.2)
rox1Δ	13 ± 0.7	91 (4.7)
mot3Δ rfx1Δ	253 ± 20		4.7	4.7
rox1Δ rfx1Δ	127 ± 9.0	9.3		9.3
rox1Δ mot3Δ	568 ± 27	2.1	2.1
rox1Δ mot3Δ rfx1Δ	1,180 ± 71

^aThe strains used were MZ233-20 (wild type), MZ233-2 (mot3Δ), MZ233-13 (rfx1Δ), RZ53-6Δrox1 (rox1Δ), RZ56-6Δr1Δm3 (rox1Δ mot3Δ), MZ233-6 (mot3Δ rfx1Δ), MZ255-6 (rox1Δ rfx1Δ), and MZ269-8 (rox1Δ mot3Δ rfx1Δ). Cells were transformed with YCp(33)RNR3Z and grown aerobically in SC-uracil.

^bThe fold repression was calculated as described in Table 3, except that the values for each of the single and double mutants indicated were divided into that for the rox1Δ mot3Δ rfx1Δ strain. The synergy factor was calculated as described in the text.

Expression of the RNR2-lacZ (Table 2) was repressed by Rfx1 and Rox1, but not by Mot3, as predicted above from the location of binding sites. The lack of a Mot3 effect was clearly evident in that the β-galactosidase activity was the same in the wild-type and mot3Δ cells and in the rfx1Δ and mot3Δ rfx1Δ mutants. Rox1 alone repressed expression of RNR2 2.3-fold, the ratio of enzyme activity in rox1Δ rfx1Δ to that in the rfx1Δ cells, and Rfx1 alone repressed expression 11-fold, the ratio of enzyme activity in rox1Δ rfx1Δ to that in the rox1Δ cells. The synergy factor was calculated as 1.8, a value that is suggestive, but not conclusive, that these repressors act synergistically at this gene.

As expected from previously reported results (15, 22), the expression of the RNR3-lacZ was most strongly repressed, showing a nearly 1,800-fold difference between the enzyme activity in wild-type cells and cells carrying the triple deletion (Table 4). This repression was achieved by synergy between Mot3 and Rfx1 (a synergy factor of 4.7) and Rox1 and Rfx1 (a synergy factor of 44) but not between Rox1 and Mot3 (a synergy factor of 1.2). Any one of the repressors alone effected weak repression as seen in the double mutants; for example, Rfx1 alone repressed RNR3 expression only about twofold (the ratio of enzyme activity in the triple deletant to that in the rox1Δ mot3Δ strain). Interestingly, the strongest repression by a single repressor is the effect by Mot3, 9.3-fold in the rox1Δ rfx1Δ mutant, and this probably resulted from the three adjacent sites close within 100 bp of the start of translation.

RNR4-lacZ expression was repressed by all three repressors. Synergy was clearly indicated between Rox1 and Rfx1, but we have some reservations about the calculation of synergy involving Mot3. The enzyme activity measured in the triple deletant was somewhat lower than that determined for the rox1Δ rfx1Δ strain, which resulted in a fold repression of less than 1, making the synergy factor calculation suspect. Although the reason for this discrepancy is not known, the fact that there is only one Mot3 site adjacent to an Rfx1 site suggests that Mot3 can act synergistically with Rfx1 but not independently; hence, no repression in the rox1Δ rfx1Δ mutant. This Mot3 effect is the same as that observed in ANB1 where Mot3 could not repress in the absence of Rox1 (27, 30).

It is interesting that despite the different effects of the three repressors, the RNR2 and -4 fusions appear to be expressed at very similar levels in wild-type cells and in the triple deletion. These genes encode the small subunit of the RNR enzyme (14, 21, 23, 49) and are probably required in equivalent concentrations in the cell. The lacZ fusions used in the present study contain the upstream region and one or six codons of the beginning of the coding sequences of the RNR genes only; the 3′ noncoding and transcriptional termination signals are those of the ANB1 gene. Thus, the levels of enzyme activity measured from these fusions may accurately represent the relative levels of transcription of the native genes.

It should be noted that, in cases in which synergy was clearly evident, the binding sites for the interacting repressors were within 30 bp of each other. When sites were 40 bp or greater apart, synergy was not seen, as in the case of the Mot3 and Rox1 sites in RNR3. However, the strength of the synergy was not predictable. Other factors, such as the intervening sequences, the location of undefined activation sequences, and the distance of the sites from the TATA box, may play roles in the strength of the synergy.

Finally, to address the concern that the results obtained with these RNR-lacZ fusions might be influenced by changes in the plasmid copy number in the different mutant strains, we integrated the RNR3-lacZ fusion into the genomic RNR3 locus in the wild-type and seven mutant strains. The β-galactosidase assays were repeated, and the results are presented in Table 6. Clearly, there were differences in the levels of expression of RNR3-lacZ. In most, but not all cases, the measured enzyme activity was lower. The levels of repression calculated were also not identical, but the synergy values calculated for the three pairwise interactions proved to be consistent. Using a centromeric plasmid, the synergies calculated for Rfx1-Mot3 and Rox1-Mot3 were 4.7 and 1.2, respectively, whereas the corresponding values obtained with the integrated plasmid were 8.6 and 1.2, respectively. While the synergy factors calculated for Rfx1-Rox1 were very different (44 for experiments with the centromeric plasmid and 152 for those with the integrated reporter), both indicate strong synergy. Overall, despite quantitative differences, the conclusions with both constructs are identical.

TABLE 6.

Repression of integrated RNR3-lacZ

Genotypea	Mean β-galactosidase activity ± SD	Fold repression (synergy factor)b
		Rfx1-Mot3	Rfx1-Rox1	Rox1-Mot3
Wild type	0.28 ± 0.1
mot3Δ	1.2 ± 0.7		486 (152)
rfx1Δ	45 ± 5.4			13 (1.2)
rox1Δ	5.2 ± 2.0	116 (8.6)
mot3Δ rfx1Δ	381 ± 41		1.6	1.6
rox1Δ rfx1Δ	89 ± 20	6.7		6.7
rox1Δ mot3Δ	298 ± 24	2.0	2.0
rox1Δ mot3Δ rfx1Δ	601 ± 14

^aThe strains used were MZ233-20 (wild type), MZ233-2 (mot3Δ), MZ233-13 (rfx1Δ), RZ53-6Δrox1 (rox1Δ), RZ56-6Δr1Δm3 (rox1Δ mot3Δ), MZ233-6 (mot3Δ rfx1Δ), MZ255-6 (rox1Δ rfx1Δ), and MZ269-8 (rox1Δ mot3Δ rfx1Δ). Each strain carried the YIp(211)RNR3Z plasmid integrated at the RNR3 locus. Cells were grown aerobically in SC-uracil.

^bThe fold repression was calculated as described in Table 3, except that the values for each of the single and double mutants indicated were divided into that for the rox1Δ mot3Δ rfx1Δ strain. The synergy factor was calculated as described in the text.

Derepression of the RNR genes results in resistance to DNA damage.

Mutations in RFX1 or other genes that increase RNR levels or activity increase cell resistance to DNA-damaging agents (5, 22). To visualize this effect for the combinations of repressor mutations, we arrayed cells on a plate containing a gradient of NQO. As seen in Fig. 2, the greatest resistance was observed with any set of mutants containing the rfx1Δ allele, which not only causes derepression of RNR2 to -4 but also other genes whose products may help protect the cell from NQO (53). Clearly, the rox1Δ allele alone also caused increased NQO resistance, and resistance was further enhanced in combination with the rfx1Δ allele. No significant affect was observed with the mot3Δ allele.

FIG. 2.

Mutations in the repressor genes increased cell resistance to DNA damage. Wild-type and mutant cells were stamped onto a gradient plate containing NQO from 0 to 2.6 μM (right to left) as described in Materials and Methods. The cells were MZ233-20 (WT), MZ233-2 (mot3Δ), MZ233-13 (rfx1Δ), RZ53-6Δrox1 (rox1Δ), RZ56-6Δr1Δm3 (rox1Δ mot3Δ), MZ233-6 (mot3Δ rfx1Δ), MZ255-6 (rox1Δ rfx1Δ), and MZ269-8 (rox1Δ mot3Δ rfx1Δ).

RNR2 and -3 are derepressed under anaerobiosis.

If Rox1 and Mot3 repress the expression of RNR2 to -4, these genes should be induced under anaerobic growth conditions when Rox1 and Mot3 are not present in the cell. To test this prediction, wild-type and mutant cells carrying either RNR2-lacZ or RNR3-lacZ fusions were grown for at least 17 h anaerobically to mid-exponential phase and assayed for the accumulation of β-galactosidase (Table 7). Comparison of the activity in wild-type cells grown aerobically and anaerobically shows that RNR2-lacZ expression was derepressed 3.1-fold and RNR3-lacZ expression was derepressed 70-fold anaerobically. Aerobic repression was due to Rox1 for RNR2 and Rox1 plus Mot3 for RNR3, as evident from the equivalent aerobic and anaerobic expression of these genes in the rox1Δ and the rox1Δ mot3Δ mutants, respectively. On the other hand, Rfx1 alone still repressed under anaerobic conditions, as determined by the 11-fold repression of RNR2-lacZ both aerobically and anaerobically (rox1Δ rfx1Δ/rox1Δ) and the 2.1-fold aerobic and 5-fold anaerobic repression of RNR3-lacZ (rox1Δ mot3Δ rfx1Δ/rox1Δ mot3Δ). These results indicate that Rfx1 is not regulated either positively or negatively by oxygen (and, by inference, Rox1 and/or Mot3); this repressor gave the same levels of repression both aerobically and anaerobically.

TABLE 7.

Anaerobic expression of RNR2 and -3

Genotypea	Mean β-galactosidase activity ± SD		Anaerobic/aerobic ratiod
	Aerobicb	Anaerobicc
RNR2-lacZ
Wild type	33 ± 5.1	101 ± 5.8	3.1
rfx1Δ	638 ± 73	947 ± 50	1.5
rox1Δ	139 ± 22	138 ± 7.5	1.0
rox1Δ rfx1Δ	1,490 ± 117	1,460 ± 7.5	0.98
rox1Δ mot3Δ rfx1Δ	1,520 ± 44	1,760 ± 46	1.2
RNR3-lacZ
Wild type	0.66 ± 0.2	46 ± 12	70
mot3Δ	2.7 ± 0.7	32 ± 3.4	12
rfx1Δ	22 ± 5.7	759 ± 112	34
rox1Δ	13 ± 0.7	65 ± 4.4	5.0
mot3Δ rfx1Δ	253 ± 20	800 ± 66	3.2
rox1Δ rfx1Δ	127 ± 9.0	835 ± 15	6.6
rox1Δ mot3Δ	568 ± 27	321 ± 20	0.56
rox1Δ mot3Δ rfx1Δ	1,178 ± 72	1,600 ± 124	1.4

^aThe strains used were MZ233-20 (wild type), MZ233-2 (mot3Δ), MZ233-13 (rfx1Δ); RZ53-6Δrox1 (rox1Δ), RZ56-6Δr1Δm3 (rox1Δ mot3Δ), MZ233-6 (mot3Δ rfx1Δ), MZ255-6 (rox1Δ rfx1Δ), and MZ269-8 (rox1Δ mot3Δ rfx1Δ). Cells were transformed with either YCp(33)RNR2Z or YCp(33)RNR3Z as indicated.

^bThe values for aerobic activity are taken from Tables 3 and 4.

^cCells were grown overnight anaerobically as described in Materials and Methods.

^dThe anaerobic/aerobic ratio was calculated by dividing the activity in cells grown anaerobically by that in the same cells grown aerobically.

If Rfx1 still repressed RNR2 and -3 under anaerobic conditions, then further induction should result from the addition of a DNA-damaging agent during anaerobic growth. This is the case as seen in Fig. 3. Cells were grown aerobically and then divided into two cultures. Anaerobic growth was initiated in both cultures by bubbling nitrogen through the flasks in the absence or presence of 1.3 μM NQO. It is clear that the addition of NQO at the onset of anaerobiosis resulted in a much more dramatic induction than anaerobiosis alone, which was barely visible on the scales of these plots.

FIG. 3.

Anaerobiosis and DNA damage induced RNR2Z and -3Z expression. MZ233-20 (wild-type) cells were transformed with either YCp(33)RNR2Z (A) or YCp(33)RNR3Z (B), and a 50-ml culture was grown aerobically to mid-exponential phase in SC-uracil supplemented with 0.2% Tween 80 and 20 μg of ergosterol/ml. A sample was taken for time zero, and then the cultures were divided into four flasks. A final concentration of 1.3 μM NQO was added to two of the flasks. Nitrogen was bubbled through the cultures, and time points were taken (see Materials and Methods). The results are plotted as Miller units versus the hours of anaerobic growth either in the absence (−NQO) or presence (+NQO) of the DNA-damaging agent.

Rox1 and Mot3 levels and activity are not regulated by DNA damage.

The apparent synergy between Rfx1 and Rox1 or Mot3 could result if these repressors indirectly regulated each other. While we ruled out regulation of Rfx1 activity by Rox1 and/or Mot3, it was still possible that the deletion of RFX1 somehow decreased the expression of Rox1 and Mot3, which would then result in derepression of RNR2-4 and the apparent loss of synergy. To test the effect of Rfx1 on Mot3 and Rox1 accumulation, we performed immunoblots with epitope-tagged forms of these proteins which are fully active in repression (27, 35). The mot3Δ and mot3Δ rfx1Δ mutants were transformed with a plasmid carrying the MOT3HA allele, and the rox1Δ and rox1Δ rfx1Δ mutants were transformed with a plasmid carrying the ROX1HA allele. As can be seen in Fig. 4, there was no difference in the levels of Mot3HA (Fig. 3A) or Rox1HA (Fig. 3B) in RFX1 or rfx1Δ cells. Furthermore, the addition of NQO to RFX1 cells did not alter the levels of either Rox1HA or Mot3HA over a 3-h period. The blots were also probed for the translation factor eIF5A as a loading control. The relative levels of Mot3HA or Rox1HA, normalized to that of eIF5A, are presented below each lane.

FIG. 4.

Rox1 and Mot3 levels were not regulated by Rfx1 or NQO. (A) An immunoblot was carried out with protein extracts from MZ233-2 (a mot3Δ strain designated wild type [WT] here because it carried the MOT3HA allele on a plasmid) and MZ233-6 (a mot3Δ rfx1Δ strain designated rfx1Δ here because it carried the MOT3HA allele on a plasmid) cells transformed with YCp(23)MOT3HA, and MZ233-2 (lane C) cells transformed with YC(22)MOT3. Cells were grown aerobically in SC-tryptophan medium. DNA damage was induced in MZ233-2 cells transformed with YCp(23)MOT3HA by the addition of NQO to a final concentration of 1.3 μM, and samples were taken at the times indicated. The blots were probed with monoclonal antibody against the HA epitope (Mot3-HA) and then stripped and reprobed with polyclonal antibody to eIF-5A. The relative levels of Mot3HA compared to the wild type are indicated below each lane, calculated as follows. The density of Mot3HA signal in each lane was divided by that for eIF5A in that lane. Then, this normalized level was divided by that for Mot3HA in the WT lane. (B) An immunoblot was carried out as in panel A with RZ53-6Δrox1 and MZ255-6 transformed with YCp(22)ROX1HA (WT and Δrfx1, respectively), and RZ53-6Δrox1 transformed with YCp(23)ROX1 (designated C). NQO induction was performed with the RZ53-6Δrox1 transformed with YCp(33)ROX1HA for the times indicated. The relative levels of Rox1HA, calculated as described for panel A, are indicated below each lane.

Although Rox1 and Mot3 levels did not fall upon induction of the DNA damage response, and although no change in the migration pattern of either protein was noted in the immunoblot after induction, it was possible that repressor activity was reduced through some undetected modification of the protein. To investigate this possibility, we measured RNR-lacZ induction by NQO in wild-type and rfx1Δ cells. If Rox1 and Mot3 retained full repression activity during NQO induction, then there would be no increase in RNR-lacZ expression in the rfx1Δ strain. As can be seen in Fig. 5, this was the case; the expression of the three genes remained unchanged in the rfx1Δ mutant and the rox1Δ mot3Δ rfx1Δ control (rox1Δ rfx1Δ for RNR2, which is not repressed by Mot3) compared to the robust induction in wild-type cells. To compare the induction profiles, we normalized the enzyme activity to the fold induction, with the uninduced levels as 1.0. (It should be noted that the un-normalized levels were quite different for each strain as seen in Tables 3 to 5, and although the fold induction levels in the rfx1Δ and double and triple mutants remained close to 1.0 throughout induction as expected, the absolute levels were quite high.)

FIG. 5.

Rox1 and Mot3 retained repression activity during DNA damage. MZ233-20 (WT), MZ233-13 (rfx1Δ), and MZ255-6 (rox1Δrfx1Δ) cells were transformed with YCp(33)RNR2Z (A), and MZ233-20 (WT), MZ233-13 (rfx1Δ), and MZ269-8 (rox1Δmot3Δrfx1Δ) cells were transformed with YCp(33)RNR3Z (B) or YCp(33)RNR4Z (C). The transformants were grown to mid-exponential phase in SC-uracil medium at 30°C with vigorous shaking. β-Galactosidase assays were performed with samples taken before (time zero) and at the indicated times after the addition of NQO to a final concentration of 1.3 μM. The β-galactosidase values were normalized to fold induction by dividing the activity at time zero into the activity at the indicated times for each strain.

A model system for characterizing synergy.

Past experience suggested that synergistic interactions in repression can be masked when a gene contains multiple repressor sites. For example, the synergy between Rox1 and Mot3 in the repression of HEM13 expression became more dramatic when the number of Mot3 sites was reduced (30). Furthermore, although we tried to simplify the analysis of synergy above by pairwise comparisons through the elimination of one repressor by mutation, we wanted to simplify the experimental system further. To this end, we generated a series of artificial constructs containing single Rox1 plus Mot3, Mot3 plus Rfx1, or Rox1 plus Rfx1 sites. These sites are shown in Table 2 and were inserted into the regulatory region of the ANB1-lacZ fusion in which all Rox1 and Mot3 sites had been deleted, and the resulting plasmids were transformed into a wild-type strain and the relevant single and double deletants. Cells were grown aerobically without DNA damaging agents, i.e., conditions under which each of the repressors were active.

The synergy between Rox1 and Mot3 is clearly evident (Table 8). Single Rox1 and Mot3 sites provided poor repression with either Rox1 (in mot3Δ cells) or Mot3 (in rox1Δ cells) but strong repression when both proteins were present (in wild-type cells). The synergy factor was more than 6. We also generated a series of constructs with the three other possible orientations of the nonpalindromic Rox1 and Mot3 sites with respect to each other, and a construct with a deletion of 5 bp between the two sites (changing the side of the helix to which the proteins bound). Synergy was evident in all cases (data not shown). This lack of orientation or spacing requirements again reinforces the conclusion that Rox1 and Mot3 do not interact directly, but through a larger, more flexible complex.

TABLE 8.

Combinatorial repression with single Rox1 or Rfx1 and Mot3 sites

Genotypea	Mean β-galactosidase activity (fold repression, synergy factor)b ± SD
	Mot3-Rox1	Mot3-Rfx1weak	Mot3-Rfx1strong
Wild type	8.8 ± 1.8 (36, 6.3)
mot3Δ	118 ± 34 (2.7)
rox1Δ	148 ± 13 (2.1)
mot3Δ rox1Δ	316 ± 17 (1.0)
Wild type		34 ± 8.7 (13, 2.3)	1.4 ± 0.8 (312, 3.4)
mot3Δ		167 ± 15 (2.6)	9.5 ± 1.4 (46)
rfx1Δ		199 ± 20 (2.2)	219 ± 7 (2.0)
mot3Δ rfx1Δ		434 ± 15 (1.0)	438 ± 21 (1.0)

^aThe strain series transformed with YCp(33)AZ-M3R1 were RZ53-6 (wild type), RZ53-6Δmot3 (mot3), RZ53-6Δrox1 (rox1Δ), and RZ53-6Δr1m3 (mot3Δ rox1Δ). The strain series transformed with YCp(33)AZ-M3Rfw and YCp(33)AZ-M3Rfs plasmids were MZ233-20 (wild type), MZ233-2 (mot3); MZ233-13 (rfx1Δ), and MZ233-6 (mot3Δ rfx1Δ).

^bThe fold repression and synergy factor were calculated as described in Table 3. The lacZ fusion plasmids used were: YCp(33)AZ-M3R1 containing the Mot3-Rox1 sites; YCp(33)AZ-M3Rfw containing the Mot3-Rfx1weak sites; and YCp(33)AZ-M3Rfs containing the Mot3-Rfx1strong sites. The inserts are presented in Table 2.

Rfx1-repressed genes contain at least one strong and one weak binding site. We constructed reporter genes containing both a single Mot3 and a single strong or weak Rfx1 site as indicated in Table 8. As evident from the data, either site alone repressed the reporter gene expression, but, as might be expected, Rfx1 repressed expression much more through the strong site either in the presence of Mot3 (312-fold repression in wild-type cells with the strong site reporter versus 13-fold repression in wild-type cells with the weak site reporter) or in its absence (46-fold repression in mot3Δ cells with the strong site reporter versus 2.6-fold repression in mot3Δ cells with the weak site reporter). In both cases, synergy was clearly evident, with a calculated synergy factor of 2.3 for the weak site and 3.4 for the strong site.

We integrated the AZ-M3Rfs reporter into the genome to ascertain that plasmid copy number effects were not influencing the measurements of synergy. The values obtained for the β-galactosidase activities were 1.2 ± 0.3 (wild type), 20 ± 5 (mot3Δ), 249 ± 25 (rfx1Δ), and 642 ± 63 (mot3Δ frx1Δ). Again, these values agree well with but are not identical to those obtained with the centromeric plasmid. The synergy factor calculated from these values is 6.3, confirming the that Mot3 and Rfx1 can act synergistically.

To test the synergy between Rox1 and Rfx1, we constructed lacZ reporter genes containing single Rox1 plus Rfx1 sites. We used a weak Rfx1 site and inserted the Rox1 site in the two different orientations with respect to the Rfx1 site. The results of β-galactosidase assays performed with wild-type, rox1Δ, rfx1Δ, and rox1Δ rfx1Δ transformants are presented in Table 9. It is evident from the data that the combined repression by Rox1 and Rfx1 was greater than that expected from additive repression alone; the synergy factor was about 3 independent of the orientation of the Rox1 site. Furthermore, although the orientation of the Rox1 site did appear to have a significant effect on repression, the fold repression was about three times greater for the Rox1 site in the inverted orientation both in the absence (rfx1Δ cells) and in the presence (wild-type cells) of Rfx1; there was no orientation dependence of the synergy factor. This observation also suggests that the two repressors do not interact directly. Thus, all of the evidence suggests that synergy involving Rox1, Rfx1, and Mot3 functions through indirect interactions.

TABLE 9.

Combinatorial repression between Rox1 and Rfx1

Genotypea	Mean β-galactosidase activity (fold repression, synergy factor) ± SDb
	Rox1-Rfx1weak	InvertedRox1-Rfx1weak
Wild type	5.1 ± 0.5 (57, 2.9)	2.0 ± 0.2 (134, 3.0)
rox1Δ	46 ± 1.8 (6.3)	58 ± 4.9 (4.6)
rfx1Δ	95 ± 11 (3.1)	28 ± 2.2 (9.6)
rox1Δ rfx1Δ	291 ± 21 (1.0)	268 ± 27 (1.0)

^aThe strain series transformed with the YCp(33)AZ-R1Rf or YCp(33)AZ-1RRf were RZ53-6 (wild type), RZ53-6Δrox1 (rox1Δ), MZ233-13 (rfx1Δ), and MZ255-6 (rox1Δ rfx1Δ).

^bThe fold repression and synergy factor were calculated as described in Table 3. The lacZ fusion plasmids used were YCp(33)AZ-R1Rf containing the Rox1-Rfx1 sites and YCp(33)AZ-1RRf containing the inverted Rox1 and Rfx1 sites. The inserts are shown in Table 2.

Neither Rox1 nor Rfx1 can function in synergy with itself.

Previous evidence indicated that neither Rox1 nor Mot3 could act synergistically with itself. While the data for Mot3 were obtained in a comprehensive study involving sites in HEM13 (30), the data for Rox1 were compiled from several different studies (9, 27, 30). To test the synergy between two Rox1 sites conclusively, we used a set of previously constructed plasmids. Each contained a deletion of OpB and either an intact OpA with its two Rox1 sites or only the 5′ or 3′ Rox1 site of OpA. These constructs were transformed into a mot3Δ strain and a rox1Δ mot3Δ strain to negate the effect of the Mot3 site present in OpA and the 3′Δ plasmids.

The results, presented in Table 10, indicate that two bound Rox1 molecules could not function synergistically. The Rox1 bound to the single 5′ site in OpAΔ3′ repressed the reporter gene expression 2.2-fold, while the Rox1 bound to the 3′ Rox1 site in OpAΔ5′ repressed reporter gene expression 3.7-fold. The product of these two effects would give an additive repression of 8.1-fold for two Rox1 molecules bound to two sites functioning independently. The measured fold repression for the wild-type OpA was 9.7-fold, which is nearly the same as that for additive repression.

TABLE 10.

Combinatorial repression between two Rox1 molecules

Genotypea	Mean β-galactosidase activity (fold repression) ± SDb
	Wild type	Δ5′	Δ3′	−5	−10
mot3Δ	38 ± 1.7 (9.7)	75 ± 2.6 (3.7)	190 ± 7.6 (2.2)	33 ± 3.8 (9.7)	100 ± 4.6 (5.7)
mot3Δ rox1Δ	374 ± 10.6	277 ± 60	418 ± 7.8	322 ± 29	573 ± 29

^aStrains RZ53-6Δmot3 (mot3Δ) and RZ53-6Δr1Δm3 (mot3Δ rox1Δ) were transformed with the YCp(33)AZΔB (wild type), YCp(33)AZΔ5′OpAΔB (Δ5′), YCp(33)AZΔ3′OpAΔB (Δ3′), YCp(33)AZΔB(−5) (−5), or YCp(33)AZΔB(−10) (−10). The number of base pairs between the Rox1 sites in these plasmids are 15 bp in the wild-type plasmid, 10 bp in the −5 plasmid, and 5 bp in the −10 plasmid.

^bThe fold repression was calculated as the enzyme activity in the mot3Δ rox1Δ cells divided by the enzyme activity in the mot3Δ cells for each plasmid.

To ensure that the lack of synergy was not due to the relative location of the Rox1 sites, we also assayed for synergy in two additional constructs that contained a 5- or a 10-bp deletion between the Rox1 sites. Besides moving these sites closer together, the 5-bp deletion altered the side of the helix to which the two Rox1 molecules bound. As seen in Table 10, these deletions did not alter the calculated synergy factor; repression from the two sites was additive. It should be noted that the 10-bp deletion did reduce repression about twofold. Perhaps the close proximity of the sites lead to an interference in binding.

Rfx1 sites are most often present in pairs, with one strong and one weak site in the upstream region of genes (for examples, see Fig. 1). However, these sites are quite far apart. To determine whether two closely spaced Rfx1 sites could act combinatorially, YCp(33)AZ-2Rf was constructed containing two weak Rfx1 sites 10 bp apart (Table 2). This plasmid was transformed into the mot3Δ and mot3Δ rfx1Δ strains. For a single Rfx1 site, the plasmid YCp(33)AZ-M3Rfw, containing the same Rfx1 site and a Mot3 site, was transformed into the same strains. Since both strains contained the mot3Δ deletion, the lacZ expression from the two plasmids would provide the fold repression for one and two Rfx1 sites. RFX1 cells carrying the plasmid with two Rfx1 sites contained 239 ± 14 Miller units, whereas rfx1Δ cells carrying the same plasmid contained 459 ± 28 Miller units, a 1.9-fold repression. The plasmid carrying only one Rfx1 site gave a 2.6-fold repression (Table 5). Thus, two weak Rfx1 sites positioned 10 bp apart could not even function additively in repression. Perhaps two Rfx1 molecules could not bind simultaneously or could not recruit additional Ssn6/Tup1 molecules. In any event these results and those with two Rox1 and two Mot3 sites indicate that synergy appears to be limited to heterologous proteins.

DISCUSSION

We previously reported two types of combinatorial repression by Rox1 and Mot3: additive and synergistic (30). In the former, a set of sites functioned independently, or additively, to repress transcription. An example of such repression was reported for HEM13, in which Mot3 could repress independently from a cluster of thee sites. Each site contributed additively to repression. Synergy was observed as an interaction between Mot3 and Rox1, in which Mot3 failed to repress independently from a single site but, in combination with Rox1, enhanced Rox1 repression. This synergy was observed both in HEM13 and in ANB1 and resulted from an indirect interaction in which Rox1 and Mot3 interacted through the Ssn6 subunit of the general repression complex. The indirect nature of synergy suggested that it might operate between Mot3 and/or Rox1 and other Ssn6/Tup1 recruiting repressors. We demonstrated here that this is indeed the case; Mot3 and Rox1 interacted synergistically with Rfx1.

We observed repression by Mot3 at RNR3 and -4 and by Rox1 at RNR2, -3, and -4. These genes encode three of the four proteins of RNR and are repressed by Rfx1. The gene encoding the fourth protein, RNR1, is not repressed by Rfx1 (22) and contains no Rox1 or Mot3 sites. Nor was the expression of an RNR1-lacZ fusion altered in the triple deletion (K. Cartmill and R. Zitomer, unpublished results). Ribonucleotide reductase is comprised of a large and a small subunit (25). Rnr2 and -4 form the small subunit in yeast, and Rnr2 contains a diferric tyrosyl radical center required for catalysis (5, 17). Formation of the tyrosyl radical requires molecular oxygen and is facilitated by Rnr4 (5, 17). Hence the reason for the regulation of RNR2 and -4 by Rox1 and Mot3 may be to increase the levels of protein to drive radical formation under hypoxic conditions. (It should be noted that there is a hypothetical 138 codon open reading frame 155 bp upstream from RNR2 which, if authentic, would share a regulatory region with RNR2.) The large subunit is composed of Rnr1 homodimers under normal growth conditions, whereas Rnr3 represents an alternative large subunit protein induced during DNA damage (15). The role that Rnr3 might play under hypoxia is unclear, but Rnr1-Rnr3 heterodimers are more active in vitro (11), and the hypoxic induction of Rnr3 may function to increase RNR activity as formation of the holoenzyme becomes compromised.

The synergy among differentially regulated repressors reported here has great biological import. It allows strong repression of a gene through the action of two (or three) weakly acting repressors, yet the loss of repression by only one allows substantial derepression. Thus, multiple repressors functioning at a single gene do not have to be coordinately regulated to achieve robust induction in response to only one signal; the loss of synergy achieves that end. For the repressors studied here, Rox1 and Mot3 syntheses are regulated positively by oxygen, while Rfx1 activity is negatively regulated by DNA damage. We presented evidence here that there was little, if any, cross-regulation; Rox1 and Mot3 were fully functional during DNA damage, and Rfx1 retained full activity during hypoxia. Nonetheless, during DNA damage, a substantial part of Rox1 and Mot3 repression of the RNR genes was lost because the synergy with Rfx1 was lost. RNR4 regulation presents a good example of this phenomenon as seen in the data in Table 5. RNR4 expression was repressed 37-fold by the combination of the three repressors (comparing the triple deletion to the wild type). Rfx1 alone could repress expression only 2.7-fold (comparing the triple deletion to the rox1Δ mot3Δ mutant). Yet the loss of Rfx1 repression that might mimic DNA damage resulted in a 22-fold induction (comparing the rfx1Δ mutant to the wild type). Because of the loss of the combined synergy between Rfx1 and Rox1 and between Rfx1 and Mot3, nearly full induction was achieved. Similarly, under hypoxic conditions, simulated by the loss of Rox1 and Mot3 in the rox1Δ mot3Δ mutant, RNR4 was induced 14-fold due to the inability of Rfx1 to strongly repress in their absence. Furthermore, since synergy is indirect, functioning through Ssn6, the evolution of synergy in this system did not even require the acquisition of direct interactions among these proteins. The only requirement for the introduction of this type of synergy in a system is the acquisition of closely spaced binding sites. Thus, we anticipate that many other examples of this biological phenomenon will be discovered. It should be noted, however, that these systems can be quite well disguised; elegant studies of Rfx1 repression of the RNR genes gave no clue that the robust repression by Rfx1 actually required the presence of two other proteins at the genes (22).

To explore the nature of synergy further, we devised artificial constructs in which we could control the number and relative positions of the repressor binding sites with respect to each other while holding the upstream activation sites, TATA box, and all other upstream sequences constant. Using single binding sites for each repressor, these constructs demonstrated the synergy between pairwise combinations. Interestingly, Mot3 acted synergistically equally well with strong and weak Rfx1 sites, suggesting that Mot3 was not simply helping Rfx1 bind better to weak sites. This conclusion fits well with our model of Rox1-Mot3 synergy in which Mot3 helps Rox1 recruit Ssn6; perhaps Mot3 is functioning in the same capacity with Rfx1. In addition, synergy is independent of the orientation of the asymmetric Rox1 and Mot3 sites and the Rox1 and Rfx1 sites, reinforcing the model that synergy does not require direct contact between the DNA-binding repressors. Interestingly, synergy was not previously observed between closely spaced Mot3 molecules or, in the present study, between closely spaced Rox1 or Rfx1 molecules. Thus, it appears that synergy is limited to heterologous repressors. This conclusion supports our previously proposed model in which the synergy arises from the sharing of Ssn6/Tup1 complexes by Rox1 and Mot3. The requirement for heterology in this model arises because the two repressors must contact different surfaces of Ssn6.

There are at least two other examples in the literature of two different Ssn6-Tup1 recruiting repressors regulating the same genes. Proft and Serrano (39) reported that the ENA1 gene was repressed by Sko1, a repressor of osmotically induced genes, and Mig1/Mig2, repressors of glucose-repressed genes. The binding sites for Sko1 and Mig1/Mig2 are 20 bp apart. While the authors did not address the question of additive or synergistic repression directly, data presented in Table 3 concerning the expression of an ENA1-lacZ fusion in wild-type, sko1Δ, mig1Δ mig2Δ, and triple sko1Δ mig1Δ mig2Δ cells allow the calculation of a synergy factor of 3.1. Thus, it appears that Sko1 can act synergistically with Mig1 and Mig2. In the second example, the Ssn6/Tup1 recruiting repressors Rim101, a repressor of meiotic genes, and Nrg1, a repressor of glucose-repressed genes, were demonstrated to repress the divergently transcribed DIT1-DIT2 genes (42). Both repressors bound to a 40-bp fragment between the two genes. This fragment was inserted upstream of the CYC1-lacZ gene, and expression studies presented in Fig. 7 of reference 42 allow the calculation of a Synergy Factor of 1, indicating that Nrg1 and Rim101 did not act synergistically. These two examples suggest that some, but not all Ssn6/Tup1 recruiting repressors can act synergistically.