Systematic analysis of tup1 and cyc8 mutants reveals distinct roles for TUP1 and CYC8 and offers new insight into the regulation of gene transcription by the yeast Tup1-Cyc8 complex

The Tup1-Cyc8 complex in Saccharomyces cerevisiae was one of the first global co-repressors of gene transcription discovered. However, despite years of study, a full understanding of the contribution of Tup1p and Cyc8p to complex function is lacking. We examined TUP1 and CYC8 single and double deletion mutants and show that CYC8 represses more genes than TUP1, and that there are genes subject to (i) unique repression by TUP1 or CYC8, (ii) redundant repression by TUP1 and CYC8, and (iii) there are genes at which de-repression in a cyc8 mutant is dependent upon TUP1, and vice-versa. We also reveal that Tup1p and Cyc8p can make distinct contributions to commonly repressed genes most likely via specific interactions with different histone deacetylases. Furthermore, we show that Tup1p and Cyc8p can be found independently of each other to negatively regulate gene transcription and can persist at active genes to negatively regulate on-going transcription. Together, these data suggest that Tup1p and Cyc8p can associate with active and inactive genes to mediate distinct negative and positive regulatory roles when functioning within, and possibly out with the complex.

when multiple mechanisms of repression used by Tup1-Cyc8 were disrupted [36]. It is conceivable that different mechanisms of action are required at different genes, or that combinations of these mechanisms can determine the transcription state of genes in response to the changing environment. Regardless, this highlights that despite years of study, a complete understanding of this complex has yet to be uncovered.
Much of the current knowledge about this complex, and the genes under its control, has come from analyses of mutants deleted for either the TUP1 or CYC8 genes [35,37]. In support of this strategy, the current model for the activity of the complex would predict that deletion of either Tup1p or Cyc8p should have the same impact in crippling complex function. However, the common consideration of these mutants as being interchangeable for the analysis of Tup1-Cyc8 complex function ignores any differences upon transcription and cell function that deletion of these genes might have.
In this study we have compared single and double mutant strains deleted for the TUP1 and CYC8 genes and show that they have distinct phenotypes and transcriptomes. Almost twice as many genes were upregulated in the cyc8 mutant compared to the tup1 mutant. By comparing the transcription data in the single mutants to that in the double mutant we show evidence of (i) genes subject to redundant repression via TUP1 and CYC8, (ii) genes which were uniquely repressed by either TUP1 or CYC8, and (iii) genes at which de-repression in a cyc8 mutant is dependent upon TUP1, and vice-versa. We also reveal that Tup1p and Cyc8p can make distinct contributions to commonly repressed genes. Furthermore, we show that Cyc8p and Tup1p can occupy promoters independently of each other to promote gene repression and can persist at active genes to negatively influence on-going transcription.
Together, these data suggest that Tup1p and Cyc8p have uncharacterised negative and positive roles when functioning both within and possibly out with the complex. Ultimately, the model for Tup1-Cyc8 functioning solely as a repressor of transcription is too simplistic. Instead, Tup1-Cyc8 should be considered as a more versatile regulator of transcription functioning to not only switch genes off, but also to modulate transcription of genes when they are active.

Results
TUP1 and CYC8 deletion mutants have largely been used interchangeably for the analysis of Tup1-Cyc8 complex function. This was justified by the current model dictating that without either Tup1p or Cyc8p, Tup1-Cyc8 complex activity should be equally disrupted [37]. We therefore compared phenotypes and transcription in tup1 and cyc8 single and double deletion mutants to determine if these mutants share the same characteristics or not.

Strains deleted for TUP1 and CYC8 show distinct growth, cell morphology, and flocculation phenotypes
Growth analysis showed that the tup1, cyc8 and tup1 cyc8 deletion mutants had progressively longer doubling times ( Fig 1A). However, following growth to saturation in YPD, all strains achieved similar final cell densities.
A striking characteristic of tup1 and cyc8 mutant strains grown in liquid media is their strong flocculation phenotype [38,39]. Flocculation is the asexual aggregation of cells due to the expression of the FLO family of genes which are known to be repressed by Tup1-Cyc8 [40][41][42]. The FLO genes encode lectin-like cell wall proteins called flocculins which bind to the mannose residues within the cell walls of neighbouring cells [43]. Flocculation can be distinguished from other forms of cell aggregation by being a calcium-dependent process [44,45]. Thus, the addition of EDTA, which chelates calcium ions and disperses the flocs, can be used as a control to confirm this phenotype.
Following growth in broth, whereas the wt showed no flocculation, the tup1 strain displayed large flocs, whilst the flocs in the cyc8 strain were smaller and more widely dispersed throughout the liquid media ( Fig 1B). The tup1 cyc8 double deletion mutant showed an intermediate flocculation phenotype in which large flocs were visible in addition to more dispersed smaller flocs. Treatment of the tup1, cyc8 and tup1 cyc8 cultures with EDTA dispersed the cells confirming the flocculation phenotypes in each mutant ( Fig 1C). Thus, deletion of TUP1 yields the greatest visible flocculation phenotype.
When cells were visualised under the microscope, the tup1 mutant cells were visible as large clumps of cells with little interstitial space evident (Fig 1D). The cyc8 mutant formed smaller clumps of cells with more visible gaps between the cells. The double mutant resembled the tup1 single mutant with large clumps of tightly packed cells being visible. These data suggest that tup1 mutant cells might form tighter associations when flocculating compared to cyc8 cells.
Profiling of the different EDTA-dispersed cell populations for cell size revealed that strains harbouring a CYC8 mutation had a larger proportion of cells with larger cell sizes compared to wt or tup1 cells (Fig 1E).

Strains deleted for CYC8 and TUP1 have different responses to stressors
Flocculation is a stress response in which cells on the inside of a floc are shielded from chemical stressors which cannot easily infiltrate the tightly packed cells [46]. We therefore investigated the ability of each strain to tolerate a variety of stressors ( Fig 1F).
When cells were exposed to ethanol, which can lead to loss of membrane integrity, wt survival was reduced to 20%. However, survival of the tup1 mutant was 2-fold less than wt, survival of the cyc8 mutant was higher than wt, and survival in the double mutant was similar to wt. Thus, the cyc8 mutant showed the greatest resistance to ethanol, despite this strain showing the weakest visible flocculation phenotype. Conversely, the highly flocculant tup1 mutant was more sensitive to ethanol than wt.
When cells were exposed to H 2 O 2 , which can cause oxidative damage, wt survival was reduced to 10% whereas all the mutants showed significantly increased survival. The tup1 mutant showed more survival than the cyc8 mutant, whilst survival was the greatest in the double mutant. Thus, resistance to H 2 O 2 was greatest in the tup1 cyc8 double mutant which showed an intermediate flocculation phenotype.
Exposure of cells to amphotericin B, an antifungal drug which targets the cell membrane, caused almost 100% cell death in wt. However, all the mutants showed significantly increased survival levels with survival in the tup1 mutant being >2-fold higher than that in the cyc8 mutant. The double mutant showed survival levels similar to the tup1 mutant.
The ability of the strains to tolerate high temperatures in liquid culture was included as a control, as temperature should affect all cells equally regardless of the extent of flocculation [46]. As expected, following exposure of cells to high temperature (50˚C), no significant difference in survival in any of the strains could be detected.
Thus, the mutants showed differences in their resistance to chemical stressors that were not always dependent upon their flocculant phenotypes.

Strains deleted for TUP1 de-repress FLO1 gene transcription to the greatest extent
The most striking phenotype displayed by the mutant strains was the flocculation phenotype which is mediated by expression of the FLO family of genes [42]. We therefore measured FLO gene transcription in each mutant and wt (Figs 2A and S1).
The mRNA levels of the FLO genes were low in wt, consistent with Tup1-Cyc8 dependent repression of transcription, and correlating with the lack of flocculation in the parent strain. Consistent with FLO1 being the dominant flocculation gene, FLO1 mRNA levels were the highest of the FLO genes tested in each of the mutants (Figs 2A and S1). FLO1 mRNA levels were greater in the tup1 mutant than in the cyc8 mutant. Interestingly, the fact that FLO1 transcription in the cyc8 single mutant was further de-repressed when TUP1 was additionally deleted (Fig 2A, compare cyc8 and tup1 cyc8) suggests that TUP1 can exert a repressive effect upon FLO1 in the absence of CYC8. Thus, the extent of FLO1 de-repression correlates with the flocculation phenotypes of each strain and TUP1 makes the major contribution to FLO1 repression.

Strains deleted for CYC8 de-repress SUC2 gene transcription to the greatest extent
Another well-characterised gene subject to Tup1-Cyc8 dependent repression is the SUC2 gene [47]. SUC2 encodes invertase which hydrolyses sucrose to yield glucose and fructose. The gene is repressed in the presence of high levels of glucose and is induced under conditions of low glucose [48,49]. We therefore analysed SUC2 transcription in the cyc8 and tup1 single and double mutants under conditions of high and low glucose (Fig 2B).
Consistent with published data, SUC2 mRNA in wt was barely detectable under conditions of high glucose (repressed), whilst the gene was significantly de-repressed under conditions of low glucose (de-repressed) (Fig 2B, wt) [49].
In the tup1 mutant, the level of SUC2 mRNA in the repressed (high glucose) condition was similar to the SUC2 mRNA present in the wt strain under the de-repressed (low glucose) condition (Fig 2B, tup1). This is consistent with the loss of glucose repression of SUC2 in the tup1 mutant due to disruption of the Tup1-Cyc8 complex [48]. However, SUC2 mRNA levels were even greater when the tup1 mutant was grown under low glucose conditions. de-repression in tup1, cyc8 and tup1 cyc8 relative to wt was 78-, 31-and 66-fold, respectively. (B) SUC2 transcript levels measured following growth on high (Repressed) and low (De-repressed) glucose levels. In all graphs, values were normalised to ACT1 mRNA and error bars reflect standard deviation from 3-4 biological replicates (** represents a p-value of p<0.005 determined by a One-way ANOVA analysis). https://doi.org/10.1371/journal.pgen.1010876.g002 In both the cyc8 and the tup1 cyc8 double mutant strains grown on high glucose, SUC2 mRNA levels were higher than levels found in both the repressed and de-repressed tup1 cells (Fig 2B). SUC2 mRNA was also further elevated in the cyc8 and the tup1 cyc8 double mutant cells when grown under low glucose conditions. Together, these data show that in the wt strain under conditions of low glucose, TUP1 and CYC8 still exert a repressive effect upon SUC2 transcription. The fact that transcription in the tup1 single mutant is further de-repressed in either the repressed or de-repressed conditions when CYC8 is additionally deleted (Fig 2B, compare tup1 and tup1 cyc8) suggests that the lower levels of SUC2 mRNA in the tup1 mutant under either glucose condition is dependent upon CYC8. Finally, these data show that even in the absence of both Tup1p and Cyc8p, SUC2 is not fully de-repressed when glucose is present.
Thus, CYC8 makes the major contribution to SUC2 repression and can exert a repressive effect upon SUC2 transcription in the absence of TUP1. Furthermore, TUP1 and CYC8 continue to exert a repressive effect upon SUC2 transcription during the de-repressing conditions associated with a low-glucose environment.

Different numbers of genes are upregulated in cyc8 and tup1 deletion mutants
We next examined global transcription in the tup1, cyc8 and tup1 cyc8 mutants to determine if there were more widespread differences in gene transcription in strains deleted for TUP1 and/ or CYC8.
Consistent with Tup1-Cyc8 having been best characterised as a co-repressor of gene transcription, 469 genes were upregulated in the tup1 mutant, 809 genes were upregulated in the cyc8 mutant, and 851 genes were upregulated the tup1 cyc8 double mutant ( Fig 3A) [2]. Conversely, only 86, 124 and 114 genes were downregulated more than two-fold in the tup1, cyc8 and tup1 cyc8 mutants, respectively. We therefore focussed our analysis on the genes that were upregulated in the tup1, cyc8 and tup1 cyc8 mutant strains compared to wt, where TUP1 and CYC8 could be inferred to be playing a role in gene repression.
According to the current model for Tup1-Cyc8 complex activity, it would be predicted that if Tup1p and Cyc8p functioned solely within the Tup1-Cyc8 complex, the same genes would be upregulated in each of the single and double mutants. However, there were 7, 114 and 131 genes exclusively upregulated in the tup1, cyc8 and tup1 cyc8 mutants, respectively (Fig 3B), suggesting unique cohorts of TUP1 and CYC8 repressed genes.

Genes showing independent repression by CYC8 and TUP1
The data showing 262 genes upregulated only in the cyc8 and tup1 cyc8 mutants compared to wt suggests that these genes are subject to unique repression by CYC8 (Fig 3B). Although a scatter plot confirms that the average upregulation of transcription of these genes in the cyc8 and tup1 cyc8 mutants was greater than transcription in the tup1 mutant (Fig 3C), to identify those genes solely repressed via CYC8 more accurately we set three, more stringent, parameters. Firstly, we proposed that the genes must show minimal transcription in the tup1 mutant compared to wt (cut off = Log 2 fold-change (FC) �0.3). Secondly, upregulation of these genes should not be higher in the tup1 cyc8 double mutant compared to the cyc8 mutant. Thirdly, we excluded genes showing very low transcription levels in the mutants (cut off average transcripts per million (TPM) �60). This analysis uncovered 13 genes which we propose are uniquely repressed via CYC8 in wt (Figs 3D, S2A and S2B). A similar analysis of the 29 genes upregulated only in the tup1 and tup1 cyc8 mutants (Fig 3B and 3E) revealed 6 genes uniquely repressed via TUP1 (Figs 3F, S2C and S2D). Cluster heatmap for the 14 genes which showed no, or minimal upregulation in the absence of TUP1, but were upregulated in the absence of CYC8. In all graphs, error bars reflect standard deviation (* represents a p-value of p<0.05, ** represents a p-value of p<0.005 determined by ANOVA analysis). Each heatmap displays Z-scores for each gene; each row represents a gene, and each column represents a deletion mutant. The colour scale indicates the standard deviations above or below the mean fold change for each gene compared to wt. https://doi.org/10.1371/journal.pgen.1010876.g003

Genes subject to redundant repression by CYC8 and TUP1
We next investigated the 131 genes that were upregulated at least two-fold only in the tup1 cyc8 double mutant (Fig 3B, 3G and 3H). This result indicates that each subunit can compensate for the absence of the other to bring about repression at these genes. Full gene de-repression is only achieved when both subunits are deleted. An example of a gene subject to potential redundant repression via TUP1 and CYC8 was FIT2 (S2E Fig).

Evidence of CYC8-repressed genes subject to positive regulation via TUP1
We next examined the genes which were de-repressed in either or both tup1 and cyc8 single mutants, but which were not de-repressed in the double mutant ( Fig 3B). For example, there were 114 genes significantly upregulated only in the cyc8 mutant (Fig 3B and 3I). This profile suggests that these genes were subject to unique CYC8-dependent repression but were TUP1dependent for transcription in the absence of CYC8. To increase the stringency for this gene cohort, genes showing very low levels of transcription (average TPM of <60) in each of the mutants were discarded, and genes showing any change in transcription in the tup1 mutant were also excluded. This revealed a cohort of 14 CYC8 repressed genes at which TUP1 was required for their transcription in the absence of CYC8 (Figs 3J and S2F).

Distinct levels of gene de-repression occur in tup1 and cyc8 mutants
We next examined the transcription levels of all the genes upregulated in the tup1, cyc8 and tup1 cyc8 mutants to determine if there were any general differences in the levels of gene derepression in any of these strains which might indicate distinct contributions to gene repression by TUP1 or CYC8.
Comparison of the levels of de-repression of the total number of genes de-repressed in the cyc8 (809), tup1 (469) and tup1 cyc8 (851) strains revealed no difference in their average levels of de-repression ( Fig 4A). However, examination of the 429 genes commonly upregulated in the tup1, cyc8 and tup1 cyc8 mutants revealed average upregulation was the least in tup1, was higher in cyc8, and was highest in tup1 cyc8 ( Fig 4B). Furthermore, the average upregulation of these commonly repressed genes in the cyc8 and tup1 cyc8 mutants was significantly greater than the average upregulation of the total number of genes de-repressed in these strains (compare Fig 4B to 4A). This indicates that the set of 429 commonly repressed genes represents a core set of genes that are subject to (i) robust repression by TUP1 and CYC8 and (ii), distinct levels of repression by TUP1 and CYC8, with CYC8, on average, making the greater contribution to their repression than TUP1.

There are distinct cohorts of genes commonly repressed by TUP1 and CYC8
Visualisation of the relative levels of transcription of the 429 commonly de-repressed genes in the different mutants via a heat map confirmed that most genes had the highest de-repression in the tup1 cyc8 mutant ( Fig 4C). However, 110 genes had the highest de-repression in the cyc8 mutant and the lowest de-repression in the tup1 mutant ( Fig 4C). An example of a gene showing this transcription profile was SUC2 (see Fig 2B). Additionally, there was a cohort of 40 genes which showed greater de-repression in the tup1 mutant compared to the cyc8 mutant, of which, FLO1 was an example (Fig 4C, see Fig 2A). There were also genes, such as RNR3, which were equally upregulated in each mutant (S3 Fig) [50].

The commonly repressed genes subject to differential TUP1 and CYC8 repression are enriched within distinct sub-telomeric regions
In a large proportion of the 429 genes commonly upregulated in the tup1, cyc8 and tup1 cyc8 mutants, transcription was de-repressed the most in a cyc8 mutant compared to tup1 (see Fig 4B and 4C). An example of this cohort of genes was SUC2. Conversely, FLO1 was representative of the smaller subset of the 429 commonly de-repressed genes which were most derepressed in the tup1 mutant. We therefore investigated these two cohorts of commonly repressed genes (SUC2-type and FLO1-type genes) to determine if there were any unique characteristics that might explain why their repression was most dependent on either CYC8 or TUP1.
Gene-ontology analysis did not reveal any distinction between these two cohorts of commonly repressed gene. Consistent with previous studies, the majority of the total TUP1 and CYC8 repressed genes were enriched near the ends of chromosomes (Fig 5A), and that genes within sub-telomeric regions were subject to the most robust repression via CYC8 and TUP1 ( Fig 5B) [27]. Further analysis revealed that the distribution of total TUP1 and CYC8 repressed genes correlated well with the Hda1-affected sub-telomeric (HAST) domains located between Scatterplots showing the log 2 fold change values of (A) all genes at least two-fold upregulated in each of the deletion strains compared to wt, and (B) the 429 genes commonly upregulated in all three deletion mutants compared to wt. In each case, log 2 �1, adjusted p-value of p�0.01, mean, and standard deviation are shown (* represents a p-value of p�0.05, ** represents a p value of p�0.005 as determined by ANOVA analysis). (C) Cluster heatmap for the 429 genes at least two-fold upregulated in all three deletion strains compared to wt. The heatmap displays Z-scores for each gene; each row represents a gene, and each column represents a deletion mutant. The colour scale indicates the standard deviations above or below the mean fold change for each gene compared to wt.
Analysis of the basal level of transcription of the SUC2-and FLO1-type genes in wt revealed that, overall, the two sets of genes showed no difference in their low TPM values in wt (S5 Fig). This suggests both gene cohorts were equally robustly repressed in wt.
Previous studies had shown that FLO1 and SUC2 were subject to long-range antagonistic chromatin remodelling by Tup1-Cyc8 and Swi-Snf in their extensive gene transcription-free upstream regions [28][29][30]. We therefore examined if the FLO1-and SUC2-type genes had different lengths of upstream or downstream gene-free regions which might influence their regulation (S6A and S6B Fig). The results showed no significant difference in the average length of up and downstream intergenic regions, or open reading frame (ORF) lengths, between the two sets of genes. However, there was a positive correlation between the length of the upstream intergenic region and the levels of gene de-repression for the FLO1-type genes in the tup1 cyc8 double mutant, and for the SUC2-type genes in the cyc8 mutant (S6C and S6D Fig). Thus, repression of both sets of genes are similarly influenced by the extent of their gene-free upstream regions.
We next analysed transcription factor (TF) consensus sequences in the promoter regions of the two gene cohorts (S1 Table). This showed that the average number of motifs found upstream of the two sets of genes was very similar, with 111.6 for the FLO1-type genes, and 111.25 for SUC2-type genes. However, at SUC2-like genes, there was an enrichment of binding sites for Nrg1p, Msn2p and Hap1p, whereas binding sites for Yap1p, Hac1p and Gcn4p were enriched at the FLO1-type genes.
Thus, the two cohorts of commonly repressed genes differ in their distribution within the HAST domains of the sub-telomeric regions and are associated with different transcription factor binding sites.

Cyc8p can occupy the SUC2 promoter in the absence of Tup1p
In the majority of the 429 genes commonly upregulated in the tup1, cyc8 and tup1 cyc8 mutants, gene transcription was de-repressed the least in a tup1 mutant compared to the derepression in the cyc8 and tup1 cyc8 double mutants (see Fig 4B and 4C). This suggests that there was a CYC8-dependent repressive effect upon transcription of these genes in the absence of TUP1. To test whether the repressive role of CYC8 in the absence of TUP1 might be direct or not, we performed chromatin immunoprecipitation (ChIP) analysis of Cyc8p and Tup1p occupancy using SUC2 as an example of this cohort of genes in the glucose-grown strains (Fig 6). We first performed ChIP analysis of RNA polymerase II (RNAP II) at the SUC2 gene to confirm the RNA-seq and RT-qPCR data (Fig 6A). Consistent with the transcription data, Distance is in kilobase (kb) pairs from each telomere grouped into 50 kb regions. (B) Scatterplot of the log 2 fold changes of the total upregulated genes and the 429 commonly upregulated genes separated into genes located within the sub-telomeric regions of the chromosomes (<25 kb from the telomeres) and the genes located throughout the rest of the chromosomes (Non sub-telomeric). Mean and standard deviation are shown, * represents a p-value of p�0.05, ** represents a p-value of p�0.005 as determined by ANOVA analysis. (C) Graph showing distribution of genes significantly upregulated in each mutant compared to wt (% of total upregulated genes) located in the first 50 kb from each telomere grouped into 5 kb regions (columns, lefthand Y axis), and the Hda1-affected sub-telomeric (HAST) domain (dashed line, right-hand Y axis). Distance is in kilobase pairs from each telomere. Right-hand Y axis indicates regions of hyperacetylation (HAST domain) in an hda1 deletion mutant; adapted from Robyr et al., 2002 [62]. (D) Distribution of the 429 commonly upregulated genes over the first 50 kb regions from the telomeres divided into 5 kb regions and separated into FLO1-type (blue), SUC2-type (green), and the remaining commonly upregulated genes (pink). The percentage of genes in each group was calculated.
Surprisingly, we could not detect significant enrichment of Tup1p or Cyc8p at the repressed SUC2 promoter in the wt strain at the previously reported site of Tup1-Cyc8 occupancy (Fig 6B and 6C; wt, repressed) [52,53]. We suggest this discrepancy is due to differences in the antibodies and ChIP signal normalisation strategies used between labs. However, supported by the abundance of literature detailing Tup1-Cyc8 repression of SUC2, we propose that Tup1p and Cyc8p are present at the repressed SUC2 promoter, but are not detectable by ChIP using our conditions [3,37,47]. Consistent with the model for Tup1-Cyc8 function, we also could not detect significant occupancy of Tup1p at SUC2 in the cyc8 mutant (Fig 6B; cyc8, repressed). However, we could detect significant enrichment of Cyc8p at the partially de-repressed SUC2 promoter in the tup1 mutant (Fig 6C; tup1, repressed). Thus, Cyc8p was present at the partially de-repressed SUC2 promoter in the absence of Tup1p where it could contribute directly to negatively influencing SUC2 transcription.

Tup1p is detectable at FLO1 in the absence of Cyc8p
We next looked at RNAP II, Tup1p and Cyc8p occupancy at the FLO1 gene which was representative of the smaller subset of the 429 commonly repressed genes (Fig 6D-6F). De-repression at these genes was greater in the tup1 and tup1 cyc8 double mutant compared to derepression in the cyc8 mutant, suggesting a TUP1-dependent repressive effect upon transcription of these genes in the absence of CYC8 (see Fig 4C). Firstly, the RNAP II ChIP results were consistent with the mRNA levels detected in the strains whereby there were low RNAP II levels in wt, and high RNAP II levels in the tup1 and tup1 cyc8 mutants (compare Figs 6D and 2A). Consistent with Tup1-Cyc8 mediated repression of FLO1, Tup1p and Cyc8p occupancy could be detected at the wt FLO1 promoter when the gene was off (Fig 6E and 6F, wt) [27,54]. Similar to what was seen at SUC2, Cyc8p was also detected at significant levels at FLO1 in the tup1 mutant (Fig 6F, tup1). Most interestingly, Tup1p could be detected at the FLO1 promoter in the absence of Cyc8p (Fig 6E, cyc8).
Therefore, ChIP analysis confirmed that Tup1p and Cyc8p could be detected at the repressed FLO1 promoter. The data also revealed that Cyc8p could be detected at high levels at FLO1 and SUC2 in the absence of Tup1p, and that Tup1p could be detected at the FLO1 promoter in the absence of Cyc8p. Thus, Cyc8p could directly contribute to repression of SUC2 transcription independent of Tup1p, and Tup1p could directly contribute to FLO1 repression in the absence of Cyc8p.

Tup1p and Cyc8p are present at the active SUC2 and FLO1 genes
Previous work had suggested that the Tup1-Cyc8 complex remains at some genes, including SUC2, following gene activation [21,23,52,53]. We therefore analysed Tup1p and Cyc8p occupancy at SUC2 following its activation in response to low glucose conditions (Fig 6B and 6C; wt, de-repressed). Consistent with previous data, we confirmed that Tup1p and Cyc8p could be detected at the SUC2 gene following its induction.
In most laboratory strains, including the ones used in this study, flocculation is an undesirable phenotype and has been attenuated by a nonsense mutation in the FLO8 gene which encodes an activator of the FLO genes [55,56]. Thus, most studies to investigate FLO1 transcription employ tup1 or cyc8 mutants in which repression of FLO1 is abolished. To examine Tup1p and Cyc8p occupancy at the active FLO1 gene under non-mutant conditions we restored the wt FLO8 gene at its genomic locus prior to performing ChIP (S7, S8A and S8B Figs). Following restoration of a functional FLO8 gene, the FLO8+ strain exhibited a flocculant phenotype and showed FLO1 mRNA levels similar to those in the cyc8 mutant (Fig 6G). Subsequent tagging of the restored FLO8 gene allowed us to confirm that Flo8p was expressed, and ChIP analysis revealed that Flo8p could specifically occupy the FLO1 promoter (Figs 6H, S8B, S8E and S8F). Together, this suggests that when Flo8p is expressed under the control of its native promoter it can bind the FLO1 promoter to activate FLO1 transcription.
ChIP analysis for Tup1p occupancy at the FLO1 promoter in the FLO8+ strain, where FLO1 is being transcribed at levels similar to that in the cyc8 mutant, revealed that Tup1p was present (Fig 6I). Furthermore, the levels of Tup1p detected at the active FLO1 gene in the FLO8+ strain were similar to those seen in the wt strain when FLO1 transcription was repressed. This suggests that Tup1p, most likely in the context of the Tup1-Cyc8 complex, can persist at the FLO1 gene when it is being actively transcribed.

Tup1p and Cyc8p have unique and shared sites of occupancy across the genome
Finally, we examined the global occupancy of Tup1p and Cyc8p using previously published data obtained via ChIP-Exo (Fig 7) [57]. The data showed there were more sites of occupancy detected for Tup1p (761) than for Cyc8p (506) and that, although there was a significant overlap in the Tup1p and Cyc8p sites of occupancy (421 genes), Tup1p and Cyc8p could be detected at a significant number of sites independently from each other (Fig 7A). Indeed, Tup1p could be found at 340 unique sites whilst Cyc8p was located at 85 unique sites.
We then examined how many of the genes up-and downregulated in a tup1 mutant harboured a Tup1p site of occupancy in their corresponding promoter region in wt. Out of the 555 genes differentially transcribed in the tup1 mutant, 23% of these genes (131) contained a Tup1p peak (Fig 7B). Similarly, when we looked at the 933 genes differentially transcribed in the cyc8 deletion mutant, 13% of these genes (125) contained a site of Cyc8p occupancy in their promoters (Fig 7C). Comparison of the unique sites of Tup1p and Cyc8p occupancy with the genes identified as being subject to unique repression by TUP1 and CYC8 respectively, revealed only a small overlap (S9 Fig). Thus, Tup1p and Cyc8p can be found at unique sites across the genome, and some of these sites correspond to those genes we previously described as being subject to unique TUP1 and CYC8 repression. However, overall, the correlation of Tup1p and Cyc8p occupancy with the genes up and down regulated in their absence, was poor.

Summary of results
In summary, our analysis has revealed that (i) CYC8 represses more genes than TUP1, (ii) some genes are uniquely repressed by either TUP1 or CYC8, (iii) other genes are subject to redundant repression by TUP1 and CYC8, and (iv), some TUP1 repressed genes require CYC8 for their de-repression and vice-versa (Fig 8A). We also show that TUP1 and CYC8 can make different contributions to commonly repressed genes (Fig 8B and 8C).
Many commonly repressed genes behave like SUC2, whereby CYC8 makes the dominant contribution to repression (Fig 8C, SUC2-type genes). A smaller cohort of commonly repressed genes behave like FLO1 and are subject to dominant repression by TUP1 (Fig 8C,  FLO1-type genes). These two cohorts of commonly repressed genes differ in their distribution within the HAST domains of sub-telomeric regions, in which they are enriched, and are associated with different transcription factor binding sites (Fig 8B and 8C). At SUC2-type genes, CYC8 can exert repression in the absence of Tup1p, whereas at FLO1-type genes, TUP1 can exert repression in absence of Cyc8p. In addition, TUP1 and CYC8 can exert repression at both SUC2 and FLO1 during gene activation.
Global ChIP data confirmed that Tup1p and Cyc8p can be found independent of each other, and at genes we had identified as being subject to unique TUP1 and CYC8-dependent repression. In addition, Cyc8p could be detected at SUC2 in the absence of Tup1p where it could directly negatively influence SUC2 de-repression (Fig 8E). Furthermore, Tup1p could be detected at FLO1 in the absence of Cyc8p, where it could directly negatively influence FLO1 de-repression ( Fig 8D). Our data also suggests that Tup1-Cyc8 persists at both the SUC2 and FLO1 genes when active, where it continues to negatively modulate transcription (Fig 8F).
Together, this suggests the potential for distinct novel regulatory roles for Tup1p and Cyc8p when functioning within, and possibly out with, the Tup1-Cyc8 complex. Furthermore, these data suggest the Tup1-Cyc8 complex can function as a molecular 'dimmer switch' to fine-tune active transcription in addition to its role as an outright repressor of transcription.

Discussion
The Tup1-Cyc8 co-repressor complex was one of the first global repressors of gene transcription identified [1]. Several mechanisms of action have been proposed for Tup1-Cyc8 repression including the formation of repressive chromatin structures, inhibiting RNA polymerase II, and blocking transcription factor activation domains [37]. These roles have been proposed to function via the WD40 domain of Tup1p and the TPR motifs of Cyc8p which offer a versatile interface for multiple interactions with a large array of transcription factors, non-acetylated histone H3 and H4 tails, various histone deacetylases, and several RNA polymerase II subunits [8,11,58].
Most work to characterise Tup1-Cyc8 function used either TUP1 or CYC8 gene deletion mutants. This approach was justified considering the general model for Tup1-Cyc8 structure and function which predicts that a tup1 mutant should lack the repressive activity of the complex, while the complex should be unable to bind target genes in a cyc8 mutant [37]. Thus, both mutants should equally inhibit Tup1-Cyc8.
Our systematic analysis of single and double mutants deleted for TUP1 and CYC8 revealed the different strains have numerous distinct phenotypes and significant differences in their transcriptomes which has offered new insight into Tup1p, Cyc8p and Tup1-Cyc8 function. We propose that there are different subsets of Tup1-Cyc8 repressed genes which are subject to distinct regulation by either Tup1p or Cyc8p. Our data also suggests that Tup1p and Cyc8p can function independently within, and possibly out with, the complex.
Phenotypically, strains deleted for CYC8 had the slowest growth and displayed a large cell morphology (Fig 1). Most strikingly, the tup1 mutant had the strongest flocculation phenotype [46]. However, the flocculation phenotypes of the mutants did not always correlate with the cell's responses to stress.
These data suggested that the wide-ranging differences in the tup1 and cyc8 mutants could be the result of altered transcription in strains deleted for TUP1 and CYC8. In support of this, TUP1 and CYC8 made distinct contributions to the repression of FLO1 and SUC2 transcription, which are two genes known to be repressed by the Tup1-Cyc8 complex (Fig 2) [27,29,47]. FLO1 transcription was de-repressed the most in the tup1 mutant, whilst SUC2 transcription was de-repressed the most in the cyc8 mutant. This suggests a greater role for Tup1p in FLO1 repression and a greater role for Cyc8p in SUC2 repression. Furthermore, since SUC2 de- repression in tup1 cyc8 was greater than that in the tup1 mutant, this suggests that CYC8 exerts a repressive effect independent of Tup1p.
To determine whether the Tup1p independent role of CYC8 at SUC2 was direct or not, we examined Cyc8p occupancy at SUC2 (Fig 6C). Surprisingly, we could not detect Cyc8p or Tup1p at SUC2 in the glucose grown wt strain, where Tup1-Cyc8 has previously been detected (Fig 6B and 6C) [52,53]. However, we propose Tup1p and Cyc8p are present at the repressed SUC2 promoter, but our ChIP protocol cannot detect them, possibly due to epitope masking by other factors present at this site. Conversely, a strong signal for Cyc8p could be detected at the SUC2 promoter in the tup1 mutant ( Fig 6C). Thus, the transcription and ChIP results at SUC2 in glucose grown cells are consistent with CYC8 making the major contribution to SUC2 repression and suggest that Cyc8p directly contributes to SUC2 repression in the absence of Tup1p. This latter result suggests that at SUC2, under conditions of high glucose in the absence Tup1p, either Cyc8p can exert a repressive effect on its own, or the presence of Cyc8p at the SUC2 promoter can enable another factor or factors to exert repression or inhibit activation. In support of this, recent work has shown that multiple proteins can interact with the Tup1-Cyc8 complex to fine-tune target gene transcription [6,7,59].
When we looked at FLO1 regulation, TUP1 played the greatest role in repression (Fig 2A). Furthermore, TUP1 exerted a repressive effect in the absence of Cyc8p since FLO1 de-repression was greater in tup1 cyc8 compared to that in cyc8. Consistent with Tup1-Cyc8 dependent repression of FLO1, Tup1p and Cyc8p were detected at the wt FLO1 promoter, when the gene is inactive (Fig 6E and 6F). In the tup1 mutant, where FLO1 transcription was de-repressed to the greatest extent, although Cyc8p occupancy was detected, our data suggests it does not contribute to repression. Surprisingly, we could detect Tup1p at FLO1 in the cyc8 mutant ( Fig 6E). This suggests that Tup1p can directly negatively influence FLO1 transcription in the absence of Cyc8p. This has not been reported before and could have been missed due to different normalisation strategies employed during ChIP analysis [27].
Analysis of SUC2 transcription in the various mutants under conditions of low glucose revealed that TUP1 and CYC8 continue to exert a repressive effect upon SUC2 transcription with CYC8 again exerting the dominant effect ( Fig 2B). Our data confirming that Tup1p and Cyc8p remain associated with the SUC2 promoter under activation conditions of low glucose suggests that the complex is present at the active SUC2 gene and acts as a 'brake' to dampen down on-going SUC2 transcription (Fig 8F).
To examine FLO1 gene activity under more physiological conditions, we reinstated the activator of FLO1 transcription, Flo8p [55,56]. In a strain expressing Flo8p (the FLO8+ strain), Flo8p occupied the FLO1 promoter, and FLO1 was transcribed at levels similar to that in a cyc8 mutant (Fig 6G and 6H). Furthermore, Tup1p was also present at the active FLO1 promoter in the FLO8+ strain (Fig 6I). Importantly, since FLO1 transcription in the FLO8+ strain was similar to that in a cyc8 mutant, this suggests that the presence of Tup1p, presumably in the context of the Tup1-Cyc8 complex, negatively influences on-going FLO1 transcription.
Together, the FLO1 and SUC2 transcription and ChIP data suggest a model of action in which Tup1-Cyc8 represses the genes under non-inducing conditions ( Fig 8C) and persists at both genes during their activation to negatively modulate on-going transcription (Fig 8F). The Tup1-Cyc8 occupancy at the actively transcribed genes could also poise genes for rapid repression.
The persistence of Tup1-Cyc8 at active genes has been reported previously and provided evidence of Tup1-Cyc8 acting as an activator of gene transcription [20,21,35,52,53]. At the repressed GAL1 gene, Tup1-Cyc8 occupancy enabled recruitment of the SAGA coactivator complex which was proposed to disrupt Tup1-Cyc8 repression and aid GAL1 transcription under inducing conditions [52]. Tup1-Cyc8 was also shown to prime the repressed matingtype specific genes under its negative control for activation via Gcn5p-dependent pre-acetylation of histones at the repressed gene promoters [60]. Interestingly, a recent study revealed that activator-dependent eviction of HO promoter nucleosomes was required for Tup1p to bind to the promoter and bring about repression [61]. Hence, Tup1-Cyc8 is emerging as being central to the dynamic interplay between gene repression and activation in which repressors are required for activation and activators are required for repression.
Our RNA-seq analysis confirmed that most genes under control of TUP1 and CYC8 were negatively regulated, consistent with Tup1-Cyc8 primarily being considered to be a repressor of gene transcription (Fig 3A) [2,3]. We also confirmed that the complex can positively regulate gene transcription, although far fewer genes require TUP1 and CYC8 for activation.
We therefore limited our analysis of the RNA seq data to the genes upregulated in the tup1, cyc8 and tup1 cyc8 mutants where it could be inferred that TUP1 and CYC8 played a role in repression. The data showed that almost twice as many genes were upregulated in the cyc8 mutant (809) compared to the tup1 mutant (469), suggesting a wider role for Cyc8p in global gene repression. Of the genes upregulated in the tup1, cyc8, and tup1 cyc8 mutants, there was a cohort of 429 genes that were commonly de-repressed in all the mutants (Fig 3B). We propose that these genes are subject to the most robust repression by Tup1p and Cyc8p functioning as the Tup1-Cyc8 complex.
Analysis of the TUP1 and CYC8 commonly repressed genes revealed two distinctly regulated classes of genes ( Fig 4C). A large number of genes behave like SUC2 and were most repressed via CYC8, whilst a smaller subset behaves like FLO1 and were most repressed by TUP1 (Fig 8C).
We showed that the 429 commonly repressed genes were enriched in gene-sparse sub-telomeric regions in an area that correlated well with the zones subject to the most Hda1p activity; the so-called HAST domains (Fig 5A and 5C) [27,62]. Interestingly, the FLO1-type genes were enriched within a single narrow peak within the HAST domain, whilst the SUC2-type genes showed a broader distribution either side of the peak of FLO1-type genes (Figs 5D and S4). This observation reinforces the link between Tup1-Cyc8 repression activity and histone deacetylases (HDACs) and might suggest a different HDAC dependency of the SUC2-type and FLO1-type genes [25,27].
In support of this hypothesis, it has been shown that Tup1p and Cyc8p have distinct interaction profiles with various class I and II HDACs [26,63]. For example, Rpd3p can interact with Cyc8p independently of Tup1p, whereas Hda1p has been shown to physically interact with Tup1p. At FLO1, repression requires both Hda1p and Rpd3p and both HDACs occupy the inactive FLO1 promoter in a Tup1-Cyc8 dependent manner [27]. Interestingly, Hda1p, but not Rpd3p, remains detectable at the partially de-repressed FLO1 promoter in a cyc8 mutant [27]. Furthermore, when hda1 is additionally deleted in a cyc8 mutant, the level of FLO1 de-repression is elevated to a level similar to that in a tup1 cyc8 mutant (S12C Fig; compare cyc8,  hda1 cyc8 and tup1 cyc8). Thus, the Tup1p that we have detected at the FLO1 promoter in the cyc8 mutant could mediate the reported retention of Hda1p which could contribute to the Cyc8p independent repression of FLO1 transcription shown here (Fig 8D).
It might therefore be predicted that a different HDAC would be detected at the SUC2 promoter in a tup1 mutant having been retained via the observed occupancy of Cyc8p at the SUC2 promoter in the absence of Tup1p (Fig 8E). Thus, it might be that the FLO1-type genes are most dependent upon Hda1p for repression via interaction with Tup1p, and the SUC2type genes might require an alternative HDAC for most repression via interaction with Cyc8p ( Fig 8C). To confirm this, HDAC occupancy in wt, tup1 and cyc8 mutant backgrounds would need to be performed, and global Tup1p and Cyc8p occupancy profiles in the respective cyc8 and tup1 mutants identified and analysed.
Previous work had shown that the FLO1 and SUC2 genes were subject to long-range chromatin remodelling in their extensive upstream intergenic regions [28][29][30]. Although we could not identify any significant difference in the length of intergenic region upstream of the SUC2and FLO1-type genes, there was a positive correlation between the level of gene de-repression and the length of the upstream regions in those mutants yielding maximal de-repression (S6C and S6D Fig). Thus, although there is an influence of the length of upstream region upon TUP1 and CYC8-dependent gene repression, this effect is common to both the FLO1-and SUC2-type genes. This supports a role for long-range chromatin remodelling in mediating Tup1-Cyc8 target gene repression [28,29].
Analysis of transcription factor binding motifs upstream of the SUC2-and FLO1-type genes, revealed that the FLO1-type genes had a larger proportion of Yap1p, Hac1p and Gcn4p binding motifs whilst Hap1p, Nrg1p and Msn2p/Msn4p were most enriched at the SUC2-type genes (S1 Table). Both Hac1p and Yap1p can recruit Tup1-Cyc8 to target genes where they play a positive role in transcription, and Hac1p has been shown to physically interact with Cyc8p [64,65]. This correlates with the higher transcription of the FLO1 type genes in the tup1 mutant strain compared to that in the cyc8 mutant strain if the Cyc8p occupancy detected at FLO1 in the tup1 mutant is also found at all the FLO1-type genes (Fig 8E). Gcn4p also plays a role in the activation of transcription. At ARG1, Cyc8p was needed for efficient binding of Gcn4p to the promoter, whereas loss of TUP1 had less of an effect on Gcn4p binding [59]. Again, this correlates with the lower de-repression of the FLO1-type genes in a cyc8 mutant compared to de-repression in a tup1 mutant and the Cyc8p occupancy detected at FLO1 in the absence of Tup1p (Fig 8D and 8E).
Together, our analysis suggests that the two cohorts of commonly repressed genes have differences in transcription factor (TF) binding and HDAC dependency. The HDACs might contribute most to gene repression, whilst the TFs might be more relevant to target genes when active. Importantly, we suggest that Tup1-Cyc8 persists at active target genes with the Tup1p and Cyc8p subunits interacting and influencing HDAC and TF occupancy and activity differently at the distinct gene cohorts to yield the transcriptional outcome.
Our analysis also found subsets of genes uniquely upregulated in either the tup1 or cyc8 mutants (Fig 3B). This suggests Tup1p and Cyc8p could have independent repressive roles at genes when residing within the complex or when functioning independently out with the complex. Indeed, global occupancy analysis reveals Tup1p and Cyc8p can be found at 340 and 85 unique sites across the genome, respectively (Fig 7A). Furthermore, the Tup1p independent occupancy of Cyc8p at SUC2, where it can exert a repressive role, supports the fact that Cyc8p can influence repression independent of Tup1p (Fig 6C). Similarly, the Cyc8p independent occupancy of Tup1p at FLO1 supports a direct role for Tup1p contributing to gene repression in the absence of Cyc8p (Fig 6E).
The correlation of Tup1p and Cyc8p occupancy with the genes under their control was poor (Fig 7B and 7C). This suggests either most genes are indirectly regulated by Tup1p and Cyc8p or, that current ChIP analysis of Tup1p and Cyc8p cannot identify all their binding sites. Application of improved techniques to measure protein occupancy may reveal their precise locations across the genome in the future. Additionally, analysing cells under dynamic growth conditions might allow the entire suite of sites occupied by Tup1p and Cyc8p, and the genes under their control, to be fully exposed [66].
By comparing transcription profiles in the single mutants to the double mutant we identified genes subject to redundant repression, and some genes at which de-repression in cyc8 was TUP1-dependent, and vice versa. (Fig 3B). Both results suggest novel functions for Tup1p and Cyc8p which are inconsistent with the model for Tup1-Cyc8 complex function. Despite the caveats described above, ChIP analysis of Tup1p occupancy in a cyc8 mutant and Cyc8p occupancy in a tup1 mutant might confirm what our global transcription analysis has suggested.
In summary, our analysis offers a compendium of data for tup1 and cyc8 mutants to be considered when studying Tup1-Cyc8 complex activity by traditional gene deletion analysis. We have shown TUP1 and CYC8 can make distinct contributions to repression and activation of specific gene cohorts. We show evidence that TUP1 and CYC8 can repress genes independently of each other and offer evidence that, at least at FLO1 and SUC2 respectively, this repression can be direct. Our data suggests that the mechanism of action of Tup1-Cyc8 is much more complex than previously thought and that Tup1p and Cyc8p can have individual roles which may be functioning when the proteins are residing within and possibly out with the complex. Further study will be required to fully elucidate the roles of the Tup1p and Cyc8p proteins and the Tup1-Cyc8 complex. Indeed, the cumulative evidence shown here and elsewhere points to the Tup1-Cyc8 complex having a much more versatile role in gene regulation which is not limited to it functioning solely as a repressor [23].

Yeast strains and growth conditions
The S. cerevisiae strains used are described in S2 Table. Yeast gene deletions and tagging were performed using polymerase chain reaction (PCR)-based methods [67,68]. All strain constructions were confirmed by PCR and/or western blot analysis and assayed for appropriate phenotypes/function where relevant. The construction and confirmation of the restored FLO8 gene and the FLO8-myc strain, are described in S7 and S8 Figs. Cells were grown at 30˚C in YPD medium unless stated otherwise.

Protein analysis
Protein lysate preparation and western blotting analysis were performed as previously described [27,54]. The antibodies and conditions used are described in S3 Table. Flocculation assay This assay was performed as previously described [54,69]. Cells with a flocculant phenotype aggregate in the absence of EDTA and are dispersed in the presence of EDTA (final concentration of 2 mM).

Cell microscopy
Exponentially growing cells were resuspended in either sterile water or EDTA (2 mM) and viewed under 100x oil immersion magnification. Leica Application Suite (LAS) software was used to capture the images.

Survival assays
The survival assays were performed as previously described with minor adaptations [46,70]. Aliquots (1 ml) of 5 x 10 8 exponentially growing cells were resuspended in 5 ml YPD and incubated for 1 hour at 30˚C. Aliquots were exposed to ethanol (15%), hydrogen peroxide (5 mM), Amphotericin B (15 μg/ml), growth at 50˚C, and growth at 30˚C (control) for 1 hour. The percent survival compared to the control was calculated by counting colony forming units (CFUs).

Determination of transcription factor binding motifs
The known DNA binding motifs for all transcription factors (TF) found in the sequences 1000 bp upstream of the FLO1-and SUC2-type genes were retrieved from YEASTRACT [71]. The number of times a motif was found upstream of each gene was calculated and this information was compiled into a table (S1 Table). For the final analysis, any TF DNA binding motif that was found at less than 40% of both the FLO1 and SUC2-type genes was omitted. A parameter of at least a difference of 10% in the proportion of genes with that motif in each of the groups (FLO1-type and SUC2-type genes) was categorised as a significant difference.

RT-qPCR
RNA extraction, cDNA preparation and RT-qPCR analysis were performed as previously described [54]. For SUC2 analysis, exponentially growing YPD cultures were divided into two equal portions. Cell pellets were washed twice and resuspended in YP containing glucose at either 2% (repressed) or 0.05% (de-repressed). Cultures were incubated for a further 120 min [29]. Values were normalised to ACT1 RNA. Primers used are shown in S4 Table. RNA-Seq RNA was extracted from exponentially growing cells using the Hot Phenol method and purified using the RNeasy Minelute Cleanup Kit (Qiagen) [72]. Total RNA was sent to Genewiz (Azenta Life Sciences) for rRNA depletion, cDNA library preparation and strand-specific RNA sequencing using the Illumina HiSeq Platform. Sequence reads were trimmed to remove possible adapter sequences and nucleotides with poor quality using Trimmomatic v.0.36. The trimmed reads were mapped to the Saccharomyces cerevisiae S288c reference genome available on ENSEMBL using the STAR aligner v.2.5.2b. and BAM files were generated. Unique gene hit counts were calculated by using featureCounts from the Subread package v.1.5.2. Genewiz preformed DGE analysis using the DESeq2 Bioconductor package. Using normalised counts, Log 2 fold change and a Benjamini-Hochberg adjusted p-value was then calculated. Conversion from BAM into BigWig files was carried out using bamCoverage from the deeptools package [73]. The Bigwig files, which represents coverage of mapped reads, were uploaded to JBrowse and are available to view at http://bioinf.gen.tcd.ie/jbrowse/?data=RNA-seq_Tup1Cyc8_ merged. The RNA-seq datasets are available in the Gene Expression Omnibus (GEO) repository (accession code GSE230732). S1 Appendix provides the RNA-seq data used to construct Figs 3, 4, 5, 7, S4, S5, S6 and S9.

Chromatin immunoprecipitation (ChIP)
Locus-specific ChIP was performed as previously described [27,54]. The antibodies and conditions used are shown in S5 Table. For ChIP, occupancy signals were determined by comparing the enrichment of DNA found in the immunoprecipitated (IP) material versus the input (IN) material. This (IP/IN) signal was then normalised to an IP/IN signal at an internal negative control region (Tel-VI for RNAP II, Int-V for Tup1p, Cyc8-Myc and Flo8-Myc) to give the 'relative occupancy'. Tup1p and Cyc8-Myc relative occupancy were further normalised to similarly processed ChIP results from a tup1 deletion and an untagged (No tag) strain, respectively. Details to show the ChIP resolution and the normalisation pathway are shown in S10 and S11 Figs. Primers used are shown in S4 Table. The global Tup1p and Cyc8p occupancy data used in Fig 7, and the Rsc1p and Gcn5p data used in S4B Fig was retrieved from a ChIP-Exo analysis performed by Rossi et al., 2021 [57], and is available in S2 Appendix. The processed ChIP-Exo peaks for Tup1p and Cyc8p occupancy were uploaded to JBrowse (peak mid-points are shown) and can be viewed at http:// bioinf.gen.tcd.ie/jbrowse/?data=RNA-seq_Tup1Cyc8_merged. Correlating FLO1-and SUC2-type gene upstream intergenic region length with gene derepression. Graphs depicting the length of the upstream intergenic length (X axis) and the change in transcription compared to wt (Y axis) for the (C) FLO1-type and (D) SUC2-type genes in tup1, cyc8 and tup1 cyc8 mutant strains. A line of best fit is shown for each strain. For FLO1-type genes (C), a two tailed Spearman correlation showed a statistically significant correlation between the upstream intergenic length and upregulation of transcription in the tup1 cyc8 strain compared to wt (P = 0.0294). For SUC2-type genes (D), a two tailed Spearman correlation showed a significant correlation in the cyc8 strain (p = 0.0449). (TIF) S7 Fig. Restoration of a functional FLO8 gene by PCR-mediated mutagenesis. (A) In the first round of mutagenesis, a PCR product containing a selectable marker integrates into the position immediately downstream of the FLO8 3' ORF by homologous recombination. The resulting strain contains a genomic copy of FLO8 immediately followed by the marker. (B) A forward primer was designed with homology to the FLO8 5' ORF that contained an A-G basepair substitution corresponding to position +425 of the FLO8 ORF. This was used in conjunction with a reverse primer with homology to an intergenic region downstream of FLO8. Using genomic DNA from the strain with the selectable marker directly downstream of the FLO8 ORF; these primers were used to generate a PCR product that contained the majority of the FLO8 ORF, but with the A-G point mutation at position +425. This product also contained the selectable marker adjacent to the FLO8 3' ORF. (C) A second transformation was carried out in a wild type BY4741 strain using the PCR product containing the point mutation at position +425 in the FLO8 ORF. This resulted in a strain with a genomic copy of FLO8 containing a G at position +425 in place of an A. This strain also contained a selectable marker immediately downstream of the FLO8 ORF. The resulting strain (YMC19, FLO8+) was sequenced to confirm the point mutation. The restored FLO8 gene in YMC19 was subsequently tagged with a 9-Myc epitope to generate strain YMC34. (TIF) S8 Fig. Confirmation of FLO8 expression, FLO8 function, and Flo8-Myc occupancy in the FLO8+ strain. (A) Transcription from the non-functional FLO8 gene in wt and cyc8, and from the restored FLO8 gene in YMC19 (FLO8+), was analysed by RT-qPCR (error bars represent SD, n = 2). (B) Expression of Flo8-Myc in YMC34 was confirmed by western blot analysis. Wt and a Cyc8-myc strain were used as controls. Actin was used as a loading control. The proteins detected for Cyc8-Myc and Flo8-Myc were of the expected size. (C) FLO1 and (D) SUC2 mRNA levels detected by RT-qPCR in wt, cyc8 and the FLO8+ strain. The result shows that FLO1 is transcribed in the FLO8+ strain, whilst SUC2 is not transcribed (error bars represent SD, n = 2). The data in C is the same data shown in Fig 6G. (E) Flo8-Myc occupancy (IP/in) at the FLO1 promoter, SUC2 promoter, and at a negative control region, Int-V. (F) Flo8-Myc occupancy data from (E) shown as 'relative occupancy' following normalisation to Int-V (error bars represent SD, n = 2). The Flo8-Myc relative occupancy data shown here for FLO1 is the same data shown in Fig 6H. The results in C-F show that the impact of Flo8-Myc expression are specific for FLO1; Flo8-Myc does not occupy the SUC2 promoter, and SUC2 remains repressed in the FLO8+ strain. (E) Cyc8-Myc occupancy at the FLO1 promoter normalized to occupancy at Tel-VI to yield 'relative occupancy'. (F) Cyc8-Myc relative occupancy following normalization to relative occupancy in the No Tag strain. In all graphs mean and standard deviation are shown, * = p�0.05, ** = p�0.005 obtained from One-Way ANOVA analysis (n = 3-6). (TIF) S12 Fig. Loss of CYC8 and HDA1 specifically abolishes repression of FLO1 transcription. (A) PCR results from genomic DNA of the strains indicated to confirm the hda1::KAN mutant strain. Lane 1: 10 bp marker (NEB), lane 2, 3: PCR using primers for ACT1 (positive PCR control), lanes 4, 5: PCR using primers upstream of HDA1 (hda1DFconF) and internal to hda1:: KAN (KanB), lanes 6, 7: PCR using primers downstream of HDA1 (hda1DFconR) and internal to hda1::KAN (KanC). (B) Western Blot analysis to confirm the deletion of CYC8 in a hda1 mutant strain (in duplicate). β-actin was used as a loading control. Bands detected were of the expected size. RT-qPCR analysis of transcription of (C) FLO1, (D) SUC2 and (E) PMA1 in the strains indicated. Values were normalised to ACT1 mRNA and error bars reflect standard deviation (n = 3-4). (RT-qPCR data of FLO1 and SUC2 in the wt, tup1, cyc8 and tup1 cyc8 strains has previously been shown in Fig 2). (F) Schematic to depict Tup1p-dependent role of Hda1p in FLO1 repression. In a cyc8 mutant strain we propose that Hda1p, in association with Tup1p, represses FLO1 in the absence of CYC8. Loss of both CYC8 and HDA1 results in high FLO1 de-repression (compare FLO1 mRNA levels in hda1 cyc8 and tup1 cyc8). This result is specific to FLO1. Transcription at SUC2 and PMA1 are not significantly affected. (TIF) S1