The yeast Cyc8–Tup1 complex cooperates with Hda1p and Rpd3p histone deacetylases to robustly repress transcription of the subtelomeric FLO1 gene

We demonstrate that the yeast flocculation gene, FLO1, is representative of a distinct subset of subtelomeric genes that are robustly repressed by the Cyc8–Tup1 complex. We have examined Cyc8–Tup1 localisation, histone acetylation and long-range chromatin remodelling within the extensive FLO1 upstream region. We show that Cyc8–Tup1 is localised in a DNase I hypersensitive site within an ordered array of strongly positioned nucleosomes around − 700 base pairs upstream of the transcription start site. In cyc8 deletion mutant strains, Tup1p localisation is absent, with concomitant histone hyperacetylation of adjacent regions at the FLO1 promoter. This is accompanied by extensive histone depletion across the upstream region and gene activation. The yeast histone deacetylases, Hda1p and Rpd3p, occupy the repressed FLO1 promoter region in a Cyc8–Tup1 dependent manner and coordinate histone deacetylation, nucleosome stabilisation and gene repression. Moreover, we show that the ATP-dependent chromatin remodelling complex Swi–Snf occupies the site vacated by Cyc8–Tup1 in a cyc8 mutant. These data suggest that distinctly bound Cyc8–Tup1 cooperates with Hda1p and Rpd3p to establish or maintain an extensive array of strongly positioned, deacetylated nucleosomes over the FLO1 promoter and upstream region which inhibit histone acetylation, block Swi–Snf binding and prevent transcription.


Introduction
Co-repressors and co-activators play a key role in eukaryotic gene regulation. Through association with DNA-bound transcription factors in the nucleus, these often large complexes determine the repressed or active state of gene promoters by altering the chromatin structure or the recruitment and competence of the RNA polymerase II holoenzyme. One of the earliest co-repressors of gene regulation identified in the yeast, Saccharomyces cerevisiae, was the Cyc8-Tup1 complex; a 1.4 megadalton complex composed of the TUP1 and CYC8 (also known as SSN6) gene products in a 4:1 ratio [1,2]. The complex is formed by the interaction of the N-terminal residues of Tup1p with tetratricopeptide repeat (TPR) motifs of Cyc8p [3][4][5]. Tup1p is related to the Groucho family of co-repressor proteins in higher eukaryotes by way of its C-terminal WD-repeat domain [6].
However, the mechanism of Cyc8-Tup1 repression is still unclear. Cyc8-Tup1 can interact directly with several components of the transcriptional machinery such as the kinase-cyclin pair Srb10/11p, the essential holoenzyme component Srb7p, and the Mediator subunit Med3p [6]. This is consistent with genetic interactions found between Cyc8-Tup1 and RNA polymerase II holoenzyme components [16][17][18][19]. Cyc8-Tup1 has also been proposed to directly block transcription activators [20][21][22][23]. Indeed, recent work suggests the primary role for Cyc8-Tup1 involves obstructing the activation domains of the DNA binding proteins responsible for its targeted recruitment [23].
The Cyc8-Tup1 complex also interacts with chromatin and can influence nucleosome positioning, histone acetylation and deposition of the histone variant, Htz1p [7,[24][25][26]. Tup1p binds the deacetylated, but not acetylated, histone tails of H3 and H4 in vitro, and the domain that is required for this interaction overlaps with the transcription repression Biochimica et Biophysica Acta 1839 (2014) 1242-1255 domain [27]. Furthermore, repression of transcription at Cyc8-Tup1 regulated genes is correlated with a reduced acetylation of histone H3 and H4 at promoters in vivo [28][29][30]. Previous studies have also demonstrated genetic and biochemical interactions between Tup1p and class I histone deacetylases (HDACs) Rpd3p, Hos1p, Hos2p and the class II HDAC, Hda1p [29,31,32]. Deletion of many of these HDACs caused hyperacetylation of histones at promoter regions [29,31,33]. Tup1passociated HDACs were found to be responsible for removal of the acetyl groups from the N-terminal tails of the core histones H2B, H3 and H4, indicating a reciprocal relationship between Tup1p binding and histone deacetylation. These findings suggested that Cyc8-Tup1 repression activity is modulated by changes in histone acetylation.
The yeast FLO1 gene is regulated by the TUP1 and CYC8 (SSN6) gene products and is the dominant member of the FLO family of genes [34,35]. FLO1 gene expression causes flocculation which is the Ca 2+dependent, nonsexual aggregation of yeast cells [36][37][38]. Flocculation has been shown to play a role in cellular resistance to external stresses such as heat, cold and various chemical reagents [39]. It is an important phenotype of brewing strains, but has become attenuated in some common laboratory strains [40]. We have previously shown that the Swi-Snf co-activator and the Cyc8-Tup1 complex influence nucleosomal arrays up to 5 kb upstream of the FLO1 transcription start site. Thus, the chromatin in which the yeast FLO1 promoter resides has a dynamic structure controlled by remodelling events which form the background in which promoter regulation takes place [41]. Hence, the FLO1 gene represents a paradigm for chromatin-mediated regulation of gene transcription, and offers an amenable model system in which to investigate the mechanism of action of Cyc8-Tup1.
In this study, we propose that the FLO1 gene represents a specific subset of genes subject to robust repression by Cyc8-Tup1 working cooperatively with the histone deacetylases, Hda1p and Rpd3p. We demonstrate that Cyc8-Tup1 localisation is focused at a hypersensitive site in the FLO1 upstream nucleosome array. Removal of Cyc8-Tup1 leading to activation of the FLO1 gene is coincident with long-range modulations of the histone H3 and H4 acetylation pattern and opening of the chromatin structure. Additionally, in the absence of Cyc8-Tup1, the occupancy of the HDACs, Rpd3p and Hda1p decreases in parallel with an increase in the occupancy of Swi-Snf at the site formerly occupied by Cyc8-Tup1. We show that Rpd3p and Hda1p are co-ordinately required for full repression but do not do so by enhancing Cyc8-Tup1 binding.
These data suggest that the long-range remodelling of the FLO1 promoter and upstream chromatin requires distinctly bound Cyc8-Tup1and Swi-Snf and is mediated by histone acetylation.

Yeast strains
S. cerevisiae strains were in the S288C background (Table S1) [42]. Yeast gene deletions and tagging were performed using PCR-based methods [43,44]. All gene deletions were confirmed by PCR or Southern blot analysis of genomic DNA and assayed for appropriate phenotypes. PCR and Western blot analysis were used to confirm that the genomic copies of CYC8, RPD3 and HDA1 were correctly tagged with a C-terminal nine Myc epitope. Epitope tagged strains were assayed to confirm appropriate wild-type phenotypes. Cells were grown at 30°C in YPD medium.

Northern blot and RT-qPCR analysis
Northern analysis of mRNA expression was performed as previously described [50]. Total RNA was prepared, resolved on 1% denaturing agarose gels, transferred to Zeta-Probe GT membranes (Bio-Rad) and hybridised using DNA probes for ACT1 and the complete FLO1 coding region [41]. Band intensities were determined by phospho-imager analysis (FujiFilm FLA2000 FlouroImager). Where appropriate, FLO1 values were normalised to ACT1. For RT-qPCR analysis of FLO1 mRNA, RNA was extracted from cells by the hot phenol method [45], treated with DNase I (Promega) and used to generate cDNA with random primers and reverse transcriptase (Applied Biosystems). Negative controls with no reverse transcriptase were included. PCR reactions were performed in triplicate using a SYBR Green Master Mix (ABI) and ABI Step One Plus PCR machine. Values were normalised to ACT1 RNA. Primers used are listed in Table S2.

Flocculation assay
Exponentially growing cells were resuspended to an equal cell density in YPD or YPD containing 100 mM EDTA (control). Equal volumes of cells were aliquoted into a tissue culture plate and agitated by shaking [51]. Five minutes after cessation of agitation, the plates were photographed. Cells displaying a flocculation phenotype aggregate in the absence of EDTA and are dispersed in the presence of EDTA (data not shown).

Analysis of Tup1p
ChIP-seq, global transcription, Hda1 ChIP and nucleosome occupancy data Expression array data generated by Chen et al. [24] were retrieved from ArrayExpress (E-GEOD-37466), RMA normalised, annotated for S. cerevisiae ORF and chromosomal locations (SGD), to calculate fold changes between wild-type and TUP1 deletion strains. Tup1 protein binding sites were derived from ChIP-seq data generated by Wong and Struhl (NCBI accession no. SRA044839.1) using the algorithms and parameters described by the authors [23]. Peak maxima with p-value = 10 −4 (in one case compared with 10 −2 ) were assigned to the closest translational start codon (ATG) of annotated ORFs. Combination of ChIP and expression data, as well as plotting of graphs, was carried out in Excel using VBA custom scripts. For the Hda1p ChIP data analysis, lists of genes scoring above the False Discovery Rate (FDR) for each of three probe locations, as well as those shared between probes, were collated from Venters et al., 2011 [52]. These were subjected to further Venn analysis with gene subsets under study and plotted as a pie chart. Nucleosomal occupancy trace data sets were kindly provided from van Bakel et al., 2013 [53]. Using local R scripts, 1000 bp regions were retrieved from the genomic nucleosomal difference traces between a tup1 mutant and wt for each subset of 90 locations, which were each plotted as an average trend over 1000 bp with standard deviation (SD). The SD for random regions (not shown for clarity) was very similar to the SD for non de-repressed genes.

Results
We previously reported that the subtelomeric FLO1 gene promoter resides within a 5 kb chromatin region under the influence of longrange remodelling by the Cyc8-Tup1 and Swi-Snf complexes [41]. To gain more insight into this gene regulatory mechanism, we queried whether the FLO1 gene was typical of a Cyc8-Tup1 repressed gene [23,24].

Tup1p repression activity is enriched at chromosome subtelomeric regions
We first compared global Tup1p occupancy data with the transcription profile of a tup1 deletion strain to examine the correlation between Tup1p promoter occupancy and gene repression attributable to the TUP1 gene [23,24]. Our analysis revealed that of the 319 genes where a Tup1p peak could be located within 1000 bp upstream of the transcription start site, the deletion of TUP1 resulted in less than two-fold increase in transcription of 251 (79%) of these genes (Fig. 1A, lanes 3 and 4). However, a subset of 68 (21%) of promoter-bound Tup1p genes is de-repressed more than two-fold in the absence of TUP1, while the low wild-type transcription levels of these genes indicate that they are indeed controlled by Tup1p repression (Fig. 1A, lanes 1 and 2). Furthermore, only 9 (3%) of the promoter-bound genes have less transcription in the tup1 deletion strain, confirming the role of Tup1p as a repressor.
Conversely, analysis of the genes showing the greatest de-repression in the absence of TUP1 (greater than five-fold de-repression) retrieved a set of 92 genes that were distinguished by very low levels of transcription in the wild-type strain (Fig. 1A, lanes 5 and 6). Interestingly, the majority of these genes most subject to Tup1p repression (74 genes) did not overlap with the promoter-bound Tup1p repressed subset of genes (Fig. 1A, Venn diagram, inset). Thus, those genes form a distinct subset under robust control of TUP1, which we suggest represents those genes directly switched on or off via the Cyc8-Tup1 complex.
We next compared the genome-wide Tup1p occupancy peak map with localised gene transcription changes in the absence of TUP1 (Fig. 1B). The analysis revealed Tup1p occupancy (Fig. 1B, ticks) has a dissimilar distribution without strong bias towards the genes highly de-repressed when Tup1p is absent (Fig. 1B, pink circle areas). Interestingly, whereas Tup1p occupancy seems evenly distributed across the genome, many of the genes most strongly de-repressed in the absence of Tup1p are clustered within subtelomeric regions (Fig. 1B).
The FLO1 gene resides within the subtelomeric region on the right arm of chromosome I and is representative of a set of subtelomeric genes that are highly de-repressed on a background of more moderately de-repressed genes when TUP1 is deleted (Fig. 1B). The high levels of de-repression in the absence of Tup1p suggest that this subset of genes is predominantly repressed by Cyc8-Tup1 without possible compensatory repression mechanisms; in the absence of the co-repressor these genes are constitutively switched on. We believe that this set of highly de-repressed genes presents good models for clarifying current problems regarding the mechanism of Cyc8-Tup1 repression with fewer confounding factors. We have focused on the FLO1 gene to address the relationship between Cyc8-Tup1, chromatin, gene repression and gene activation.

3.2.
Cyc8-Tup1 is localised in the −500 to −1000 bp FLO1 upstream region and is required for repression of FLO1 gene transcription To investigate the molecular mechanism of FLO1 gene repression by the Cyc8-Tup1 complex, we mapped Tup1p and Cyc8p localisation over a 3.5 kb chromatin region upstream of the FLO1 coding sequence ( Fig. 2A). The results confirmed that Tup1p and Cyc8p were concentrated in the upstream region between −500 and −1000 bp in wild-type (wt) strains ( Fig. 2B and C, wt) [23,54,55]. To test the contribution of the individual Cyc8 and Tup1 protein subunits to localisation of the complex, we constructed CYC8 and TUP1 deletion mutants (cyc8 and tup1) and measured the respective Tup1p and Cyc8p abundance and occupancies.
Western blot analysis of Tup1p and Cyc8p levels in the respective cyc8 and tup1 deletion mutants confirmed the stability of each subunit was unaffected by the absence of the other ( Fig. 2D and Supplemental Fig. S1). ChIP analysis revealed that the peak of Tup1p around − 700 bp upstream of FLO1 in wt strains was completely absent in cyc8 strains (Fig. 2B, cyc8). This suggests Tup1p is recruited in the context of the Cyc8-Tup1 complex confirming observations at the RNR2 and STE6 genes [30]. Surprisingly, the peak of Cyc8p was significantly increased in tup1 strains. Thus, Cyc8p remains bound to the FLO1 promoter in the absence of Tup1p where it either increases in occupancy or is more amenable to ChIP analysis.
We next confirmed the role of Cyc8-Tup1 in FLO1 repression by measuring FLO1 transcription in cyc8 and tup1 single and double deletion mutant strains. The data revealed that de-repression of FLO1 transcription was greatest in cyc8 strains compared to tup1 strains and wt (Fig. 2E). Analysis in a tup1 cyc8 double mutant showed that FLO1 de-repression was similar to that seen in the cyc8 single mutant. Taken together, the results suggest that Cyc8p occupancy in the absence of Tup1p either directly or indirectly imparts partial FLO1 gene repression. We conclude that the cyc8 mutant, and not the tup1 mutant, is representative of a strain defective for Cyc8-Tup1 complex binding and activity at FLO1. Furthermore, Tup1p is representative of Cyc8-Tup1 complex occupancy at FLO1.

Cyc8-Tup1 localises within a DNase I hypersensitive site and is required for chromatin organisation at the repressed FLO1 promoter and upstream region
Cyc8-Tup1 has been proposed to promote gene repression by stabilising promoter nucleosomes to form repressive chromatin structures [41,50,[55][56][57][58]. We therefore analysed the DNase I sensitivity in yeast nuclei over a distance of 3 kb of upstream chromatin and 1.5 kb of FLO1 coding region (not shown), using the indirect end-labelling technique. Fig. 3A shows the digestion patterns from the vantage point of two restriction sites in the upstream region. We discovered a striking DNase I hypersensitive site around − 700 bp upstream of  [41]. We conclude that Cyc8-Tup1 is recruited to a distinct opening within the nucleosomal array. In the immediate vicinity of this opening, DNase I cuts a nucleosomal ladder pattern, indicative of a well-spaced ordered array of nucleosomes ( Fig. 3A and B). In contrast, the chromatin further upstream appears more inaccessible to DNase I, which could be explained by sterical hindrance from close spacing of nucleosomes [59].
Interestingly, when Cyc8-Tup1 repression of FLO1 is relieved in cyc8 strains, the DNase I digestion pattern shows a dramatic expansion of the hypersensitive site in the chromatin, which now covers a space several nucleosomes wide ( Fig. 3A and B, cyc8). Again, DNase I cuts a nucleosomal ladder pattern immediately upstream of this widened hypersensitive region, demonstrating further opening of the adjacent chromatin structure. This result is consistent with our previous micrococcal nuclease analysis, which showed a gross disruption of the nucleosomal array in this region flanked by more widely spaced nucleosomes [41]. These data suggest a role for Cyc8-Tup1 in the organisation of repressive regions of chromatin or in antagonising chromatin disruption [23,41,50,55].
3.4. Histone H3 occupancy at the FLO1 promoter and upstream region is reduced in the absence of Cyc8-Tup1 The indirect end-labelling analysis experiment indicates the gross disruption of the nucleosomal array at the FLO1 promoter region when Cyc8-Tup1 repression is relieved. However, the severe disruption in the micrococcal nuclease pattern corresponding to the expansive DNase I hypersensitive site was not accompanied by a reversal to the naked DNA digestion pattern [41]. Furthermore, restriction enzyme accessibility assay results were not consistent with digestion of naked DNA (Supplemental Fig. S2), suggesting the remodelled region was not completely devoid of nucleosomes. To directly determine the fate of nucleosomes at the de-repressed FLO1 promoter and upstream region, we examined histone H3 occupancy in cyc8 mutant strains by ChIP (Fig. 4A).
In wild-type cells, where FLO1 transcription is repressed, histone H3 occupancy at the FLO1 promoter and upstream region was higher than at the control intergenic region to which the H3 data were normalised (Fig. 4A, wt). This suggests that there is a relatively high histone density at FLO1 compared to the control non-transcribed region considered representative of the global chromatin background. Strikingly, in the absence of Cyc8p, when FLO1 transcription is de-repressed, there is a general decrease in histone H3 occupancy across the entire region tested (Fig. 4A, cyc8). However, the most dramatic histone density decrease was in the region adjacent to the site previously bound by Cyc8-Tup1, around − 0.5 kb to − 1 kb upstream of the FLO1 transcription start site. Western blot analysis shows that histone levels in the cyc8 mutant are the same as wt suggesting that the dramatic histone depletion is specific to the FLO1 promoter region and not due to a general loss of histones in this mutant (Fig. 4B).
The results over the proximal promoter region are consistent with our previous finding that the nucleosome array in the FLO1 promoter region is severely disrupted in cyc8 strains with a likely loss of nucleosome density, rather than complete nucleosome removal, due to histone eviction [41]. Overall, the data suggest that Cyc8-Tup1 acts to establish or maintain a highly organised chromatin structure over the repressed wild-type FLO1 gene promoter and upstream region.
3.5. Cyc8-Tup1 is required for histone H3 and H4 deacetylation at the FLO1 promoter region The association of Cyc8-Tup1 with chromatin in vivo has been linked to low levels of histone acetylation [28][29][30]. We anticipated that in wildtype strains, the histone tails co-localising in the chromatin with Cyc8-Tup1 in the FLO1 promoter region would be hypoacetylated. We therefore performed ChIP to analyse histone acetylation in this region by using antibodies against acetylated histone H3 (lysine 9), acetylated histone H4 (lysine 12, not shown), or all acetylated isoforms of histone H4 (lysines 5, 8, 12 and 16). At each position tested, histone acetylation levels were normalised to the respective histone H3 occupancy at that site ( Fig. 4C and D).
In the wild-type strain, where FLO1 transcription is repressed, levels of acetylated histone H4 (H4ac) and acetylated lysine 9 of histone H3 (H3K9ac) were low over the entire 3.5 kb FLO1 promoter and upstream region ( Fig. 4C and D, wt). Conversely, in cyc8 strains where FLO1 transcription is de-repressed, increased H4ac between −2 kb and the FLO1 transcription start site was evident, with a sharp peak of increased acetylation present in the −0.6 kb to −0.4 kb area immediately adjacent to where Tup1p localises in wild-type strains (Fig. 4C, cyc8). Deletion of CYC8 resulted in lysine 9 of histone H3 being broadly hyperacetylated over the entire 3.5 kb FLO1 upstream region, but with the greatest increase in acetylation levels again evident across the −1.5 kb upstream FLO1 region (Fig. 4D, cyc8). These results suggest a role for Cyc8-Tup1 in either promoting histone deacetylation or blocking histone acetylation across the region extending up to −3.5 kb upstream of the FLO1 transcription start site.
3.6. Disruption of the HDA1 and RPD3 histone deacetylase genes results in partial de-repression of FLO1 gene transcription Previous reports have connected Cyc8-Tup1 to class I histone deacetylases (HDACs) Rpd3p and Hos2p [31], and Tup1p has been linked to the class II HDAC Hda1p [33]. Flocculation was also reported to be induced during high-gravity fermentation in an hda1 mutant [60]. In order to identify which HDACs, if any, were involved in FLO1 gene repression in our system, mutations in various HDACs were prepared and screened for FLO1 transcription. If the deletion of an HDAC gave rise to FLO1 transcription it could be inferred to play a role in FLO1 gene repression. We therefore made deletions of class I, II and III HDAC genes either individually, or in combination, and assayed for flocculation and FLO1 transcription.
As shown in Fig. 5A, single gene deletions of various HDAC-encoding genes did not de-repress FLO1 transcription. However, a double deletion of both the HDA1 and RPD3 HDAC-encoding genes caused a partial de-repression of FLO1 transcription. Indeed, FLO1 de-repression in the absence of both Hda1p and Rpd3p was measured at 21% of that detected in the absence of CYC8. The resultant flocculation phenotype exhibited by the rpd3 hda1 double deletion mutant cells is shown in Fig. 5B. These data suggest the class I and II HDACs, Hda1p and Rpd3p respectively, both contribute to repression of FLO1 gene transcription.

Disruption of HDA1 and RPD3 results in reduced histone H3 occupancy at the FLO1 promoter
The de-repression of FLO1 transcription in the absence of Cyc8-Tup1 coincides with a dramatic reduction of histone occupancy across the FLO1 promoter and upstream region, potentially as a result of histone eviction (Fig. 4A, cyc8). We therefore analysed histone occupancy in the rpd3 hda1 double deletion mutant compared to rpd3, hda1, cyc8 single mutants and wild-type, to determine if a histone occupancy reduction was associated with the partial levels of FLO1 de-repression detected in the rpd3 hda1 double mutant (Fig. 6A).
Using ChIP we could detect no difference in histone H3 levels in the rpd3 single mutant compared to wild-type, which correlated with the absence of FLO1 transcription in this strain (Fig. 6A, rpd3). Surprisingly, histones were found to be depleted from the −0.5 to -1.0 kb FLO1 promoter region in the hda1 single mutant, despite there being no detectable FLO1 transcriptional de-repression in this strain (Fig. 6A, hda1). In the absence of both the RPD3 and HDA1 genes however, where FLO1 is partially de-repressed, there was a similar decrease in histone occupancy at the FLO1 proximal promoter region as seen in the hda1 single mutant, but the decrease in histone occupancy extended further to cover the − 2.5 kb FLO1 upstream region (Fig. 6A, compare rpd3 hda1 and hda1). Importantly, the loss of histones in the rpd3 hda1 double mutant was not as dramatic as the histone depletion evident in the absence of Cyc8p where maximal FLO1 de-repression occurs (compare Fig. 6A, rpd3 hda1 and Fig. 4A, cyc8). Overall, the data suggest that both Hda1p and Rpd3p are required for the establishment or maintenance of histone occupancy at the repressed FLO1 gene.

Disruption of HDA1 and RPD3 results in histone H3 and H4 hyperacetylation at the FLO1 promoter
If the role of Rpd3p and Hda1p in FLO1 repression is to deacetylate histones at the FLO1 promoter and upstream region, it would be predicted that the histones in this region would be hyperacetylated in the absence of both Hda1p and Rpd3p. We therefore analysed histone H3 lysine 9 and histone H4 acetylation levels in hda1, rpd3 and rpd3 hda1 mutants compared to wild-type by ChIP. Histone acetylation levels were normalised to histone levels in each strain to account for differences in histone occupancy ( Fig. 6B and C). In the rpd3 mutant we could not discern any changes in histone H4 or H3 acetylation levels compared to wild-type over the entire FLO1 upstream region tested ( Fig. 6B and C, rpd3). In contrast, in hda1 cells, we observed a distinct increase in H4 acetylation levels focussed within the −1 kb FLO1 promoter proximal region (Fig. 6B, hda1). Additionally, the pattern of H3K9ac in hda1 cells showed an increase that was consistently higher than the wild-type profile across the 2 kb FLO1 upstream region (Fig. 6C, hda1). It is important to note that neither single RPD3 nor HDA1 deletion de-represses the FLO1 gene, as shown by lack of a flocculation phenotype and the absence of signal in Northern blots as well as microarray data [61,62].
However, when both the HDA1 and RPD3 genes were deleted, a strain in which FLO1 transcription is partially de-repressed, the pattern of H4 acetylation was generally higher and broader than in the hda1 single mutant, encompassing almost 2 kb of the FLO1 upstream region (Fig. 6B, rpd3 hda1). Histone H3K9ac levels in the rpd3 hda1 double deletion mutant were also reproducibly higher than the acetylation levels observed in the hda1 single mutant, with the greatest increase in the region between −1.0 kb and −0.3 kb upstream of the FLO1 transcription start site (Fig. 6C, rpd3 hda1). These data suggest that there is a biased redundancy between Hda1p and Rpd3p histone deacetylation activities at the wild-type FLO1 promoter and upstream region which contributes to histone deacetylation and gene repression: Hda1p can compensate for the absence of Rpd3p deacetylase activity, whereas Rpd3p may only partially compensate for loss of Hda1p.
Importantly, histone H4 and H3 acetylation levels at the FLO1 upstream region in the rpd3 hda1 double mutant were lower than those in the absence of either Tup1p or Cyc8p, but showed a similar profile (compare rpd3 hda1 in Fig. 6B and C with Fig. 4C and D, cyc8). This result correlates with the lower level of FLO1 de-repression in the rpd3 hda1  Fig. 2A.
mutant compared to the cyc8 mutant. These data do not exclude that other HDACs may also contribute to FLO1 promoter deacetylation and gene repression. However, the absence of previously reported Cyc8-Tup1 associated HDACs Hos1p and Hos2p [31] had no effect on FLO1 transcription, either alone or in combination with an rpd3 deletion ( Fig. 5A and data not shown).

Hda1p and Rpd3p bind the repressed FLO1 promoter and upstream region
Our results suggest that Hda1p and Rpd3p histone deacetylase activities contribute to histone deacetylation at the FLO1 promoter and upstream region and are required for FLO1 gene repression. To test if this effect was direct, we used ChIP to determine if we could detect Hda1p and Rpd3p occupancy at the repressed FLO1 promoter and upstream region. In the wild-type strain, the occupancy of both epitope-tagged Hda1p and Rpd3p HDACs was confirmed across the repressed FLO1 promoter proximal region (Fig. 7A, wt and Supplementary Fig. S3). Interestingly, the data showed a reproducible enrichment of both HDACs centred at the site of Cyc8-Tup1 occupancy at − 585 bp proximal to the FLO1 start site. This finding confirms that both Hda1p and Rpd3p are bound at the repressed FLO1 promoter.

Regulatory interplay between Cyc8-Tup1 and Hda1p and Rpd3p
One model for the activity of the Cyc8-Tup1 complex is that it recruits HDACs to deacetylate histones at target promoters. The resultant histone deacetylation has been proposed to promote further recruitment of Cyc8-Tup1 thereby reinforcing its occupancy to strengthen gene repression [7,31]. Considering our findings of a potential overlap in the sites of occupancy between Cyc8-Tup1 and Rpd3p and Hda1p at the FLO1 promoter, we investigated if there was evidence to support the proposed regulatory interplay between the HDACs and Cyc8-Tup1 at FLO1. The model would predict that HDAC occupancy would be reduced in the absence of Cyc8-Tup1, and that Cyc8-Tup1 occupancy would be reduced in the absence of the HDACs. We first examined Hda1p and Rpd3p occupancy at FLO1 in the absence of CYC8. Fig. 7A shows that the occupancy of both Hda1p and Rpd3p was reduced in a cyc8 mutant. This result supports a role for Cyc8-Tup1 in promoting HDAC occupancy at the repressed FLO1 promoter.
Next, we determined if the HDACs Hda1p and Rpd3p modulated the binding of Tup1p at the FLO1 promoter. Using ChIP, we measured the occupancy of Tup1p in the rpd3 hda1 double mutant where we had previously shown histone hyperacetylation across the FLO1 promoter region and partial FLO1 de-repression. However, no difference in Tup1p occupancy at the FLO1 promoter region was detected (Fig. 7B, compare  rpd3 hda1 and wt). Thus, the data indicate that Cyc8-Tup1 regulates the occupancy of Hda1p and Rpd3p at the repressed FLO1 promoter, however Hda1p and Rpd3p and their associated HDAC activities, do not regulate the level of Cyc8-Tup1 occupancy. The data also demonstrate that FLO1 transcription can occur in the presence of Cyc8-Tup1.

Snf2p binds the de-repressed FLO1 promoter
We have previously shown that in the absence of Cyc8-Tup1, FLO1 transcription and the concomitant promoter and upstream region nucleosome rearrangements are dependent upon the Swi-Snf complex [41]. We therefore determined if Swi-Snf was enriched at the FLO1 promoter in the absence of Cyc8p when transcription is de-repressed. Using ChIP analysis we measured the occupancy of the Snf2p subunit of Swi-Snf at the FLO1 promoter region in the repressed wild-type strain, and in the cyc8 and rpd3 hda1 mutants. The results showed that Snf2p occupancy was enriched at the FLO1 promoter in the cyc8 mutant strain compared to wild-type ( Fig. 7C and Supplemental Fig. S4). Furthermore, Snf2p occupancy was enriched relative to wild-type levels in the rpd3 hda1double mutant, but was present at lower levels than that detected in the cyc8 mutant. Interestingly, Snf2p was recruited to the same region previously occupied by Cyc8-Tup1 in the cyc8 mutant, whilst in the rpd3 hda1 mutant Snf2p potentially co-localised with fully bound Cyc8-Tup1. Thus, Snf2p occupancy at the FLO1 gene promoter correlates with the level of FLO1 transcription, and shows the greatest binding in the absence of Cyc8-Tup1 at its former binding site.
3.12. The FLO1 gene is representative of genes that are strongly de-repressed in the absence of TUP1 The pivotal role that HDA1 plays in repression at FLO1 (Figs. 5-7) raises the question of whether this mechanism is employed at other gene promoters. A yeast genome-wide study of the regulator cooccupancy network identified a 'lowly transcribed' cluster of genes that contained generally repressive chromatin remodellers and histone deacetylases, such as Isw1p, Isw2p, Cyc8p/Ssn6p, Hos1p, and Hda1p [52]. This study assessed binding between distal versus proximal promoter regions as well as the 3' ends of genes by ChIP. HDA1 has widespread functions in the yeast genome [52,63]. We therefore asked whether the set of genes that is strongly de-repressed in the absence of TUP1 has a specific Hda1p binding profile compared with the whole genome Hda1p binding profile. Analysis of Hda1p ChIP data over the three sites demonstrated an 81% enrichment (p = 0.005) of Hda1p on the more distal − 320 to − 260 upstream activator sequence (UAS) site for this set of genes, while 20% more Hda1p binding was detected overall (Fig. 8A). This significantly different upstream region binding profile of Hda1p compared with whole genome is consistent with its specific involvement at this set of genes. Furthermore, Rpd3p, Cyc8-Tup1, and Swi-Snf were amongst factors found to preferentially occupy this upstream promoter region [52].
The extensive nucleosome depletion observed at FLO1 in the absence of TUP1 (Figs. 3 and 4) additionally poses the question of whether this is also part of a mechanism that occurs at other promoters. A compendium study assessed the respective roles of a large collection of yeast genes in genome-wide nucleosome positioning and transcription in loss of function gene mutants. This study identified the tup1 and cyc8 single deletion mutants as having a particularly strong relationship between changes in nucleosome occupancy in the −200 to transcription start site (TSS) region and the expression of their target genes [53]. We therefore analysed data for the −1000 bp to TSS region, where nucleosome depletion is observed at the de-repressed FLO1 gene, for changes in nucleosome occupancy in the set of genes that is strongly de-repressed in the absence of TUP1. The set shows a high incidence of negative nucleosome occupancy changes in mutant chromatin compared with wild-type, resulting in a significantly lower average trace compared with the average for non de-repressed genes, or for random  N = 90, right). Hda1p presence over the −320 to −260 upstream region ('UAS') is 81% increased in this set (p = 0.005) compared with whole genome, while it is lower in the −90 to −30 proximal region ('TSS') and the 3' end 'ORF' region. ORFs scoring in two regions are plotted in the areas between major categories. Only ORFs scoring above the FDR are included in the pie chart, while detectable Hda1p was 20% increased overall (p = 0.05). (B) TUP1 de-repressed upstream regions have a high incidence of nucleosome depletion. Average of nucleosomal occupancy differences between tup1 and wt strains is plotted over the upstream 1000 bp region. Genes N5× de-repressed (N = 90) in a tup1 mutant show a lower average trace than non de-repressed genes (N = 90), or a set of random genomic 1000 bp regions (N = 90). Neither control shows significant changes in nucleosome occupancy between wt and mutant. De-repressed gene changes (tup1) show a wide variance ranging from wt occupancy to depletion occurring with high incidence across a wide upstream region (error bars, 1 SD). (C) Model for the role of Cyc8-Tup1 in FLO1 gene repression. In wt, Cyc8-Tup1 binds to a distinct site at the FLO1 promoter and is associated with an array of strongly positioned, deacetylated nucleosomes covering the promoter and upstream region. The histone deacetylases (HDACs), Rpd3p and Hda1p bind the promoter in a Cyc8-Tup1 dependent manner and contribute to nucleosome positioning, histone deacetylation and gene repression. In the absence of Cyc8-Tup1 (cyc8 mutant), there is a gross remodelling of FLO1 promoter and upstream chromatin involving nucleosome acetylation, rearrangement and eviction which accompanies FLO1 gene derepression. The occupancy of Rpd3p and Hda1p (to a lesser extent) is reduced. Swi-Snf is recruited to the site previously occupied by Cyc8-Tup1 and potentially directs nucleosome disruption. The increased chromatin acetylation in the cyc8 mutant suggests that histone acetyltransferase (HAT) recruitment and activity occur in this region. In the absence of the HDACs Rpd3p and Hda1p (rpd3 hda1 double mutant), Tup1p occupancy persists. Swi-Snf occupancy, nucleosome rearrangement and eviction also occur at the FLO1 promoter and upstream chromatin but at a reduced level compared to cyc8, and FLO1 is only partially de-repressed (compared to cyc8). Histone acetylation over the region is also lower than in the cyc8 mutant suggesting decreased HAT occupancy or activity in this strain. regions of the genome, respectively (Fig. 8B). These control sets showed no nucleosomal occupancy change overall, suggesting that nucleosome depletion is specific to this de-repressed set of genes. Nucleosome loss at individual genes occurs at various locations over this upstream region, resulting in a wide variance in occupancy around the calculated lower average at any particular position. These dramatic changes in occupancy are not commonly observed for single chromatin mutants, as they may only manifest their effects under conditions of stress, or may be subject to inherent redundancy of factors involved in chromatin homeostasis [53]. The study of FLO1 de-repression in tup1 and cyc8 single deletion mutants can therefore lead to unique insights into the de-repression mechanism while being a representative model for other Cyc8-Tup1 repressed genes.

Discussion
Despite considerable research, the precise mechanism of repression by the Cyc8-Tup1 complex has proved elusive. The complex can form a perplexing range of contacts with numerous binding partners mediated through the WD40 domain of Tup1p and the TPR domain of Cyc8p [64]. Cyc8-Tup1interacts with non-acetylated histone H3 and H4 tails, histone deacetylases and several different RNA polymerase II holoenzyme components, as well as with a large variety of DNA binding factors. This has led to the proposals of several, not mutually exclusive, mechanisms of action. One model proposes that Cyc8-Tup1 alters chromatin structure into a repressive form; another that it inhibits the RNA polymerase II holoenzyme by contacting it; and a third that it blocks transcription activators [7,8,23].
Our analysis of public Tup1p occupancy and activity data suggests that different subsets of Cyc8-Tup1 repressed genes may exist, subject to different extents of repression by the various overlapping mechanisms proposed for Cyc8-Tup1 activity. We propose that the FLO1 gene is representative of a distinct subset of subtelomeric genes strongly repressed by the Cyc8-Tup1 complex. Furthermore, our analysis may be an underestimate of this novel subset of genes since a recent study identified 36 new genes subject to Cyc8-Tup1 repression, 14 of which were located in subtelomeres [53].
We previously mapped a 30-nucleosome array and demonstrated a range of changes in the extended gene-free upstream region of the FLO1 gene in addition to extensive remodelling at the proximal promoter [41]. These changes, which were attributed to the absence or presence of the Swi-Snf or Cyc8-Tup1 complexes, go beyond current promotercentred models and raise questions as to how such long-range chromatin remodelling effects are achieved.
Two modes of association have been proposed for the Cyc8-Tup1 complex: a continuous polymerisation along the chromatin fibre [65], or localisation at distinct foci [55]. We confirmed the occupancy of Tup1p and Cyc8p across the FLO1 promoter and upstream region concentrated at a single location around − 700 bp upstream of the transcription start site within a region of 3.5 kb of upstream sequence analysed [24,55]. Recruitment of Tup1p to this site required the Cyc8p component of the complex. However, Cyc8p remained associated with the FLO1 promoter in the absence of Tup1p where it either directly or indirectly imparts partial repression. Interestingly, the peak of Cyc8p detected in the absence of Tup1p was greater than in wt. This could be due to either increased recruitment of Cyc8p in the absence of Tup1p or, most likely, increased epitope availability. Importantly, the data suggest that only the cyc8 strain is representative of a true Cyc8-Tup1 null mutant at FLO1.
Analysis of the FLO1 promoter DNA sequence on which Cyc8-Tup1 was enriched revealed the presence of potential binding sites of several proteins known to interact or recruit either Tup1p or Cyc8p including Mcm1p, Phd1p, Skn7p and Yap6p [66,67]. Individual deletion of these genes, or mutation of the Mcm1p DNA binding site, failed to derepress FLO1 transcription (data not shown). This suggests that recruitment of Cyc8-Tup1 to FLO1 either is redundant, involves a scaffold of multiple proteins or occurs via other proteins.
Our indirect end-labelling analysis revealed that Cyc8-Tup1 colocalises with a chromatin hypersensitive site situated within an array of six strongly positioned promoter proximal nucleosomes. ChIP analysis indicated a higher than background nucleosome density at the repressed FLO1 promoter and upstream region. However, in cyc8 mutants, the promoter nucleosome positioning is lost resulting in a dramatic expansion of the hypersensitive site and increased restriction enzyme accessibility to the underlying DNA. This is accompanied by a marked decrease in histone occupancy at the proximal promoter immediately adjacent to where Cyc8-Tup1 is normally bound. Furthermore, significant loss in histone density was seen to extend over the entire 3.5 kb upstream region analysed. A general depletion of histones was not the cause of the extensive histone loss as Western blot analysis showed that global histone levels were unaffected in the cyc8 mutant. However, it is possible that upregulation of unannotated non-coding transcripts mapped upstream of FLO1 in the cyc8 mutant could contribute to the upstream histone loss shown [68].
Overall, the changes are consistent with a randomisation or mobilisation of the array of positioned proximal promoter nucleosomes and altered positions and loss of more distal nucleosomes [41]. These data suggest that when bound at the FLO1 promoter, the Cyc8-Tup1 complex is required to establish or maintain a highly organised repressive chromatin structure encompassing the promoter and upstream region.
Cyc8-Tup1 repression has been linked with histone deacetylation enzymatic activity as well as deacetylated histone binding in vitro [7]. Our data show that Cyc8-Tup1 does not bind to any particular region of deacetylated histones at the repressed FLO1 promoter. However, in the absence of Cyc8-Tup1 we observed increased acetylation of histone H3 and H4 residues across chromatin covering the FLO1 promoter and upstream region. This suggests that extensive histone acetylation over the FLO1 promoter and upstream region is counteracted by Cyc8-Tup1.
From a panel of candidate histone deacetylases (HDACs), we identified Hda1p and Rpd3p as contributing redundantly to FLO1 promoter and upstream region deacetylation and gene repression. Single deletion of Hda1p also increased promoter acetylation but FLO1 was not derepressed in this strain. However, deletion of both Rpd3p and Hda1p led to greater promoter histone acetylation compared to levels in either single mutant, and importantly, yielded FLO1 gene de-repression. The histone hyperacetylation evident in the rpd3 hda1 double mutant was lower than the hyperacetylation in the absence of Cyc8-Tup1, which correlates with the lower level of de-repression in the rpd3 hda1 mutant. This suggests that histone deacetylation by Hda1p and Rpd3p at the repressed FLO1 promoter could be redundant with the activity of other HDACs if a threshold level of histone acetylation is required to enable maximum FLO1 transcription. Alternatively, the increased histone acetylation in the single hda1 and double hda1 rpd3 mutants may lead to poised chromatin, requiring additional promoter de-repression or gene activators to potentiate full transcription.
In the mutant deleted for both Rpd3p and Hda1p, where FLO1 transcription is partially de-repressed, we observed significant histone depletion across the FLO1 promoter and upstream region. However, histone loss was greatest in the cyc8 mutant in which FLO1 transcription is fully de-repressed. These data suggest that Cyc8-Tup1 and both the HDACs, Rpd3p and Hda1p, contribute to the ordered array of nucleosomes at the repressed FLO1 gene promoter and upstream region. Interestingly, significant histone loss was detected in the absence of FLO1 gene de-repression in the hda1 single mutant. This result highlights that repression of FLO1 transcription can occur even when the promoter is significantly depleted of histone H3, as well as acetylated. Moreover, these data indicate that nucleosome loss is not a consequence of transcription, but may precede it. Together, the data suggest that a threshold level of nucleosome loss and/or histone acetylation may be required across the FLO1 promoter and upstream region before maximal FLO1 transcription can occur.
One model for Cyc8-Tup1 activity purports that it recruits HDACs to target promoters which deacetylate adjacent histones to bring about gene repression [29,63]. The promoter histone deacetylation has been proposed to enhance further binding of Cyc8-Tup1 thereby reinforcing transcription repression of target genes [31]. In accordance with separate involvements at the ENA1 and STE6 promoters [29,30], we find that the histone deacetylases Hda1p and Rpd3p were centred on the Cyc8-Tup1 binding site at the FLO1 promoter. Importantly, Hda1p and Rpd3p occupancies at FLO1 were reduced in a cyc8 mutant, suggesting that their association with the promoter was Cyc8-Tup1 dependent. Surprisingly, Cyc8-Tup1 occupancy was not reduced in the rpd3 hda1 double mutant. Although seemingly at odds with the proposed histone deacetylase-mediated enhancement of Cyc8-Tup1 binding, the result may also reflect persistence of the complex at the promoter during gene activation in the rpd3 hda1 mutant, within the sensitivity limits of the Tup1p ChIP analysis where FLO1 transcription is only partially de-repressed.
Overall, our data support a role for Cyc8-Tup1 in organising an ordered array of deacetylated nucleosomes at the wild-type FLO1 promoter and upstream region which is required for gene repression. Our data also indicate a Cyc8-Tup1 dependent role for Hda1p and Rpd3p in establishing or maintaining this repressive chromatin structure.
In our previous study of chromatin remodelling at FLO1, the changes in cyc8 chromatin relative to wild-type were attributed to the actions of the Swi-Snf complex [41]. We therefore examined Swi-Snf occupancy at the FLO1 promoter and revealed increased levels of the Snf2p subunit of Swi-Snf at the de-repressed FLO1 promoter in a cyc8 mutant compared to the repressed wild-type. Additionally, we detected intermediate levels of the Snf2p co-activator at the FLO1 promoter in the rpd3 hda1 double mutant where FLO1 transcription is partially derepressed, suggesting a correlation between transcription and coactivator binding at FLO1. Furthermore, we demonstrate that Snf2p binds maximally to the same region vacated by Cyc8-Tup1 in the cyc8 mutant, and may show partial co-occupancy with Cyc8-Tup1 in the rpd3 hda1 mutant.
Based on our data and that of others, one model for regulation of FLO1 transcription might involve a dynamic equilibrium between the abundant HDACs, HATs, Swi-Snf and Cyc8-Tup1 complexes which can be biased one way or another to either promote gene repression or activation. In wt cells, Cyc8-Tup1 dependent enrichment of Rpd3p and Hda1p at FLO1 would promote strongly positioned deacetylated nucleosomes across the promoter and upstream region and repress gene transcription (Fig. 8C, wt). In the absence of Cyc8-Tup1, reduced Rpd3p and Hda1p occupancy would allow HAT activity to yield histone hyper-acetylation. The high histone acetylation would enable Swi-Snf to bind the FLO1 promoter and initiate extensive histone eviction leading to transcription (Fig. 8C, cyc8) [69][70][71][72].
In the rpd3 hda1 mutant, the equilibrium between HDAC and HAT binding would again be upset, allowing increased HAT occupancy at FLO1. The subsequent increased histone acetylation would enable Swi-Snf to drive histone eviction and de-repress FLO1 transcription. However, other HDACs could partially compensate for the loss of Rpd3p and Hda1p. In conjunction with Cyc8-Tup1 which remains bound at FLO1, full HAT binding at FLO1 would therefore be restricted resulting in partial histone acetylation, reduced Swi-Snf binding, lower histone eviction and reduced transcription de-repression (Fig. 8C, rpd3 hda1).

Concluding remarks
In this study, we confirmed that Cyc8-Tup1 binds to a distinct site at the FLO1 promoter where it is associated with an array of strongly positioned, deacetylated nucleosomes covering the promoter and upstream region. We revealed that in the absence of Cyc8-Tup1, there is a gross remodelling of FLO1 promoter and upstream chromatin involving nucleosome acetylation, rearrangement and eviction which accompanies FLO1 gene de-repression. We demonstrate direct involvement of Swi-Snf in this remodelling whereby Swi-Snf is recruited in the cyc8 mutant to the same site previously occupied by Cyc8-Tup1. Hence, the extensively remodelled FLO1 promoter and upstream chromatin region is under the control of chromatin remodellers bound at a single discrete site. These findings rule out a propagating mechanism for the activities of Swi-Snf and Cyc8-Tup1, and suggest that the long-range impact on modulation of FLO1 promoter and upstream chromatin may be due to the large size of the participating complexes, or involves the threedimensional packing of the nucleosomal array.
Intriguingly, the histone acetylation which accompanies FLO1 transcription emanates extensively from the Tup1p peak where it marks the region in which the extensive changes in chromatin structure are observed. It is therefore possible that part of the antagonistic activity of the Cyc8-Tup1 and Swi-Snf complexes is acted out at the level of histone acetylation, and that the rapid reversibility of this modification allows for the dynamic switch between gene activation and repression.
Our analysis of genomic data suggest that genes strongly de-repressed in the absence of TUP1 share a significantly increased involvement of Hda1p in the UAS, and also have in common a high incidence of nucleosome depletion upon de-repression in an extensive upstream region. The Cyc8-Tup1 mechanism for repression and de-repression observed at FLO1 is therefore representative of this larger subset. Further investigation of the various classes of Tup1-Cyc8 bound and regulated promoters might reveal alternative repression mechanisms and help to reconcile the different mechanisms that have been proposed.