Antisense non-coding transcription represses the PHO5 model gene at the level of promoter chromatin structure

Pervasive transcription of eukaryotic genomes generates non-coding transcripts with regulatory potential. We examined the effects of non-coding antisense transcription on the regulation of expression of the yeast PHO5 gene, a paradigmatic case for gene regulation through promoter chromatin remodeling. A negative role for antisense transcription at the PHO5 gene locus was demonstrated by leveraging the level of overlapping antisense transcription through specific mutant backgrounds, expression from a strong promoter in cis, and use of the CRISPRi system. Furthermore, we showed that enhanced elongation of PHO5 antisense leads to a more repressive chromatin conformation at the PHO5 gene promoter, which is more slowly remodeled upon gene induction. The negative effect of antisense transcription on PHO5 gene transcription is mitigated upon inactivation of the histone deacetylase Rpd3, showing that PHO5 antisense RNA acts via histone deacetylation. This regulatory pathway leads to Rpd3-dependent decreased recruitment of the RSC chromatin remodeling complex to the PHO5 gene promoter upon induction of antisense transcription. Overall, the data in this work reveal an additional level in the complex regulatory mechanism of PHO5 gene expression by showing antisense transcription-mediated repression at the level of promoter chromatin structure remodeling.


Introduction
The canonical view of eukaryotic transcription has evolved from being considered a highly regulated process initiated from specialized genomic regions, such as gene promoters, to a process that permeates the entire genome [1]. In addition to gene promoters, transcription often initiates from intergenic and intragenic regions, as well as regulatory regions such as gene enhancers. Most of the transcripts originating from these regions are non-coding RNAs that are usually rapidly degraded after synthesis, suggesting that the act of transcription has more potential to exert important biological functions compared to the transcripts themselves [2].
In eukaryotic cells, promoter activation occurs in the context of a repressive chromatin structure, i.e. the packing of DNA with histone proteins into nucleosomal arrays [3]. Since chromatinized DNA is not accessible for interaction with the transcriptional machinery, activators work in concert with chromatin-modifying and -remodelling factors to expose regulatory sites and allow promoter activation. Chromatin modifiers catalyze covalent modifications of histones, such as acetylation, methylation, and phosphorylation, whereas chromatin remodelers use the energy of ATP hydrolysis to slide histones along the DNA or evict them from the DNA [4,5]. Dynamic changes in chromatin conformation also require histone chaperones, which act as histone acceptors and bind them when they are not incorporated into nucleosomes [6]. In gene-dense genomes such as that of yeast, transcription often initiates at the 3' end of genes, leading to the production of antisense (AS) non-coding transcripts [2]. AS readthrough transcription invades the promoter region of the corresponding gene, where it can exert regulatory effects that are usually repressive to transcription of the coding gene [7][8][9][10]. Genome-wide and single gene studies have shown that promoters invaded by AS transcription read-through have high nucleosome occupancy and narrow nucleosome-depleted regions (NDRs) [11,12]. Our recent genome-wide study showed that induced elongation of non-coding antisense transcription into coding gene promoters results in increased deacetylation of promoter nucleosomes by Rpd3. Histone deacetylation leads to decreased recruitment of the major chromatin remodeler RSC and consequently to NDR closure, which represses transcription [13]. However, there are still few examples of bona fide effects of specific AS RNAs on transcriptional regulation of their respective genes, such as the yeast PHO84 gene.
Studies with the PHO84 gene have been highly instructive in elucidating the mechanisms of transcriptional regulation through AS non-coding RNAs [14][15][16]. These studies converged on a model in which PHO84 AS transcription is rapidly terminated in wild-type cells by the NNS (Nrd1-Nab3-Sen1) complex and degraded by the activity of the Rrp6-containing nuclear RNA exosome. Inactivation of any of these crucial factors, such as in rrp6Δ mutant cells, leads to transcriptional read-through of PHO84 AS transcripts, allowing recruitment of histone deacetylases (HDACs) Hda1 or Rpd3 to the PHO84 promoter. Histone deacetylation is thought to lock the chromatin structure of the promoter in a repressed conformation, thereby negatively regulating transcription of the sense transcript, i.e. PHO84 mRNA. This mechanism was subsequently explored genome-wide in yeast, which revealed a group of genes that accumulate AS RNAs in the absence of Rrp6 and are silenced in an HDAC-dependent manner [15]. Genes of this class are characterized by AS transcripts that span the entire gene length, extend beyond the TSS and are enriched for so-called 'closed' promoters. These promoters are typical of inducible or stress-activated genes, and are characterized by precisely positioned nucleosomes whose remodeling is a prerequisite for transcriptional activation [17,18]. A paradigmatic closed promoter that also belongs to this gene class is that of the PHO5 gene, which is a member of the same (PHO) regulon as PHO84 [19].
The PHO5 gene encodes the secreted non-specific acid phosphatase which is located in the periplasmic space and has a role in phosphate metabolism. Accordingly, expression of the PHO5 gene is regulated in response to intracellular phosphate concentration through the PHO signalling pathway, so that it is repressed when the intracellular concentration is abundant and induced under phosphate starvation conditions [19]. This regulation is primarily achieved through phosphorylation of the specific activator Pho4. Under a high phosphate concentration Pho4 undergoes phosphorylation by the cyclin-dependent-kinase Pho80-Pho85, preventing its accumulation in the nucleus and transcriptional activation of the PHO5 gene. In low phosphate, Pho4 is imported into the nucleus and activates transcription. From the early days of chromatin research in the 1980s until now, the PHO5 gene promoter has been and continues to be instrumental in the discovery of numerous fundamental principles and mechanisms of chromatin structure remodeling (reviewed in [19]). In the repressed state, the PHO5 promoter features five precisely positioned nucleosomes, which upon induction are remodelled into a broad nucleosome-depleted region of *600 bp [20]. This massive chromatin transition requires the concerted action of a large network of chromatin-modifying and -remodeling complexes as well as histone chaperones. The repressive chromatin conformation is maintained by H3K4 methylation catalyzed by Set1, a mark that recruits the histone deacetylase Rpd3 to the PHO5 promoter [21,22]. Another histone deacetylase, Hda1, plays a minor role in this process [23]. When the intracellular phosphate concentration is limited, signal transduction via the PHO signaling pathway leads to the accumulation of the unphosphorylated transcriptional activator Pho4 in the nucleus [19,24]. The first step in transcriptional activation of the PHO5 gene is the binding of Pho4 to the UASp1 (Upstream activating sequence phosphate 1) site, which is located in the short nucleosome-depleted region between nucleosomes -2 and -3 of the PHO5 gene promoter. Pho4 recruits histone acetyltransferases, such as the Gcn5-containing SAGA complex, which establish a hyperacetylated promoter configuration [25,26]. Acetylated histones are read by chromatin-remodeling complexes containing bromodomains [27,28]. Alternatively, these remodelers can be recruited to the PHO5 promoter by direct interaction with Pho4 [29]. Five remodelers (SWI/SNF, RSC, INO80, Isw1, Chd1) from all four yeast remodeler families cooperate to catalyze the chromatin opening at the PHO5 promoter [30,31], with the most abundant remodeler, RSC, providing the crucial share of the remodeling activity required for this transition [31]. Histone eviction allows Pho4 to bind to the UASp2 site otherwise covered by nucleosome -2, which is ultimatively required for full transcriptional activation [32][33][34].
Another level of PHO5 promoter regulatory complexity was revealed upon mapping of the PHO5 AS transcript, CUT025 [35,36]. This transcript initiates from the 3' region of the PHO5 ORF and extends through its promoter region, spanning *2.4 kb in size. It is produced only in cells growing under repressive (phosphate-rich) conditions and is more abundant in rrp6Δ mutants compared to wild-type cells, indicating its degradation by the nuclear RNA exosome [35]. AS transcription is generally associated with a repressive effect on transcription of the corresponding genes, and the PHO5 gene is among the rare examples for which AS transcription is proposed to have a positive effect [35]. In this work, we examined the effect of non-coding AS transcription on PHO5 gene expression by enhancing or impairing elongation of the PHO5 AS transcript. In both cases, our results argue in favour of antisense transcription having a negative effect on PHO5 gene expression. Moreover, we provide evidence that this negative effect occurs through a chromatin-remodeling based mechanism mediated by AS transcription which decreases the accessibility of the chromatin structure at the PHO5 gene promoter.

Materials and methods
Yeast strains and primer sequences used in this study are listed in S1 and S2 Tables, respectively.

Strains, media, plasmids and strain construction
Yeast Saccharomyces cerevisiae strains used in this study are listed in S1 Table. All strains were grown at 30˚C. For repressive conditions (high phosphate, +P i ), yeast strains were grown in YNB medium supplemented with 1 g/l KH 2 PO 4 (YNBP) with or without lack of amino acids for plasmid selection. For gene induction by phosphate starvation (-P i ), cells were washed in water and resuspended in the phosphate-free synthetic medium with or without lack of amino acids for plasmid selection [30,31,37]. Anchor-away of Nrd1-AA and Sth1-AA was induced by adding 1 μg/ml of rapamycin (Sigma) to the medium. The RRP6 gene was deleted using a disruption cassette generated by PCR with the primer pairs RRP6-Kan1 and RRP6-Kan2 or RRP6hph_fwd and RRP6hph_rev and the BMA41 rrp6::KanMX4 genomic DNA or the hphcarrying pYM16 plasmid from [38] as template, respectively. The GCN5 gene was deleted using a disruption cassette generated by PCR with the primer pair gcn5HIS_fwd and gcn5HIS_rev and the SpHIS5-carrying pKT101 plasmid from [39] as template. Transformants were selected on G-418 (0.2 mg/ml, Sigma), Hygromycin B (0.3 mg/ml, Sigma) or -His plates, depending on the marker, and gene deletion was confirmed by PCR. The BMA41 TEF1-PHO5 AS strain was constructed by transformation with a cassette generated by PCR with primers TEF1PHO5AS_fwd and TEF1PHO5AS_rev and the pYM-N18 plasmid from [38] as template. Transformants were selected on G-418 plates, and correct insertion of the cassette was confirmed by PCR. The pP5Z reporter plasmid is centromeric vector that carries a PHO5 promoter-lacZ gene fusion and is described in [40]. The pCEN-RRP6 plasmid was previously constructed by Gateway cloning from the pAG416GPD backbone [41]. Plasmid pTDH3-dCas9 (pFS3891) [42] was obtained from Addgene (Plasmid #46920). Plasmid pFS3892, which contains the guide RNA scaffold, was generated by one-step isothermal Gibson assembly reaction (New England BioLabs) using two fragments, one obtained by PCR on pRPR1_gRNA_handle_RPR1t (Addgene Plasmid #49014) using OFS_2869 and OFS_2870 oligonucleotides, the other by PCR on YCpLac33 using OFS_2871 and OFS_2872 oligonucleotides. Plasmid PHO5 AS gDNA-URA3 was then obtained by Gibson assembly reaction (NEB) using OFS_2886 and OFS_2887 to amplify pFS3892 backbone and OFS_2888 and OFS_3095 for gDNA cloning. To test the putative in trans activity of the PHO5 AS RNA, strains were designed as following. First, the PHO5 ORF was replaced by the URA3 marker in either a MATA FSY6857 or a MATα FSY5439 strain (see S1 Table). This was performed by amplification of the URA3 marker from the pUG72 plasmid with OFS5084 and OFS5085 primers and the resulting amplicon was transformed in FSY6857 and FSY5439 strains following selection in a -Ura medium. The MATA and MATα strains deleted for PHO5 were named FSY9286 and FSY9287. We then amplified the wild-type PHO5 gene with either the OFS5086 and OFS5087 primer pair or OFS5088 and OFS5089 primer pair in order to insert a terminator for the PHO5 mRNA (sense) or the AS transcript, respectively. The PCR products targeting either the sense or the antisense transcription were transformed in the FSY9286 and FSY9287 and counter-selected on a 5-FOA medium. The strains targeting the PHO5 mRNA or AS RNA were named FSY9288 and FSY9291. Finally, the 3 different diploids (AS in cis, AS blocked and AS in trans) were generated after crossing FSY6857 with FSY9287, FSY9286 with FSY9291 and FSY9288 with FSY9291, respectively, and selection on -His-Trp medium.

Enzyme activity assays, RNA isolation, Nothern blot and RT-qPCR
Acid phosphatase and beta-galactosidase activity assays were done with intact yeast cells, exactly as described in [31]. Total RNA was extracted by the hot phenol method [43], treated with RNAse-free DNAse I (New England Biolabs) and purified by phenol/chloroform extraction. Strand-specific reverse transcription was performed using 1 μg of RNA and strand-specific oligonucleotides (0.1 μM each) with the ProtoScript First Strand cDNA Synthesis Kit (New England Biolabs) supplemented with actinomycin D (Sigma) to final concentration 5 μg/ml to ensure strand specificity. cDNAs were amplified in Roche LightCycler 480 with the Maxima SYBR Green qPCR Master Mix detection kit (Thermo Scientific). The qPCR datasets were analysed using the ΔΔCt method, and the results were normalized to either PMA1, ACT1 or SCR1 RNAs amplification, which were used as internal controls. To test the putative in trans activity of the PHO5 AS RNA, OFS2522 and OFS2523 were used to measure PHO5 mRNA and AS RNA levels. Amplifications were done in duplicate for each sample, and three independent RNA extractions were analysed. For the Northern blot, total RNA (10 μg for each sample) was run on a 1% denaturing formaldehyde agarose gel and transferred to nylon membranes (Amersham Hybondtm-N+). Membranes were crosslinked and incubated overnight at 60˚C with 100μg/ml boiled salmon sperm DNA in 50% formamide, 5x standard saline citrate (SSC), 20% dextran sulfate sodium, 1% sodium dodecyl sulfate (SDS). Subsequently, membrane wered hybridized with 32P-labeled SP6/T7 riboprobes in 50% formamide, 7% SDS, 0.2 M NaCl, 80 mM sodium phosphate (pH 7.4), and 100 μg/ml boiled salmon sperm DNA for 6h. All blots were washed with 2X SSC and 0.1% SDS for 5 minutes at60˚C and then with 0.5X SSC and 0.1% SDS for 45 minutes at 60˚C. Riboprobes were obtained by SP6/T7 in vitro transcription of gene-specific PCR fragments containing an SP6/T7 promoter. Quantifications were performed with a Phosphor Imager machine.

Chromatin analysis
For anti-histone H3 chromatin immunoprecipitation (ChIP), forty millilitres of cells were fixed with 1% formaldehyde for 20 min. After quenching with 400 mM glycine to stop the reaction, the cells were washed and lysed with glass beads to isolate chromatin. Sonication of cell lysates was performed with a Vibra-Cell sonicator in 1.2 mL of FA150 buffer (50 mM Hepes pH 7.5, 150 mM NaCl, 1 mM EDTA, 1% Triton X-100, 0.1% Sodium deoxycholate and 0.1% SDS) to reduce average fragment size to approximately 500 base pairs. The samples were centrifuged at 2500 g and the supernatant recovered. Chromatin fractions of 400 μl were taken for each immunoprecipitation reaction and incubated with 4 μl of anti-histone H3 antibodies (ab1791, Abcam) at 4˚C overnight. After incubation, 40 μl of protein G PLUS-agarose beads (sc-2002, Santa Cruz Biotechnology) were added and incubated at 4˚C for 2 h. The beads were washed extensively by successive washing steps: 3 times with FA150 lysis buffer, 3 times with FA500 lysis buffer (similar to FA150 but with 500 mM NaCl), 1 time with washing buffer 1 (10 mM Tris-HCl pH 8, 250 mM LiCl, 1 mM EDTA, 0.5% NP-40, 0.5% Sodium deoxycholate) and 1 time with washing buffer 2 (10 mM Tris-HCl pH 8, 1 mM EDTA, 1% SDS). Chromatin was eluted at 80˚C in elution buffer (50 mM Tris-HCl pH 8, 10 mM EDTA, 1% SDS) during 20 minutes. Samples (regardless of Input or IP) were reverse cross-linked at 65˚C overnight. Eluted supernatants (output) and the input controls were hydrolysed with Pronase (0.8 mg/ml final concentration, Sigma) at 42˚C for 2 h, followed by incubation at 65˚C for 7 h to reverse cross-linked DNA complexes. DNA was extracted using the Macherey Nagel Nucleospin Gel & PCR Cleanup Kit. The immunoprecipitated DNAs (output) were quantified by qPCR in Roche LightCycler 480 with the Maxima SYBR Green qPCR Master Mix detection kit (Thermo Scientific). Amplifications were done in triplicate for each sample. Immunoprecipitated samples (output) were normalized to input and to a PHO5-adjacent control region which does not show chromatin signatures similar to the PHO5 gene promoter, as described in [33]. Chromatin analysis of yeast nuclei by restriction nuclease accessibility assay was done as described previously [31,37,44]. 120 U of the ClaI restriction enzyme (New England Biolabs) was used for chromatin analysis of nuclei and 40 U of HaeIII (New England Biolabs) was used for secondary cleavage. Probe for hybridization was as described previously [31,37,45]. Quantification of the percentage of cleaved DNA was done by PhosphorImager analysis (Fuji FLA3000). ChIP of dCas9 was essentially perfomed as in [13] without addition a S. pombe spike-in. An anti-Cas9 antibody (Diagenode #C15310258) was used for the immunoprecipitation step.

AS transcription is involved in regulation of PHO5 gene expression
The product of antisense transcription at the PHO5 model gene locus, CUT025 (hereafter referred to as PHO5 AS), is initiated at the 3' end of the gene ORF in the antisense direction and extends through the PHO5 promoter region ( Fig 1A). The 3'-5' exoribonuclease Rrp6, which is the catalytic subunit of the nuclear RNA exosome complex, degrades this transcript in wild-type (wt) cells, consistent with the increased level of this transcript in rrp6Δ mutant cells (S1A Fig). We confirmed the increased level of the PHO5 AS transcript at the PHO5 promoter region in rrp6Δ compared to wt cells by strand-specific reverse-transcription quantitative PCR (RT-qPCR) upon shifting the cells from repressive (phosphate-rich, +P i ; YNB with additional 1 g/l KH 2 PO 4 ) to inducing (no phosphate, -P i ) conditions. PHO5 AS accumulation in rrp6Δ was most pronounced under repressive conditions ( Fig 1A, 0 h of induction), consistent with [35]. After shifting to inducing conditions, the level of PHO5 AS gradually decreased in both wild-type and rrp6Δ cells, however the increased level in rrp6Δ cells was still present at an early time point of gene induction ( Fig 1A). The PHO5 AS transcript has a much lower steady-state level than the corresponding PHO5 mRNA transcript, as observed by RNA-seq, which measures steady-state RNA levels, i.e., takes into account both the level of nascent transcription and RNA degradation. However, the RNAPII photoactivatable ribonucleoside-enhanced crosslinking and immunoprecipitation (PAR-CLIP) signal, which measures only nascent transcription, is comparable or even higher for the AS transcript than for the mRNA transcript under same growth conditions (4 mM P i ), showing that the AS transcript is being produced to a potentially significant level ( Fig 1B).
Whole-genome tiling array datasets revealed production of another non-coding transcript at the PHO5 gene locus, SUT446, transcribed in the sense direction through the PHO5 promoter region, which appears not to be accumulated in rrp6Δ mutant cells and is weakly expressed ( [15,36]; S1A Fig). It was determined by RT-qPCR that the level of SUT446 was not significantly increased in rrp6Δ compared to wild-type cells neither in repressive nor inducing conditions (S1B Fig), arguing against its gene-regulatory function. Overall, these data support locus with thick straight red lines denoting the the regions used for qPCR (left) and AS RNA levels at the PHO5 promoter region in wild-type BMA41 (wt) and corresponding rrp6Δ mutant cells upon induction through phosphate starvation, monitored by strand-specific reverse transcription quantitative PCR (RT-qPCR) (right). RT-qPCR values were normalized to PMA1 RNA and expressed relative to transcript abundance in wild-type cells under repressive conditions (0 h of induction), which was set to 1. (B) The left panel shows RNA-seq signal from an Nrd1-AA strain in the absence of rapamycin (wild-type equivalent) at the PHO5 locus. The right panel represents RNAPII PAR-CLIP signal or nascent transcription signal in the same conditions. Data were retrieved from [46] and [47], respectively. (C) Levels of PHO5 mRNA in wild-type BMA41 (wt) and corresponding rrp6Δ mutant cells upon induction through phosphate starvation. RT-qPCR values were normalized to PMA1 RNA and expressed relative to transcript abundance in wild-type cells at repressive conditions (0 h of induction), which was set to 1. (D) Same as (C), but acid phosphatase induction kinetics were monitored by measuring acid phosphatase activity with whole cells. Reported values represent the means and standard deviations of three independent experiments (n = 3). https://doi.org/10.1371/journal.pgen.1010432.g001

PLOS GENETICS
a possible regulatory role of the CUT025 AS non-coding transcript, but not the SUT446 promoter non-coding transcript, in regulation of PHO5 gene transcription.
We further investigated whether the increased level of the PHO5 AS transcript under repressive conditions and during early gene induction in rrp6Δ cells correlates with a change in PHO5 mRNA level. PHO5 mRNA was quantified by RT-qPCR upon gene induction and a strong delay in its expression was observed in rrp6Δ cells compared to wild-type cells (Fig 1C). This delay persisted during the first hours of gene induction and corresponded to a delay in expression of the Pho5 acid phosphatase, as determined by measuring its enzymatic activity ( Fig 1D). However, after prolonged induction, the level of acid phosphatase in rrp6Δ cells reached that of wild-type cells (Fig 1D). The observed delay in gene expression was dependent on the catalytic activity of Rrp6, because the catalytically dead rrp6Y361A mutant cells also exhibited delayed PHO5 gene expression, and acid phosphatase activity was brought to wildtype levels when a functional RRP6 gene was expressed from a centromeric plasmid in rrp6Δ cells (S2A Fig). A similar delay was also measured with rrp6Δ cells of two other genetic backgrounds (S2B and S2C Fig), showing that it is not specific to the W303-derived strain used in these experiments.
We also performed a control experiment to test whether the observed kinetic delay in PHO5 expression in rrp6Δ cells is caused by an indirect effect due to compromised signal transduction through the PHO signaling pathway. We made use of a construct in which expression of the lacZ reporter gene was driven by the PHO5 promoter and monitored its expression by measuring beta-galactosidase activity upon induction (no phosphate, -P i ) in wild-type and rrp6Δ cells (S2D Fig). Expression kinetics of the PHO5 promoter-lacZ construct were similar in wild-type and rrp6Δ cells, arguing that PHO signaling is not compromised in rrp6Δ cells. This result demonstrates that the kinetic delay in PHO5 expression observed with the rrp6Δ strain (Fig 1C and 1D) was not an indirect effect caused by compromised induction strength and consequently impaired PHO5 transcriptional activation. Additionally, this result speaks in favour of a possible regulatory role of the AS transcript originating from the PHO5 ORF.

PHO5 gene expression kinetics are delayed upon induction in mutants related to RNA exosome function
Rrp6 is the nuclear-specific catalytic subunit of the RNA exosome complex. To determine the involvement of other RNA exosome subunits and cofactors in the regulation of PHO5 gene expression, we examined the kinetics of PHO5 gene expression using appropriate mutant cells. Deletion mutants for the monomeric cofactors of the nuclear exosome, Rrp47 and Mpp6, also showed delayed acid phosphatase expression kinetics (Fig 2A). The TRAMP complex is another cofactor of the nuclear RNA exosome and consists of a non-canonical poly(A) polymerase (Trf4 or Trf5), an RNA-binding subunit (Air1 or Air2), and the essential helicase Mtr4 [48,49]. Interestingly, single air1Δ and air2Δ mutant cells showed no delay, whereas the air1Δ-air2Δ double mutant showed an even greater delay than the rrp6Δ mutant (Fig 2A), consistent with a high degree of redundancy between homologous TRAMP subunits [50]. Somewhat surprisingly, acid phosphatase activity measured after overnight induction was increased in some mutants compared with wild-type cells. It is possible that this may reflect specialized cofactor requirements that support the specific conditions of prolonged gene induction and was not pursued further. The mutant for the exonuclease activity of the essential RNA exosome catalytic subunit Dis3 (dis3Δ + pDis3-exo -) also showed delayed kinetics compared with the corresponding wild-type cells (dis3Δ + pDis3) and with the mutant for its endoribonuclease activity (dis3Δ + pDis3-endo -) ( Fig 2B). These results demonstrate the involvement of the second catalytic subunit of the RNA exosome, Dis3, as well as the nuclear RNA exosome cofactors Rrp47, Mpp6 and the TRAMP complex in the regulation of PHO5 gene expression.
In rrp6Δ and other RNA exosome deletion mutant backgrounds, AS transcription is constitutively induced due to sequestration of the NNS (Nrd1-Nab3-Sen1) termination complex by stabilised non-coding RNAs. The NNS complex cannot be efficiently recycled to sites of transcription, inducing termination defects at non-coding RNA loci and resulting in their increased elongation frequency [51]. To rule out possible indirect effects on transcription of the PHO5 gene due to gene deletion mutant backgrounds in which AS transcription is constitutively elongated, we turned to a system in which AS elongation is inducible. To this end, we used the Anchor Away (AA) system to rapidly deplete Nrd1 protein from the nucleus by rapamycin treatment [52]. Since Nrd1 belongs to the NNS surveillance system, its removal is expected to trigger transcriptional read-through of non-coding RNAs [51]. Indeed, treatment of Nrd1-AA cells with rapamycin resulted in rapid induction of the PHO5 AS transcript production, clearly demonstrating that the NNS complex is important for its early termination in wild-type cells. Importantly, even under PHO5 repressive conditions, induction of PHO5 AS transcript production through the Nrd1-AA system was accompanied by downregulation of PHO5 mRNA levels, as shown by Northern blot (Fig 3A). Furthermore, with the Nrd1-AA system, it was possible to induce elongation of AS transcription by adding rapamycin simultaneously when shifting the cells to PHO5 inducing conditions (i.e. phosphate free medium) ( Fig  3B) or an hour before the shift (Fig 3C). Consistently, the kinetics of Pho5 expression monitored by measuring acid phosphatase activity showed a kinetic delay which was dependent on the timing of rapamycin addition during cultivation (Fig 3B and 3C). The results of this experiment demonstrated that the negative correlation between PHO5 AS and mRNA transcript levels is not an indirect consequence of gene deletion mutant backgrounds, since it is also seen upon induced Nrd1 depletion.

Transcription of PHO5 AS RNA regulates PHO5 gene expression in cis
To increase the transcription level of the PHO5 AS transcript without using RNA degradation/ termination mutant backgrounds, we inserted the strong constitutive TEF1 promoter in the antisense configuration downstream of the PHO5 gene ORF (Fig 4A). We confirmed that this resulted in the TEF1 promoter driving AS transcription at the PHO5 gene locus by RT-qPCR, as the level of PHO5 AS transcript in these mutant cells was �20 times higher than in the corresponding wild-type cells. Impressively, even under +P i conditions in which the PHO5 mRNA is only basaly expressed, TEF1-induced overexpression of the AS transcript caused a severalfold decrease in PHO5 mRNA level, confirming the negative correlation between PHO5 AS and mRNA transcript levels. What is more, Pho5 expression in these cells was delayed compared with wild-type cells and did not reach full expression level even after overnight induction (Fig 4B). This result indicates that an artificially induced high constitutive level of AS transcription at the PHO5 locus drives repression of the PHO5 gene even after prolonged induction.
Furthermore, we tested whether the PHO5 AS transcript can regulate PHO5 gene expression when expressed in trans, i.e. whether the AS transcript itself has a regulatory function. We constructed diploid strains (as in [10]) in which only one copy of the PHO5 AS transcript was expressed either in cis (from the same chromosome as PHO5 mRNA), in trans (from the opposite chromosome) and another one in which AS transcription in cis was blocked (Fig 4C). Insertion of a terminator sequence to block AS transcription in cis resulted in only partial downregulation of PHO5 AS level as shown by RT-qPCR. However, there was a marked increase in PHO5 mRNA level in this diploid strain compared with the strain with native PHO5 AS levels expressed in cis (Fig 4C). Crucially, when PHO5 AS was expressed in trans in addition to downregulation of its level in cis, PHO5 S expression was higher than for the native locus indicating no repressive effect of the AS. These results argue that the act of AS transcription, rather than the AS RNA transcript itself, represses transcription of the PHO5 gene.

Block of AS transcription through dCas9 enhances the kinetics of PHO5 gene expression
Given that accumulation of the PHO5 AS transcript negatively affects PHO5 gene transcription kinetics, blocking AS transcript production should enhance it. To specifically target PHO5 AS transcription, we undertook a CRISPRi approach in which a catalytically dead Cas9 protein (dCas9) is directed by a guide RNA (gRNA) to interfere with AS transcription at the PHO5 gene locus. The CRISPRi system blocks transcription due to physical collision between the elongating RNA Polymerase and the dCas9:gRNA complex [53]. Furthermore, this system was shown to function in a strand-specific manner, by blocking transcription only when the nontemplate DNA strand of a transcription unit is targeted [53,54]. Therefore, we targeted dCas9 to the nontemplate strand of the AS transcription unit at the PHO5 gene locus to block only AS transcription. First, we confirmed the presence of the dCas9 protein at the PHO5 ORF by anti-Cas9 chromatin immunoprecipitation (ChIP). Notably, a strong peak of dCas9 binding at the PHO5 gene ORF compared to a control strain not expressing the gRNA was observed ( Fig  5A), while no dCas9 binding could be detected at the PHO5 promoter region covered by nucleosomes -4 and -1 (Fig 5A). RNA levels in the Nrd1-AA strain with the active CRISPRi

PLOS GENETICS
Antisense RNA represses PHO5 gene Chromatin immunoprecipitation (ChIP) analysis of dCas9 binding at the PHO5 gene locus. Immunoprecipitated DNA was quantified by qPCR with primers specific for different regions of the PHO5 promoter (Nucleosomes -4 and -1) and ORF regions. Both strains were transformed with a dCas9 expressing plasmid, while the CRISPRi strain was additionaly transformed with a plasmid expressing a gRNA targeted to strand-specifically block PHO5 AS transcription and the control strain with the corresponding empty plasmid. Nucl-nucleosome, No Ab-no antibody ChIP control. (B) Levels of PHO5 AS transcribed at the PHO5 promoter and ORF regions at 0 h of induction in Nrd1-AA strain with or without addition of rapamycin (for 1 hour; to deplete Nrd1 and induce AS transcription) and an active CRISPRi system. RT-qPCR values were normalized to SCR1 RNA. Reported values represent the means and standard deviations of three independent experiments (n = 3). Indicated differences show the significant differences using an unpaired Student's t test. Two ( �� ) and four ( ���� ) asterisks denote a p-value lower than or equal to 0.01 and 0.0001, respectively. (C) Levels of PHO5 AS and mRNA transcripts in the CRISPRi and the corresponding control strain, monitored by RT-qPCR at 0 h of induction as in (B). Strains are Nrd1-AA with the absence of rapamycin (wild-type equivalent). Right: Scheme of the CRISPRi strategy used to block PHO5 AS transcription. Strains were transformed with two plasmids, one expressing dCas9, and the other expressing or not a gRNA system were monitored by RT-qPCR and showed a highly reproducible decrease in PHO5 AS levels compared to the control strain (Fig 5B). This decrease was significant at the PHO5 promoter and ORF regions without rapamycin addition or with rapamycin (i.e., depletion of Nrd1 which induces AS transcription). These results are consistent with a dCas9-mediated transcriptional roadblock of AS transcription at the PHO5 gene locus. After the addition of rapamycin, PHO5 AS levels were increased in both the CRISPRi Nrd1-AA strain and the control Nrd1-AA strain. However, its levels in the CRISPRi strain remained significantly lower, maintaining the difference in levels already observed without the addition of rapamycin ( Fig  5B). These results demonstrated that the dCas9-mediated roadblock of AS transcription at the PHO5 gene locus is robust and maintained after global induction of AS transcription, although AS transcription was not completely abolished.
Importantly, impairment of PHO5 AS RNA elongation led to an increase in PHO5 mRNA levels (Fig 5C), clearly demonstrating the direct role of AS transcription in PHO5 gene repression. Also, it argues in favour that the CRISPRi system strand-specifically blocked only AS transcription without significantly impacting mRNA transcription. We further tested if impairment of AS transcription with use of the CRISPRi system would result in enhanced kinetics of PHO5 gene expression. As expected, the kinetics of PHO5 gene expression upon gene induction were slightly faster when AS transcript production was impaired by dCas9 than in the control strain (Fig 5D and 5E). This effect was noticed only at very early timepoints of gene induction (30 min for mRNA levels and 1,5 h for acid phosphatase levels), possibly due to the dCas9 protein losing its roadblock function past a certain level of ongoing transcription.

AS RNA elongation affects PHO5 promoter chromatin structure
Since transcriptional activation of the PHO5 promoter requires a large transition of its chromatin structure, we investigated whether the kinetics of PHO5 promoter chromatin opening upon gene induction also inversely correlate with PHO5 AS transcription. To this end, we examined the chromatin structure at the PHO5 promoter with anti-histone H3 ChIP at nucleosome -2, which covers the high-affinity Pho4 binding site and is considered the critical nucleosome for PHO5 chromatin remodeling [19]. A higher histone occupancy was observed in rrp6Δ compared to wild-type cells already under repressive conditions (Fig 6A). Accordingly, histone removal from the PHO5 promoter was slower in rrp6Δ than in wild-type cells during the first hours of gene induction and reached a similar final level after 5 hours (Fig 6A). To confirm the delayed kinetics of chromatin opening in rrp6Δ cells, we took advantage of the ClaI restriction enzyme accessibility assay, which quantifies the efficiency of cleavage by ClaI enzyme at nucleosome -2 of the PHO5 promoter (Fig 6B). Consistent with the anti-histone H3 ChIP, the accessibility of the ClaI site at the PHO5 promoter was lower in rrp6Δ and air1Δ-air2Δ than in wild-type cells during the first hours of gene induction (Fig 6C). These results show that AS transcription mediates a negative effect on PHO5 transcriptional activation by influencing the chromatin structure at its promoter region.
Our results suggest that AS transcription at the PHO5 gene locus locks the chromatin structure of the PHO5 promoter in a more repressive configuration that is harder to remodel ( Fig  6). This could be due to the activity of HDACs, which have been shown to negatively affect targeting the non-template strand of the PHO5 AS transcription unit. (D) Levels of PHO5 mRNA in the CRISPRi and the corresponding control strain upon induction through phosphate starvation monitored by RT-qPCR as in (B). Strains are same as in (C). Reported values represent the means and standard deviations of three independent experiments (n = 3). Indicated differences show the significant differences using an unpaired Student's t test. One ( � ) and two ( �� ) asterisks denote a p-value lower than or equal to 0.05 and 0.01, respectively. (E) Same as (D), but acid phosphatase induction kinetics were monitored by measuring acid phosphatase activity with whole cells. https://doi.org/10.1371/journal.pgen.1010432.g005

PLOS GENETICS
chromatin structure at the PHO5 promoter [22,23]. Remarkably, inactivation of the HDAC Rpd3 in the rrp6Δ mutant background does not affect the level of the PHO5 AS RNA, but it restores transcription activation of the PHO5 gene to the level or even higher than in wild-type cells, as shown by tiling arrays and RT-qPCR with single and double deletion mutant cells ( [15]; Fig 7A). Accordingly, the expression kinetics of acid phosphatase measured with the rpd3Δ rrp6Δ double mutant cells are not delayed compared to wild-type cells, in contrast to the corresponding rrp6Δ single mutant cells (Fig 7B). Consistent with this, expressing the PHO5 AS-blocking CRISPRi system leads to faster gene expression kinetics in wild-type and rrp6Δ, but not in rpd3Δ and rpd3Δ rrp6Δ double mutant cells (S3 Fig). These results demonstrate that the PHO5 AS transcript acts via a pathway that involves histone deacetylation. Gcn5, the catalytic subunit of the SAGA and ADA complexes, is known to be the major histone acetyltransferase that enables physiological gene induction kinetics at the PHO5 promoter https://doi.org/10.1371/journal.pgen.1010432.g006 [25,26]. We reasoned that in the absence of Gcn5, i.e. when the majority of histone acetylation normally present at the PHO5 gene promoter is reduced, an rrp6Δ strain should have no additional effect on the kinetics of PHO5 gene expression. Indeed, the kinetics of acid phosphatase expression in gcn5Δrrp6Δ double mutant strain are the same as in the gcn5Δ single mutant  [15] at the PHO5 locus for the Watson (W, upper half) and the Crick (C, lower half) strands. Three replicates of each strain are represented. A darker signal depicts a higher score of RNA expression. The red vertical lines represent the inferred coding and non-coding genes boundaries. Below: Levels of PHO5 AS transcript measured by RT-qPCR with the same strains at +Pi conditions. Values were normalized to ACT1 RNA. (B) Acid phosphatase induction kinetics in wild-type W303 (wt) and the corresponding rrp6Δ, rpd3Δ and rrp6Δrpd3Δ cells upon induction through phosphate starvation. Reported values represent the means and standard deviations of three independent experiments (n = 3). (C) As in (B), but for wild-type BY4741 (wt) and the corresponding rrp6Δ, gcn5Δ and rrp6Δgcn5Δ mutant cells.

AS transcription negatively affects recruitment of RSC to the PHO5 promoter
Histone acetylation plays two important roles in transcriptional activation. It neutralizes the positive charge of lysine groups, thereby weakening histone-DNA interactions, and it also provides docking sites for the bromodomains of proteins involved in transcriptional regulation. RSC (Remodels Structure of Chromatin) complex is the most abundant and the only essential remodeler in yeast and contains seven of the fourteen bromodomains identified in S. cerevisiae [28,55]. RSC was found to be the major remodeler among the five chromatin remodelers involved in the chromatin remodeling process at the PHO5 promoter [31]. Its partial depletion, achieved by a temperature-sensitive mutant of its catalytic subunit sth1 td , resulted in a strong delay in promoter chromatin structure opening and, consequently, delayed kinetics of acid phosphatase expression upon PHO5 gene induction.
To first test the hypothesis that histone acetylation recruits RSC to the PHO5 gene promoter upon induction, we used the anchor away system to deplete its catalytic subunit Sth1 from the nucleus in Sth1-AA and corresponding gcn5Δ mutant cells. Because Sth1 is essential for cell viability, we first attempted to induce its depletion in parallel with the induction of the PHO5 gene. Addition of rapamycin upon shifting the cells to phosphate-free medium caused a delay in acid phosphatase expression kinetics similar to the partial depletion through sth1 td (Fig 8A). In gcn5Δ mutant cells, this partial depletion leads to an additive effect on acid phosphatase expression kinetics. However, since RSC is very abundant, it is possible that the partial depletion of Sth1 still leaves a lot of active RSC complex in the nucleus in the first hours of gene induction. We therefore added rapamycin two hours before PHO5 gene induction to achieve more extensive RSC depletion before shifting the cells to phosphate-free medium. Addition of rapamycin two hours before gene induction resulted in an epistatic effect of the Sth1 depletion. Upon simultaneous inactivaton of Gcn5 and RSC, acid phosphatase expression kinetics were severely delayed, but reached overnight levels comparable to wild-type (Sth1-AA -Rap) cells ( Fig 8A). This result positions Gcn5 and RSC in the same pathway of PHO5 gene transcriptional activation and speaks in favour of a link between RSC recruitment and Gcn5-mediated acetylation upon induction.
To test the effect of Sth1-AA depletion in rrp6Δ mutant cells, we monitored acid phosphatase expression kinetics upon addition of rapamycin (S4 Fig). Even when rapamycin was added two hours before induction to achieve more complete inactivation of RSC, it resulted in an additive effect on acid phosphatase expression kinetics with the rrp6Δ mutation. It is possible that Rrp6 and RSC regulate PHO5 gene expression through at least partially independent pathways. However, because these cells barely induced the PHO5 gene, as indicated by the levels of acid phosphatase activity measured after overnight induction, we cannot rule out the possibility that this additive effect is due to the severely impaired cell viability because of the Sth1-AA depletion in the slow-growing rrp6Δ background.
To directly answer the question of whether AS-induced deacetylation of the PHO5 promoter may inhibit the recruitment of RSC, resulting in a more closed chromatin conformation, we took advantage of our genomic analyses recently performed with Nrd1-AA and Nrd1-AA rpd3Δ cells with and without the addition of rapamycin [13]. We examined the PHO5 gene locus in the Micrococcal Nuclease sequencing (MNase-seq), Assay for Transposase-Accessible Chromatin using sequencing (ATAC-seq) and Sth1-Chromatin Endogenous Cleavage-sequencing (Sth1-ChEC-seq) datasets, which give us information about the

PLOS GENETICS
chromatin conformation and Sth1 binding at the PHO5 promoter upon induction of AS transcription (+Rap/-Rap) and depending on the presence of Rpd3 (Fig 8B). The ChEC-seq data show the fold change in association of Sth1, the ATP-ase subunit of RSC, with chromatin upon induction of AS transcription (+Rap/-Rap). In addition, the ATAC-seq data provide us with information about chromatin accessibility under the same conditions. In Nrd1-AA cells there is a negative fold change, i.e. a decrease in Sth1 binding, associated with a decrease in chromatin accessibility upon addition of rapamycin, for the region encompassing nucleosome -2 of the PHO5 promoter (the position of which was determined using MNase-seq data in -Rap) (Fig 8B). Conversely, in isogenic rpd3Δ cells, addition of rapamycin has a much smaller effect on Sth1 binding or chromatin accessibility in this region (Fig 8B). When comparing the two biological experimental replicates, the log2 values for the change Sth1 binding (+Rap/-Rap) were consistently lower in Nrd1-AA compared to isogenic rpd3Δ cells (-0,3947 and -0,566 vs. -0,2057 and -0,2841, respectively, calculated over the middle 40 bp region of nucleosome -2). These data argue in favour of a model in which read-through of AS transcription acts via recruitment of histone deacetylases to the PHO5 gene promoter, the activity of which results in decreased recruitment of the RSC complex (Fig 8C).

Discussion
The role of non-coding RNAs in regulation of gene expression could not be appreciated until recent advances in high-throughput methods facilitated their detection and characterization. From a gene-centered view, non-coding RNAs can be transcribed in tandem with genes, i.e. from the same strand as the gene, or from the opposite strand, resulting in production of antisense (AS) non-coding RNAs. Apart from a few isolated examples, production of AS non-coding RNAs is generally thought to have a repressive cis-regulatory effect on the expression of associated mRNAs [7,9,56]. This seems to be particularly the case when transcription of AS non-coding RNAs invades promoters of coding genes [10,15,57]. In light of this current view, we felt compelled to reexamine the role of AS transcription at the model yeast PHO5 gene locus, which was originally suggested to support gene activation [35]. In this work, we show a clear negative role for AS transcription in PHO5 gene expression. By leveraging mutant backgrounds in which AS transcription is constitutively enhanced or inducible and artificially driving its expression from a strong promoter in cis, we show that increased PHO5 AS elongation frequency correlates with decreased expression of the corresponding mRNA. Furthermore, we demonstrate that the use of a CRISPRi system that specifically blocks AS transcription at the PHO5 gene locus increases the level of PHO5 mRNA and enhances its induction kinetics upon phosphate depletion. Importantly, these observations show that AS RNA transcription has an impact on PHO5 gene expression in wild-type cells, and not only upon enhanced AS RNA stabilisation in strains mutant for RNA degradation factors. We also show that AS RNA transcription regulates expression of the PHO5 gene only when transcribed in cis, and not in trans. The role of PHO5 AS transcription is therefore reminiscent of the role of AS transcription in maintaining the tight repression of quiescence-related transcripts during the exponential growth phase, recently demonstrated by Nevers et al. [10]. A previous study suggesting a positive regulatory role for PHO5 AS transcription achieved AS inactivation by incorporating a full-length marker gene sequence with its promoter region in the middle of the PHO5 gene ORF [35]. This major perturbation of the PHO5 gene locus may have resulted in experimental artefacts, highlighting the need for precise interventions, such as those achieved by the CRIS-PRi system, to perform functional analyses of AS transcripts [58].
There are now several well-described examples of yeast gene loci at which either antisense or upstream non-coding transcription that extends through a coding sense promoter has an inhibitory effect on its transcription initiation [14,[59][60][61][62]. In most cases, it is likely that elongation of non-coding transcription leads directly to displacement of transcription factors (TFs) and/or the preinitiation complex (PIC) or that the recruitment of TFs or the PIC to these gene promoters is decreased as a consequence of a more repressive chromatin configuration established at the promoter region due to elongation of non-coding transcription (see [63] for a review). This model is supported by whole-genome analyses showing that invasion of gene promoters by AS transcription leads to increased histone occupancy and altered recruitment of chromatin-modifying and -remodeling complexes [11,13,64]. At the tandemly transcribed SRG1 lncRNA/SER3 protein-coding gene locus, non-coding transcription has been shown to cause nucleosome deposition at the gene promoter, thereby repressing SER3 transcription [65]. As another example, we have shown that AS transcription at the PHO84 gene locus silences the corresponding gene by recruiting HDACs to its promoter region [14]. The AS RNA does not recruit the HDACs directly, but the act of its transcription promotes a histone methylation-based mechanism to restore the repressive chromatin structure in the wake of the elongating RNA Pol II. The histone methyltransferase Set2 associates with the elongating RNA Pol II and catalyses H3K36 methylation, a mark read by the Eaf3 chromodomain of the HDAC Rpd3 [66]. Consistent with this, our recent genome-wide study in yeast has shown that AS transcription leads to deacetylation of a subpopulation of -1/+1 nucleosomes associated with increased H3K36 methylation, which in turn leads to decreased binding of the RSC chromatin-remodeling complex and sliding of nucleosomes into previously nucleosome-depleted regions [13]. We have now shown that elongation of PHO5 AS under repressive conditions leads to increased histone occupancy at the PHO5 gene promoter and slower histone removal upon gene induction. Moreover, the negative effect of AS RNA elongation on PHO5 gene activation is mitigated by inactivation of Rpd3, suggesting a histone acetylation-based regulatory mechanism that may affect the recruitment of RSC, a chromatin remodeler that plays an important role in PHO5 gene promoter opening [31]. This is supported by ChEC-seq of Sth1, the catalytic subunit of RSC, showing a decrease in its recruitment to the PHO5 gene promoter upon induction of antisense transcription, that is suppressed by inactivation of Rpd3.
PHO5 belongs to a group of~100 genes that are more transcribed in AS direction as a noncoding transcript than in the sense orientation as an mRNA in a standard medium (Fig 1B). In such culture conditions, the Pho4 transcriptional activator is rarely located in the nucleus [19]. Thus, as we proposed in [46] for the SAGA-dependent gene class to which PHO5 belongs, the steady-state chromatin structure of the promoter NDR might be maintained tightly closed by ongoing AS transcription. What may also be relevant to this mechanism is the recently discovered autoregulatory mechanism of the SAGA complex, which is induced in response to environmental changes such as phosphate starvation conditions [67]. The SAGA catalytic subunit Gcn5 has been shown to acetylate the Ada3 subunit, which promotes dimerization of the SAGA complex and in turn leads to higher efficiency of SAGA-catalysed histone acetylation. PHO5 expression was shown to correlate negatively with decreasing levels of Ada3 acetylation and consequently lower efficiency of histone acetylation by Gcn5. The same was also found for SUC2 transcription, which is induced during growth in sucrose-containing media. Of importance to our work is the finding that of the 8 known histone deacetylases, the Ada3 subunit is deacetylated only by Rpd3, but the mechanism of its recruitment to SAGA remains to be elucidated. Therefore, the enhanced recruitment of Rpd3 mediated by AS transcription may play a dual role in regulating PHO5 gene expression, considering that Rpd3 deacetylates promoter histones and Gcn5, both of which contribute to transcriptional repression. It remains to be investigated whether such a regulatory mechanism of AS transcription-mediated repression could be a common mechanism for AS transcription-induced repression of stress-inducible and SAGA-dependent genes regulated by promoter chromatin structure remodeling.
The regulatory roles of non-coding RNAs are intertwined with that of chromatin structure. Not only does non-coding transcription affect chromatin structure, but chromatin structure also determines where and how often non-coding RNAs are transcribed. This fact is increasingly appreciated with respect to the directionality of transcription at promoters of coding genes. Specifically, chromatin modifiers such as the HDAC Hda1, and chromatin remodelers such as RSC, have been shown to dictate promoter directionality by attenuating divergent non-coding transcription [68,69]. Furthermore, chromatinization of DNA limits aberrant transcription that would otherwise occur on naked DNA, as was recently demonstrated through in vitro experiments by the Kornberg group [70]. In this study, a chromatinized PHO5 gene locus fragment was transcribed seven times more from the physiological transcription start site than the same naked DNA locus, and also resulted in transcription patterns more similar to those seen in vivo. Although only chromatin was considered in this study, it would be interesting to also investigate non-coding transcription using a similar in vitro transcription system.
Chromatin remodeling complexes and non-coding RNAs are important regulators of gene expression, and therefore dysregulation of either of these factors may affect the development and progression of various cancers. The SWI/SNF family of chromatin remodeling complexes includes the SWI/SNF complex with its catalytic subunits BRG1 or BRM in humans, and the SWI/SNF and RSC complex with their catalytic subunits Snf2 and Sth1, respectively, in yeast. Numerous associations between chromatin remodelers of this family and long non-coding RNAs have been detected in human cancers (reviewed in [71]). These complexes and the corresponding regulatory non-coding RNAs therefore represent promising diagnostic and therapeutic targets. Transcription of long non-coding RNAs is particularly important for the yeast genome, which has a very high gene density, such that many of them overlap coding gene ORFs or promoter regions. Another reason why budding yeast is a good model for studying the transcription of such long (�200 nt) non-coding RNAs is that it exclusively synthesizes this non-coding transcript class since its divergence from other yeasts and the loss of the RNAi system that produces small non-coding RNAs [72]. In addition, extensively studied gene loci, such as the yeast PHO5 gene, are invaluable for mechanistic studies of gene regulation. Studies of the PHO5 gene and its promoter region made an immense contribution to deciphering the mechanisms of gene regulation through chromatin remodeling [19] and our study now opens the possibility to focus on non-coding transcription in this system. Supporting information S1 Fig. Non-coding transcripts CUT025 and SUT446 are transcribed at the PHO5 gene locus. (A) A heatmap summarising tiling array expression data at the PHO5 gene locus in wild-type W101 (wt) and corresponding rrp6Δ cells. Data is from [36] and is visualized with the SGV Genomics Viewer [73]. (B) Levels of SUT446 in wild-type BMA41 (wt) and corresponding rrp6Δ mutant cells upon induction through phosphate starvation. RT-qPCR values were normalized to PMA1 RNA and expressed relative to transcript abundance in wild-type cells at repressive conditions (0 h of induction), which was set to 1.