RSC and GRFs confer promoter directionality by limiting divergent noncoding transcription

The directionality of gene promoters - the ratio of protein-coding over divergent noncoding transcription - is highly variable and regulated. How promoter directionality is controlled remains poorly understood. Here, we show that the chromatin remodelling complex RSC and general regulatory factors (GRFs) dictate promoter directionality by attenuating divergent transcription. At gene promoters that are highly directional, depletion of RSC leads to a relative increase in divergent noncoding transcription and thus a decrease in promoter directionality. We find that RSC facilitates nucleosome positioning upstream in promoters at the sites of divergent transcription. These highly directional promoters are also enriched for the binding of GRFs such as Reb1 and Abf1. Ectopic targeting of divergent transcription initiation sites with GRFs or the dCas9 protein suppresses divergent transcription. Our data suggest that RSC and GRFs play a pervasive role in limiting divergent transcription. We propose that any DNA binding factor, when stably associated with cryptic transcription start sites, form barriers for repressing divergent transcription. Our study provides an explanation as to why certain promoters are more directional than others.


Introduction
Transcription is highly pervasive and results in intergenic and intragenic noncoding transcription events. Noncoding transcription and the produced noncoding RNAs play diverse roles in gene and genome regulation (Ard et al 2017, Gil & Ulitsky 2020. Transcriptionally active, protein-coding gene promoters are a major source of noncoding transcription. At these genomic locations, noncoding transcription initiates in the divergent direction, a process known as divergent or bidirectional transcription, which generates upstream transcripts from a distinct core promoter in the opposing direction to the coding gene (Neil et al 2009, Seila et al 2008, Sigova et al 2013. Divergent noncoding transcription is an intrinsic property of coding gene transcription, present across eukaryotes, and widespread across actively transcribed regions in the genome (Seila et al 2009). Divergent and coding gene transcription share the same promoter sequence and the ratio (divergent over coding transcription) gives insights on the directionality of promoters (Jin et al 2017).
Our understanding of the function of divergent transcription is still incomplete. There is evidence that the noncoding transcripts emanating from divergent transcription events can help to promote gene transcription in the coding direction and to facilitate cell fate control (Frank et al 2019, Luo et al 2016, Yang et al 2021. Other studies have suggested that divergent transcription may consitute transcriptional noise of active promoters (Seila et al 2009, Struhl 2007. Accordingly, promoters with a high ratio of divergent over coding transcription may have evolved fortuitously, and these promoters have been proposed to represent a transcriptional ground-state (Jin et al 2017, Wu & Sharp 2013. Conversely, highly directional promoters (i.e. promoters with little divergent transcription but high levels of sense coding gene transcription) are likely more evolved.
Aberrant noncoding transcription, including divergent transcription, can negatively impact cell fitness. For example, induced noncoding transcription can cause R-loop formation and DNA damage (Nojima et al 2018). In addition, aberrant divergent transcription events can affect coding gene transcription leading to mis-regulation of gene expression (Chiu et al 2018, du Mee et al 2018. This especially impacts species with gene-dense genomes such as Saccharomyces cerevisiae. Hence, there are various molecular mechanisms that limit the accumulation of noncoding RNAs or noncoding transcription itself. Indeed, to reduce the accumulation of divergent RNAs, these noncoding transcripts are often rapidly degraded via the RNA degradation machinery (Flynn et al 2011, Malabat et al 2015, Neil et al 2009, van Dijk et al 2011. Divergent transcription units are also typically short, due to enrichment in transcription termination signals (Schulz et al 2013). Additionally, divergent transcription can be repressed by chromatin remodellers and histone modifying enzymes (Gowthaman et al 2021, Marquardt et al 2014. Lastly, general regulatory factors (GRFs) can repress initiation of aberrant noncoding transcription events (Challal et al 2018. In previous work, we showed that the GRF and pioneer transcription factor Rap1 represses divergent transcription in the promoters of highly expressed ribosomal protein and metabolic genes (Challal et al 2018. At these highly directional promoters, Rap1 is key for both promoting transcription in the coding direction and for limiting transcription in the divergent direction (Challal et al 2018. We proposed that Rap1 limits divergent transcription by interfering with recruitment of the transcription machinery. In the same study, we also identified a role for RSC, a major chromatin remodelling complex important for chromatin organization (Cairns et al 1996). We showed for two gene promoters that RSC promotes divergent transcription in Rap1 depleted cells. RSC activity increases nucleosome positioning, which, in turn, is important for maintaining nucleosome depleted regions (NDRs) in promoters (Badis et al 2008, Hartley & Madhani 2009, Lorch et al 1998, Parnell et al 2008. RSC also acts with GRFs to stimulate transcription in the protein-coding direction (Brahma & Henikoff 2019, Kubik et al 2017. Specifically, RSC positions the +1 nucleosome in gene promoters to allow pre-initiation complex (PIC) assembly and transcription start site Here, we examined the role of RSC and GRFs in controlling promoter directionality.
Our analysis reveals that RSC depletion leads to relative in increase in divergent noncoding transcription. These promoters tend to be highly directional and are enriched for GRFs such as Abf1 and Reb1. Consistent with the role of RSC in chromatin organization, its depletion affects nucleosome positioning in upstream regions of directional promoters. Finally, we demonstrate that ectopic targeting of GRFs and dCas9 to divergent core promoters can also repress divergent noncoding transcription. We propose that nucleosomes positioned by RSC and GRFs constitute physical barriers which limit aberrant divergent transcription in promoters, thereby increasing promoter directionality.

RSC represses divergent transcription independently of Rap1
To investigate how RSC and GRFs promote directionality of transcripiton more closely, we performed RNA-seq on nascent transcribed RNA (nascent RNA-seq) and on polyadenylated RNAs (mRNA-seq) after depletion of both RSC and Rap1. We determined the levels of nascently transcribed RNA by measuring RNA Pol IIassociated transcripts using an adapted native elongating transcript sequencing protocol (Churchman & Weissman 2011). In short, we affinity-purified RNA Pol II using Rpb3-FLAG and quantified the Pol II-associated RNAs in wild-type (WT) cells, and after depletion of RSC and/or Rap1 ( Figure S1A). To deplete RSC and Rap1 we used auxin-inducible degron alleles (RAP1-AID and STH1-AID) and treated cells with indole-3-acetic acid (IAA) (Nishimura et al 2009). Rap1 and Sth1 were efficiently depleted in cells harbouring either one or both degron alleles after treatment ( Figure   1A, top panel). For mRNA-seq we added a spike-in control of S. pombe cells in a defined ratio and used the S. pombe mRNA-seq signal to normalize mRNA-seq data for S. cerevisiae ( Figure 1A, bottom panel).
First, we examined how a subset of regulated protein-coding transcripts were affected by Rap1 and RSC depletion. We focused on a set of 141 Rap1-regulated gene promoters described previously . As expected, almost all Rap1-regulated genes decreased in expression after Rap1 depletion (RAP1-AID + IAA vs RAP1-AID + DMSO) ( Figure 1B). Expression of Rap1-regulated genes also decreased in Sth1-depleted cells as detected by mRNA-seq ( Figure 1B). This decrease was lesser in nascent RNA-seq likely, in part, because internal normalization was used for the analysis. Second, we examined two previously wellcharacterized promoters (RPL43B and RPL40B) at which Rap1 is known to repress divergent transcription (Figures 1C and S1B) . As expected, transcription of divergent noncoding transcripts (IRT2 and iMLP1) was upregulated upon Rap1 depletion, while coding transcription decreased. Upon co-depletion of Rap1 and Sth1, we observed that IRT2 and iMLP1 levels were reduced in mRNAseq. While Sth1 depletion by itself did not affect IRT2 levels, levels of iMLP1 were already upregulated in Sth1-depleted cells ( Figure S1B). This suggests that RSC may repress divergent transcription at this locus. Indeed, the RPS10A gene showed increased divergent transcription in both Rap1-and Sth1-depleted cells ( Figure 1D).
Interestingly, the divergent transcription in Sth1-depleted cells initiated further upstream from the coding gene than in Rap1-depleted cells, suggesting that Rap1 and RSC limit divergent transcription through different mechanisms ( Figure 1D).
Next, we examined Rap1-regulated gene promoters more closely for changes in divergent transcription ( Figure 1E). As expected, Rap1-regulated gene promoters displayed increased levels of divergent transcripts upon Rap1 depletion ( Figure 1E).
In Sth1-depleted cells we also observed a notable increase in expression of divergent RNAs in nascent RNA-seq (see Figure 2C for quantification) and, to a lesser extent, in the mRNA-seq ( Figure 1E). The changes in divergent RNA transcription after Sth1 and Rap1 co-depletion were comparable to Rap1 depletion, suggesting that at Rap1-regulated genes in Rap1 depleted cells RSC does not further promote or repress divergent transcription. Notably, we observed increased RNA expression in the protein-coding direction in Rap1-depleted and Rap1/Sth1 codepleted cells ( Figure S1C). Indeed, it is known that Rap1 depletion can lead to inappropriate transcription initiation in both the protein-coding and divergent directions (Challal et al 2018). These data suggest that RSC depletion caused a relatively greater in divergent transcription compared to protein coding transcription at Rap1 regulated gene promoters, which we further explored in Figures 2 to 5.

RSC controls promoter directionality at Rap1 regulated promoters
While our analysis using nascent RNA-seq cannot determine absolute effects in transcription changes, it is possible to dissect relative changes. Promoter directionality score is a suitable relative measurement as it is defined by the ratio between transcription in the sense direction over the divergent direction ( Figure 2A) (Jin et al 2017). We computed the directionality score for tandem gene promoters, based on the TSS annotation of a dataset described previously (Park et al 2014). We focused our analysis on Rap1 regulated gene promoters (141 genes) and all tandem, non-overlapping gene promoters where the upstream gene is positioned in the tandem orientation (n=2609 promoters), to avoid confounding signals from divergent coding genes ( Figures 2B and S2). As expected, Rap1-regulated genes are highly directional when compared to all tandem genes ( Figure 2B). Upon Rap1 depletion, directionality was strongly reduced due to both the increased in the divergent transcription and a decrease in sense transcription ( Figures 2C and 2D).
Upon Sth1 depletion, the promoter directionality scores were reduced. Indeed, Sth1depleted cells displayed a relative increase in divergent transcription (Figures 2B and 2D). Thus, at the highly directional Rap1-regulated genes, RSC controls promoter directionality by a repressive effect on divergent transcription.
Employing the directionality analysis described in Figure 2A, we found that more than 900 displayed a relative increase (FC> 2) in divergent transcription, out of the 2609 tandem and non-overlapping genes ( Figure 3B and 3C). Consequently, more than 1000 gene promoters showed a decrease (FC> 2) in directionality score upon Sth1 depletion (STH1-AID + IAA) compared to WT or control cells (STH1-AID +DMSO) ( Figure 3C). Notably, control cells showed a decreased directionality score compared to the WT, indicating that the AID tag on Sth1 had a small negative impact on RSC function, possibly underestimating the overall effect of RSC ( Figure 3C, S3A and S3B). We conclude that RSC is important for limiting divergent noncoding transcription and promoting sense coding transcription at least one third of tandem gene promoters.

RSC-mediated nucleosome organization may limit TBP association upstream of directional promoters
To further explore RSC's role in controlling divergent transcription, we examined whether there are features that could explain the effect of RSC on promoter directionality. For example, it is known that RSC acts on highly transcribed gene promoters and coding sequences (Biernat et al 2021, Kubik et al 2015, Rawal et al 2018. Our data indicate that the ratio of divergent over sense transcription levels change upon Sth1 depletion, indicating that RSC's role in promoter directionality is, at least in part, mediated through controlling divergent transcription ( Figure 3B and S3A). We divided gene promoters into quintiles (Q1-Q5) based on directionality score in WT cells, wherein Q5 represents the group of promoters with the highest directionality score and Q1 the lowest. Subsequently, we computed the changes in divergent and sense transcription, as well as changes in directionality score for each quintile ( Figure 4A). Promoters with the highest directionality score (Q5) in WT cells showed the largest relative increase in divergent transcription after Sth1 depletion ( Figure 4A and S4A, left panel), suggesting that RSC-mediated repression of divergent transcription is more prominent at highly directional promoters (Figures 4A and S4A right panel). RSC depletion also affected directionality of gene promoters with the highest sense transcription levels (Q5, sense transcription in WT), albeit less compared to the promoters with highest directionality (Q5, directionality score in WT) ( Figure S4B). Control cells (STH1-AID +DMSO) displayed a marginal increase in divergent transcription in the most highly expressed and most directional promoters (Q5 sense transcription levels and Q5 directionality in WT, respectively) compared to WT cells, which was consistent with the observation that the AID tag partially affected RSC activity and had a small negative impact on promoter directionality ( Figures 3D and S4C). Thus, RSC mainly represses divergent transcription at promoters that are highly directional.
RSC plays a prominent role in positioning nucleosomes in promoters, which in turn affects PIC recruitment and TSS scanning (Hartley & Madhani 2009, Klein-Brill et al 2019. To investigate the mechanisms underlying these effects of RSC on promoter directionality, we assessed the chromatin structure Sth1 depleted cells ( Figure 4B and 4C). Specifically, we compared profiles of MNase sensitivity in cells treated with high and low amounts of MNase using a published dataset (Kubik et al 2015). The We observed the RSC-dependent differences in chromatin structure at individual gene promoters. For example, the PDE2 and YBR238C promoters display increased divergent transcription and a small but notable decrease in MNase-seq signal for the -1 nucleosomes in Sth1 depleted cells ( Figure 4B). Notably, the YBR238C promoter also harbours a fragile nucleosome ( Figure 4B, labelled "FN"). The broad effect of Sth1 depletion on nucleosome positioning (-2, -1, and +1) in directional promoters may explain why both divergent and sense transcription are often affected upon RSC depletion.
To determine whether the effect at the PDE2 and YBR238C gene promoters renders genome-wide changes, we analysed nucleosome positioning in the promoter classes sorted by directionality score. First, we found that the group of gene promoters with the highest directionality score (Q5, WT) displayed more defined nucleosome peaks, narrower in width at the +1, -1 and -2 positions compared to gene promoters with the lowest directionality score (Q1) ( Figure 4C, left panel). The -1 and -2 nucleosomes were present in the regions where divergent transcription initiates, while +1 nucleosome was present in the region where transcription initiation from the coding TSS occurs. Second, upon depletion of Sth1, nucleosome positioning became broader and less defined for the -1 and -2 positions, while the +1 nucleosome shifted more upstream. Third, the +1, -1 and -2 nucleosomes were more sensitive to MNase concentration in Q5 promoters than Q1 promoters ( Figure 4C, right panel), suggesting a prevalence for partially unwrapped nucleosomes in this group. We also compared the effect of promoter directionality on a subset of promoters with comparable sense coding gene transcription levels ( Figure S4D and S4E). This analysis revealed that the upstream -1 nucleosome is more defined and sensitive to MNase concentrations for the promoters with the highest directionality score (highest 20%) compared to group of promoters in the lowest 20% directionality score group within this subset of promoters with comparable coding gene transcription, further supporting the idea that positioning of upstream nucleosomes mediates promoter directionality ( Figure S4E).
The increase in divergent transcription, as observed in Sth1 depleted cells, is possibly a consequence of altered PIC formation at divergent promoters. We assessed whether association of TATA-binding protein (TBP) was affected upon Sth1 depletion in the group of gene promoters with highest directionality (Q5) using a published TBP ChIP-seq dataset . Despite that the ChIP-seq data does not have high spatial resolution, we found that upon Sth1 depletion, TBP binding was less affected, perhaps slightly increased, in the region approximately 400 bp upstream of the coding gene TSS which overlaps with the region where core promoters of divergent transcripts are found ( Figure 4D). TBP enrichment near the coding TSS decreased, which is expected because RSC promotes TBP recruitment to coding gene promoters (Kubik et al 2015).
Lastly, we examined whether DNA sequence motifs associated with RSC activity are differentially enriched and/or distributed between promoters with low and high directionality. Both polyA stretches and the CG(C/G)G motif are associated with RSC binding (Badis et al 2008, Krietenstein et al 2016, Lorch et al 2014. We examined the distribution of these sequences in both the sense and antisense strand ( Figure 4E and S4F). We found that polyA stretches and the CG(C/G)G motif were enriched in the divergent antisense direction in the group of promoters with the highest directionality (Q5) compared to the group of promoters with lowest directionality (Q1) (Table S1). Interestingly, the A track on the antisense strand showed two peaks in promoters with highest the directionality, possibly indicating that multiple RSC localisations may facilitate the positioning of FNs and -1 nucleosomes at highly directional promoters ( Figure 4E). Altogether, these data further support a model where RSC is recruited to distinct positions to promote positioning of upstream nucleosomes in promoters, thereby limiting divergent transcription initiation. The highly positioned nucleosomes, in turn, can form physical barriers which inhibit recruitment of the PIC and RNA Pol II.

RSC and GRFs are enriched at highly directional promoters
In addition to nucleosome remodeling by RSC, GRFs such as Rap1 are key players for controlling promoter architecture and regulating gene transcription (Brahma & Henikoff 2019, Kubik et al 2017. To assess how RSC and the GRFs Abf1 and Reb1 associate with respect to promoter directionality, we analysed a published dataset and compared groups of gene promoters with the lowest or highest directionality score (Q1 vs Q5, Figure 4A) (Brahma & Henikoff 2019). In line with the role of RSC in controlling divergent transcription, we found that RSC was more enriched in highly directional gene promoters (compare Q5 to Q1, Figure 5A and S5). Moreover, we found that Abf1 and Reb1 were also more enriched for the group of promoters with the highest directionality score (compare Q5 to Q1, Figure   5A and S5). This suggests that, apart from Rap1, other GRFs may contribute to repressing divergent transcription. To further investigate this hypothesis, we examined a published dataset that measured RNA crosslinked to RNA Pol II (RNA Pol II CRAC) upon Reb1 depletion (Candelli et al 2018b, Zentner et al 2015. We found that depletion of Reb1 increased levels of divergent RNA associated with Pol II notably at the BAP2 and CSG2 promoters ( Figure 5B).
Targeting GRFs or dCas9 to divergent core promoters is sufficient to repress divergent noncoding transcription.
Our analysis suggests that Abf1 and Reb1 can repress divergent transcription, possibly using a similar mechanism as Rap1 ( Figure 5A). As we previously demonstrated, Rap1 limits initiation of divergent noncoding transcription by occupying target motifs nearby or adjacent to cryptic divergent core promoters . To investigate whether Reb1, Abf1, and transcription factors (TFs) including GRFs can repress divergent transcription, we introduced ectopic transcription factor binding sites adjacent to a divergent core promoter. We used an established fluorescent reporter assay based on the PPT1 promoter which normally has a noncoding transcript SUT129 in the divergent direction, with both the coding and the divergent noncoding core promoters driving the transcription of fluorescent reporter genes encoding mCherry and YFP, respectively (Marquardt et al 2014. We cloned the binding site sequences of several transcription factors 20 base pairs (bp) upstream of the SUT129/YFP TSS in the PPT1 reporter ( Figure 6A).
We selected GRFs (Cbf1, Abf1, and Reb1) and TFs (Gal4, Gcn4, Cat8, Gcr1) that have well-defined DNA binding sequence motifs. To establish that the changes in the reporter signal were dependent on the transcription factor's presence in the cells and not solely due to the alteration in the underlying promoter DNA sequence, we also measured the reporter signal after deleting or depleting the same transcription factors. For example, for the reporter construct with Reb1 binding sites we measured YFP/mCherry levels in Reb1 control and depleted cells using the auxin induced degron (REB1-AID +IAA), whereas for the reporter construct with Cbf1 binding sites we measured the YFP/mCherry levels in WT and cbf1 cells.
Our data indicate that several GRFs were able to repress transcription in the divergent direction and increase promoter directionality, suggesting that they behave similarly to Rap1 ( Figure 6B) . Specifically, cbf1 cells and depletion of Abf1 and Reb1 (ABF1-AID or REB1-AID +IAA) resulted a reduced signal in divergent noncoding transcription compared to control cells ( Figure 6B). Introduction of binding site motifs for one transcription factor, Gcr1, did not affect noncoding (YFP) direction relative to the WT, but the YFP signal increased in gcr1 cells suggesting that Gcr1 also can repress divergent transcription ( Figure 6A and 6B).
The TFs that did not negatively affect the divergent noncoding transcript signal (Gal4, Gcn4, and Cat8) were possibly less active under the growth conditions where we performed the reporter assay ( Figure 6B). For example, Gal4 and Gcn4 are most active under galactose containing medium and under amino acid starvation, respectively. Notably, Gal4 and Gcn4 showed increased signal in the coding direction, suggesting that they are still somewhat active in the conditions tested but perhaps not sufficiently active for repressing divergent transcription ( Figure S6A).
We conclude that, as we reported previously for Rap1 , other transcription factors can repress divergent transcription when targeted to regions near a divergent gene promoter.
Our data indicate that RSC-mediated nucleosome positioning and GRFs can repress initiation of divergent transcription, and thereby promote directionality of gene promoters (Challal et al 2018. One explanation is that these proteins are barriers for PIC formation and Pol II recruitment. If true, then any protein stably associated with DNA near divergent TSSs should be able to interfere with divergent transcription. To test this, we targeted an exogenous protein to a divergent TSS. We used the catalytically inactivated version of S. pyogenes Cas9 (dCas9), which is a bulky protein (over 150 kDa) and has been widely used as tool to modulate transcription and chromatin . We induced divergent transcription by depleting Rap1 and used gRNAs to target dCas9 to the core promoter of IRT2, which is divergently transcribed from RPL43B and strongly induced upon Rap1 depletion, Figures 6C-E). As expected, Rap1 depletion induced expression of IRT2 ( Figure 6D and 6E). Importantly, we did not detect IRT2 expression when dCas9 was targeted to the IRT2 TSS ( Figure 6F, 6G and S6B). Expression of IRT2 was still detectable in cells where dCas9 was targeted to the TEF1 the promoter or upstream in the MLP1 promoter (iMLP1) ( Figure 6G and S6B). We conclude that steric hindrance by DNA-binding proteins (such as dCas9) is sufficient for repression of divergent transcription.

Discussion
Transcription from promoters is bidirectional, leading to sense coding and divergent noncoding transcription. However, the directionality (sense over divergent transcription) of promoters greatly varies indicating that there are mechanisms in place to control directionality of promoters (Jin et al 2017). Why some promoters are more directional than others is not well understood. Here, we showed that RSC mediated nucleosome positioning represses divergent noncoding transcription.
Furthermore, we provided evidence that a wide range of DNA-binding proteins, including nucleosomes, GRFs, and dCas9, are capable of repressing divergent transcription. We propose that such RSC-positioned nucleosomes and other DNAbinding factors form physical barriers at promoters which prevent transcription in the divergent direction, thereby promoting promoter directionality.

RSC-mediated nucleosome positioning represses divergent transcription
The RSC complex largely contributes to generating NDRs by 'pushing' the nucleosomes positioned at +1 and -1 positions 'outwards', widening the NDR (Ganguli et al 2014, Hartley & Madhani 2009, Parnell et al 2008. RSC also binds directly to DNA via A-track sequence motifs (Krietenstein et al 2016, Lorch et al 2014. How does RSC mediate the position of the -1 nucleosome to repress divergent transcription? Interestingly, we found that A-tracks were abundant in promoters on both strands indicating that directional positioning of nucleosomes upstream in promoters may contribute to preventing divergent transcription. RSC also promotes the partial unwinding of nucleosomes, which has been also referred to as "fragile" nucleosomes as these nucleosome particles are sensitive to MNase concentration (Brahma & Henikoff 2019. These sub-nucleosome complexes are also bound by GRFs. We found that at directional promoters the -1 nucleosome and FN are more defined compared to promoters with less directionality, suggesting that the formation of sub-nucleosome complexes controls promoter directionality ( Figure 6H). We further found that depletion of RSC resulted in less defined consensus nucleosome patterns and reduced sensitivity to MNase. We propose that nucleosomes strongly positioned by RSC inhibit PIC formation and Pol II recruitment for the divergent direction ( Figure   6C). Upon RSC depletion, the positioning of the -1 nucleosome and FN is less strong and PIC formation can occur.  6H). How other GRFs act with RSC in controlling divergent transcription remains to be dissected further. Based on our analysis, it appears more likely that RSC and GRFs act together to repress divergent transcription and promote transcription in the protein-coding direction.

Model for promoter directionality
Nucleosomes form barriers for transcription. In the context of repressing divergent transcription, chromatin assembly factor I (CAF-I) plays a widespread role in yeast, demonstrating that chromatin assembly upstream in promoters is important in repressing divergent transcription (Marquardt et al 2014). On the other hand, opening of chromatin in promoters increases divergent transcription. For example, increased histone lysine 56 acetylation (H3K56ac) leads to more divergent transcription in yeast (Marquardt et al 2014). Conversely, deacetylation of histone H3 by Hda1 deacetylase complex (Hda1C) facilitates repression of divergent transcription (Gowthaman et al 2021). The role of RSC at promoters is likely different from CAF-I, H3K56ac, and Hda1C. We propose that RSC-mediated positioning of NDR-flanking nucleosomes in promoters generates barriers for limiting RNA Pol II recruitment and initiation of divergent transcription. In line with this, differential TSS usage has been observed in RSC-depleted cells (Kubik et al 2019).
In a previous work we showed that Rap1 can repress divergent transcription at gene promoters likely by occupying cryptic divergent promoters . Our new analysis suggests that other GRFs have potentially the same ability in a model system. Moreover, we found that Abf1 and Reb1 were more enriched at highly directional promoters, suggesting that these GRFs also perform this role in vivo.
dCas9 has the same capacity to repress aberrant noncoding transcription when targeted to divergent core promoters. We propose that proteins physically interfere with divergent transcription when bound to DNA. With this view, RSC activity is important to position crucial nucleosomes and to ensure that cryptic divergent promoters are 'protected' and transcriptionally repressed. Thus, GRFs and positioned nucleosomes constitute essential components for promoter organization, which promote sense coding transcription and limit divergent noncoding transcription ( Figure 6H). While our studies focussed on tandem gene promoters. It is worth nothing that sequence specific DNA binding factors, such as Tbf1 and Mcm1, have the ability insulate two independently regulated divergent gene promoter pairs from each other, thus effectively repress divergent transcription (Yan et al 2015).
Additionally, in mammalian cells the multifunctional transcription factor CTCF directly represses initiation divergent noncoding transcription at hundreds for promoters, indicating that the mechanism of DNA binding factors repressing divergent transcription is likely conserved and widespread (Luan et al 2021).
RSC is part of the SWI/SNF family of chromatin remodellers related to the mammalian PBAF and BAF complexes (Mohrmann & Verrijzer 2005). The BAF complex in mouse embryonic stem cells (esBAF), has related functions in repressing noncoding transcription to some extent (Hainer et al 2015). Like RSC, esBAF positions the nucleosomes flanking the NDR. Depletion of esBAF leads to less stable nucleosome positioning, and consequently to widespread increased noncoding transcription, of which some are divergent noncoding transcription events. We anticipate that the role of RSC and BAF complexes in controlling divergent transcription and promoter directionality is likely conserved.
Further work will be required to dissect the interplay between GRFs and chromatin states in controlling aberrant noncoding transcription and promoter directionality.
These regulatory complexes are present and conserved across eukaryotes. Thus, our study may provide important insights on how promoter directionality is controlled in multi-cellular organisms.

Strains and plasmids and growth conditions
Strains isogenic to the Saccharomyces cerevisiae BY4741 strain background (derived from S288C) were used for this investigation. A one-step tagging procedure was used to generate endogenous carboxy (C)-terminally tagged auxin-inducible degron (AID) alleles of Rap1 and Sth1 (Nishimura et al 2009). The RAP1-AID strain was previously described, which contains three copies of the V5 epitope and the IAA7 degron . For tagging RPB3 with the FLAG epitope, we used a one-step tagging procedure using a plasmid (3xFLAG-pTEF-NATMX-tADH1, gift from Jesper Svejstrup). The single copy integration plasmids for expression of Oryza sativa TIR1 (osTIR1) ubiquitin E3 ligase were described previously .  Table S2, S3, and S4, respectively.

RNA extraction
Yeast cells were harvested from cultures for RNA extraction by centrifugation, then washed once with sterile water prior to snap-freezing in liquid nitrogen. RNA was

RT-qPCR
Reverse transcription and quantitative PCR for IRT2 performed as described previously (Tam & Table S4.

Western blot
Western blots were performed as described previously (Chia et al 2017, Chiu et al 2018. Protein extracts were prepared from whole cells after fixation with trichloroacetic acid (TCA). Samples were pelleted by centrifugation and incubated with 5% w/v TCA solution at 4 °C for at least 10 minutes. Samples were washed with acetone, pelleted, and dried. Samples were then resuspended in protein breakage buffer (50 mM Tris (pH 7.5), 1 mM EDTA, 2.75 mM dithiothreitol (DTT)) and subjected to disruption using a Mini Beadbeater (Biospec) and 0.5 mm glass beads.
SDS-PAGE (polyacrylamide gel electrophoresis) was performed using 4-20% gradient gels (Bio-Rad TGX) and samples were then transferred onto PVDF membranes by electrophoresis (wet transfer in cold transfer buffer: 3.35% w/v Tris, 14.9% w/v glycine, 20% v/v methanol). Membranes were incubated in blocking buffer (1% w/v BSA, 1% w/v non-fat powdered milk in phosphate buffered saline with 0.01% v/v Tween-20 (PBST) buffer) before primary antibodies were added to blocking buffer for overnight incubation at 4 °C. For probing with secondary antibodies, membranes were washed in PBST and anti-mouse or anti-rabbit IgG horseradish peroxidase (HRP)-linked antibodies were used for incubation in blocking buffer (1 hour, room temperature). Signals corresponding to protein levels were detected using Amersham ECL Prime detection reagent and an Amersham Imager 600 instrument (GE Healthcare).

Antibodies
The following antibodies were used for western blotting.

Nascent RNA sequencing (Nascent RNA-seq)
For nascent RNA sequencing (nascent RNA-seq), RNA fragments associated with RNA polymerase II subunit Rpb3 endogenously tagged with 3xFLAG epitope at the C-terminus were isolated by affinity purification as described previously (Churchman & Weissman 2011;, Moretto et al 2021. Small batch cultures of yeast cells grown in YPD media were collected by centrifugation, the supernatant was aspirated, and cell pellets were immediately snap-frozen by submerging in liquid nitrogen to minimise changes in nascent transcription activity (e.g. in response to cell resuspension in cold lysis buffer with high concentrations of salts and detergents).
Frozen cell pellets were dislodged from centrifuge tubes and stored at -80 °C. Cells were subjected to cryogenic lysis by freezer mill grinding under liquid nitrogen (SPEX 6875D Freezer/Mill, standard program: 15 cps for 6 cycles of 2 minutes grinding and 2 minutes cooling each). Yeast "grindate" powder was stored at -80 °C.

RNA-seq data analysis
Adapter trimming was performed with cutadapt (version 1.
Adapters were trimmed from MNase-seq reads using cutadapt as described above.

Stratification of promoters and sequence analysis
Gene promoters were stratified in quintiles (Q1 to Q5) using Matt (

Data plotting and visualisation
Bar plots, scatter plots, and volcano plots were generated using GraphPad Prism

Publicly available datasets used in this study
MNase-seq and TBP-ChIP seq data were obtained from GSE73337 (

Competing Interests
The authors declare no competing interests.

Data Availability
The accession number for the RNA sequencing data reported in this paper is GSE179256. Table S1 to S4

Data Files
Gel scans, contains the western and northern blot scans used for assembling the figures in this study. Promoters were clustered on antisense strand signal using k-means clustering (k = 3, (c1, c2, and c3)) based on previous analysis for Rap1-regulated gene promoters . Differences with respect to the DMSO control for each depletion strain are displayed.        Same analysis as in Figure 5A, except that the two biological repeats are displayed.
Metagene analysis and heatmaps of CUT&RUN data of Sth1 (RSC), and the GRFs Abf1 and Reb1 (top). In addition, signals of Sth1 CUT&RUN followed by histone H2B immunoprecipitation are displayed. The data were obtained from (Brahma & Henikoff 2019). Compared Q1 (blue) being the group of gene promoter with lowest directionality score, and Q5 (green) being the group with the highest directionality score. Corresponding heatmaps for Q1 and Q5 are displayed (bottom).     Figure S3 A B sense div. sense div. sense div.      Wu et al. Figure 6 -+ -+ E