The Paf1 complex factors Leo1 and Paf1 promote local histone turnover to modulate chromatin states in fission yeast

Abstract

The maintenance of open and repressed chromatin states is crucial for the regulation of gene expression. To study the genes involved in maintaining chromatin states, we generated a random mutant library in Schizosaccharomyces pombe and monitored the silencing of reporter genes inserted into the euchromatic region adjacent to the heterochromatic mating type locus. We show that Leo1–Paf1 [a subcomplex of the RNA polymerase II‐associated factor 1 complex (Paf1C)] is required to prevent the spreading of heterochromatin into euchromatin by mapping the heterochromatin mark H3K9me2 using high‐resolution genomewide ChIP (ChIP–exo). Loss of Leo1–Paf1 increases heterochromatin stability at several facultative heterochromatin loci in an RNAi‐independent manner. Instead, deletion of Leo1 decreases nucleosome turnover, leading to heterochromatin stabilization. Our data reveal that Leo1–Paf1 promotes chromatin state fluctuations by enhancing histone turnover.

Introduction

Boundaries between different chromatin states must be maintained for stable gene expression patterns 1, 2. Although many different chromatin states have been described, the two most fundamental categories are active euchromatin and silent heterochromatin 3. Constitutive heterochromatin is associated with H3K9me2/3, HP1, and low histone turnover 4, 5. Although generally inactive, heterochromatin may be transcribed during defined periods of the cell cycle, but the resulting transcripts are degraded 6, 7, 8. The fission yeast Schizosaccharomyces pombe uses several alternative heterochromatin formation pathways in different regions that may substitute for one another. The RNAi pathway, which involves the proteins Dcr1 and Ago1, is the predominant mechanism used to nucleate heterochromatin 9, 10. RNAi‐independent heterochromatin formation depends on transcription and RNA surveillance by factors such as Mlo3‐TRAMP 11. The constitutive heterochromatin regions in S. pombe are large and consist of well‐defined regions where nucleation occurs. The heterochromatin subsequently spreads from these regions 12, first by propagation of the H3K9me2/3 mark and then via the binding of HP1^Swi6 and deacetylation of H4K16 by Sir2 13, 14. The binding of HP1^Swi6 is dynamic. As RNA polymerase II (RNAPII) transcribes through H3K9me2, HP1^Swi6 sequesters the nascent RNA and delivers the RNA to the TRAMP component Cid14 for degradation 8, 15. HP1^Swi6 binds to H3K9me2/3 and recruits the H3K9 methyltransferase Clr4 to act on adjacent nucleosomes. This results in a chain reaction with self‐propagation of heterochromatin via histone modification.

The spreading of heterochromatin stops at any of the defined boundary elements. In S. pombe, the tRNA barriers that surround the pericentric repeats, the inverted repeats (IR) that surround the mating type locus and most centromeres, and the rDNA at the telomeres on chromosome 3 are three different examples of boundaries between euchromatin and heterochromatin. However, the boundaries of facultative heterochromatin islands and telomeres are less well understood. TFIIIC is associated with most boundaries, including both RNAPIII‐transcribed tRNAs and RNAPII‐transcribed IR elements 16. Several factors have been identified that maintain the chromatin boundaries at IR elements (notably Epe1, Mst1, and Bdf2) 17, 18. Epe1 physically interacts with Bdf2, which in turn binds to the acetylated H4 tails and protects them from deacetylation by Sir2 18. H4K16ac is added by Mst1 18, and H4K16ac and H3K9ac are removed from the heterochromatin by Sir2 14, 19, 20. To date, Epe1 is the only protein that has been described to possess anti‐silencing activity in S. pombe chromatin. Epe1 is recruited to heterochromatin 21 but is degraded from the interior through Ddb1 22. However, Epe1 also plays a role at the heterochromatin boundaries that is independent of H4K16ac and RNAi 18, 23.

In addition to fission yeast, RNAi‐mediated heterochromatin formation has also been described in plants and Drosophila. However, other pathways of heterochromatin formation appear to play a role in metazoans 24. Alternative mechanisms of heterochromatin nucleation also exist in S. pombe 15, 25, 26, 27, 28. For example, facultative H3K9me2/3 heterochromatin formation at meiotic genes is RNAi‐independent and requires transcription and a complex consisting of Red1, Rrp6, and Mmi1 29, 30, 31, 32. These H3K9me2/3 islands are dynamically regulated in response to environmental signals 30. Red1 is a cofactor for both RNAi‐ and exosome‐mediated RNA degradation and degrades meiosis‐specific mRNAs 33. This type of silent chromatin is characterized by the presence of both H3K9me2/3 and HP1^Swi6. Additional distinct types of silent chromatin form over developmental genes and retrotransposons (“HOODs”) 7 and the central domain of the centromere where the RNA is rapidly degraded 6, 34. In contrast to human cells in which LTR retrotransposons (HERVs) are embedded in H3K9me3‐associated heterochromatin, the S. pombe LTR retrotransposon Tf2 is not associated with H3K9me2/3 but is readily transcribed by RNAPII. However, the nascent Tf2 transcripts are targeted to the exosome independent of Red1 35. The siRNA pathway may take over in the event of failure of these posttranscriptional mechanisms to degrade the Tf2 transcripts, leading to the formation of H3K9me2 7. Heterochromatin is strictly confined to the HOOD regions even in the absence of known boundary elements.

There are clearly gaps in our knowledge in terms of how a boundary controls the spread of chromatin states, especially in the absence of tRNA barriers. In this study, we set out to uncover mechanisms that determine heterochromatin confinement. We identified Leo1 and Paf1 as factors that reverse chromatin silencing at IR border regions and heterochromatin islands. Both factors are part of the Paf1C. Additionally, we found that Leo1 acted by destabilizing nucleosomes in a general manner and thereby modulating chromatin states.

Results

Paf1C components inhibit heterochromatin propagation across the IR‐L boundary element

To identify factors required for heterochromatin boundary maintenance in S. pombe, we applied a random mutagenesis assay and screened the resulting mutants for the silenced reporter genes ade6 ⁺ and ura4 ⁺, which were integrated at regions of normally open chromatin. The ade6 ⁺ reporter gene was positioned at the inverted repeat boundary to the left (IR‐L) of the silent matK region, whereas the ura4 ⁺ gene was integrated 1.2 kb further into the euchromatic matL region (Fig 1A). The cells show no phenotype when grown on complete media (YEA or YES) under non‐selective conditions. However, the cells will be resistant to 5‐fluorotic acid (5FOA) when ura4 ⁺ is silenced 36, and the colonies will be red when grown on media containing a low concentration of adenine (YE) when ade6 ⁺ is silenced. Previously, Epe1 was demonstrated to prevent heterochromatin spread across this boundary 17. We used the Hermes transposon mutagenesis system to generate mutants 37 (Fig EV1A) and isolated a clone that was both 5FOA‐resistant and red on low‐adenine (YE) plates; thus, this mutant had an extended silenced region surrounding the matK region (Fig 1B). Upon sequencing of the DNA flanking the Hermes integration site, the clone was found to contain the transposable element inserted within the SPBC13E7.08c (leo1 ⁺) open reading frame 507 nucleotides from the start codon. We confirmed a single integration site of the Hermes retrotransposon by Southern blotting (Fig EV1B). As expected, the heterochromatin mark H3K9me2 was found in the IR‐L and matL regions in the leo1∆ and leo1::Hermes strains as determined by chromatin immunoprecipitation (Fig 1C), demonstrating that the effect was due to de facto heterochromatin formation and not merely posttranscriptional degradation of ura4 ⁺ and ade6 ⁺. To rule out secondary effects of the transposon integration, we constructed an HA‐tagged version of the truncated Leo1^1–169 protein by PCR‐directed mutagenesis of the leo1 ⁺ gene, thereby mimicking the transposon‐induced truncation. Also this strain had H3K9me2 at the IR‐L and matL regions. Two regions are conserved between the S. pombe and human Leo1 proteins that correspond to amino acids (aa) 16–46 and 87–253 in S. pombe Leo1 (Fig 1D). The Leo1^1–169 protein contains a conserved N‐proximal domain that may be responsible for H3 binding and almost the entire conserved Paf1 interaction domain (aa87–176) 38. Schizosaccharomyces pombe Leo1 is physically and functionally associated with Paf1 39. The C‐proximal conserved region at aa176–253 has an unknown function. Interestingly, the disrupted leo1 ⁺ alleles (leo1::Hermes and leo1 ^1–169 ::HA) exhibited a severe growth defect not present in the full gene deletion (leo1∆), suggesting a dominant effect of the truncated protein. The minor phenotypic effects of leo1∆ on growth and morphology were noted previously 39. The strain with the leo1::Hermes allele and the strain with the truncated Leo1^1–169 grew more slowly than the WT and leo1∆ (Fig 1E) and also displayed a distinct cellular morphology (Fig 1F).

Figure 1. Identification of Leo1 in a screen for factors counteracting heterochromatin spreading across the IR‐L boundary.

1. Scheme of the mating type region in Schizosaccharomyces pombe. Integration positions of reporter genes (ura4 ⁺ or ade6 ⁺) are indicated.
2. Spotting assay using strains with ura4 ⁺ and ade6 ⁺ integrated as in (A) on plates with added adenine (YEA), with low adenine (YE), or with 5‐fluorotic acid (5FOA).
3. ChIP–qPCR analysis of H3K9me2 levels at the ura4 ⁺ or ade6 ⁺ insertion sites depicted in (A). ChIP was performed in three independent experiments (n = 3); error bars indicate the standard deviation (SD).
4. Schematic representation of the protein domains of Leo1; Sp: S. pombe; Hs: human. The different colors represent conserved domains: orange: Paf1 interaction; green: conserved, potentially H3 interacting; pink: conserved region of unknown function, potentially RNA binding.
5. Growth curves of WT cells and cells with different leo1 ⁺ alleles (as indicated).
6. Bright‐field microscopy images of the indicated strains.

Figure EV1. The transposon mutagenesis system.

• A
  Cartoon of the Hermes transposon screen.
• B
  Southern blot of the Hermes strain. Probes were complementary to the kanMX region (top panel) present in both leo1::Hermes and leo1∆, leading to a band of length 7–8 kb (Hermes‐kanMX, lower panel, upper arrow) and a band of length 4–5 kb (leo1∆::kanMX) (lower panel, lower arrow).
• C–E
  The Ddb1 and epe1::GFP effect on other components of Paf1C. Genetic interaction between ddb1 and cdc73 and tpr1. (C) Spotting assay with ura4 ⁺ at its endogenous locus. (D) Spotting assay with strains in ddb1∆ background. (E) Spotting assay in epe1 ⁺ ::GFP‐derived strains.

Leo1 and Paf1 mediate chromatin state fluctuation but not the entire Paf1C

Leo1 is a conserved protein that participates in transcription elongation by RNAPII. The protein is a component of the similarly conserved RNA polymerase II‐associated factor 1 complex (Paf1C) 40, 41. In S. pombe, Paf1C is minimally composed of four subunits: Paf1, Leo1, Cdc73, and Tpr1 39 (Fig 2A). Given the role of Leo1 in Paf1C, we compared the effects on IR‐L silencing following the deletion of genes encoding the other components of the protein complex. In addition to leo1∆, paf1∆ enhanced silencing within IR‐L and extended the silent domain into the matL region. However, deletion of tpr1 ⁺ or cdc73 ⁺ had little or no effect on chromatin silencing (Fig 2B). These observations demonstrate that the Leo1–Paf1 heterodimer rather than the complete Paf1C prevents the propagation of the repressed state across the IR‐L boundary element.

Figure 2. Spreading of heterochromatin across IR‐L is specifically affected by Paf1 and Leo1.

• A
  Model of Paf1C with its subunits and interactors.
• B
  Spotting assay using four strains with ura4 ⁺ and ade6 ⁺ integrated as in Fig 1A, carrying deletions of genes encoding different subunits of the Paf1 complex (as indicated).
• C
  Position variegation effect by Leo1 and Paf1. Spotting assay of strains regrown from –Ura or 5FOA plates with the relevant genotypes indicated.
• D
  Immunoblot of H2Bub1 and H4 levels in WT, htb1‐K119R, and leo1∆ cells. Numbers indicate the H2Bub1/H4 ratio for each lane.
• E, F
  Growth assay performed on a low‐adenine plate (YE) (E) and a 5FOA‐containing plate (F) using mutants with ura4 ⁺ or ade6 ⁺ integrated as in Fig 1A. The mutants in this panel lack different factors in the H3K4me (set1∆) or H2Bub1 (rfp1∆, shf1∆, or htb1‐K119R) pathways.

Stability of the chromatin state

We noted irregular patterns that manifested as large and small colonies on 5FOA and red‐white sectoring on the low‐adenine plates for the leo1∆ and paf1∆ clones (Figs 1B and 2B), suggesting an unstable chromatin silencing phenotype. The propagation of the repressed state across IR‐L was observed in only a subset of cells within the colony. This phenotype has previously been reported in epe1 mutants 17 and is caused by position effect variegation (PEV) of the inserted ade6 ⁺ or ura4 ⁺ reporters. To test whether variegation could also be observed in the Leo1–Paf1 mutants, the cells were grown on non‐selective medium following selection (–Ura) and counterselection (5FOA) and then restreaked onto selective plates (Fig 2C). The results showed that once established, the repressed state of ade6 ⁺ and ura4 ⁺ in the matL region was stably maintained under non‐selective conditions for more than 30 generations in the leo1 and paf1 mutant backgrounds. These results demonstrate that Leo1 and Paf1 (similar to Epe1) function to destabilize the repressed chromatin state.

The effects of mutations that affect histone modifications and RNA processing

In addition to the intrinsic function of Paf1 and Leo1 in promoting RNAPII‐mediated chromatin transcription 40, these components of the Paf1 complex are involved in multiple activities that facilitate transcription elongation. In budding yeast, Paf1C interacts with Set1 to methylate Lys4 on histone H3 and with the ubiquitin ligase Bre1 to monoubiquitinate Lys120 on histone H2B 42, 43. In S. pombe, Set1‐mediated H3 methylation is conserved, and the corresponding H2B residue Lys119 is monoubiquitinated by the Bre1 homologs Brl1 and Brl2 44, 45. We confirmed a previous observation that H2Bub1 was partially dependent on Leo1 because H2Bub1 was reduced in leo1Δ (61% of WT levels), albeit not to the same extent as the H2Bub1‐deficient htb1‐K119R (Fig 2D).

To determine whether paf1Δ and leo1Δ indirectly caused heterochromatin propagation across IR‐L by impairing H2Bub1 or H3K4me, we examined cells with deficient H2B ubiquitination (rfp1Δ, shf1Δ, and htb1‐K119R) or H3K4 methylation (set1Δ) with regard to chromatin repression (Fig 2E and F). In contrast to paf1Δ or leo1::Hermes, the mutations negatively affecting H2Bub1 or H3K4me did not induce gene silencing across the heterochromatin boundary at IR‐L. This result is in agreement with a functional separation between the Leo1–Paf1 heterodimer and the rest of the Paf1 complex containing Rfp1 39. Together, these data indicate that although Paf1C is involved in recruiting factors to generate H2Bub1 and H3K4me, the role of the Leo1–Paf1 subcomplex in heterochromatin spreading across the IR‐L boundary is distinct from the functions of its known interactors. Thus, heterochromatin propagation is not related to the transcriptional marks H2Bub1 and H3K4me.

Leo1–Paf1 participates in Epe1‐mediated anti‐silencing activity that counteracts heterochromatin assembly

Because leo1::Hermes and paf1∆ promote silencing propagation across IR‐L in a similar manner as epe1∆ (Fig 1B) and Epe1 acts as an anti‐silencer 17, 21, 23, we hypothesized that Epe1 interacted genetically with the two Paf1C components to impair gene silencing. The Cul4‐Ddb1^Cdt2 ubiquitin ligase promotes the degradation of Epe1 within heterochromatin domains 22. Consequently, in ddb1∆ cells, Epe1 is not targeted for degradation and thus is stabilized. This stabilized Epe1 impairs heterochromatin silencing 22. To test whether Epe1‐dependent expression in the ddb1∆ mutants was dependent on Leo1 and Paf1, we determined the effects of Epe1 stabilization on paf1∆‐ and leo1∆‐dependent ura4 ⁺ silencing in the matK region. As in previous reports, we concomitantly deleted ddb1 ⁺ and its other known target spd1 ⁺ as not to confound data by cell cycle defects caused by stabilized Spd1 22. The silencing of this region was enhanced by both paf1∆ and leo1∆ as determined by growth on 5FOA (Fig 3A) and by measuring the levels of the heterochromatin mark H3K9me2 in the leo1∆ ddb1∆ strain at matK (Fig 3B) and the pericentric dhI region (Fig 3C) by ChIP. Endogenous ura4 ⁺ expression was not affected by the inactivation of Paf1 or Epe1 (Fig EV1C). The interaction between epe1 and leo1/paf1 was specific to the Leo1–Paf1 subcomplex because deletion of the other components of Paf1C (cdc73 ⁺ or tpr1 ⁺) did not rescue the ddb1∆ phenotype (Fig EV1D). These results demonstrate a genetic interaction between epe1 and leo1/paf1 that cooperatively counteracts chromatin silencing.

Figure 3. Leo1–Paf1 and Epe1 mediate chromatin anti‐silencing.

1. Scheme of the mat locus in S. pombe. Integration position of ura4 ⁺ is indicated (upper panel). Spotting assay on YEA and 5FOA plates (lower panel).
2. ChIP–qPCR across the XbaI site in the matK domain.
3. ChIP–qPCR across the pericentric dh repeat of centromere 1.

Data information: (B, C) ChIP was performed in three independent experiments; error bars indicate SD.

To rule out the contributions of other Ddb1 targets, we specifically increased the protein levels of Epe1 by adding a peptide tag to Epe1 that stabilized the protein 46. As expected, compromising the Leo1–Paf1 subcomplex by deletion of paf1 ⁺ or leo1 ⁺ or in cells with the leo1::Hermes allele reversed the silencing defect in an epe1::GFP strain (Fig EV1E), confirming the specific genetic interaction between epe1 and paf1/leo1.

Facultative heterochromatin islands are suppressed by Leo1 and Paf1

To determine whether Leo1 imposes a general anti‐silencing effect in the genome, we performed genomewide mapping of the H3K9me2 marks in WT and leo1∆ cells. The results with subnucleosomal resolution were obtained through histone ChIP–exo 5, 47. The constitutive heterochromatin at pericentric regions was contained in the leo1∆ cells, but H3K9me2 spreading could be observed at the mat locus and at telomeres to varying degrees (Fig 4A). In addition to these constitutive heterochromatin regions, some H3K9me2‐enriched regions appeared in the leo1∆ strain compared with the WT cells. These regions overlapped the previously identified islands of facultative heterochromatin to a large extent 30, including mcp7, mei4, and ssm4 (Fig 4B). Most of the genes at these islands are required for meiosis 30, 32.

Figure 4. Genomewide distribution of heterochromatin in leo1∆ cells.

1. A browser view of ChIP–exo signals in the WT and leo1∆ strains using an antibody against H3K9me2 showing the constitutive heterochromatin domains of chromosome 2 (indicated in red in the lower panel). The ratios of RPKM (reads per kilobase per million mapped reads) over the indicated region are shown. Two independent experiments were performed.
2. Browser views of H3K9me2 ChIP–exo signals in WT and leo1∆ cells at five facultative heterochromatin islands. The RPKM values are shown for each gene and strain.
3. ChIP–qPCR of H3K9me2 at Tf2s.
4. ChIP–qPCR of Swi6 in WT, leo1∆, and leo1::Hermes strains at four loci.
5. Spotting assay of WT and leo1∆ with ura4 ⁺ integrated in the Tf2‐3 locus. The top row shows a WT strain with functional endogenous ura4 ⁺; all other strains have the non‐functional ura4‐D18 allele.
6. Spotting assay using strains with ura4 ⁺ integrated close to the SPAC23H3.14 locus.
7. ChIP–PCR of H3K9me2 at the SPAC23H3.14 locus. Numbers indicate ratios for the signal SPAC23H13.14/leu1.

Data information: (C, D) ChIP was performed in three independent experiments; error bars indicate SD.

Next, we specifically examined the retrotransposable element Tf2, which is normally posttranscriptionally silenced by exosomal degradation 7. We observed a twofold increase and a shift in the profile of H3K9me2‐marked nucleosomes at these elements in the leo1∆ cells (Fig EV2A). We confirmed this increase in H3K9me2 by ChIP–qPCR at Tf2s (Fig 4C). As demonstrated by the partial suppression of leo1∆ by ddb1∆ in the matK region above, stabilizing Epe1 by removing ddb1 ⁺ (ddb1∆) in the leo1∆ background resulted in restoration of WT levels of H3K9me2 at the Tf2s. To confirm heterochromatin formation, we also quantified the change in Swi6 association at the Tf2 chromatin. Similar to the H3K9me2 data, the Swi6 levels at Tf2 were also increased in the leo1∆ cells. This result was in contrast to the similar levels of Swi6 detected at constitutive heterochromatin (dhI and matK) and the undetectable levels at the expressed spd1 ⁺ gene (Fig 4D). Silencing of the ura4 ⁺ reporter gene integrated into the Tf2‐3 locus further confirmed heterochromatin formation in the leo1∆ background (Fig 4E). The leo1∆ cells also exhibited the delayed chromatin state switch characteristic of PEV at the Tf2‐3 locus (Fig EV2B).

Figure EV2. Heterochromatin formation over retrotransposable element Tf2 pericentric region of chromosome 1.

1. ChIP–exo for H3K9me2 at Tf2.
2. Position variegation at Tf2 effect by leo1∆. Spotting assay on –Ura and FOA plates of strains previously grown on –Ura or FOA plates.
3. Spotting assay of indicated strains on YEA and 5FOA media with ura4 ⁺ integrated at the left pericentric region of chromosome 1.
4. Spotting assay of indicated strains on YEA and 5FOA media with ura4 ⁺ integrated at the right pericentric region of chromosome 1.

To further characterize the contribution of Leo1–Paf1 to facultative heterochromatin assembly, we compared the effects of leo1::Hermes and paf1∆ on the expression of ura4 ⁺ from an ectopic locus adjacent to SPAC23H3.14. Expression of this locus has been reported to be dependent on Epe1 activity 21, 46. The expression of this locus was partially dependent on Leo1 because leo1::Hermes inhibited the expression of ura4 ⁺ from SPAC23H3.14 (Fig 4F). Remarkably, while leo1::Hermes and epe1∆ had similar effects on gene silencing and the H3K9me2 levels at SPAC23H3.14, leo1∆ and paf1∆ had little to no effect on 5FOA sensitivity. To determine whether the repressive effects of leo1::Hermes reflected a change in chromatin structure at SPAC23H3.14, we conducted ChIP analysis with anti‐H3K9me2 antibodies. The results show an elevated level of H3K9me2 association with SPAC23H3.14 in the leo1::Hermes strain (Fig 4G).

In summary, Leo1–Paf1 (similar to Epe1) suppresses facultative heterochromatin formation at heterochromatin islands, meiotic genes, and transposable elements. Leo1 also functions to confine heterochromatin at the mat locus and telomeres but plays a minor role in the formation or spreading of constitutive pericentric heterochromatin.

Paf1C regulates RNAi‐independent chromatin silencing at pericentric heterochromatin domains

In contrast to heterochromatin formation at the mat locus and facultative heterochromatin islands, which are predominantly RNAi‐independent 25, 32, heterochromatin formation at the pericentric domains is severely impaired by inactivation of the RNAi machinery 9. First, we investigated whether leo1::Hermes (similar to epe1∆ 4, 23) could rescue the ago1∆ heterochromatin defect at pericentric loci. Pericentric heterochromatin is important to ensure proper segregation of the chromosomes. As a consequence, cells that do not form pericentric heterochromatin are sensitive to the microtubule destabilizing drug thiabendazole (TBZ), as observed in ago1∆ cells (Fig 5A). The double mutants ago1∆ leo1::Hermes and ago1∆ paf1∆ reduced TBZ sensitivity. A related readout for pericentric heterochromatin loss is chromosome segregation defects. We can monitor the chromosome segregation process by measuring the mitotic stability of the non‐essential minichromosome Ch16. The high level of Ch16 loss in ago1∆ cells, as determined by both the completely red colonies and red‐white sectored colonies, was reversed by the leo1::Hermes allele (Fig 5B).

Figure 5. Leo1 aggravates the effects of siRNA loss.

1. Spotting assay to test TBZ sensitivity of the indicated strains. Tenfold serial dilutions of the indicated cultures were grown on rich medium (YEA) in the presence (17 μg/ml) or absence of TBZ.
2. Loss of minichromosome Ch16. Two independent experiments were performed; error bars show the range.
3. Representation of centromere 1 in S. pombe. Integration sites of ura4 ⁺ are indicated with a blue vertical arrow (upper panel). Spotting assay of strains with the relevant genotypes indicated (lower panel).
4. Spotting of strains lacking HP1 (swi6∆) or SHREC (clr3) with ura4 ⁺ integrated at the imr1.
5. Spotting of strains lacking FACT (pob3∆) with ura4 ⁺ integrated at the imr1.

By combining the reporter gene ura4 ⁺ integrated at different positions in the left pericentric region of centromere 1 with leo1 mutations, we demonstrated that the heterochromatin defects observed in the RNAi‐deficient mutants were rescued by the leo1::Hermes allele as heterochromatin was restored (Fig 5C). Interestingly, the complete deletion of leo1 ⁺ only resulted in a partial rescue, again stressing the previous observation that the truncated version of Leo1 had a more severe phenotype (Fig EV2C). Surprisingly, leo1::Hermes and epe1∆ did not restore the ago1∆ heterochromatin defect at the otr1R(Sph1) locus (Fig EV2D) in contrast to the results from the left pericentric region.

In contrast to the independence of the RNAi machinery, the loss of Leo1 did not bypass the requirement for other complexes (i.e., HP1 proteins, SHREC, or FACT), as determined by crossing the mutants with swi6∆, clr3∆ (Fig 5D), or pob3∆ (Fig 5E), respectively. Thus, inactivation of Leo1 and Paf1 (similar to the inactivation of Epe1) promoted RNAi‐independent heterochromatin formation.

The distribution of small RNA in WT and leo1∆ cells

Next, we determined the effects of the leo1 genotype on the distribution of sRNA. To this end, small RNA (sRNA) populations were isolated and mapped to the genome. The sRNA mapping to the pericentric dg/dh repeats (otr) constituted the fourth largest sRNA population (8.4%) in the WT cells, representing transcription in both directions. This population was severely diminished in the Leo1‐deficient cells (1.3%) (Fig 6A). In both the WT and leo1∆ cells, a small proportion of sRNAs mapped to the subtelomeric regions and then only to the tel1L‐ and tel2R‐containing centromere‐like repeats and tlh1/2. Few reads mapped to the facultative heterochromatin islands or the regions surrounding the mat locus in either the WT or leo1∆ strains (Fig EV3A and B). These results are consistent with the indication that heterochromatin stabilization in leo1‐deficient cells is not mediated by RNAi, at least at the pericentric region.

Figure 6. Small RNAs generated from heterochromatin are strongly reduced but still present without Leo1.

1. Pie charts illustrating proportions between different sRNA populations in WT and leo1∆.
2. sRNAs mapped across centromere 3. The number of reads from two independent experiments was normalized to the reads mapping to tRNAs. Red dots in the upper panel indicate positions of tRNAs.
3. Growth of strains with ade6 ⁺ integrated at IR‐L on a YE (low‐adenine) plate.

Figure EV3. Small RNAs and heterochromatin at telomeres and centromeres.

• A
  sRNA distribution at tel1L and the centromere of chromosome 1.
• B
  sRNA distribution at tel2L, the centromere of chromosome 2, the mat region, and tel2R.
• C
  Distribution of sRNA populations in ago1∆.
• D–F
  H3K9me2 ChIP–exo over centromere 1 (D), centromere 2 (E), and centromere 3 (F). Green boxes indicate the location of the BORDERLINE transcript 2013 and the IRC3 elements 2008.
• G
  Spotting of strains with ura4 ⁺ at the euchromatic (inner centromeric) side of tRNA barrier at cen1R.
• H, I
  H3K9me2 ChIP–qPCR over the IRC1R locus (H) and the outer repeats (otr2) and central core (cnt2) of centromere 2 (I). Experiments were done in three independent experiments; error bars show the SD.

Data information: (A–F) Two independent experiments were performed, and the profiles show the average.

The bulk of the sRNAs (54% in WT and 78% in leo1∆) were single‐stranded and mapped to the rDNA locus. The relative fractions of RNAPIII‐transcribed tRNA and gene populations were decreased in leo1∆, possibly due to the increased rRNA. We also isolated sRNAs in an RNAi‐deficient strain (ago1∆) (Fig EV3A–C). The ago1∆ results demonstrated that only the otr and subtelomeric sRNAs were RNAi‐dependent.

Surprisingly, because the major loci of new heterochromatin in leo1∆ cells were outside the pericentric regions, the most substantial changes in sRNA populations mapped to the pericentric regions, and especially to the IR boundaries (Fig 6B). Although sRNA transcription from the IRC boundaries was affected, the spreading of pericentric heterochromatin was limited in cells without fully functional Leo1 (Fig EV3D–F), suggesting that the tRNA borders were unaffected. We tested the integrity of the tRNA boundary and found that it was intact in leo1‐deficient cells (Fig EV3G). The right pericentric border of centromere 1 lacks tRNAs and relies solely on the IRC element 29, 48. A slight increase in H3K9me2 at this IRC1R element was observed (Fig EV3H).

The fact that the pericentric heterochromatin levels were not reduced in leo1∆ but conversely were somewhat augmented in the leo1::Hermes strain (Fig EV3I) indicated that heterochromatin formation is intact, by RNAi or an alternative heterochromatin nucleation mechanism 15, 25, 26, 27, 28. Additionally, nucleosomes in heterochromatin may have been stabilized in leo1∆ cells independent of RNAi.

Relationship of heterochromatin stabilization to the loss of TRAMP and exosomes

To relate to previous studies, we investigated the effect of leo1∆ on pathways that have been described to stabilize heterochromatin. The loss of heterochromatin in RNAi‐deficient cells has been shown to be restored by the concomitant loss of the TRAMP or exosome complexes 7, 11. The sRNA profile in leo1∆ was reminiscent of results from cells lacking the TRAMP subunit Cid14. In cid14∆ cells, the sRNAs at the dg/dh repeats are reduced and the IRC sRNAs are undetectable; this phenomena is coupled with an increase in rRNAs 49. Conceivably, we hypothesized that Cid14 may be involved in the heterochromatin stabilization in leo1∆ cells. Indeed, deletion of cid14 ⁺ rescued the leo1::Hermes phenotype (Fig 6C). Silencing of ade6 ⁺ was also observed at the otr1 and mat2 loci and was dependent on cid14 ⁺ regardless of the leo1 genotype (Fig EV4A).

Figure EV4. Effect of strain genotype on gene expression (or histone retention) at various chromosomal locations.

• A, B
  Retrotransposable element Tf2 is silenced by an siRNA‐independent mechanism in leo1∆. (A) Growth of indicated strains, with ade6 ⁺ integrated at two locations, on a low‐adenine plate. (B) Distributions of sRNA mapping to the 13 Tf2 retrotransposons. Normalized reads are aligned relative to the transcription start site, sense (top panel) and anti‐sense direction (bottom panel) in WT (blue), leo1∆ (red), and rrp6∆ (green) cells.
• C, D
  Histone turnover and the effects of acetyltransferases on heterochromatin formation. (C) ChIP–qPCR using antibody against HA in cells with the RITE construct, using primers for dh1, tf2, matK, rad50 ⁺, and spd1 ⁺. ChIP was performed in two independent experiments. (D) Position variegation effect by Mst2 and Paf1. Spotting assay of strains preselected as being Ura⁺ or FOA^R as indicated.
• E
  Spotting assay of strains with ura4 ⁺ integrated at matL, and ade6 ⁺ at IR‐L.
• F
  Spotting assay of strains with ura4 ⁺ integrated at matK.

It has also previously been demonstrated that silencing double‐stranded sRNAs are generated at HOODs (many of which comprised Tf2 elements) in exosome‐depleted rrp6∆ cells 7. We confirmed this finding, but in contrast to rrp6∆ cells, in WT and leo1∆ cells sRNAs were primarily detected in the sense direction only (Fig EV4B), suggesting that most Tf2 sRNAs in WT and leo1∆ were RNAi‐independent degradation products. Taken together with the previously observed increase in heterochromatin at Tf2s in the leo1∆ cells (Fig 4C–E), these data supported the existence of an alternative mechanism for heterochromatin stabilization.

A role for Leo1 in histone turnover

The function of Leo1 in shifting chromatin states led us to suspect that Leo1 played a role in histone turnover. To test this possibility, we investigated the effect of the leo1 genotype on new histone incorporation into the chromatin using the recombination‐induced tag exchange (RITE) assay 5, 50. In this system, the hormone β‐estradiol is used to swap the epitope tag of histone H3 (hht2 ⁺) from HA (“old” histone) to T7 (“new” histone) using a Cre/Lox recombinase (Fig 7A). H3 expression is regulated under the control of its endogenous promoter. The RITE assay was used to measure histone turnover in unsynchronized WT and leo1∆ cells. Samples were collected before and 2 h after the induction of the genetic switch. ChIP–qPCR was performed using HA and T7 antibodies to determine the loss of “old” histones (produced before the switch) and the incorporation of “new” histones (produced after the switch) by comparing the HA and T7 signals at 0 h and 2 h. We observed a decrease in the new incorporation of histone H3 into leo1∆ cells compared with the WT at the pericentric regions (dhI), the silent matK region, and the Tf2 elements based on the lower increase in T7 at 2 h after the genetic switch (Fig 7B). The HA signal was reduced to background levels at the heterochromatin loci dhI, matK, and tf2 for WT and leo1∆, but retention of the HA signal was observed in leo1∆ at the coding regions rad50 ⁺ and spd1 ⁺ (Fig EV4C). We also studied the effects of Ddb1 on histone turnover. As expected from the Leo1‐antagonizing role we observed for this factor in heterochromatin formation (Figs 3A and B, and EV1), deleting ddb1 ⁺ in the leo1∆ background indeed suppressed the histone turnover reduction in leo1∆. These data clearly demonstrated a role for Leo1 in maintaining an open chromatin state by promoting histone H3 exchange.

Figure 7. Leo1 mediates histone turnover.

1. Diagram of the RITE system. Green triangles represent LoxP sites.
2. ChIP of H3‐T7 at silent chromatin regions.
3. ChIP of H3‐T7 at expressed euchromatic regions transcribed by RNAPII (IRC1R, spd1 ⁺, rad50 ⁺, pyk1 ⁺, and scm3 ⁺) or RNAPIII (tRNA).
4. Spotting assay using strains with ura4 ⁺ at the matK region.
5. Spotting assay using strains with ade6 ⁺ integrated at the IR‐L and ura4 ⁺ integrated at matL domains.
6. Spotting assay using strains with ura4 ⁺ at the matK region.

Data information: (B, C) ChIP was performed in three independent experiments; error bars indicate SD.

If the mechanism whereby Leo1 elicits its chromatin effect occurs through transcription‐coupled histone turnover in general, we would expect turnover to decrease at all RNAPII‐transcribed loci in the leo1∆ cells. To this end, we determined the turnover at five additional RNAPII‐transcribed loci: the boundary element IRC1R, rad50 ⁺, spd1 ⁺, scm3 ⁺, and pyk1 ⁺. We also calculated the turnover at the RNAPIII‐transcribed gene for the alanine tRNA. The results showed that the RNAPII‐transcribed loci were associated with a similar reduction in histone turnover in the leo1∆ strain, whereas the RNAPIII‐transcribed locus was not affected (Fig 7C); this result was in concordance with a general role for Paf1C in transcription elongation.

To determine whether the modification of histone turnover in and of itself could lead to the modulation of the chromatin state, we determined whether a known modifier of histone turnover (i.e., Mst2 51) could also influence heterochromatin propagation across the IR‐L and matK loci of the ddb1∆ mutant. The results revealed that reducing histone turnover by deleting the histone acetyltransferase Mst2 partially restored the matK heterochromatin, albeit to a lower level than the deletion of Paf1 (Fig 7D). Also, heterochromatin was propagated across IR‐L, and a PEV phenotype was observed (Fig EV4D). We also tested heterochromatin propagation when deleting another histone acetyltransferase, Hat1 52, that has been shown to promote histone turnover in S. cerevisiae 53. In contrast to mst2∆, hat1∆ in S. pombe did not show a heterochromatin‐spreading phenotype at matK or across IR‐L (Fig EV4E and F).

The role of transcript termination

Recently, Paf1C was shown to negatively affect siRNA‐directed gene silencing through transcription termination. Mutations that inactivate Paf1C components or the Ctf1 and Res2 termination factors impair accurate transcription termination, leading to siRNA production and siRNA‐mediated gene silencing in strains expressing synthetic hairpin RNA 54. To explore the possibility that impaired RNA processing in the leo1::Hermes mutants promoted heterochromatin propagation across a boundary element and suppression of the silencing defect in the ddb1∆ background, we determined the effects of ctf1∆ or res2∆ on the expression state of ura4 ⁺ inserted at the euchromatic matL region or at the matK region of the ddb1∆ mutants. The results indicated that in contrast to the inactivation of Paf1, the inactivation of Ctf1 or Res2 did not promote silencing propagation across IR‐L (Fig 7E) or the suppression of the silencing defect at the matK region of the ddb1∆ mutants (Fig 7F). These data demonstrate that faulty processing of RNA was not the cause of stabilization of the heterochromatic state in the leo1 and paf1 mutants observed here.

Discussion

Chromatin state dynamics is regulated at two levels: One involves the interplay between activities that modify or de‐modify nucleosomal histones, and the other involves mechanisms that enhance or suppress histone turnover. The two modes of regulation are not mutually exclusive. In fact, some proteins may function at both levels. For example, Epe1 is a putative demethylase that negatively affects H3K9 stability 55, 56. However, it may also up‐regulate histone turnover at heterochromatin boundaries, leading to replacement of chromatin‐associated methylated H3K9 by non‐methylated H3 histones. Likewise, Mst2 may affect chromatin structure by acetylating H3K14, but it also affects heterochromatin stability by up‐regulating histone turnover 51. Paf1 and Leo1 promote chromatin transcription 41, most likely by binding H3 and destabilizing nucleosomal structure. A related activity of these Paf1C components may modulate heterochromatin spreading by up‐regulating histone turnover at boundary elements. Notably, Leo1 may also antagonize heterochromatin spreading by facilitating Mst1‐mediated acetylation of H4K16 57.

Histone turnover and stability of the heterochromatin state

Heterochromatin regions are associated with stable nucleosomes 4 that are recycled during transcription 5, 58. During replication, Swi6 associates with a Clr6‐HDAC complex and HIRA, which incorporate new histones that readily acquire the posttranslational modifications of silent chromatin 59, 60, 61.

A causal relationship between up‐regulation of histone turnover and modulation of heterochromatin stability has been proposed for Epe1 4 and Mst2 51. The observations that inactivation of Epe1, Leo1, Paf1 and, to a lesser extent, Mst2 has an effect on propagation of heterochromatin across IR‐L suggest that these proteins may modulate heterochromatin stability by similar mechanisms. Furthermore, mst2∆, like leo1∆ or paf1∆, suppresses silencing impairment at heterochromatin domains by ddb1∆. We therefore postulate that Leo1 and Paf1, like Epe1 and Mst2, inhibit heterochromatin propagation across boundary elements by enhancing histone turnover.

The differences we observed between mst2∆ and hat1∆ have previously been observed at telomeric heterochromatin 62, 63 and may be explained by the distinct mechanisms of actions by the two acetyltransferases. Hat1 acting mainly on histones in the cytoplasm before chromatin incorporation 64, whereas Mst2 belongs to the MYST family and acts on the chromatin‐bound fraction 62. This implies a distinction between (i) turnover initiated through incorporation of new histones and (ii) turnover primarily due to eviction of old histones. Consequentially, these results suggest that Leo1–Paf1 promotes turnover by eviction of histones present in the chromatin.

A direct role for Leo1–Paf1 in the destabilization of nucleosomes has previously been proposed in other organisms. Leo1–Paf1 was shown to promote chromatin transcription 40, most likely by destabilizing human nucleosomes. Leo1–Paf1 has also been shown to exhibit H3 binding affinity in human cells and S. cerevisiae 38. In S. cerevisiae, Paf1 is required for the activation of stress‐inducible genes 65, which is consistent with a role in histone exchange 5. Other examples implicate Leo1–Paf1 in transcriptional changes during development and inflammation 66.

A role of Leo1–Paf1 in histone turnover during RNAPII‐mediated transcription elongation on chromatin templates is complementary to the role of the entire Paf1C in maintaining heterochromatin by transcription termination 54. Our data also shed additional light on the mechanism by which Leo1 was recently described to promote histone acetylation 57.

The interaction of Leo1–Paf1 and Epe1

HP1^Swi6 recruits Epe1 to heterochromatin, where it interacts with Clr3 HDAC activity to suppress histone turnover 4. Both epe1 and leo1 mutants exhibited PEV that implied defects in chromatin state fluctuations; we showed that this effect was mediated through histone turnover. Based on our observations, it is likely that Epe1 and Leo1–Paf1 act in the same pathway in this process. The question remains regarding the nature of the interaction between Leo1–Paf1 (the subunits of a transcription elongation complex) and Epe1 (a JmjC‐containing enzyme). Epe1 is part of the JHDM1 family. Other members of this family act as histone demethylases 67, but the S. pombe homolog Epe1 lacks conserved residues within the JmjC domain 23, 67. Although no direct evidence for Epe1 demethylase activity has been reported, recent studies have proposed a putative demethylase activity 55, 56.

Relationship between Leo1–Paf1 and H2Bub1

The role of Leo1 in histone turnover together with the partial requirement of Leo1 for H2Bub1 is intriguing (Fig 2D). The reduced H2Bub1 levels in leo1∆ cells may be a reflection of slower RNAPII in Paf1C‐compromised cells 54 because H2Bub1 is normally associated with fast elongation by RNAPII 68.

In contrast to the results obtained with the leo1 and paf1 mutants in this study, we previously observed the appearance of heterochromatin at the central domain of the centromere in H2Bub1‐deficient cells 34. This finding was not attributed to spreading across tRNA barriers but to de novo heterochromatin formation. Here, we observed that spreading across IR barriers was not affected by the loss of H2Bub1 (Fig 2E). H2Bub1 was directly involved in transcription elongation, and the sRNA profile of htb1‐K119R mimicked that of leo1∆ (Fig EV5A and B). However, the rescue of ago1∆ did not depend on H2Bub1 (Fig EV5C). These observations further detail the interplay between transcription and histone turnover and the consequences for heterochromatin spread.

Figure EV5. The contribution of H2Bub1 on silencing of the pericentric chromatin.

1. sRNA profile over centromere 3. The number of reads from two independent experiments was normalized to reads mapping to tRNAs. Red dots in top panel indicate positions of tRNAs.
2. sRNA profile in htb1‐K119R.
3. Spotting assay of strains with ura4 ⁺ integrated at imr1R.

Chromatin state spreading across defective boundaries

The leo1∆ rescue of pericentric heterochromatin in RNAi‐deficient mutants (ago1∆) can be explained as an indirect effect of reducing turnover and thereby stabilizing the heterochromatin formed by RNAi‐independent mechanisms 69. Because the tRNA barriers were not affected by Leo1–Paf1 (Fig EV3G), most of the pericentric heterochromatin was not challenged by the spreading of active chromatin from the surrounding euchromatin regions. In cells with defective Leo1–Paf1, only the integrity of the IR boundaries was compromised; thus, the surrounding euchromatin may invade into the pericentric region of otr1R lacking tRNAs. This is consistent with our results showing that the increased heterochromatin stability in leo1::Hermes cells is overruled at the right pericentric region with a defective chromatin border (Fig EV2D). Euchromatin spreading across defective boundaries has previously been described in S. pombe 70. This suggests that the boundary defects we observe in Leo1–Paf1 deficient cells are bidirectional because they prevent both heterochromatin and euchromatin from spreading.

In summary, here, we present a function for Leo1–Paf1 (a subcomplex of Paf1C), which acts together with Epe1 to promote chromatin state fluctuations by enhancing histone turnover. In cells deficient in Leo1–Paf1, the increase that we observe in H3K9me2 can be explained by reduced histone turnover. Replacement of the histone H3 at the boundary element through Leo1–Paf1 or Epe1 will reverse the heterochromatin formation and maintain an open chromatin conformation. In the absence of Leo1–Paf1, Clr4 and HP1^Swi6 from the surrounding H3K9me nucleosomes propagate chromatin silencing.

Materials and Methods

Strains

Strains used in this study are described in Table EV1. Standard fission yeast medium was used. Growth curve was determined by collecting exponentially growing cells at different time points. Partial deletion of leo1 was performed by PCR‐based gene targeting with G418 resistance genes as selection marker. In the spotting experiments, the dilutions were tenfold.

Chromatin immunoprecipitation (ChIP)

Chromatin from cultures was extracted in duplicate and subjected to immunoprecipitation. Samples were immediately subjected to formaldehyde cross‐linking (final concentration 1%), and chromatin was isolated according to Durand‐Dubief et al 71. ChIP was performed using 4 μg anti‐H3K9me2 (Abcam ab1220), 1.5 μg anti‐HA (Abcam ab9110), 1.5 μg anti‐T7 (EMD Chemicals 69522), or 3 μg anti‐Swi6 (Abcam ab188276) antibodies per 30 μl of chromatin extract. DNA was recovered with QIAquick PCR Purification columns (Qiagen). qPCR was performed on Chip samples with SYBR Master Mix (Life Technologies) using the Applied Biosystems 7500 Real‐Time PCR System. A list of the primers used is provided in Table EV2.

Western blotting

Histones were extracted using H₂SO₄. A total of 10 μg of histones was run on the 12% SDS–PAGE. Gel was blotted on PVDF membrane. Membrane was incubated with anti‐H2Bub1 (Millipore 05‐1312) (1:1,000) and anti‐H4 (Abcam ab61255) (1:1,000) antibodies overnight at 4°C and detected by ECL kit.

RITE

Cells were cultured in YES media to mid‐log phase (~5 × 106 cells/ml) at 30°C. For induction of genetic switch, exponentially growing cells were treated with 1 μM β‐estradiol for 2 h at 30°C.

ChIP–exo

ChIP–exo was performed as described in Svensson et al 5. Schizosaccharomyces pombe strains were grown in rich media and subjected to formaldehyde cross‐linking and then processed through the ChIP–exo assay 72, 73. Briefly, cells were disrupted using glass beads, and chromatin was fragmented by sonication (Bioruptor Pico, 20 cycles, 30 s on, 30 s off). Fragmented chromatin was immunoprecipitated using antibodies directed against either HA or T7. Protein A‐coated magnetic beads (NEB) were used to bind the antibodies. Beads were washed, and with the immunoprecipitate still on the beads, the DNA was polished, A‐tailed, and ligated to an Illumina sequencing library adaptor. Digestion lambda exonuclease (final concentration 2.5 U/reaction) removed nucleotides from 5′ ends of double‐stranded DNA until blocked by the formaldehyde‐induced protein–DNA cross‐link. The cross‐links were reversed (4 h at 65°C), and DNA was eluted from the beads. The single‐stranded DNA was subsequently made double‐stranded by primer annealing and extension. A second sequencing adaptor was ligated. Using indexing primers, the fragments were PCR‐amplified and gel‐purified (Qiagen MinElute). Samples were quantified on using Qubit (HS dsDNA) and sequenced on Illumina Hiseq 2000 (50 cycles, single‐end sequencing) at the BEA facility (Huddinge, Sweden) following the manufacturer's instructions. Raw data from the Hiseq (fastq files) were aligned to ASM294v2 using Bowtie2 using default parameters. The ASM294v2.24 annotation was downloaded from pombase.org and used in Podbat. The aligned data (sam files) were imported and normalized to million reads. Data from independent biological duplicates were averaged. Identical reads were discarded to remove PCR artifacts. Signals were calculated as averages over 150 nt were taken.

Data analysis

Genomewide data were analyzed mainly by Podbat 74 and R. RPKM (reads per kilobase per million) was calculated for all features of the ASM294v2 annotation.

Small RNA library

Cells were grown to mid‐log phase at 30°C. Total RNA was extracted with hot acid phenol and chloroform. A total of 1 μg of total RNA from each sample was used to prepare library with TruSeq Small RNA Library preparation kit (Illumina). Prepared small RNA library was sequenced using MiSeq platform at BEA facility. Small RNA libraries were sequenced on MiSeq (36 cycles, single end) at the BEA facility (Huddinge, Sweden).

Southern blotting

Genomic DNAs from strains AP2647 (Hermes) and Hu2689 (leo1Δ) were purified from 50 ml of overnight cultures using standard molecular biology method. Genomic DNA (10 μg) from AP2647 and Hu2689 strains was digested with restriction enzymes EcoRI or EcoRI and KpnI for Hu2689 at 37°C overnight. The digested DNA was resolved on a 1% agarose gel for 5 h at 10 V/cm Tris–acetate–EDTA (TAE) buffer. In‐gel depurination, denaturation, and neutralization were carried out before capillary transfer to Hybond‐N+ membrane (0.45 μm) using 20× SSC for 20 h. The membrane was cross‐linked by UV transilluminator (245 nm/1,600 J). Forward primer, SB‐F (25 pmol) was end‐labeled using γ‐P³² (PerkinElmer) with Optikinase (USB). PCR was carried out with the end‐labeled forward primer and reverse primer (SB‐R) to amplify a region within KanMx6 cassette. The probe was purified using Qiagen PCR purification kit, denatured together with 2.5 mg/ml of sheared salmon sperm DNA at 90°C for 5 min followed by snap cooling, and added to the membrane after prehybridization with buffer containing 1.5× SSC, 5× Denhardt's solution (USB), and 0.1% SDS at 64°C for 1 h. Hybridization was carried out for 18 h, and the membrane was washed with three times with 2× SSC and 0.1% SDS and exposed for phosphorimaging.

Data availability

Data from ChIP–exo sequencing and sRNA sequencing were deposited at GEO (accession number: GSE66941).

Author contributions

LS performed experiments (ChIP, smallRNA‐seq, spotting, Tf2, turnover) and wrote the manuscript; PP performed experiments (Southern blot); KE provided funding, contributed with discussing results, and wrote the manuscript; JPS performed experiments (ChIP, turnover) and wrote the manuscript; AC initiated the project, performed experiments (initial screen, most strain crosses, and spotting assays), and wrote the manuscript.

Conflict of interest

The authors declare that they have no conflict of interest.

Supporting information

Expanded View Figures PDF

Table EV1

Table EV2

Review Process File

EMBO Reports (2015) 16: 1673–1687