Defects in RNA quality control factors reveal RNAi-independent nucleation of heterochromatin.

Heterochromatin assembly at Schizosaccharomyces pombe centromeres involves a self-reinforcing loop mechanism wherein chromatin-bound RNAi factors facilitate targeting of Clr4-Rik1 methyltransferase. However, the initial nucleation of heterochromatin has remained elusive. We show that cells lacking Mlo3, a protein involved in mRNP biogenesis and RNA quality control, assemble functional heterochromatin in RNAi-deficient cells. Heterochromatin restoration is linked to RNA surveillance because loss of Mlo3-associated TRAMP also rescues heterochromatin defects of RNAi mutants. mlo3Δ, which causes accumulation of bidirectional repeat-transcripts, restores Rik1 enrichment at repeats and triggers de novo heterochromatin formation in the absence of RNAi. RNAi-independent heterochromatin nucleation occurs at selected euchromatic loci that show upregulation of antisense RNAs in mlo3Δ cells. We find that the exosome RNA degradation machinery acts parallel to RNAi to promote heterochromatin formation at centromeres. These results suggest that RNAi-independent mechanisms exploit transcription and non-coding RNAs to nucleate heterochromatin.

Heterochromatin is linked to numerous cellular functions including transcriptional regulation, chromosome segregation, and suppression of recombination[1,2]. Heterochromatic regions show a distinctive pattern of histone modifications. In addition to deacetylation of histones, heterochromatin in many eukaryotes including the fission yeast Schizosaccharomyces pombe is marked by methylation of histone H3 at lysine 9 (H3K9me)[3]. These modifications are critical for recruitment of heterochromatin factors including HP1 proteins and for the assembly of repressive chromatin structures[1,4].

In S. pombe, heterochromatin is preferentially enriched across large chromosomal domains at centromeres, subtelomeres and the mating type locus[3]. These heterochromatin domains contain dg and/or dh repeats that are transcribed by RNA polymerase II (RNAPII)[1,2]. Transcripts generated from dg/dh repeats are processed into siRNAs by the RNAi machinery including Argonaute (ago1), Dicer (dcr1) or RNA-dependent RNA polymerase (rdp1)[1,2]. Mutations in RNAi factors cause defects in H3K9me at centromeres and loss of siRNAs[1]. siRNAs are bound by Ago1, a subunit of the RNA-induced transcriptional silencing (RITS) complex that is composed of two additional proteins: Chp1 and Tas3[1]. siRNA-bound Ago1, together with binding of the H3K9me by Chp1, assist in the localization of RITS to heterochromatin[5,6]. RITS facilitates targeting of Clr4, a homolog of mammalian SUV39h, which methylates H3K9 at heterochromatic loci[7,8]. Clr4 exists in a multisubunit complex, ClrC, which among other factors contains Rik1 that is critical for RNAi-dependent loading of the complex onto transcribed repeats[8]. The results showing the involvement of RNAi in targeting H3K9me and the requirement of heterochromatin factors for generation of siRNA have revealed the existence of a positive feedback loop in the assembly of heterochromatin[5,9].

H3K9me also recruits HP1 family proteins Chp2 and Swi6, which in turn promote the localization of various effectors including factors involved in chromosome dynamics and gene silencing[1]. Swi6 is required for the localization of cohesin-loading complex involved in proper segregation of chromosomes[10]. Chp2 and Swi6 also associate with Snf2–HDAC Repressor Complex (SHREC), Asf1–HIRA histone chaperone and Clr6 histone deacetylase (HDAC) complexes involved in transcriptional silencing[10-13]. Heterochromatin precludes RNAPII accessibility at target loci, but paradoxically, RNAPII transcription of dg/dh repeats is required to generate siRNA precursors. Centromeric repeats are transcribed preferentially during the S phase of the cell cycle when heterochromatin is more amenable to transcription[14,15]. Apart from generating siRNA precursors, RNAPII transcription may have more direct roles in heterochromatin formation. Indeed, mutations in RNAPII subunits and RNA splicing factors impair heterochromatic silencing[16-18].

RNAPII transcription has been shown to integrate multiple aspects of nuclear metabolism. Elongating RNAPII recruits chromatin-modifying activities to help remodel chromatin[19]. RNAPII machinery also recruits RNA processing factors, including factors which promote mRNP biogenesis and RNA quality control[20,21]. RNA quality control in the nucleus is monitored by multiple factors such as the TRAMP complex. TRAMP containing Cid14, a member of the Trf4 family of polyA polymerases, channels RNAs into degradation pathways including the exosome with 3′–5′ exonucleolytic activity[22]. Both the exosome and TRAMP function to degrade centromeric transcripts in S. pombe[23,24].

Despite major advances, the exact cascade of events that leads to the initial nucleation of heterochromatin at centromeric repeats has remained unclear. It has been argued that a new class of small RNAs, termed primal RNAs, which require Ago1 for their biogenesis, nucleate heterochromatin, and that ago1Δ shows H3K9me levels comparable to clr4Δ cells[25]. In this study, we sought to explore mechanisms that nucleate heterochromatin structures. Our analyses have uncovered an RNAi-independent pathway that exploits RNAPII transcription and non-coding RNAs (ncRNAs) to nucleate heterochromatin at centromeres and other parts of the genome. Factors involved in co-transcriptional processes including that affect RNAPII processivity influence RNAi-independent heterochromatin assembly. We provide evidence that the exosome RNA degradation machinery acts parallel to the RNAi to mediate heterochromatin formation.

RESULTS

Loss of Mlo3 restores centromeric silencing in RNAi mutants

Mutations in RNAPII lead to defective heterochromatic silencing at centromeres[16,17]. Given that RNAPII is linked to heterochromatin modifications, we investigated if RNAPII coupled processes affect heterochromatin assembly. Loss of Mlo3, a homolog of S. cerevisiae Yra1 and metazoan Ref or Aly that acts at the interface of RNAPII transcription and RNA metabolism[21,26], restores centromeric silencing in RNAi deficient cells. Whereas ago1Δ alleviated silencing of a ura4+ reporter inserted at an outer centromeric repeat region (otr1R::ura4+), simultaneous deletions of mlo3 and ago1 restored centromeric silencing (). The observed suppression required heterochromatin machinery because mlo3Δ failed to suppress the silencing defect in clr4Δ mutant ().

These results suggested that loss of Mlo3 promotes heterochromatic silencing in RNAi deficient cells. We next investigated whether mlo3Δ restores transcriptional repression and localization of heterochromatin factors at centromeres. Chromatin immunoprecipitation (ChIP) analyses across centromere 2 (cen2) showed that the observed changes in silencing correlated with marked reduction in RNAPII occupancy in mlo3Δ ago1Δ cells, as compared to ago1Δ (). More importantly, mlo3Δ restored H3K9me and Swi6 localization at otr1R::ura4+ and endogenous centromeric repeats in ago1Δ mutant (). mlo3Δ also restored silencing and heterochromatin formation at centromeres in dcr1Δ mutant (). The rescue of H3K9me in RNAi mutants was not due to changes in histone H3 occupancy () or restoration of siRNAs (). Together, these results demonstrate that mlo3Δ suppresses heterochromatin defects of RNAi mutants.

mlo3Δ restores functional heterochromatin in RNAi mutants

Heterochromatin facilitates the loading of cohesin essential for sister chromatid cohesion[10,27]. Defective heterochromatin in RNAi mutants causes impaired centromere cohesion, resulting in chromosome missegregation and sensitivity to the microtubule destabilizing drug thiabendazole (TBZ)[28,29]. To test whether loss of Mlo3 in RNAi mutant cells restores functional heterochromatin, we measured the TBZ sensitivity of single and double mutants. As expected, cells carrying ago1Δ or dcr1Δ showed severe sensitivity to TBZ, consistent with defective chromosome segregation in these mutants[28,29]. Combining mlo3Δ with ago1Δ or dcr1Δ suppressed the TBZ sensitivity of RNAi mutants (). ChIP analyses showed that loss of TBZ sensitivity correlates with a partial restoration of cohesin localization at centromeres in mlo3Δ ago1Δ mutant (). Therefore, heterochromatin assembled upon loss of Mlo3 is functional, capable of supporting proper segregation of chromosomes in RNAi deficient cells.

Mlo3 interacts with centromeric transcripts

To determine whether Mlo3 directly functions at centromeres, we tested if it interacts with centromeric transcripts. As expected for a factor involved in mRNP formation[26,30], Mlo3 interacted with a euchromatic gene (fbp1) transcript (). Importantly, Mlo3 also interacted with dh transcript (), consistent with results of ChIP analyses showing Mlo3 enrichment at transcribing centromeric repeats[30]. This interaction was greatly enhanced in ago1Δ cells. Therefore, in addition to euchromatic genes, Mlo3 targets heterochromatic repeat transcripts. These data argue that Mlo3 functions at centromeric repeats and that restoration of heterochromatin in RNAi mutant cells may be coupled to its role in co-transcriptional processing of centromeric transcripts[30].

TFIIS modulates RNAi-independent heterochromatin assembly

We wondered whether loss of Mlo3, like Yra1, causes defective RNAPII transcription. To test this, we constructed mlo3Δ clr3Δ double mutant. Mutant cells lacking SHREC subunit Clr3 show marked increase in RNAPII occupancy at centromeric repeats[10,11,31]. Combining mlo3Δ with clr3Δ resulted in small decrease in RNAPII as compared to clr3Δ, albeit levels of H3K9me or Swi6 at repeat elements were comparable in clr3Δ and clr3Δ mlo3Δ (). The change in RNAPII levels led us to wonder if defective RNAPII transcription is partially responsible for restoration of heterochromatin in RNAi mutants. We tested this possibility by deleting tfs1 gene encoding the S. pombe homolog of TFIIS, a factor known to affect RNAPII processivity[32]. Deletion of tfs1, which led to 6-azauracil (6-AU) sensitivity ()[33] and changes in the distribution of RNAPII at body of genes (), resulted in variegated suppression of silencing defects in ago1Δ and dcr1Δ mutants ( and ). This suppression was more pronounced in M mating-type (mat1-Msmt0) cells and resulted in decreased RNAPII occupancy at centromeric repeats (). tfs1Δ-dependent silencing required Clr4 () and was accompanied by an increase in levels of H3K9me and Swi6 at both otr1R::ura4+ and centromeric repeats ( and . The increase in H3K9me was not due to changes in histone H3 occupancy or restoration of siRNAs (). tfs1Δ suppressesed TBZ sensitivity and partially restored cohesin localization at centromeres in RNAi mutants ( and ).

These results suggest that cells with compromised RNAPII transcription rescue heterochromatin defects caused by loss of RNAi machinery. Consistently, we found that loss of deubiquitylating enzyme Ubp3, which causes elevated levels of RNAPII ubiquitylation and degradation[34], partially suppressed silencing and heterochromatin defects of ago1Δ mutant (

mlo3 and tfs1 differentially affect heterochromatin assembly

Loss of Clr3 and RNAi factors causes a dramatic loss of H3K9me across pericentromeric domains ( and )[31]. To gain more insight into effects of mlo3Δ and tfs1Δ on heterochromatin formation in the absence of RNAi, we investigated their effects on heterochromatin modifications in clr3Δ ago1Δ double mutant cells. tfs1Δ failed to restore H3K9me at otr1R::ura4+ in clr3Δ ago1Δ cells (). In contrast, mlo3Δ resulted in considerable restoration of H3K9me at centromeres in clr3Δ ago1Δ cells (). These differences in H3K9me were not limited to otr1R::ura4 ChIP-chip analyses across centromere 2 showed that mlo3Δ, but not tfs1Δ, restored H3K9me across the entire pericentromeric region in clr3Δ ago1Δ cells (, ). mlo3Δ also decreased in RNAPII occupancy across pericentromeric domains in clr3Δ ago1Δ cells (), while tfs1Δ had only a minor effect ().

Differences in mlo3Δ and tfs1Δ mutants were also evident in their effects on hairpin RNA-triggered heterochromatin formation. Despite the prominent role of RNAi in silencing of centromeric repeats, this pathway is constrained and rarely targets detectable levels of heterochromatin in trans[35,36]. Expression of hairpin complementary to the trp1+ (trp1-HP) failed to elicit H3K9me in tfs1Δ cells but induced H3K9me both in cis and in trans in mlo3Δ cells, dependent on RNAi (). Treating cells with 6-AU, which affects RNAPII transcription, caused increased H3K9me both in mlo3Δ and tfs1Δ cells but again the effect was stronger in mlo3Δ (). Together, these data highlight differential effects of tfs1Δ and mlo3Δ on heterochromatin formation at centromeres and at an ectopic site. Moreover, these analyses suggest that additional RNAi and SHREC independent mechanism(s) nucleate heterochromatin at centromeres.

cid14Δ rescues heterochromatin defects of ago1Δ mutant

Mlo3 forms a complex with TRAMP, which is involved in RNA surveillance and degradation of centromeric transcripts[22,24,30]. We therefore investigated whether mlo3Δ-mediated suppression of heterochromatin defects in RNAi mutants is linked to defects in RNA surveillance. Remarkably, loss of Cid14 subunit of TRAMP affects RNAi-independent heterochromatin formation in a manner similar to mlo3Δ. Loss of Cid14 restored H3K9me at otr1R::ura4+ and centromeric repeats in ago1Δ mutant (). Moreover, cid14Δ suppressed the silencing defect caused by ago1Δ, as indicated by reduction in the levels of dg/dh transcript in cid14Δ ago1Δ as compared to ago1Δ (). These results suggest that defects in RNA surveillance mechanisms involving Mlo3-associated TRAMP, promote heterochromatin formation independent of RNAi.

Exosome acts parallel to RNAi to nucleate heterochromatin

As mentioned above, Mlo3–TRAMP channels centromeric transcripts to downstream-acting RNA degradation factors including the exosome[24,30]. However, these factors share a complex genetic relationship indicative of their diversified functions. Unlike single mutants, double mutants containing null alleles of Mlo3–TRAMP and the exosome subunit rrp6 show synthetic lethality (our unpublished data). In light of these observations and previous results showing that aberrant RNAs sequestered near site of transcription are degraded by the exosome[37,38], it was of interest to investigate whether the loss of Rrp6 affects centromeric heterochromatin. Northern and RT-PCR analyses using single and double mutants showed a large increase in centromeric repeat transcripts in rrp6Δ ago1Δ mutant as indicated by both Northern blot and RT-PCR analyses (). This result is in marked contrast to the restoration of silencing observed in cid14Δ ago1Δ mutant (). We next investigated whether rrp6Δ alone or in combination with ago1Δ affects centromeric heterochromatin assembly. Unlike mlo3Δ and cid14Δ, combining rrp6Δ with ago1Δ largely abolished H3K9me levels at otr1R::ura4+ and dg repeats (). Furthermore, ChIP-chip showed severe cumulative loss of H3K9me across the entire pericentromeric domain in rrp6Δ ago1Δ mutant, as compared to rrp6Δ or ago1Δ (). Together, these data reveal the nuclear exosome acts parallel to RNAi to mediate heterochromatin assembly at centromeres.

De novo heterochromatin assembly in mlo3Δ cells

mlo3Δ restored H3K9me both in ago1Δ and clr3Δ ago1Δ cells. We next tested whether mlo3Δ, which causes accumulation of transcripts in the nucleus (), can trigger de novo heterochromatin assembly in the absence of RNAi. For this purpose, we employed strains that carry either clr4Δ alone or in combination with ago1Δ or ago1Δ mlo3Δ. Since Clr4 is the sole H3K9 methyltransferase in S. pombe, these mutant strains lack H3K9me (). We transformed the mutant strains with a plasmid containing clr4+ and monitored H3K9me levels at centromeres by ChIP. Introduction of clr4+ led to restoration of H3K9me at centromeres in clr4Δ single mutant but not in clr4Δ ago1Δ double mutant (). Remarkably, H3K9me could be readily detected at centromeres in clr4Δ ago1Δ mlo3Δ cells upon introduction of the clr4+ (). These data demonstrate that loss of Mlo3, involved in processing of centromeric transcripts[30], triggers de novo heterochromatin assembly at centromeres, independent of RNAi.

We next tested if mlo3Δ affects localization of ClrC, which requires RNAi for its targeting to centromeric repeats[7,8,14]. In particular, ClrC subunit Rik1 is recruited to centromeres during S-phase, when both forward and reverse strands of centromeric repeats are transcribed[14]. We probed the effect of mlo3Δ on Rik1 enrichment at cenH, which is homologous to dg/dh and serves as an RNAi-dependent heterochromatin nucleation center at mat locus[39,40]. cenH was selected because defects in RNAi abolish Rik1 ChIP enrichment at this site without affecting heterochromatin structure nucleated by redundant mechanisms[8]. This regime ensures that Rik1 enrichment is not indirectly altered by changes in heterochromatin modifications in mlo3Δ. As expected, ago1Δ abolished Rik1 ChIP enrichment at cenH (). However, we found that simultaneous deletion of mlo3 and ago1 restored Rik1 enrichment at cenH (). Given the connection between transcript levels and Rik1 localization[8], we tested if mlo3Δ affects cenH transcripts. Levels of forward and reverse transcripts originating from cenH were elevated considerably in mlo3Δ cells (). Together, these results demonstrate that mlo3Δ, which causes accumulation of bidirectional cenH transcripts, bypasses the RNAi requirement for targeting Rik1.

mlo3Δ induces H3K9me at euchromatic loci

Since cells lacking Mlo3 accumulate antisense RNAs at euchromatic loci[30], we wondered whether heterochromatin could be assembled at euchromatic loci in mlo3Δ cells. Our analyses showed that loss of Mlo3 induces H3K9me at selected euchromatic genes (), which show increased accumulation of antisense transcripts in mutant cells (. Notably, H3K9me at these loci was not abolished when mlo3Δ cells was combined with ago1Δ (). Therefore, the targeting of H3K9me occurs even in the absence of RNAi (). We also analyzed a locus showing overlapping sense and antisense transcripts that were unaffected by mlo3Δ. However, H3K9me could not be detected at this site (). One possibility is that the retention of transcripts at transcribed loci such as observed in RNA surveillance mutants[37,38], is necessary to generate signals for RNAi-independent heterochromatin formation. Regardless, it is interesting that mlo3Δ, which causes accumulation of antisense transcripts, results in heterochromatin modifications at euchromatic loci.

DISCUSSION

Heterochromatin assembly is a complex multistep process that involves a variety of factors[1]. Studies from diverse systems have suggested a prominent role for transcription and ncRNAs in heterochromatin assembly[1,41,42]. In S. pombe, RNAPII transcription of centromeric repeats provides scaffolds for heterochromatin formation. This process involves RNAi, which not only processes repeat transcripts but also mediates the loading of heterochromatin factors[7,8]. Despite the prominent role for RNAi in heterochromatin assembly, loss of RNAi factors does not completely abolish heterochromatin modifications such as H3K9me at centromeric repeats[31,43,44]. Similarly, evidence from other systems suggest that defects in RNAi has no major effects on heterochromatin modifications[45,46], although in some of these cases transcription and ncRNAs could be involved[42,47]. Our analyses suggest that RNAPII transcription and ncRNAs function to target heterochromatin via an RNAi-independent mechanism.

We demonstrate that defects in cotranscriptional RNA surveillance or factors that affect RNAPII processivity bypass the RNAi requirement to assemble functional heterochromatin. Besides loss of Mlo3 or Cid14, tfs1Δ restores heterochromatin in cells lacking the RNAi machinery. Distinct mechanisms are likely responsible for mlo3Δ- or tfs1Δ-mediated suppression of heterochromatin defect in RNAi mutants, despite similarities in phenotypes. We note that mlo3Δ, but not tfs1Δ, rescues H3K9me in clr3Δ ago1Δ double mutant deficient in heterochromatin modifications at centromeres[31]. Also, mutations in these factors differentially affect hairpin-induced H3K9me at an ectopic site. Given that the RNAi machinery interacts with RNAPII and modulates transcription in other organisms[48,49], defective RNAPII elongation could directly contribute to bypassing of RNAi. Tfs1 may affect heterochromatin by influencing processing of transcripts and/or release of RNAPII. Indeed, cells lacking Ubp3, which affects RNAPII stability[34], show partial rescue heterochromatin defects in ago1Δ cells. Another possibility is that impaired transcription affects chromatin dynamics by precluding elongation-coupled turnover of histones and/or their modification state[50,51]. The binding of factors, such as Clr4[8], to residual histones decorated with H3K9me could shift the dynamic equilibrium and stabilize and/or propagate heterochromatin in cis[40].

Mlo3 interacts with TRAMP[30], and these factors are required for processing centromeric transcripts and antisense RNAs[23,24,30]. It is therefore interesting that loss of either Mlo3 or TRAMP suppresses the heterochromatin defects in ago1Δ or dcr1Δ mutants which are deficient in siRNA production. cid14Δ causes severe reduction in siRNAs without altering H3K9me levels, leading to suggestion that low level of siRNAs are sufficient to nucleate heterochromatin[24]. Our results suggest that cid14Δ activates RNAi-independent heterochromatin assembly pathway(s), which might also be triggered by the accumulation of RNAs produced by multiple copy sequences in other systems[52]. We show that the targeting of ClrC subunit Rik1 to dg/dh repeats, which normally requires RNAi[8], is restored in mlo3Δ cells showing elevated levels of forward and reverse repeat transcripts. Furthermore, mlo3Δ triggers de novo targeting of heterochromatin to centromeric repeats independent of RNAi. Remarkably, loss of Mlo3 also causes RNAi-independent targeting of H3K9me at selected euchromatic loci, which show accumulation of antisense transcripts in mlo3Δ mutant. These results argue that ncRNAs generated by centromeric repeats and certain euchromatic loci assemble heterochromatin by mechanism(s) independent of RNAi. Consistent with the existence of such pathway(s) that relies of accumulation of RNAs, we have found that the exosome involved in degradation of aberrant RNAs[22], acts in a pathway parallel to RNAi to mediate heterochromatin formation at centromeres.

How do accumulations of transcripts caused by loss of Mlo3–TRAMP impact heterochromatin assembly in RNAi mutants? One possibility is that ncRNAs accumulated at the sites of transcription recruit the exosome degradation machinery that in turn facilitates loading of heterochromatin factors in a manner similar to the RNAi, in which targeting of ClrC is linked to the processing of repeat transcripts in cis[5]. RNAi-dependent nucleation of heterochromatin involves RITS, which interacts with ClrC[7,8]. Indeed, RITS tethering to transcripts can induce heterochromatin formation[53]. However, this process still requires Dcr1, suggesting that additional siRNA-dependent steps, perhaps the generation of double stranded RNA (dsRNA) or other RNA structures, is necessary for nucleating heterochromatin. In this regard, loss of Mlo3–TRAMP, acting cotranscriptionally, could bypass RNAi requirement by generating dsRNA or yet undefined RNA signals capable of loading Clr4–Rik1. Such signals would be distinct from the recently described primal RNAs that require Ago1 for their biogenesis[25]. It is possible that mechanism(s) that trigger H3K9me in mlo3Δ or cid14Δ mutant are activated during S-phase when both strands of dg/dh repeat are transcribed, correlating with targeting of Rik1[14]. Regardless of the mechanism, it is clear that an RNAi-independent pathway(s) exists that relies on ncRNAs to target heterochromatin.

Since transcription and ncRNAs have been linked to epigenetic chromatin modifications in multiple organisms including mammals[42,47,54-56], our results may have general significance. In several instances, transcription and ncRNAs can modify chromatin independent of RNAi. For example, RNAPII transcription and ncRNAs trigger chromatin modifications and parental imprinting in mammals[54,57] In such cases, RNAs retained in the nucleus act largely in cis. The retention of bidirectional transcripts near their transcription sites might facilitate the localization of chromatin modifying activities. To this end, widespread transcription of eukaryotic genomes[41] might allow RNAPII and ncRNAs to function as a molecular sensors that specify certain genomic regions as preferential targets for repressive chromatin assembly. Such domains might include transposons that when uncontrolled can lead to genome instability.