DNA–RNA hybrid formation mediates RNAi-directed heterochromatin formation
Communicated by: Yasushi Hiraoka
Abstract
Certain noncoding RNAs (ncRNAs) implicated in the regulation of chromatin structure associate with chromatin. During the formation of RNAi-directed heterochromatin in fission yeast, ncRNAs transcribed from heterochromatin are thought to recruit the RNAi machinery to chromatin for the formation of heterochromatin; however, the molecular details of this association are not clear. Here, using RNA immunoprecipitation assay, we showed that the heterochromatic ncRNA was associated with chromatin via the formation of a DNA–RNA hybrid and bound to the RNA-induced transcriptional silencing (RITS) complex. The presence of DNA–RNA hybrid in the cell was also confirmed by immunofluorescence analysis using anti-DNA–RNA hybrid antibody. Over-expression and depletion of RNase H in vivo decreased and increased the amount of DNA–RNA hybrid formed, respectively, and both disturbed heterochromatin. Moreover, DNA–RNA hybrid was formed on, and over-expression of RNase H inhibited the formation of, artificial heterochromatin induced by tethering of RITS to mRNA. These results indicate that heterochromatic ncRNAs are retained on chromatin via the formation of DNA–RNA hybrids and provide a platform for the RNAi-directed heterochromatin assembly and suggest that DNA–RNA hybrid formation plays a role in chromatic ncRNA function.
Introduction
The higher-order chromatin structure is an important determinant for epigenetic gene regulation and maintenance of genome integrity. A recent study showed that a specific combination of histone modifications determines chromatin higher-order structures. One important issue of chromatin higher-order structure is its temporal and spatial regulation. One of the mechanisms involved in this regulation involves sequence-specific DNA-binding proteins, which recruit histone-modifying enzymes. In addition to DNA-binding proteins, recent analyses showed that many noncoding RNAs (ncRNAs) are transcribed from the eukaryotic genome and that some are involved in the regulation of chromatin higher-order structure (Bernstein & Allis 2005; Cam et al. 2009), such as dosage compensation in mammals (Chang et al. 2006) and fruit flies (Rea & Akhtar 2006). In both cases, relatively long ncRNAs (i.e., xist RNA in mammals and rox RNA in flies) are transcribed from the X chromosome and associate with chromatin to set up chromatin higher-order structure for transcriptional inactivation and activation, respectively. Similarly, the Kcnq1ot1 and Air RNAs associate with chromatin and are essential for the onset of silent chromatin at imprinting regions (Nagano et al. 2008; Pandey et al. 2008). These ncRNAs are assumed to function as a scaffold for silent chromatin assembly.
RNAi-mediated transcriptional gene silencing has been reported in fission yeast and plants (Bühler & Moazed 2007; Grewal & Elgin 2007). In these systems, ncRNAs transcribed from silenced loci appear to be important for gene silencing. During RNAi-mediated heterochromatin formation in fission yeast, ncRNAs are transcribed from heterochromatic repeats by RNA polymerase II (Pol2) (Djupedal et al. 2005; Kato et al. 2005). ncRNAs bind to the RNA-induced transcriptional silencing (RITS) complex, which is an RNAi effector complex that comprises an argonaute family protein, Ago1, and a small interference RNA (siRNA) derived from the ncRNA(Noma et al. 2004; Verdel et al. 2004). siRNAs are thought to be used as guiding molecules for the binding of RITS to ncRNAs. RITS complex bound to ncRNA recruits the RNA-dependent RNA polymerase complex (RDRC) to produce double-stranded RNA (Motamedi et al. 2004), which is processed to siRNA by the Dicer RNase (Verdel et al. 2004; Colmenares et al. 2007). In parallel, RITS recruits Clr4 histone methyl transferase to methylate lysine 9 of histone H3 (H3K9), which attracts the heterochromatic protein Swi6 and Chp2, fission yeast homologues of HP1, and results in the formation of heterochromatin (Verdel et al. 2004; Zhang et al. 2008; Bayne et al. 2010). Binding of RITS to ncRNA is a key step for the RNAi-directed formation of heterochromatin. This is shown by the fact that the artificial recruitment of the RITS complex using an RNA-binding protein to ura4 mRNA induces the generation of the ura4-siRNA and heterochromatin formation at the ura4 locus (Bühler et al. 2006). As siRNA generation seems to be coupled with transcription (Kato et al. 2005) and with the localization of RITS at heterochromatin (Verdel et al. 2004), heterochromatic ncRNAs are assumed to remain on chromatin and function as a platform for the RNAi machinery (Cam et al. 2009). Similarly, during RNAi-mediated transcriptional gene silencing in plants, ncRNAs work on chromatin as a platform for the assembly of RNAi factors, which include argonaute family proteins (e.g., see Wierzbicki et al. (2009)). However, there is no direct evidence of the association of ncRNAs with chromatin in these systems, and the mechanisms that underlie the association are not understood.
In this report, we used an RNA precipitation assay that was targeted to histones to confirm that ncRNAs transcribed from heterochromatic repeats in fission yeast bound to chromatin. Surprisingly, we found that the ncRNAs were retained on chromatin through the formation of a DNA–RNA hybrid. Immunofluorescence analysis using an antibody specific for the DNA–RNA hybrid detected nuclear foci that colocalized with the heterochromatic regions and the foci were diminished by the RNase H treatment, which confirmed the formation of DNA–RNA hybrid in vivo. This hybrid formation did not depend on heterochromatin, whereas it required the RNAi machinery. Over-expression or depletion of RNase H caused a decrease and an increase in hybrid formation and chromatin association of ncRNA, respectively. Both situations disrupted heterochromatin formation, which suggests a dosage effect of the DNA–RNA hybrid on heterochromatin formation. Similarly, a DNA–RNA hybrid was formed in artificial heterochromatin induced by tethering of RITS to the ura4 RNA, and overproduction of RNase H inhibited the establishment and maintenance of artificial heterochromatin. These results showed the importance of the retention of ncRNAs on chromatin via the DNA–RNA hybrid to form the RNAi-directed heterochromatin.
Results
Detection of chromatin-bound ncRNA by histone-RIP assay
To investigate the association of ncRNAs with chromatin, we performed a RIP assay using an antibody against histone. In the RIP assay, association of specific RNAs with a protein of interest is detected by performing RT-PCR analysis on immunoprecipitates from in vivo formaldehyde-cross-linked cells (Gilbert et al. 2004). The RIP assay using antibodies against histone (the histone-RIP assay) detects chromatin-associated RNAs (Fig. S1 in supporting information) It should be noted that, in the RIP assay performed previously (Gilbert et al. 2004; Motamedi et al. 2004), cell extracts prepared from cross-linked cells were treated with DNase I to degrade DNA before the immunoprecipitation of target proteins not to detect the indirect RNA–protein interaction mediated by DNA, whereas in this study the samples were treated with DNase I after immunoprecipitation of target proteins (Fig. S1 in Supporting Information), because we wanted to precipitate the intact chromatin–RNA complexes and DNase I treatment before precipitation would disrupt the complexes.
Heterochromatic ncRNA was transcribed from both strands of centromeric dh repeats (Fig. 1A) (Volpe et al. 2002; Dutrow et al. 2008), and the ncRNA was shown to bind to the RITS complex via the RIP assay (Motamedi et al. 2004). To confirm the reliability of our RIP assay, we precipitated myc-tagged Chp1, which is a component of the RITS complex (Verdel et al. 2004) that was shown to bind ncRNA using the previous RIP assay (Motamedi et al. 2004) and an anti-myc antibody. We detected the association of Chp1 with heterochromatic ncRNA (Fig. 1B; Motamedi et al. 2004). It should be noted that the high background signal obtained in the RT-PCR performed without reverse transcriptase (−RT), which also observed in other RIP assays in Fig. 1, was the result of DNA–RNA hybrid formation, as shown below.
Histone-RNA immunoprecipitation (RIP) assay detected the association of heterochromatic ncRNAs with chromatin. (A) Schematic diagram showing the structure of the centromere region of chromosome 2 (cen2), which includes the outer repeats (dg, dhI, dhII), the innermost repeats (imr2) and the center region (cnt2). The positions of the forward and reverse ncRNAs (dhfor and dhrev) transcribed from the dh repeats are indicated. The box indicates the position of a primer set used for the detection of the dh ncRNA. (B–D) RIP assays in the presence (+) or absence (−) of RT were performed to detect ncRNA associated with Chp1 tagged with myc tag (B), histone H3 (C) and Pol2 (D). The association of act1 and cdc2 mRNAs with H3 and Pol2 was also analyzed in (C) and (D). It should be noted that the relative amounts of act1 mRNA, cdc2 mRNA and dhfor ncRNA in the input fraction were 1, 0.03 and 0.005, respectively. (E) Schematic indication of the mating-type locus and the subtelomeric region of the left arm of chromosome 1. The mat2 and mat3 genes, the cenH region (orange box), which is highly homologous to the centromeric repeats, IR-L and IR-R repeats, which form the boundary of heterochromatin, and a subtelomeric gene (SPAC212.11, white boxes) containing cenH-like sequences (orange boxes) are indicated. The boxes and arrows indicate the positions of the primer sets used and the direction of transcripts detected by qRT-PCR in (F), respectively. (F) Results of histone-RIP assays using antibodies against H3 (H3-RIP) in the presence (+RT) or absence (−RT) of RT. The proportion of precipitated RNA to input RNA in +RT samples calculated from quantitative RT-PCR data was plotted, with error bars showing the standard error of the mean (n = 3).
The H3-RIP assay effectively detected both forward and reverse transcripts from dh repeats (Fig. 1C), which showed that both ncRNAs associate with heterochromatin. Under the same conditions, the act1 and cdc2 mRNAs used as control were poorly precipitated with histone H3 (Fig. 1C), which indicated that mRNAs did not associate with chromatin. As the cdc2 mRNA is transcribed constitutively but is less abundant than the act1 mRNA (the relative amounts of act1 mRNA, cdc2 mRNA and dhfor ncRNA in the input fraction were 1, 0.03 and 0.005, respectively), the poor association of mRNA with chromatin was not affected by the efficiency of transcription. A RIP assay using the anti-Pol2 antibody, which can recognize elongating polymerase (Das et al. 2007), indicated that a fraction of heterochromatic ncRNA, act1 mRNA and cdc2 mRNA, which probably represent elongating transcripts, was associated with Pol2 (Fig. 1D). Interestingly, a larger fraction of heterochromatic ncRNA than act1 or cdc2 mRNAs associates with Pol2, which suggests that heterochromatic ncRNAs tend to remain close to Pol2. Because cdc2 mRNA that is far less abundant than act1 mRNA more weakly associated with Pol2, the weak transcription of dhfor ncRNA may not cause the higher association of the transcripts with Pol2. Thus, heterochromatic ncRNA appeared to remain at transcription sites, where it is associated with chromatin after transcription, whereas most mRNA was promptly released from chromatin after transcription.
In addition to the centromere, other RNAi-mediated heterochromatin exists at the mating-type region (mat locus) and at the subtelomeres in fission yeast. The mat locus contains sequences that are highly homologous to centromeric repeats (cenH), and the subtelomeric region of chromosomes 1 and 2 also contains cenH-like sequences (Fig. 1E). Both sequences are transcribed and act as nucleation sites for RNAi-mediated heterochromatin assembly (Noma et al. 2004; Cam et al. 2005; Kanoh et al. 2005). The H3-RIP assay detected chromatin-associated ncRNAs at the mat locus and at the subtelomeric region (Fig. 1F). Thus, the association of heterochromatic ncRNA with chromatin was generally observed in RNAi-mediated heterochromatin formation.
Formation of an DNA–RNA hybrid by heterochromatic ncRNA
Using H3-RIP assays of heterochromatic ncRNA, we observed a high background signal in the reaction lacking reverse transcriptase (−RT) (Fig. 1C,F), which typically represents DNA contamination. Similar signals in −RT reactions were also observed in the Chp1- and Pol2-RIP assays of dhfor ncRNA, but not in the Pol2-RIP assay of act1 mRNA (Fig. 1B,D). Therefore, the high background detected in −RT reactions seemed to be specific for heterochromatic ncRNA. We suspected the presence of DNA–RNA hybrids that are resistant to DNase I (Sutton et al. 1997). To test this possibility, we treated RNA-IP samples with RNase H, which specifically degrades the RNA in DNA–RNA hybrids, before DNase I treatment. If DNA–RNA hybrids existed in the precipitates of the RIP assay, sequential treatment with RNase H and DNase I would decrease both +RT and −RT signals. RNase H treatment decreased the signal of the H3-RIP assay of dhfor RNA to 40–50% of the signal in the absence of RNase H treatment in both +RT and −RT reactions (Fig. 2A), which suggests that a significant portion of the RNA detected by the H3-RIP assay formed DNA–RNA hybrids. A RIP assay using an anti-Swi6 antibody indicated that the dhfor ncRNA associated with Swi6 and that most of it was also RNase H-sensitive (Fig. 2B), which confirmed the formation of a DNA–RNA hybrid at heterochromatin. Importantly, a fraction of the ncRNA bound to Chp1 or Pol2 was also RNase H-sensitive, whereas the act1 mRNA associated with Pol2 was not (Fig. 2C). This suggested that DNA–RNA hybrids are located close to Pol2 and RITS. The RIP assay against Rdp1 did not show such a high background signal (Fig. S2 in Supporting Information), suggesting that RDRC associates with ncRNA at a position away from the DNA–RNA hybrid. Similar DNA–RNA hybrid formation was observed at the mating locus and at subtelomeric heterochromatin (Fig. 2D), which suggests that hybrid formation is a common feature of ncRNAs transcribed at heterochromatic regions. It should be noted that although a 8- to 9-bp-long DNA–RNA hybrid is formed at the Pol2 complex during RNA synthesis (Gnatt et al. 2001), this hybrid was not detected by our RIP assay, even at the highly transcribed act1 gene (Fig. 2C). As we used primer sets that produced RT-PCR products that were ∼150 bp long, the length of the hybridized region is likely to be greater than 150 bp. These results suggest that the formation of a DNA–RNA hybrid by ncRNA is coupled with transcription and that the resulting ncRNA on chromatin becomes a target for the RITS complex.
Heterochromatic ncRNA formed a DNA–RNA hybrid. (A–D) Heterochromatic ncRNA formed a DNA–RNA hybrid. A RNA immunoprecipitation (RIP) assay was performed with (+) or without (–) RNase H treatment and in the presence (+RT) or absence (−RT) of RT to detect centromeric dh forward ncRNA associated with H3 (A), Swi6 (B), and Chp1 and Pol2 (C). In the Pol2-RIP assay (C), the association of act1 mRNA with Pol2 was also examined. (D) ncRNAs transcribed from the mating-type locus and subtelomere associated with chromatin. A RIP assay was performed with (+) or without (−) RNase H treatment. (E, F) DNA–RNA hybrid formation of RNAs transcribed from the marker genes inserted into centromeric heterochromatin. (E) Schematic diagram showing the structure of the centromere region of chromosome 1 (cen1), which includes the outer repeats (dg, dhI, dhII ), the innermost repeats (imr1), the center region (cnt1) and the inserted marker genes (ura4 and ade6). The positions of forward and reverse ncRNAs (dgfor and dgrev) transcribed from the dg and imr1 repeats are indicated by arrows. The box indicates the position of the primer set used for the detection of the ura4 and ade6 ncRNAs. (F) Results of histone-RIP assays for the inserted marker genes using antibodies against H3 (H3-RIP) with (+) or without (−) RNase H treatment. The proportion of precipitated RNA to input RNA calculated from quantitative RT-PCR data was plotted, with error bars showing the standard error of the mean (n = 3). The proportion of precipitated RNA to input RNA calculated from strand-specific qRT-PCR was plotted, with error bars showing the standard error of the mean (n = 3). P values were determined in (A–D) using a two-sided Student’s t-test: **P < 0.01, n = 3; ***P < 0.001, n = 3.
We tested for the presence of DNA–RNA hybrids at four transcriptionally active loci (Dutrow et al. 2008) on centromeric dh repeats (Fig. S3 in Supporting Information) using an antibody against methylated histone H3K9 that represents heterochromatin. At all four sites examined, 40–60% of the chromatin-associated ncRNAs were sensitive to RNase H, which suggests that the DNA–RNA hybrids are formed entire transcribed-region in the centromeric dh repeats, although it is not clear whether the results represent the formation of small DNA–RNA formed at each region or the formation of long DNA–RNA hybrid encompassing entire region.
When euchromatic genes such as ade6 or ura4 are inserted into centromeric heterochromatin, the heterochromatin structure (H3K9me and Swi6/Chp2) spreads into the inserted genes and suppresses gene expression. However, a small amount of RNA is transcribed from the inserted genes and is converted to siRNA (Bühler et al. 2007), which indicates that the RNAi system is operative at the inserted transcription units. Therefore, we analyzed the association between the ade6 and ura4 mRNAs inserted into centromeric repeats (otr::ade6 and imr::ura4; Fig. 2E) using the H3-RIP assay. The ade6 and ura4 mRNAs did not associate with chromatin at native loci, such as the act1 or cdc2 mRNAs (1, 2), whereas the insertion of the genes into centromeric repeats led to the coprecipitation of a portion of the ade6 or ura4 mRNAs with histone H3 (Fig. 2F). In addition, as observed with centromeric ncRNA, this fraction was sensitive to RNase H treatment (Fig. 2F), which suggests that the local environment of centromeric repeats induces the formation of DNA–RNA hybrids.
To confirm the formation of the DNA–RNA hybrid in vivo, immunofluorescence analysis was performed using an antibody specific for the DNA–RNA hybrid (S9.6) (Boguslawski et al. 1986) (Fig. 3). Chp1-GFP was used as a control, which formed a few (1–3) nuclear foci that represent heterochromatin (Sadaie et al. 2004). The anti-DNA–RNA hybrid antibody detected signals in the cytoplasm of almost all cells, and the number of signals varied from cell to cell. The cytoplasmic signals may arise from mitochondrial DNA because mitochondrial DNA is thought to harbor an R-loop at its origin of replication (Brown et al. 2008) and DAPI-staining signals in the cytoplasm colocalized with the foci (Fig. S4 in Supporting Information). Nuclear foci that mainly localized to the periphery of the nucleus were also observed, and some of these foci colocalized with one of the Chp1-GFP signals (Fig. 3, arrows) in approximately 5% of the cells, showing the existence of DNA–RNA hybrids in some of the heterochromatic regions. One of the reasons for the low percentage of colocalization would be the S-phase-specific transcription of heterochromatic ncRNA (Chen et al. 2008). The nuclear foci that did not colocalize with heterochromatin (Fig. 3, arrowheads) may indicate DNA–RNA hybrids formed at euchromatic regions. Importantly, both nuclear and cytoplasmic foci detected by the DNA–RNA hybrid antibody were diminished by the treatment of the cells with RNase H before immunostaining (Fig. 3, lower panels), which confirmed that the antibody detected the DNA–RNA hybrid. These results showed that the DNA–RNA hybrid is formed in the heterochromatic region in vivo.
Detection of DNA–RNA hybrids by immunofluorescence using a monoclonal antibody against DNA–RNA hybrid. (Upper panels) The localization of DNA–RNA hybrid and heterochromatin was visualized by immunofluorescence with anti-DNA–RNA hybrid (S9.6, Boguslawsky et al. 1986) and anti-GFP antibodies, respectively, using cells expressing Chp1 tagged with GFP (Chp1-GFP). DAPI staining indicates nuclear DNA. Cells observed by Normarski optics are indicated in the right panels, Arrows and arrowheads indicate the nuclear DNA foci overlapped with Chp1 foci and not overlapped with Chp1 foci, respectively. (Bottom panels) Cells were treated with RNase H before immunodetection (Experimental procedures). Both cytoplasmic and nuclear signals detected by the anti-DNA–RNA hybrid antibody are diminished by RNase H treatment.
Requirements for DNA–RNA hybrid formation of heterochromatic ncRNA
To examine whether heterochromatin is required for chromatin association or DNA–RNA hybrid formation, we performed the H3-RIP assay on yeast strains that carry defects in heterochromatin formation (Fig. 4A). Compared with the wild-type strain, the increased level of ncRNA was retained on chromatin in deletion strains of swi6 or chp2 (both of which encode major structural components of heterochromatin). Similarly, ncRNA bound to chromatin in clr4Δ cells, in which heterochromatin was completely erased by the loss of the H3K9-specific methyltransferase Clr4. Hence, heterochromatin per se was not required for DNA–RNA hybrid formation and chromatin binding and hybrid formation are intrinsic properties of heterochromatic ncRNA. Interestingly, the RNase H-sensitive fraction of chromatin-associated ncRNA was significantly increased in swi6Δ, chp2Δ, and clr4Δ cells, suggesting that heterochromatin decreased DNA–RNA hybrid formation. Notably, clr4 deletion did not influence chromatin association of otr::ade6 or imr::ura4 RNAs; rather, it increased the RNase H-sensitive fraction of the RNAs (Fig. 2F).
Requirements for the formation of DNA–RNA hybrids. (A) H3-RNA immunoprecipitation (RIP) assays for DNA–RNA hybrid formation of dh forward ncRNA in the indicated mutants. The relative amount of signal obtained from each mutant to that of the wild type was plotted. Error bars represent SEM (n = 2). (B) Schematic indication of the artificial heterochromatin formation by RNA-induced transcriptional silencing (RITS) tethering to the ura4 mRNA. In this system, RITS was tethered artificially to the ura4 RNA via binding of the λN protein fused to Tas3 (which is a subunit of RITS) to its recognition sequence, boxB, five copies of which are inserted into 3′ UTR of the ura4 mRNA. This induced siRNA generation and heterochromatin at the ura4 locus in an RNAi-dependent manner. (C) Gene silencing of ura4 via tethering of RITS. Strains harboring the RITS-tethering system (ura4-5boxB, tas3λN) or ura4-5boxB alone were grown to a density of 1.0 × 107 cells/mL and serial dilutions (1:5) of the cultures were spotted onto nonselective (N/S), counterselective [fluoro-orotic acid (FOA)] and selective (−URA) plates for silencing of the ura4 gene. A strain harboring clr4Δ was also included, as a control. A FOA-resistant colony was picked, grown in nonselective medium for 10 generations and respotted on the plates to analyze the stability of the silent state (box with an arrow and lower panel). (D) H3-RIP assays in the presence or absence of RNase H treatment were performed to detect DNA–RNA hybrid formation in the RITS-tethering system, using cells grown in nonselective condition (in which only a small fraction of cells formed heterochromatin) and cells selected for FOA resistance (in which heterochromatin was stably formed) (see panel C).
Next, we analyzed the effect of mutation of the RNAi machinery (Fig. 4A). The deletion of genes encoding a subunit of the RITS complex (chp1 or ago1), RDRC (rdp1 or hrr1) or an siRNA-generating RNase (dcr1) resulted in a significant decrease in the association of ncRNA with chromatin. Similarly, the rpb2-m203 mutation, which is a point mutation in Pol2 that disturbs siRNA generation but does not affect ncRNA transcription (Kato et al. 2005), also resulted in a decrease in the association of dhfor ncRNA with chromatin. These results suggest that the RNAi machinery is required for the efficient association of heterochromatic ncRNA with chromatin via DNA–RNA hybrid.
Binding of the RITS complex to ncRNA is a key step for the RNAi-directed formation of heterochromatin; it induces H3K9 methylation (for heterochromatin formation) and processing of ncRNA (for siRNA generation). This was shown by the fact that the tethering of RITS to the ura4 RNA induces RNAi- and heterochromatin-dependent gene silencing of the ura4 gene (Bühler et al. 2008). Tethering of RITS is achieved by the fusion of Tas3, which is a subunit of RITS, with the λN protein, which binds to the 5BoxB sequence inserted at the 3′ UTR region of the ura4 RNA (Fig. 4B). In this system, the ura4 RNA should remain on chromatin to recruit RITS close to chromatin. Thus, we analyzed the chromatin binding and DNA–RNA hybrid formation of the ura4 RNA in the RITS-tethering system using the H3-RIP assay. The expression of the Tas3-λN protein in cells harboring the ura4-5BoxB gene rendered only a small portion (approximately 0.1–0.5%) of the cells resistant to 5-fluoro-orotic acid (FOA), which indicates silencing of ura4 gene expression. Once the silent state was established, it was stably maintained for generations under nonselective conditions; approximately 90% of cells were FOA resistant after 10 generations (Fig. 4C). We compared the chromatin association of ura4-5BoxB RNA in Tas3-λN-expressing cells that were not selected or selected for FOA resistance. In nonselected cells, only a small portion of the ura4-5boxB RNA bound to histone H3, which is similar to what was observed for native ura4 RNA (<0.1% of the input RNA) (Fig. 2F). In contrast, FOA-resistant cells exhibited a significant amount of ura4-5boxB RNA bound to chromatin and most of the RNA was RNase H-sensitive (Fig. 4D). These results suggest that simple association of RITS complex with RNA did not induce DNA–RNA hybrid formation; rather, DNA–RNA hybrid formation correlated with heterochromatin formation induced by RITS tethering.
Requirement for a specific amount of DNA–RNA for hybrid heterochromatin formation
To test the importance of the DNA–RNA hybrid in the formation of heterochromatin, we first tried over-expressing RNase H in cells, as the over-expression of RNase H reduces the amount of DNA–RNA hybrids in human cells and budding yeast (Huertas & Aguilera 2003; Li & Manley 2005). Over-expression of rnh201, which encodes the catalytic subunit of RNase H2, neither affected the growth rate (results not shown, Fig. S6 in Supporting Information, N/S) nor the DNA content of cells (Fig. S5B in Supporting Information), whereas it caused a decrease in the level of RNase H-sensitive H3-associated ncRNA at heterochromatin (Fig. 5A). This showed a reduction in the amount of DNA–RNA hybrids. Under the same conditions, the level of H3K9me was decreased to that observed in dcr1Δ cells (Fig. 5B), in which the RNAi-mediated heterochromatin formation system is defective. Swi6, as measured using the chromatin immunoprecipitation (ChIP) assay, was also significantly reduced (Fig. 5B), although the decrease was less than that of H3K9me. These results indicate that the reduction in the amount of DNA–RNA hybrids caused a disturbance in heterochromatin structure.
Effect of over-expression and deletion of RNase H in vivo. (A) The catalytic subunit of RNase H2, which is encoded by rnh201, was over-expressed from a strong nmt1 promoter and its effects on chromatin association and DNA–RNA hybrid formation of dhfor RNA were analyzed using the H3-RNA immunoprecipitation (RIP) assay. (B) Heterochromatin status was also analyzed by measuring the level of H3K9me and Swi6 using the ChIP assay. (C, D) Effects of deletion of rnh1 and rnh201, which encode the catalytic subunit of RNase H1 and H2, respectively, on the chromatin association and DNA–RNA hybrid formation of dhfor ncRNA (C) and on heterochromatin status (D). The proportion of precipitated RNA to input RNA calculated from the quantitative PCR data was plotted, with error bars showing the standard error of the mean (n = 3). P values were determined using a two-sided Student’s t-test: *P < 0.05, n = 3; **P < 0.01, n = 3; ***P < 0.001, n = 3. (E) Strains harboring imr::ura4 and a plasmid for the over-expression of RNase H or the indicated deletions were grown to a density of 1.0 × 107 cells/mL and serial dilutions (1:5) of the cultures were spotted onto nonselective (N/S), counterselective [fluoro-orotic acid (FOA)] and selective (−URA) plates for silencing of the ura4 gene.
The decrease in the quantity of DNA–RNA hybrids after over-expression of RNase H raised the possibility that RNase H is involved in the regulation of DNA–RNA hybrid formation at heterochromatin. Thus, we analyzed the effect of deletion of two genes that encode RNase H, rnh201 and rnh1 (the latter encodes the catalytic subunit of RNase H1). As expected, the amounts of RNase H-sensitive dhfor ncRNA increased in cells harboring a deletion of rnh201 or rnh1, by approximately two- and threefold, respectively (Fig. 5C). This showed that both gene products contributed to the decrease in the levels of DNA–RNA hybrids. Interestingly, the levels of H3K9me and Swi6 were decreased in these mutants (Fig. 5D). Moreover, the greater the increase in DNA–RNA hybrid levels in rnh1Δ cells, the greater the decrease in H3K9me. It should be noted that the reduction in Swi6 was more obvious than that of H3K9me in rnh201Δ cells. Together with the result showing that Swi6 tended to remain on chromatin in rnh201-over-expressing cells (Fig. 5B), these findings led us to suspect that the DNA–RNA hybrids may inhibit the association of Swi6 with heterochromatin, independently of H3K9me.
We next analyzed the effect of overproduction or depletion of RNase H on the silencing of the ura4 gene inserted into heterochromatin (imr::ura4) (Fig. 2E). Consistent with the reduction of H3K9me and Swi6, we observed that both over-expression of rnh201 and deletion of rnh201 or rnh1 caused the decrease in the silencing of imr::ura4, which was shown by the enhanced growth on the plates lacking uracil (-URA, Fig. 5E). The results suggest that the proper turnover of the DNA–RNA hybrid is important for heterochromatin formation and that RNase H is involved in the regulation of this process.
To test the importance of DNA–RNA hybrid formation in RITS-tethering-induced heterochromatin formation, we analyzed the effect of overproduction of RNase H on the establishment and maintenance of the silent state (Fig. 6A). Over-expression of Rnh201 in RITS-tethering cells led to a significant decrease in the number of FOA-resistant cells (Fig. 6A, upper panel); the conversion rate from the FOA-sensitive to the FOA-resistant state per generation, which represents the efficiency of heterochromatin establishment, was reduced from 1.4% to 0.05% (Fig. 6B). Therefore, over-expression of RNase H inhibited the formation of RITS-tethering-mediated heterochromatin. We picked FOA-resistant colonies from control and Rnh201-over-expressing cells, cultured them in nonselective medium for more than 10 generations and then analyzed the number of FOA-resistant cells. For control cells, the number and size of the colonies on FOA plates were similar to those grown on nonselective plates (Fig. 6B, lower panel, and Fig. S6 in Supporting Information), which suggests the stable maintenance of the silent state. For Rnh210-over-expressing cells, the number of colonies was decreased significantly and their size was variable on FOA plates, which showed the instability of the silent state (Fig. 6A and Fig. S6 in Supporting Information). The conversion rates from the FOA-resistant to the FOA-sensitive state in control cells and in Rnh201-over-expressing cells were 1.0% and 12%, respectively (Fig. 6B, right panel). This indicates that over-expression of Rnh201 destabilized RITS-tethering-mediated silencing. These results suggest that the association of ura4-5BoxB RNA with chromatin via DNA–RNA hybrid formation is a necessary step for the establishment and maintenance of RITS-tethering-induced heterochromatin.
Effect of the over-expression of RNase H on gene silencing by artificial heterochromatin induced by RNA-induced transcriptional silencing (RITS) tethering was analyzed using a spot assay. (A) Spot assays were performed using cells harboring RITS-tethering system, a plasmid for the over-expression of RNase H, or vector alone. Indicated strains that did not harbor the plasmid were also included as a control. Two fluoro-orotic acid (FOA)-resistant colonies from each strain were picked, and stability of the FOA resistance was analyzed after growth for 10 generations in nonselective medium (arrow and lower panel). (B) Conversion rates from FOA-sensitive (FOAS) to FOA-resistant (FOAR) or FOAR to FOAS were measured after 10 generations of growth in nonselective medium. The former represents the efficiency of heterochromatin establishment, whereas the latter represents the stability of heterochromatin. The conversion rates were calculated in at least three independent experiments, and the standard error was indicated.
Discussion
Here, we showed that ncRNA transcribed from heterochromatin was associated with chromatin via the formation of a DNA–RNA hybrid. Importantly, hybrid formation was involved in heterochromatin formation. We propose a model for the association of heterochromatic ncRNA with chromatin and its function in heterochromatin formation (see Fig. 7I–III).
Model for the association of heterochromatic ncRNA with chromatin (I) DNA–RNA hybrid is formed at the heterochromatic repeats by transcription-coupled mechanisms. (II) RNA-induced transcriptional silencing (RITS) complex targets to the chromatin-associated RNA (or the single-stranded DNA) produced by DNA–RNA hybrid to generate siRNA and heterochromatin. (III) The amount of DNA–RNA hybrid is regulated by RNase H and possibly by DNA/RNA helicases.
As shown in Fig. 7(I), the ncRNAs transcribed by Pol2 from centromeric repeats form DNA–RNA hybrids, probably via a transcription-coupled mechanism. In budding yeast, transcription-coupled DNA–RNA hybrid formation at normal gene loci is induced by the depletion of HPR1, which functions at the interface of transcription and mRNA metabolism as a component of the THO/TREX complex, or Sen1, which functions in transcription termination as a component of NRD complex (Huertas & Aguilera 2003; Mischo et al. 2011). Similarly, depletion of the ASF/SF2 (alternative splicing factor 2) in avian lymphoid cells or senataxin (homologue of Sen1) in human cells leads to DNA–RNA hybrid formation (Li & Manley 2005; Skourti-Stathaki et al. 2011). These results suggest that a mechanism that includes mRNA export/processing factors is involved in the prevention of DNA–RNA hybrid formation at normal protein-coding genes. Therefore, we speculate that an active mechanism induces DNA–RNA hybrid formation. Recently, several splicing factors were shown to be involved in RNAi-directed heterochromatin formation (Bayne et al. 2008; Chinen et al. 2010). Considering the requirements of RNAi machinery for DNA–RNA hybrid formation, we suggest that hybrid formation involves the cooperation of the RNAi machinery and the RNA processing machinery (Fig. 7Ia).
Our results suggest that there are two distinct mechanisms of DNA–RNA hybrid formation, heterochromatin-dependent and heterochromatin-independent mechanisms. As heterochromatin ncRNA transcription exhibited an intrinsic property to form DNA–RNA hybrid, independent of heterochromatin (Fig. 4), cis element(s) may confer this property to the transcription apparatus. This might be a similar mechanism to DNA–RNA hybrid formation at the DNA replication origin of the colE1 plasmid in Escherichia coli, in which a specific secondary structure of the transcribed primer RNA triggers DNA–RNA hybrid formation via an interaction with RNA polymerase (Itoh & Tomizawa 1979; Masukata & Tomizawa 1986). Alternatively, convergent transcription may be a signal for DNA–RNA hybrid formation, as a couple of recent reports suggested that the convergent transcription triggers RNAi-directed heterochromatin formation (Gullerova & Proudfoot 2008; Iida et al. 2008). In both situations, transcribed RNA should be retained on chromatin and become a target for RITS for the formation of heterochromatin. We found that transcripts derived from the euchromatic genes inserted into heterochromatic region (otr::ade6 and imr::ura4) formed DNA–RNA hybrids in a heterochromatin-independent manner (Fig. 2F). We assumed that the observed DNA–RNA hybrid was derived from read-through transcripts of heterochromatic ncRNA. This type of read-through product was reported previously and was suggested to be necessary for the spreading of heterochromatin onto inserted genes (Irvine et al. 2006). Alternatively, read-through transcripts from the downstream of the marker genes may generate the convergent transcription, which could induce the DNA–RNA hybrid formation.
In addition to the ncRNA-specific hybrid formation, there might be another mode of DNA–RNA hybrid formation, a heterochromatin-dependent system (Fig. 7Ib), because the hybrid was formed only when heterochromatin was formed in an RITS-tethering system (Fig. 4B–D). In this system, the ura4 gene was transcribed from its native promoter, and there was no heterochromatic DNA/RNA element except 5BoxB sequence. Importantly, simple tethering of RITS to ura4 RNA did not induce DNA–RNA hybrid formation; we could detect DNA–RNA hybrid after cells established heterochromatin. Therefore, heterochromatin formation on the ura4 gene itself seemed to induce DNA–RNA hybrid formation. We propose, hence, that there are two modes of DNA–RNA hybrid formation: the heterochromatin-independent mode, which requires the cis element present in heterochromatic repeats, and the heterochromatin-dependent mode, which does not require the cis element. The latter mode of hybrid formation might contribute to the spreading and maintenance of heterochromatin by inducing siRNA generation from transcription units that are newly embedded in the heterochromatin.
We assumed that the RITS complex is targeted to the ncRNA that is tethered on the chromatin via partial DNA–RNA hybrid formation (Fig. 7IIc). Alternatively, single-stranded DNA generated in a R-loop, which might be covered by single-stranded DNA proteins, such as RP-A, may become a target of the RITS complex (Fig. 7IId), as it is by the recombination protein AID during immunoglobulin class switch recombination(Reaban & Griffin 1990; Yu et al. 2003). In both cases, the chromatin-associated RITS complex promotes siRNA production and heterochromatin formation. It is noteworthy that Drosophila Piwi, which is an argonaute family protein implicated in heterochromatin formation (Pal-Bhadra et al. 2004; Huisinga & Elgin 2009), was shown to associate with the heterochromatic region and to be released from heterochromatin by RNase H treatment (Brower-Toland et al. 2007). This finding suggests that the Piwi complex may thus target DNA–RNA hybrids formed at heterochromatin, as is the case in fission yeast.
The produced DNA–RNA hybrid must be removed quickly because the R-loop would inhibit subsequent transcription, which would prevent the supply of ncRNA for heterochromatin formation. This conclusion is supported by the observation that the deletion of RNase H genes resulted in an increase in DNA–RNA hybrid levels and in the disruption of heterochromatin (Fig. 5C,D). In addition to the S-phase-specific transcription of the heterochromatic ncRNA, the dynamic turnover of the DNA–RNA hybrid could explain why only a small fraction of cells showed the DNA–RNA hybrid signal during the immunofluorescence analysis (Fig. 3). Our results suggested that RNase H is involved in such a dynamic regulation of DNA–RNA hybrid. It is possible that DNA/RNA helicases like Sen1 (Senataxin) helicase, which is recently suggested to resolve DNA–RNA hybrid in the cell (Mischo et al. 2011; Skourti-Stathaki et al. 2011), are also involved in the regulation of DNA–RNA hybrid formed in heterochromatin. As RDRC binding to ncRNA occurs at a position away from DNA–RNA hybrid (Fig. S2 in Supporting Information), it is attractive to speculate that a putative helicase, Hrr1, in RDRC unwinds DNA–RNA hybrid and the released RNA is used as a substrate for dsRNA synthesis by RDRC.
Our immunofluorescence analysis using an antibody against the DNA–RNA hybrid sometimes detected nuclear signals that did not colocalize with Chp1 (Fig. 3 arrowheads), suggesting that the DNA–RNA hybrids exist in the euchromatic region. Considering the recent reports that show that some ncRNA transcribed at regulatory regions is involved in the regulation of transcription (Hirota et al. 2008; Wang et al. 2008; Kim et al. 2010; Bertani et al. 2011), we speculate that the DNA–RNA hybrids also play regulatory roles in the euchromatin. Indeed, ncRNA transcribed in the promoter region of the cyclin D1 gene in human cells and the intergenic region of Hoxa6 and Hoxa7 genes in mouse cells thought to form DNA–RNA hybrids and recruit the RNA-binding transcriptional regulatory proteins TLS and MLL1, respectively, to regulate the gene expression (Wang et al. 2008; Bertani et al. 2011). Because DNA–RNA hybrids inhibit nucleosome formation (Dunn & Griffith 1980), it is also possible that hybrid formation can be used to remove or reorganize nucleosomes around regulatory regions. Judging from the nuclear signals in the immunofluorescence analysis, euchromatic DNA-RNA hybrid regions might form clusters like heterochromatin. Because R-loop provides ssDNA region that can be used for pairing of homologous sequence, it is possible that the hybrids are formed at repeated sequences and contribute to cluster formation. In this sense, it is noteworthy that Tf2 retroposons make clusters into ‘Tf bodies’ in fission yeast (Cam et al. 2008).
The R-loop is hyper-recombinogenic, as has been shown for the induction of DNA–RNA hybrids by the mutation/depletion of the RNA processing/transport factors mentioned above (Huertas & Aguilera 2003; Li & Manley 2005; Mischo et al. 2011). It is noteworthy that deletion of swi6 or clr4, which exhibited a significant increase in hybrid formation (Fig. 4A), shows synthetic growth defects with deletion of several genes involved in DNA repair/recombination, whereas deletion mutants of dcr1 or chp1, which exhibited a decrease in hybrid formation (Fig. 4A), do not (Roguev et al. 2008). This may reflect the fact that an increase of the R-loop in heterochromatin in clr4Δ or swi6Δ cells in the cells deficient in DNA repair/recombination causes an increase in hyper-recombination at the R-loop to a level that is toxic for cell growth. Similar genetic interaction between SEN1 and DNA repair/recombination genes has been reported recently in budding yeast (Mischo et al. 2011). We want to point out that the hyper-recombinogenic property of the R-loop raised the possibility that DNA–RNA hybrid formation provides a hot spot for recombination or homologous chromosome pairing during meiosis. Therefore, DNA–RNA hybrid formation of ncRNA may regulate chromatin and/or DNA metabolism in the eukaryotic genome.
Experimental procedures
Schizosaccharomyces pombe strains and culture media
All strains used in this study are listed in Table S1 in Supporting Information. The media and genetic methods used in S. pombe experiments were as described (Moreno et al. 1991). Yeast cells were cultured in YES, EMMS at 30 °C. For deletion or epitope-tagging of the target genes, the PCR-based module method (Krawchuk & Wahls 1999) was used.
Antibodies and oligonucleotides
The following antibodies were used in this study: anti-c-myc (9E11; Santa Cruz), anti-Pol2 (8WG16; Abcam), anti-H3K9me monoclonal antibodies (Kato et al. 2005), anti-Swi6 polyclonal antibodies (Sadaie et al. 2004) and anti-DNA–RNA hybrid monoclonal antibodies (Boguslawski et al. 1986). Oligonucleotides used in this study are listed in Table S2 in Supporting Information.
Over-expression and depletion of RNase H
For the expression of RNase H in S. pombe cells, the rnh201 (SPAC4G9.02) cDNA was amplified by PCR from genomic DNA and ligated between the BamHI and SalI sites of pREP1, which contains the thiamine-repressible nmt1 promoter (Maundrell 1993). For deletion of the rnh1 (SPBC336.06c) and rnh201 genes, the PCR-based module method (Krawchuk & Wahls 1999) was used.
Chromatin immunoprecipitation (ChIP)
ChIP analysis was performed as described previously (Nakagawa et al. 2002), using the anti-Swi6 rabbit polyclonal antibody and the anti-H3K9me monoclonal antibody.
RNA immunoprecipitation
RNA immunoprecipitation was performed as described (Motamedi et al. 2004), with the following modifications: to prepare whole-cell extracts, cells were cross-linked with 1% formaldehyde, suspended in chilled lysis buffer (50 mm HEPES-KOH [pH 7.5] containing 140 mm NaCl, 1 mm EDTA, 1% Triton X-100 and 0.1% Na-deoxycholate) with proteinase inhibitors (Nacalai), RNase inhibitors (Promega) and 10 mm phenylmethylsulfonyl fluoride (PMSF) and disrupted with glass beads. Nucleic acids in whole-cell extracts were fragmented by sonication to an average length of 0.5 kb. The sample was immunoprecipitated using antibodies against target proteins and Dynabeads M-280 anti-mouse or anti-rabbit IgGs (Invitrogen). Precipitated protein–nucleic acid complexes were washed twice with lysis buffer, twice with lysis/NaCl buffer (50 mm HEPES-KOH [pH 7.5], 500 mm NaCl, 1 mm EGTA, 1% Triton X-100, 0.1% Na-deoxycholate, proteinase inhibitors, RNase inhibitors and 10 mm PMSF) and twice with wash buffer (10 mm Tris [pH 8.0], 250 mm LiCl, 1 mm EGTA, 0.5% NP-40, 0.5% Na-deoxycholate, RNase inhibitors and 10 mm PMSF). After washing, the beads were resuspended in diethylpyrocarbonate (DEPC)-treated distilled water. Samples were adjusted to 0.05% SDS and 100 μg/mL proteinase K and incubated for 45 min at 45 °C and then at 65 °C for reversal of cross-linking. Samples were then extracted once with phenol–chloroform and once with chloroform–isoamyl alcohol. After ethanol precipitation, samples were resuspended in a suitable volume of DEPC-treated distilled water.
Samples were treated with DNase I (150 U per 100 μL; Invitrogen) at 37 °C for 60 min before RT-PCR analysis. When necessary, nuclease treatment with RNase H and/or RNase T1 was performed before DNase I treatment: RNase T1 (1 U per 50 μL; Ambion) at 25 °C for 30 min and/or RNase H (1 U per 50 μL; Takara) at 37 °C for 60 min. RNA samples were reverse-transcribed into cDNA using PrimeScript Reverse Transcriptase (Takara) and strand-specific primers. We performed quantitative PCR on an ABI7500 real-time PCR machine (Applied Biosystems) using SYBR Premix ExTaq (Takara). Centromeric ncRNA, mating-type locus-specific ncRNA, subtelomeric ncRNA and act1 mRNA were detected using the primers listed in Table S2, each of which generated 150- to 180-bp-long PCR products. The ratio of precipitated RNA to input RNA in the samples treated with reverse transcriptase was calculated from quantitative RT-PCR data.
It should be noted that the difference in immunoprecipitation efficiency between experiments using the same antibody is caused by the difference in the lots of the antibodies.
Immunofluorescence analysis
Immunofluorescence analysis was performed as described in the study by Dohke et al. (2008).
Acknowledgements
We thank R. Allshire for the strains, J-I. Nakayama for the Swi6 antibody and S. H. Leppla for the S9.6 antibody. We also thank K. Ishii and our laboratory members for helpful discussions and support. Y. M. was supported by a Grant-in-Aid for Scientific Research (A) from the Japan Society for the Promotion of Science and by a Grant-in-Aid for Scientific Research on Priority Areas from the Ministry of Education, Culture, Sports, Science and Technology of Japan.
