RNA-interference-directed chromatin modification coupled to RNA polymerase II transcription
RNA interference (RNAi) acts on long double-stranded RNAs (dsRNAs) in a variety of eukaryotes to generate small interfering RNAs that target homologous messenger RNA, resulting in their destruction. This process is widely used to 'knock-down' the expression of genes of interest to explore phenotypes 1-3 . In plants 3-5 , fission yeast 6-8 , ciliates 9,10 , flies 11 and mammalian cells 12,13 , short interfering RNAs (siRNAs) also induce DNA or chromatin modifications at the homologous genomic
... us, which can result in transcriptional silencing or sequence elimination 14 . siRNAs may direct DNA or chromatin modification by siRNA-DNA interactions at the homologous locus 4,5 . Alternatively, they may act by interactions between siRNA and nascent transcript 15, 16 . Here we show that in fission yeast (Schizosaccharomyces pombe), chromatin modifications are only directed by RNAi if the homologous DNA sequences are transcribed. Furthermore, transcription by exogenous T7 polymerase is not sufficient. Ago1, a component of the RNAi effector RISC/RITS complex, associates with target transcripts and RNA polymerase II. Truncation of the regulatory carboxy-terminal domain (CTD) of RNA pol II disrupts transcriptional silencing, indicating that, like other RNA processing events    , RNAi-directed chromatin modification is coupled to transcription. In plant and mammalian cells siRNAs homologous to the open reading frame of a gene results in post-transcriptional silencing, degrading transcripts by means of RNAi. However, siRNAs homologous to a gene's promoter can induce transcriptional silencing, resulting in the modification of DNA and/or chromatin 3-5 . siRNAs may hybridize to DNA and thereby recruit DNA/chromatin modifying activities that effect silencing 4,5,14 . Alternatively, the RNAi machinery may target nascent transcripts and cause chromatin modification on templates homologous to loaded siRNAs 15,16 . At fission yeast centromeres and the silent mating type locus, noncoding RNAs are generated by the transcription of both strands of related repeats 6,20 . These form dsRNAs, which are cleaved by Dicer (Dcr1) into siRNAs and then loaded into the Ago1 (Argonaute)containing RITS (for RNA-induced initiation of transcriptional silencing) complex, which mediates RNAi 20,21 . Nascent transcripts may direct the RNAi machinery to the homologous locus, induce dimethylation of the surrounding chromatin on lysine 9 of histone H3 (H3K9me2) through Clr4, recruiting Swi6 (HP1) and thereby silencing transcription 15,16 . Components of RITS and RNAdependent RNA polymerase (Rdp1) are known to associate with, and act in cis on, this silent chromatin 6,21,22 . However, Chp1 (a RITS component) and Swi6 bind H3K9me2 (ref. 23), and Rdp1 (and the RDRC (RNA-directed RNA polymerase complex)) interacts with RITS and requires Swi6 for chromatin association 24,25 . Because of this and the inherent self-enforcing nature of the process, it is difficult to determine whether nascent transcripts are required to mediate RNAi-directed chromatin modifications, and what additional interactions are involved. Expression of a synthetic hairpin RNA homologous to a 280-basepair (bp) region located within the ura4 þ gene (sh-ura4-280) induces Dicer-dependent transcriptional silencing of ura4 þ along with H3K9me2 of ura4 chromatin and recruitment of Swi6 (ref. 7). To determine whether this hairpin-induced chromatin modification requires a homologous transcript, a strain containing a modified ura4 þ gene with an efficient transcription terminator module immediately upstream of the 280-bp ura4 target region was used 26 . In this strain (Ter þ ura4) the transcriptional terminator is inserted within an artificial intron, so that more than 99% of transcripts are terminated before the 280-bp region homologous to the sh-ura4-280 trigger. A second strain (Ter-M5 ura4) contains a cis-acting mutation within the terminator, allowing 75% of transcripts to traverse the downstream 280-bp region of the gene 26 (Fig. 1a ). Both strains also contain ura4-DS/E at the ura4 þ locus, which is fully transcribed but lacks the 280-bp region homologous to the sh-ura4-280 trigger, thus providing a convenient internal control 7 . The construct expressing sh-ura4-280 was introduced into both strains, and transcription of ura4 þ , H3K9me2 modification of ura4 chromatin and recruitment of Swi6 were assessed in wild-type strains in the presence and absence of the sh-ura4-280 hairpin. In cells containing the Ter þ ura4 gene, truncated (ura-T), but not full-length (ura4-FL), transcript is detected in the presence or absence of sh-ura4-280 (Fig. 1b) . In Ter-M5 cells, full-length ura4 transcript is lost in the presence of sh-ura4-280 but expression of ura4-DS/E remains unaffected (Fig. 1b) . Thus, expression of a hairpin target homologous to downstream DNA sequences does not affect Ter þ ura4, whereas Ter-M5 ura4 transcripts are repressed by sh-ura4-280 expression. Chromatin immunoprecipitation (ChIP) was used to assess the levels of H3K9me2 modification on the Ter þ and Ter-M5 ura4 genes relative to ura4-DS/E. H3K9me2 was detected only on Ter-M5 ura4, and only in strains expressing sh-ura4-280 (Fig. 1c) . ura4 siRNAs were detected only in strains containing the sh-ura4-280 construct (Fig. 1d) . Thus, RNAi can induce chromatin modifications at a homologous locus only if transcripts traverse a region identical in sequence to the hairpin trigger and the resultant siRNAs. It is possible that the passage of RNA polymerase II (pol II) during transcription itself, by opening chromatin, provides access for siRNAs to underlying DNA sequences, thus allowing siRNA-DNA interactions 4,5,14 . Alternatively, Ago1-bearing siRNAs may bind homologous nascent transcripts and in so doing recruit chromatin-modifying activities through the Ago1-containing RITS and/or Rdp1-containing RDRC complex 15, 16 . If opening the two DNA strands is sufficient, then an LETTERS exogenous RNA polymerase might allow siRNAs access to homologous chromatin. To test this, the ura4 transcription unit was placed downstream of the bacteriophage T7 promoter (Fig. 2) . T7 polymerase was constitutively expressed from the adh1 promoter in the presence or absence of sh-ura4-280 (Fig. 2a , and Supplementary Fig. S1a ). T7:ura4 transcripts are detected only in cells expressing T7 polymerase. The expression of sh-ura4-280 did not reduce the level of T7:ura4 transcripts significantly (Fig. 2b) , although RNAi is active because ura4 siRNAs are readily detected (Fig. 2d ). In addition, no H3K9me2 or Swi6 could be detected on T7:ura4 chromatin (Fig. 2c , and Supplementary Fig. S1b ) in cells expressing sh-ura4-280 homologous siRNAs (Fig. 2d) , although histone H3 is present on the T7:ura4 gene ( Supplementary Fig. S1c ). Lack of RNAi-directed chromatin modification of the T7:ura4 template may reflect the absence of features normally associated with endogenous RNA pol II transcription. T7 and RNA pol II transcription and the resulting transcripts differ in many respects; regardless of this, transcription of target chromatin alone is not sufficient to mediate RNAi-directed chromatin modifications on homologous chromatin. This indicates that transcripts generated by, or associated with, a specific RNA polymerase might be required. RNA pol II is responsible for the generation of fission yeast centromere repeat transcripts that are processed by RNAi into homologous siRNAs. A mutation (rpb7-1) in Rpb7, a small pol II subunit, leads to a loss of these transcripts and siRNAs (K.E., unpublished observations). The CTD of the large subunit of pol II (Rpb1) contains multiple conserved YSPTSPS heptad repeats, the phosphorylation state of which regulates the binding of various mRNA processing factors, thus coupling mRNA processing to transcription 17-19 . In Saccharomyces cerevisiae the deletion of up to Figure 1 | Transcription of siRNA target is required to effect silent chromatin assembly. a, Strains containing the Ter þ -ura4 gene with an efficient Ter þ or defective Ter-M5 terminator upstream of the 280-bp ura4 target region. b, RT-PCR analysis of ura4 þ transcripts on oligo(dT)-primed cDNA from RNA samples of the indicated strains; expression of ura4 þ -terminated (ura-T) and fulllength (ura4-FL) transcripts relative to the ura4-DS/E minigene in the same strains with or without sh-ura4-280 hairpin. c, ChIP analyses with anti-H3K9me2 and Swi6 antibodies over the ura4 gene in the indicated strains. T, total extract; IP, immunoprecipitate. d, Detection of sh-ura4-280generated siRNAs by northern blotting. nt, nucleotides. Figure 2 | Transcription by T7 polymerase is not sufficient for RNAidirected chromatin modification. a, The ura4 þ promoter was replaced with the bacteriophage T7 promoter (T7:ura4). T7 polymerase was constitutively expressed from the adh1 promoter in the presence or absence of sh-ura4-280. Cells were grown at 25 8C. b, RT-PCR analysis of ura4 þ transcripts was performed on oligo(dT)-primed cDNA from RNA samples from the indicated strains expressing (þ) or not (2) sh-ura4280-generated siRNAs. c, ChIP analyses with anti-H3K9me2 and anti-Swi6 antibodies over the ura4 gene in the indicated strains. T, total extract; IP, immunoprecipitate. d, Detection of sh-ura4-280 generated siRNAs by northern blotting. LETTERS NATURE|Vol 435|30 June 2005 1276 © 2005 Nature Publishing Group 16 of the 26 CTD heptad repeats from pol II results in compromised RNA polymerase functions 27 . If pol II has a specific function in mediating RNAi-mediated chromatin modification, then cells bearing a defective pol II might display aberrant silencing of marker genes at centromeres. To examine this, a strain was constructed with 17 of the 28 CTD heptad repeats deleted and simultaneously epitopetagged (rpb1-11, see Methods; Fig. 3a ). This strain was slow-growing but viable at all temperatures tested and was clearly defective in its ability to silence centromeric ura4 þ and ade6 þ markers as revealed by increased growth on plates lacking uracil (2ura) and the appearance of white ade þ colonies, respectively (Fig. 3b, left) . Consistent with this was the detection of increased levels of cen1:ura4 þ transcripts (Fig. 3b, right) and decreased levels of H3K9me2 associated with centromere repeats (Fig. 3c) . However, centromeric transcripts do not accumulate appreciably in rpb1-11 compared with dcr1D, and centromeric siRNAs are readily detected as in the wild type (Fig. 3d ). This indicates that although RNAi remains active it is unable to induce chromatin modifications efficiently on homologous sequences. The phenotype of rpb1-11 is clearly distinct from that of rpb7-1, which is defective in centromeric transcription and siRNA production, causing a failure of transcriptional silencing (K.E., unpublished observations). Microarray expression profiling indicated that none of the known genes involved in RNAi-directed chromatin silencing are significantly affected in rpb1-11 cells in comparison with the wild type, and few genes were affected to any great extent (Supplementary Table 1 ). Thus, the CTD truncation does not seem to cause a substantial general defect in transcription. This indicates that the CTD of pol II might act downstream of RNAi to stabilize interactions between RNAi components, the nascent transcript and possibly the pol II holoenzyme to induce chromatin modifications. RNAi components might require intact pol II to fully engage a chromatin-associated nascent transcript, or intact pol II Figure 3 | RNA pol II CTD truncation affects centromeric silent chromatin. a, rpb1-11 retains 11 of 28 heptad repeats. b, Left: growth assay of indicated strains with ade6 þ or ura4 þ genes in cen1 outer repeat on non-selective (YES), limiting adenine (YES (low ade)) or without uracil (PMG (2ura)). Right: RT-PCR analysis of cen1otr1:ura4 þ transcripts performed on RNA from the indicated strains. WT, wild-type. c, ChIP analyses of H3K9me2 and Swi6 over the cen1 outer repeat region (otr1) relative to the actin gene (act1). T, total extract; IP, immunoprecipitate. d, Detection of centromeric repeat transcripts and siRNAs by RT-PCR (top) and northern blotting (bottom), respectively. e, Immunoprecipitation and detection of HA-Ago1 and Rpb1 by western blot analysis. NATURE|Vol 435|30 June 2005 LETTERS 1277 © 2005 Nature Publishing Group might be specifically required to synthesize a transcript in a form that can effectively associate with RNAi components. Argonaute (PAZ/PIWI domain) proteins enter RISC (or RITS) complexes and use loaded siRNAs to guide RISC/RITS to target RNAs. Immunoprecipitates of HA-Ago1 were found by western blot analyses to contain pol II (Rpb1); reciprocal to this, HA-Ago1 was detected in immunoprecipitates of pol II. This interaction also required siRNAloaded RITS because Ago1 and pol II do not immunoprecipitate together from cells lacking Dicer (Fig. 3e) . To determine whether Ago1 associates with chromatin targeted for silencing by RNAi, ChIPs were performed with anti-Ago1 antibodies or HA-Ago1. Ago1 associated with centromeric outer repeats in wild-type, but not dcr1D, cells (Fig. 4a, top) . Ago1 also showed sh-ura4-280 siRNA and transcription-dependent association with the ura4 gene (Fig. 4a, bottom) . Ago1, but not Rad21, associated with centromeric otr transcripts, but not with control transcripts (act1), in wild-type cells, and not in cells lacking siRNAs (dcr1D) (Fig. 4b) . The association of Ago1 with centromeric chromatin was sensitive to RNase (Fig. 4c ). In addition, this association was reduced in strains carrying a truncated pol II CTD (rpb1-11) (Fig. 4d) . Consistent with previous reports 20 , immunolocalization shows that HA-Ago1 is concentrated at centromeres in the nucleus, as shown by localization with centromere-specific CENP-A Cnp1 (Fig. 4e , and Supplementary Fig. S2 ). Taken together, these data show that, in fission yeast, RNAi requires the transcription of a homologous target to direct chromatin modifications. Opening DNA by T7 pol transcription does not allow modification of the target chromatin to occur. T7 pol might deal with impeding nucleosomes differently, or T7 pol transcripts might not be packaged or processed in the same manner as RNA pol II transcripts, rendering them immune to RNAi. The fact that truncation of the pol II CTD affects RNAi-directed chromatin modifications and association of Ago1 with centromeric repeats, without noticeably affecting centromere repeat transcription or siRNA generation, indicates that pol II transcription might facilitate the conversion of RNAi signals into chromatin modification. Many different factors associate with pol II through its CTD during distinct stages of transcription    ; the pol II complex might provide a scaffold that promotes interactions between Ago1/RITS-borne siRNA and target pol II transcripts, leading to the efficient modification of occupied chromatin (see model in Supplementary Fig. S3 ). Indeed, it is known that RNA processing and export seem to be orchestrated with respect to ongoing transcription    . Our data indicate that RNAi-directed chromatin modification is another example of an RNA processing event that occurs co-transcriptionally, and offer an explanation for the apparent paradox that RNA pol II is not only required for transcriptional activity but is pivotal in transcriptional silencing. Moreover, plants have even evolved a distinct RNA polymerase (pol IV) required for RNAi-dependent chromatin silencing of certain repeat sequences 28, 29 .