Cell, Vol 111, 159-162, October 2002 Minireview A Chromosome RNAissance Abby F. Dernburg 31 and Gary H. Karpen 32 1 Life Sciences Division, Lawrence Berkeley National Lab, One Cyclotron Road, Be rkeley, CA 94720 USA 2 MCBL, The Salk Institute, 10010 North Torrey Pines Road, La Jolla, CA 92037 US A Correspondence: Abby F. Dernburg [email protected] Summary In RNA-mediated interference (RNAi), double-stranded RNAs (dsRNAs) target comple mentary mRNAs for degradation. New work demonstrates that essential chromosomal functions are mediated through RNAi protein components and short RNAs, which alt er chromosome function at specific DNA loci via histone modification. Main Text Summary Main Text Selected Reading New roles for RNA in regulating gene expression have recently been revealed by t he discovery of RNA interference (RNAi), which has rapidly become a standard exp erimental tool for targeted destruction of mRNAs in worms, flies, plants, and ma mmals. The molecular mechanism of RNAi involves proteins of the Dicer family, wh ich cleave dsRNAs to generate small interfering RNAs (siRNAs) that range from 21 to 28 nucleotides, depending on the organism (reviewed by Hannon, 2002 ). siRNA s target the RNA-induced silencing complex (RISC) to degrade homologous mRNAs, r esulting in loss of specific protein expression. siRNA production can also be am plified by RNA-dependent RNA polymerases (RdRPs), which synthesize new dsRNAs us ing mRNA templates and siRNA primers. Proteins essential for the degradation pat hway, including Dicer and Argonaute, are highly conserved in fungi, plants, and animals, although absent in budding yeast. These proteins have been shown to pla y essential roles in organismal development, germline fate, and host defenses ag ainst transposable elements and viruses. The inheritance of genetic traits and the organization of genomic information in eukaryotes have been generally regarded as DNA-mediated functions. However, sur prising new results demonstrate that RNA plays essential roles in epigenetic inh eritance and chromosome function in Schizosaccharomyces pombe and Tetrahymena th ermophila. In each of these processes, small RNAs appear to target regulatory pr oteins, including enzymes that modify histone tails, to specific, homologous chr omosomal loci, using protein components of the RNAi machinery. Establishment of Silencing in S. pombe Genes inserted into centromeric regions or at the silent mating type locus (mat 2/3) are silenced in S. pombe by the local recruitment and spreading of proteins including Clr4, Swi6, and Rik1. Swi6 is recruited to these loci by binding of i ts chromodomain to a particular posttranslational mark on a histone tail, methyl ated lysine at position 9 of histone H3 (Me(Lys9)H3). This modification has been linked to heterochromatin establishment and function in many organisms. Swi6 ca n multimerize and also interacts directly with Clr4, the histone methyltransfera se (HMT) that adds methyl groups to H3 K9, providing a mechanism for spreading t he methylated heterochromatin. A clear link between centromeric silencing and ch romosome segregation was forged by the demonstration that Swi6 is necessary for the recruitment of cohesin, a protein that maintains contact between sister chro matids from S phase until anaphase (Bernard et al., 2001 ; Nonaka et al., 2002 ) . A role for RNAi machinery and siRNAs in fission yeast centromeric silencing has been demonstrated in two papers recently published in Science. Martienssen, Grew al, and coworkers demonstrated that mutations in Dicer (Dcr1), Argonaute (Ago1), and RdRP (Rdp1) eliminated silencing of a ura4+ gene inserted into the outer an d inner repeats that flank the central core of the centromeres, as previously ob served for swi6 and clr4 mutations (Volpe et al., 2002 ). A corresponding reduct ion in Me(Lys9)H3 was observed in the outer repeats of these mutant cells. These authors searched for candidate RNAs that might mediate silencing and made the r emarkable discovery that large noncoding RNAs (~1.4 and 2.4 kb) homologous to th ese centromeric repeats accumulated in dcr1, ago1, and rdp1 mutant cells but wer e not detected in wild-type cells. Nuclear run-on assays demonstrated that these RNAs were transcribed in wild-type cells but were highly unstable. This result suggests that the long RNAs are synthesized independently of the RNAi machinery and are rapidly diced into siRNAs. Direct evidence for this idea was provided by Reinhart and Bartel (2002 ), who cloned small RNAs of the structure predicted f or Dicer products from S. pombe and recovered abundant species homologous to the centromeric repeats. Volpe et al. also provided evidence that Rdp1 protein is p hysically bound to the outer repeats, using a chromatin immunoprecipitation (ChI P) assay. Taken together, these results suggest that Rdp1 within the outer repea t chromatin may transcribe the second strand from nascent centromeric transcript s, generating dsRNAs. Dcr1 would process these dsRNAs into siRNAs, resulting in local H3-K9 methylation and silencing (Figure 1). Methylation could be targeted to the repeats by direct recruitment of the Clr4 HMT or a histone deacetylase ac tivity (HDAC) such as Clr3, which would remove the acetyl modifications that inh ibit methylation. Further experiments should illuminate the pathway leading from siRNAs to histone methylation, and also the contribution of Ago1 to this proces s. Figure 1. Models for the Establishment and Maintenance of Silencing in S. pombe Silencing initiates at the mat loci and centromere flanks in response to dsRNA a nd siRNA production, which may be amplified by the tethering of Rdp1. The siRNAs lead to methylation of H3-K9 at sites of homologous DNA, possibly by direct rec ruitment of Clr4 (as shown), or perhaps more indirectly by recruitment of other activities. The RNAi machinery is dispensable for maintenance of silencing. Epig enetic inheritance and spreading of the silent state is most likely mediated by segregation of modified nucleosomes to daughter strands during replication, recr uitment of Clr4 (and Rik1) by Swi6 binding to Me(Lys9)H3, and modification of ne w, adjacent nucleosomes in the region. Spreading proceeds until an ''insulator'' or ''boundary element'' is encountered. View larger version: [In this window] [In new window] A fundamental question generated by this work is whether the RNAi machinery is n ecessary to establish silencing in S. pombe, to maintain it once established, or both. In a related paper, Grewal and colleagues (Hall et al., 2002 ) demonstrat e that normal silencing at the mating-type locus (mat 2/3) also relies on the RN Ai proteins Dcr1, Rdp1, and Ago1, as well as cis-acting DNA sequences homologous to the centromere repeats. Nevertheless, a previously silenced mat locus retain s high levels of Me(Lys9)H3 and silencing through many rounds of replication and division in RNAi-deficient cells. Thus, maintenance of the silenced state at th e mat locus does not require RNAi machinery and is most likely inherited through propagation of Me(Lys9)H3 and/or Swi6 during replication ( Figure 1; Nakayama e t al., 2000 ). Hall et al. also demonstrate that RNAi-deficient cells cannot eff ectively establish silencing at the mat locus even with normal levels of Clr4 an d Swi6. In the absence of Swi6, Me(Lys9)H3 is restricted to the immediate vicini ty of the putative RNA template, suggesting that it acts as a silencing ''initia tion'' site. Normally, silencing then spreads over a 20 kb region and is halted by ''insulators'' or ''boundary elements.'' In summary, this work shows that Swi 6 is necessary for reinforcement and spreading of the Me(Lys9)H3 mark but is dis pensable for its initial establishment. Conversely, RNAi is required at the mat locus to establish silencing, but not to maintain it. A difference between centromeric and mat silencing is that at the mat locus ther e are no obvious defects in RNAi-deficient cells unless silencing is experimenta lly erased, either by genetic introduction of a Me(Lys9)H3-free copy of the locu s or by treatment of cells with trichostatin A (TSA), a histone deacetylase inhi bitor. By contrast, centromeres become desilenced in RNAi-deficient cells, raisi ng the interesting possibility that reestablishment of silencing is more critica l on an ongoing basis at the centromere than at mat. DNA Elimination in Tetrahymena Unlike most other single-celled organisms, Tetrahymena and other ciliates have t wo types of nuclei. Germline micronuclei (MICs) contain a complete copy of the g enome but do not express genes. Somatic macronuclei (MACs) are transcriptionally active, but their development from micronuclei involves massive chromosome rear rangements through DNA elimination, the excision of abundant dispersed sequences called internal eliminated sequences (IES) and breakage eliminated sequences (B ES) (reviewed by Yao et al., 2002 ). During the sexual cycle (conjugation) of Te trahymena, two cells, each containing a MIC and a MAC, fuse. The MICs in each ce ll undergo meiosis, and the cells exchange haploid meiotic products, which fuse to form a zygotic nucleus. The zygotic nucleus divides twice to produce four nuc lei; two are retained as MICs, and two undergo DNA elimination leading to the de velopment of new MACs. Several genes required for normal MAC development have pr eviously been identified, including three PDD (programmed DNA degradation) genes . The Gorovsky group (Mochizuki et al., 2002 ) found that Twi1p, a Piwi family pro tein related to Argonaute, is also essential for successful conjugation, IES eli mination, and MAC development. They hunted for RNAs of a size similar to siRNAs and detected them at the stages of conjugation preceding DNA elimination. These ''scnRNAs'' (or ''scan RNAs,'' see below and Figure 2) are not detected in TWI1 knockout cells and are delayed in their appearance in PDD1 knockouts. The sequen ces of the small RNAs have not been directly determined, but they hybridize far more strongly with MIC DNA than with the MAC genome, indicating that they corres pond to IES and/or BES sequences. Some of these RNAs may be derived from IES-con taining transcripts detected during early MIC meiosis (Chalker and Yao, 2001 ). During conjugation, Twi1p is first observed in the cytoplasm, then in the parent al MAC, and finally in the new MAC. Together with the observation of small RNAs and results from previous studies, this localization pattern suggested that Twi1 p could be transmitting RNA-encoded information from the old to the new MAC to d esignate the sequences to be eliminated. Figure 2. A Model for DNA Elimination in Tetrahymena This is an adaptation of the model of Mochizuki et al. (2002 ), in which dsRNAs homologous to the entire genome are produced in the micronuclei (MIC) and transf erred through the cytoplasm to the parental MACs. A suggested function for Twi1p is to mediate a ''scanning'' process in the parental MAC, in which RNAs corresp onding to MAC sequences are recognized and degraded. Intact scnRNAs homologous t o the MIC-specific IESs (and BESs) are then transported to the new MAC, where sc nRNAs or the resulting Me(Lys9)H3 mark recuits Pdd1p and other components requir ed for elimination to the IES/BES sequences. View larger version: [In this window] [In new window] In a complementary study by Allis and coworkers (Taverna et al., 2002 ), elimina tion of the IESs was shown to be associated with Me(Lys9)H3, the identical modif ication targeted to S. pombe centromeres and the mat locus by the RNAi machinery . Pdd1p and Pdd3p, two gene products essential for elimination, contain chromodo mains, which were shown to bind in vitro to Me(Lys9)H3. Western blotting and imm unolocalization demonstrated that Me(Lys9)H3 is only present at the time and pla ce where elimination occurs. ChIP analysis revealed that Pdd1p and Me(Lys9)H3 ar e associated preferentially with eliminated sequences in the developing new MAC. Furthermore, abnormal PDD1 expression strongly reduces but does not eliminate M e(Lys9)H3. Finally, the sufficiency of Pdd1p to promote elimination was demonstr ated in an elegant experiment in which the protein was artificially tethered to an ectopic genomic locus. We do not yet know whether Pdd1p mediates K9 methylati on, or if Pdd1p localization is normally downstream of methylation. Mochizuki et al. propose a model (Figure 2) in which RNAs originating in the MIC are first transported to the parental MAC. These RNAs somehow ''scan'' the geno me for homologous DNA. RNAs matching DNA sequences are degraded, so that only RN As not homologous to the genome of the old MAC are retained and transported to t he new MAC. Taking the two papers' results together, we can incorporate the idea that the scnRNAs mediate K9 methylation and Pdd1p recruitment, ultimately leadi ng to elimination. Me(Lys9)H3 is not detected in vegetative cells, indicating th at this epigenetic mark is implemented de novo during conjugation. The requireme nt for the RNAi machinery to establish but not to maintain silencing at the mat locus in S. pombe is thus consistent with its proposed role in IES elimination i n Tetrahymena. siRNA Recruitment of Silencing and Elimination Proteins Fundamental questions remain about the role of siRNAs in targeting histone modif ication enzymes to particular loci. Substrates for modification are likely speci fied by base-pairing interactions. Small RNAs could be targeted to chromosomes b y base-pairing with nascent transcripts, much as they are targeted to processed mRNAs. If so, the targeting machinery might share components with RISC. Alternat ively, small RNAs could associate directly with DNA. These associations could be mediated by proteins containing chromodomains, some of which bind both RNA and histone tails in vitro (Akhtar et al., 2000 ; Muchardt et al., 2002 ) and requir e an RNA component for heterochromatin localization in vivo (Maison et al., 2002 ). However, some chromodomain proteins, including Swi6, act downstream of RNA t argeting in silencing pathways. Base-pairing interactions between small RNAs and complementary RNA and/or DNA might also be mediated by the enigmatic Argonaute/ Piwi family. Events downstream of H3-K9 methylation could be specified by recruitment of othe r activities directly by RNA, by differences in the Me(Lys9)H3 binding proteins present in different nuclei, by sequestration of particular loci into specialize d nuclear compartments, or by combination of Me(Lys9)H3 with other modifications that specify the ''histone code'' (Jenuwein and Allis, 2001 ). Future studies s hould elucidate how an identical chromatin mark—Me(Lys9)H3—can mediate very di stinct outcomes, including silencing and DNA elimination. Transcription of siRNA Precursors A paradox arises from observations that the silencing of specific DNA sequences is caused by production of transcripts from within these very regions. It seems that chromatin-mediated repression would eventually prevent production of siRNA precursors and thereby silence the silencing mechanism. However, Volpe et al. de monstrate that transcription continues from the centromere repeats of silenced c ells. Other RNAi-mediated phenomena, particularly cosuppression, in which genes are silenced by the presence of extra copies of homologous DNA sequences (review ed by Zamore, 2002 ), also appear to require ongoing transcription from ''silent '' loci. The continued production of regulatory transcripts from silenced region s could simply result from a feedback mechanism, consistent with indications tha t Swi6 may repress forward strand transcription of the S. pombe outer repeats (V olpe et al., 2002 ). Perhaps only occasional nascent transcripts are necessary f or RNAi-mediated silencing. Evidence for physical association of an RdRP with th e S. pombe centromeric repeats suggests that rare transcripts could be immediate ly utilized to make dsRNA, thus amplifying siRNA production (Figure 1). A critic al unanswered question is how RdRPs are recruited to specific DNA sequences and whether recruitment requires Me(Lys9)H3. Finally, regulatory transcripts destine d for processing into siRNAs might also be synthesized by RNA polymerases, such as Pol I or III, which may be relatively insensitive to chromatin-mediated silen cing mechanisms. Identification of the promoters and polymerases that regulate t ranscription of centromeric repeats in S. pombe and at other silent loci should address these issues. Repetitive Sequences and RNAi Diverse silencing mechanisms are induced by repetitive DNA, including cosuppress ion in nematodes, repeat-induced point mutation (RIP) in Neurospora, and positio n-effect variegation (PEV) in Drosophila. In cases where RNAi components have be en implicated in repeat-induced silencing, it has been hypothesized that the tri gger might be aberrant coding mRNAs (reviewed by Zamore, 2002 ). Alternatively, transcripts might be shunted into the RNAi pathway through the recruitment of Rd RPs to repetitive DNA, if indeed this is a general phenomenon. If repetition is a trigger for silencing, how do essential repeated genes escape silencing? The ribosomal RNA genes are very actively transcribed by Pol I, but it is notable that Pol II genes inserted into these repeats are silent (Huang, 2 002 ). Repeated genes could bypass silencing by sequestration in specialized com partments (e.g., the nucleolus) or by utilizing promoters for polymerases that m ay be insensitive to silencing. Essential noncoding transcripts may also be desi gned to avoid processing by the RNAi machinery, perhaps by differences in 5' cap ping or polyadenylation. How did repeat-associated silencing mechanisms evolve? The centromeric repeats o f S. pombe and the IESs of Tetrahymena are thought to have originated from trans posable elements; for example, the S. pombe centromeric repeats bind to cellular proteins with homology to transposases (Nakagawa et al., 2002 ). It is likely t hat the maintenance of tandem and dispersed repeats as heterochromatin is an ada ptation of defense mechanisms that protect the host from transposon invasion or mobilization. In some cases, these silencing mechanisms appear to be specific to the germline, where transposon mobilization and ectopic recombination between r epeats are especially hazardous. The recent work reviewed here has illustrated h ow eukaryotes may have coopted defensive mechanisms to mediate essential functio ns at specialized repeats, including silencing of genetic information, sister-ch romatid cohesion, and programmed DNA rearrangements. What properties of transposons and other repeats trigger silencing? It is possib le that the silencing mechanisms recognize unusual structures present in repeate d DNA as ''dangerous,'' rather than the abundance of sequences per se. Although transposons and retrotransposons are usually present as dispersed repeats, the i nverted or direct repeats at their termini may flag them for silencing. Alternat ively, promoters within the transposons may trigger silencing. A role for abunda nce is suggested by the observation that multiple, dispersed copies of ''normal' ' genes induce RNAi-mediated silencing in Drosophila (Pal-Bhadra et al., 2002 ). However, silencing in this case could be instigated by the P element ends flank ing the integrated constructs. Additional studies are necessary to distinguish b etween the effects of repeat structure and sequence abundance in RNAi-mediated s ilencing. RNAi and Chromosome Function These recent publications link RNAi machinery and siRNAs to epigenetic chromatin modifications and chromosome behavior. They represent major advances in our und erstanding of how particular sites are marked for heterochromatin formation and will likely enhance the status of ''junk'' DNA by demonstrating unexpected roles for specific noncoding sequences and repeated DNA. It is easy to imagine that R NAi-related mechanisms might contribute to a wide diversity of chromosome behavi ors, including X chromosome inactivation in mammals, the ''spreading'' of silenc ing through the pericentromeric repetitive DNA of higher eukaryotes, telomere ma intenance, sex chromosome silencing in the germline of many species, hybrid dysg enesis in Drosophila, chromatin diminution in Ascarid nematodes, and nucleolar d ominance in plants. These exciting new developments should stimulate a ''RNAissa nce'' of investigations into these and many other enigmatic chromosome phenomena . Selected Reading Summary Main Text Selected Reading Akhtar A., Zink D. and Becker P.B. (2000) Nature, 407:405-409. [Medline] Bernard P., Maure J.F., Partridge J.F., Genier S., Javerzat J.P. and Allshire R. C. (2001) Science, 294:2539-2542. [Medline] Chalker D.L. and Yao M.C. (2001) Genes Dev., 15:1287-1298. [Medline] Hall I.M., Shankaranarayana G.D., Noma K.I., Ayoub N., Cohen A. and Grewal S.I. (2002) Science, 297:2232-2237. [Medline] Hannon G.J. (2002) Nature, 418:244-251. [Medline] Huang Y. (2002) Nucleic Acids Res., 30:1465-1482. [Medline] Jenuwein T. and Allis C.D. (2001) Science, 293:1074-1080. [Medline] Maison C., Bailly D., Peters A.H., Quivy J.P., Roche D., Taddei A., Lachner M., Jenuwein T. and Almouzni G. (2002) Nat. Genet., 30:329-334. [Medline] Mochizuki K., Fine N.A., Fujisawsa H. and Gorovsky M.A. (2002) Cell, 110:689-699 . [Medline][Summary][Full Text] Muchardt, C., Guilleme, M., Seeler, J.S., Trouche, D., Dejean, A., and Yaniv, M. (2002). EMBO Rep. Nakagawa H., Lee J.K., Hurwitz J., Allshire R.C., Nakayama J., Grewal S.I., Tana ka K. and Murakami Y. (2002) Genes Dev., 16:1766-1778. [Medline] Nakayama J., Klar A.J. and Grewal S.I. (2000) Cell, 101:307-317. [Medline][Summa ry][Full Text] Nonaka N., Kitajima T., Yokobayashi S., Xiao G., Yamamoto M., Grewal S.I. and Wa tanabe Y. (2002) Nat. Cell Biol., 4:89-93. Pal-Bhadra M., Bhadra U. and Birchler J.A. (2002) Mol. Cell, 9:315-327.[Summary] [Full Text] Reinhart B.J. and Bartel D.P. (2002) Science, 297:1831. [Medline] Taverna S.D., Coyne R.S. and Allis C.D. (2002) Cell, 110:701-711. [Medline][Summ ary][Full Text] Volpe T.A., Kidner C., Hall I.M., Teng G., Grewal S.I. and Martienssen R.A. (200 2) Science, 297:1833-1837. [Medline] Yao M.-C., Duharcourt S. and Chalker D.L. (2002) Genome-wide rearrangements of D NA in ciliates. In: Gellert M. (Ed.) Mobile DNA II. Washington, DC: ASM Press Zamore P.D. (2002)