Whether in animals, plants or fungi, the discovery of epigenetic gene silencing (heritable changes in gene expression without changes in DNA sequence) is a fascinating story. Like many other discoveries in genetics, it began as an obscure observation in an unlikely organism (the petunia) and gradually worked its way up to model organisms as powerful as Schizosaccharomyces pombe1,2. The post-transcriptional form of epigenetic silencing is known as RNA interference (RNAi). The key RNAi proteins are Dicer (Dcr1), an enzyme that cleaves double-stranded RNA (dsRNA) into small fragments known as short interfering RNAs (siRNAs); RNA-dependent RNA polymerase (Rdp1), which uses siRNAs to polymerize additional dsRNA; and RNA-induced silencing complex (RISC), which mediates the degradation of RNAs homologous to RISC-associated siRNAs. In contrast, transcriptional silencing occurs at the chromatin level. Silent chromatin is almost always associated with a methylated form of histone H3 (H3-Lys9) and a conserved H3-Lys9 binding protein, Swi6.

For many years, these two forms of silencing were viewed as separate and distinct. But links between transcriptional and post-transcriptional silencing began to appear in the late 1990s (ref. 3), and in 2002, two key discoveries in S. pombe made the connection irrefutable4,5. These studies showed that without Dcr1, Rdp1 or the presumed RNase component6 of RISC known as Argonaute (Ago1), pericentromeric heterochromatin was abolished4 and a transcriptionally inactive mating type locus (mat) could no longer be silenced5. A search for proteins mediating the interaction between RNAi and transcriptional silencing led to the identification of the RNA-induced initiation of transcriptional gene silencing (RITS) complex, which contains Ago1, a heterochromatin-binding protein known as Chp1, Tas3 and siRNAs7. On page 1174 of this issue, Noma et al.8 present convincing evidence that RITS is an integral component of the epigenetic silencing pathway in S. pombe.

Marking a silent domain

Like other silencing factors in S. pombe, RITS seems to stably associate with all main heterochromatic domains: the pericentromeres, the telomeres and the mat locus. Pericentromeres and mat share a similar nucleating domain known as cenH, which generates dsRNA4. The histone H3-Lys9 methylase Clr4 is initially recruited to the 4.3-kb cenH sequence and then fans out over flanking chromatin to cover the entire 20-kb mat region. The spreading reaction is mediated in part by Swi6 (ref. 5), and Noma et al.8 now show that it involves RITS as well.

Their interpretations are based on several observations. First, they found that a mutation in the chromodomain of Chp1 abolishes RITS binding to chromatin. They also found a loss of RITS binding in clr4− mutants, which lack methylated H3-Lys9 binding sites for RITS. In addition, they found that RITS spreads in a Swi6-dependent manner. These data support the view that Swi6 mediates the recruitment of Clr4, which provides the necessary H3-Lys9 binding sites for RITS. The authors note that siRNAs and Dcr1 are not required for spreading or subsequent maintenance of RITS at heterochromatic loci. Instead, the Chp1 chromodomain is primarily responsible for the affinity of RITS for silent chromatin. The sequence-independence of this spreading reaction presumably allows RITS to exert control over sequences incapable of initiating the RNAi response.

Enforcing the silence

Drawing on the analogy of RNAi as a genetic immune response9, several authors have hypothesized that RNAi might guide chromatin-silencing complexes back to the gene that produced the dsRNA3,4,5,7. Further, if components of the RNAi machinery remain bound to transcriptionally silenced loci, rare transcripts that escape silencing could be recognized and destroyed before transcription is completed. In a key set of experiments, Noma et al. show that the free, chromatin-unbound form of RITS does not associated with siRNAs (with the caveat that small quantities of siRNAs might have been missed). When bound to chromatin through H3-Lys9, however, RITS seems to become especially receptive to siRNAs. RITS may acquire its siRNAs from post-transcriptional (trans-acting) RNAi activity or more locally from aberrant RNAs processed at the source of transcription. Once bound to chromatin and primed with siRNAs, RITS probably becomes the center of localized RNA destruction, ensuring that stray transcripts from silent chromatin are cleaved in cis (presumably by Ago1)6. Thus, although siRNAs are not necessarily required to target RITS to specific chromatin sites, siRNAs and RITS both seem to be involved in the predicted cis-acting RNAi function (Fig. 1).

Figure 1: A simplified view of how RITS fits into the RNAi pathway.
figure 1

RNAi is initiated by transcription from both strands of a DNA sequence. The resulting dsRNA is degraded by Dcr1 to produce siRNAs. Both RISC and RITS use siRNAs and Ago1 to cleave RNA. RISC cleaves RNA in trans, whereas RITS binds to chromatin and is thought to destroy newly produced RNAs in cis.

Full size image

Initiating events

The authors also discuss briefly the issue of what factors might be involved in the earliest events of Dcr1-dependent, sequence-specific heterochromatin formation. It is important to incorporate Dcr1 and siRNAs into any model for the initiation of silencing, because the genetic data for this at pericentromeres and other loci and quite good4,10. Although silencing at mat is genetically more complex than at pericentromeres, there are good indications that the two silent domains are regulated in a similar manner. Aside from the facts that pericentromeres and mat contain nearly identical sequences and bind the same set of proteins4,5, a dcr1− swi6− double mutant fails to initiate and maintain mat silencing8. Given that swi6− alone does not have this effect5, siRNAs are probably required in some form to designate mat as a silent domain.

The discovery that RITS binds efficiently to chromatin without siRNAs8 puts a damper on early predictions that RITS might serve as the primary initiator of chromatin silencing. An alternate possibility is that a larger silencing complex of some sort8 interacts with Clr4 (ref. 10). However this question is ultimately resolved, it's a good bet that the answer will come from S. pombe, where the combination of genetics and biochemistry has been very productive.