Hairpin RNA induces secondary small interfering RNA synthesis and silencing in trans in fission yeast

Abstract

RNA interference (RNAi) is widespread in eukaryotes and regulates gene expression transcriptionally or post-transcriptionally. In fission yeast, RNAi is tightly coupled to template transcription and chromatin modifications that establish heterochromatin in cis. Exogenous double-stranded RNA (dsRNA) triggers seem to induce heterochromatin formation in trans only when certain silencing proteins are overexpressed. Here, we show that green fluorescent protein (GFP) hairpin dsRNA allows production of high levels of Argonaute-associated small interfering RNAs (siRNAs), which can induce heterochromatin formation at a remote locus. This silencing does not require any manipulation apart from hairpin expression. In cells expressing a ura4⁺–GFP fusion gene, production of GFP siRNAs causes the appearance of ura4 siRNAs from the target gene. Production of these secondary siRNAs depends on RNA-dependent RNA polymerase Rdp1 (RDRP^Rdp1) function and other RNAi pathway components. This demonstrates that transitivity occurs in fission yeast and implies that RDRP^Rdp1 can synthesize RNA from targeted RNA templates in vivo, generating siRNAs not homologous to the hairpin.

Introduction

Small interfering RNA (siRNA)-mediated silencing mechanisms have been observed in most eukaryotes. The RNA interference (RNAi) machinery frequently reduces gene expression by cleaving homologous transcripts or by inhibiting their translation. However, RNAi can also act at the transcriptional level, leading to chromatin modification and heterochromatin formation.

The crucial features of RNA-mediated silencing include the production of 21–25 nt siRNAs by the ribonuclease III enzyme Dicer (Dcr1), and the formation of an Argonaute (Ago)-containing complex into which these siRNAs are incorporated (Farazi et al, 2008). The Ago complex is guided to transcripts homologous to the encapsulated siRNA, to block their translation or to use its inherent endonuclease activity (known as ‘slicing') to cleave the cognate transcripts.

In fission yeast (Schizosaccharomyces pombe) the RNAi machinery is required for the formation of heterochromatin at pericentromeric repeats, the mating-type locus, and subtelomeric repeats. In these regions, chromatin is underacetylated and methylated on lys 9 of histone H3 (H3K9me) and the transcription of underlying genes is repressed (reviewed in Buhler & Moazed, 2007; Grewal & Jia, 2007). The formation of heterochromatin at these loci is a multi-step process triggered by double strand (ds) RNA production. Non-coding transcripts derived from repetitive elements form dsRNA and are processed into siRNAs by Dcr1. The siRNAs are loaded into Ago1, which directs the RNA-induced initiation of transcriptional gene silencing (RITS) complex to homologous repeats. The RITS complex is composed of Ago1, Tas3 and Chp1, and binds H3K9me chromatin through the chromo-domain of Chp1, allowing the recruitment of the RNA-dependent RNA polymerase complex (RDRC) to chromatin. The RDRC contains Rdp1 and is thought to synthesize dsRNA by using the RNA-dependent RNA polymerase (RDRP) activity of Rdp1, thereby amplifying the siRNA pool. The action of the RNAi machinery recruits the crucial histone methyltransferase Clr4 to siRNA homologous loci. The H3K9 methylation by Clr4 allows binding of the chromo-domain proteins Swi6, Chp1, Chp2 and Clr4 itself to chromatin, forming a nucleation site from which heterochromatin components can spread outwards along the chromatin fibre, repressing underlying genes.

In most organisms, the expression of exogenous hairpin dsRNAs is sufficient to generate siRNAs that, in turn, silence target genes by inducing post-transcriptional RNA cleavage and degradation, demonstrating that the RNAi machinery can act in trans to inhibit expression of any gene (Kennerdell & Carthew, 2000; Smith et al, 2000; Tavernarakis et al, 2000; Paddison et al, 2002). In plants, hairpin RNA-triggered trans-silencing is known to be converted into transcriptional silencing owing to the modification of homologous DNA and chromatin that mediates repression (Baulcombe, 2004). By contrast, in fission yeast, several observations indicate that the RNAi machinery is more constrained and unable to efficiently promote transcriptional silencing through heterochromatin formation in trans. The siRNAs derived from a silenced marker gene can occasionally repress a second copy of the same gene only if the nuclease Eri1 is deleted (Buhler et al, 2006). Furthermore, expression of a long GFP hairpin (GFP-HP) RNA can post-transcriptionally reduce the expression of a homologous target, without heterochromatin-associated chromatin modifications (Sigova et al, 2004). Finally, ura4 hairpins (U-HPs) were shown to silence ura4⁺ gene expression only when the chromo-domain protein Swi6 (heterochromatin protein 1; HP1) was overexpressed and when antisense transcription occurred across the target gene (Iida et al, 2008).

Here, we show that exogenous hairpin RNAs can, in fact, induce heterochromatin formation in wild-type fission yeast. In contrast to previous reports, silencing does not necessarily require Swi6 overexpression and/or antisense transcription. Analysis of the hairpin-mediated trans-silencing reveals two important features of RNAi-mediated gene silencing in S. pombe. First, on silencing, a class of siRNAs, not derived directly from the hairpin but corresponding to the target gene, is detected; RDRP^Rdp1 is required for the production of these additional siRNAs in vivo. Second, the presence of Ago1-associated siRNAs does not guarantee robust silencing; Ago1-mediated slicing and degradation of a target RNA might only be a marginal activity in fission yeast.

Results And Discussion

U-HP silences a target adjacent to heterochromatin

To assess whether the fission yeast RNAi machinery can induce heterochromatin formation in trans, we constructed a hairpin complementary to 200 bp of ura4⁺ expressed from the nmt1 promoter (U-HP), and integrated at ars1 on chromosome 1 (Fig 1A). The hairpin generates detectable ura4 homologous siRNAs (Fig 1B). To assess the influence of chromatin context on the ability of U-HP to repress homologous target genes, we tested whether the hairpin silences the normal ura4⁺ gene and ura4⁺ that is in close proximity to pericentromeric heterochromatin of centromere 1 (cen1; otr1L XhoI:ura4⁺). Silencing of ura4⁺ results in restricted growth on plates lacking uracil (−URA) and good growth on counter-selective plates containing 5-fluorootic acid (5-FOA; Allshire et al, 1995). Expression of U-HP did not affect the growth of cells with ura4⁺ at its normal chromosomal location. However, growth on 5-FOA plates revealed that ura4⁺ adjacent to centromeric repeats is more sensitive and silenced by this same hairpin (Fig 1C). This observation raises the possibility that in S. pombe, exogenous siRNAs only silence efficiently in trans when the target locus is near endogenous sites of heterochromatin. In fact, H3K9me was detected on ura4⁺ at this site, but does not result in transcriptional repression (Allshire et al, 1995; Trewick et al, 2007). Expression of U-HP in these cells clearly tips the balance in favour of robust heterochromatin formation and silencing. Other U-HPs have also been shown to tighten weak silencing at other locations in heterochromatin domains (Iida et al, 2008).

Figure 1.

A U-HP silences ura4⁺ inserted close to centromere 1, but not the endogenous ura4⁺ gene. (A) Schematic diagram of the U-HP construct, the hairpin RNA and the ura4⁺ gene. The U-HP includes nucleotides 442–641 of the ura4⁺ ORF. A 62 bp spacer containing the first intron from the rad9 gene separates the inverted ura4 DNA composing the hairpin. (B) Northern analysis of siRNAs in wild-type cells±U-HP. The ura4 siRNAs from the U-HP are detected. Detection of centromeric siRNAs provides a positive control and the snRNA58 probe is a loading control. (C) Growth assay to monitor silencing. Serial dilutions of cells were spotted on non-selective (NS), lacking uracil (−URA) and 5-FOA plates. Silencing of the ura4⁺ gene results in reduced growth on selective −URA plates and good growth on counter-selective plates containing 5-FOA. The U-HP does not silence the endogenous ura4⁺ gene but can silence ura4⁺ adjacent to centromere 1 (cen1; otr1L XhoI: ura4⁺) allowing growth on 5-FOA. 5-FOA, 5-fluorootic acid; cen, centromere; ORF, open reading frame; siRNA, small interfering RNA; snRNA, small nuclear RNA; U-HP, ura4 hairpin.

GFP siRNAs produce secondary siRNAs

It has been reported previously that a GFP-HP can post-transcriptionally silence a target gene, resulting in decreased levels of GFP messenger RNA independently of the RITS components Tas3 and Chp1, and the heterochromatin-associated protein Swi6 (Sigova et al, 2004). However, in that study the target GFP gene was expressed from the exceptionally strong adh1 promoter and the GFP-HP RNA was expressed from a high copy, mitotically unstable plasmid. These features might not be conducive for heterochromatin formation at the target locus. Therefore, we developed a moderately expressed target for GFP-HP-generated siRNAs. A ura4⁺–GFP carboxy-terminal fusion gene was constructed, expressed from the ura4 promoter, and integrated at the arg3 locus. The GFP-HP construct, expressed from the nmt1 promoter, is composed of an RNA with two GFP open reading frames arranged in an inverted orientation around the first intron from the rad9 gene (Sigova et al, 2004). To promote uniform expression, the GFP-HP construct was integrated at ars1; it produces clearly detectable siRNAs that are incorporated into Ago1 (Fig 2A,B). Equivalent levels of cen siRNAs are present in cells with and without the hairpin, suggesting that GFP siRNA production does not interfere with cen siRNA production.

Figure 2.

Detection of primary and secondary siRNAs. (A) Schematic diagram of the GFP-HP construct and the target fusion genes ura4⁺–GFP and ade6⁺–GFP. Constructs are integrated at ars1, arg3 and ade6 loci, respectively. A 62 bp spacer containing the first intron from the rad9 gene separates the inverted GFP DNA composing the hairpin. (B) Northern analysis of total and FLAG-Ago1-associated siRNAs from cells expressing or not expressing the GFP-HP. GFP siRNAs produced from the GFP-HP are loaded into FLAG-Ago1. (C) Sequenced FLAG-Ago1-associated siRNAs aligned to GFP. y-axes represent the log2 of the number of sequence reads that align to each nucleotide of GFP. (D) Detection of ura4 secondary siRNAs. Total siRNA was isolated, subsequently the 10–30 nt fraction was gel purified and subjected to northern analysis. The ura4 siRNAs are detected in wild-type cells expressing GFP-HP, but not in dcr1Δ cells. The asterisk marks degradation products purified from the gel slice. Ago, Argonaute; cen, centromere; GFP, green fluorescent protein; GFP-HP, GFP hairpin; siRNA, small interfering RNA; wt, wild type.

Illumina/Solexa sequencing of FLAG-Ago1-associated siRNAs showed that both sense and antisense GFP homologous siRNAs are generated, which cover most of the length of GFP (Fig 2C). The GFP siRNAs were calculated to be approximately fourfold more abundant than siRNAs originating from centromeric dg repeats, but at levels equivalent to the centromeric dh element siRNAs (see supplementary information online).

In plants, it is known that the production of exogenous siRNAs for one part of a target gene allows the generation of secondary siRNA homologous to regions 5′ and 3′ to the initial primary siRNA target (Baulcombe, 2007). Northern blot analyses with a ura4 probe allowed weak detection of siRNAs homologous to ura4 5′ to the GFP target (data not shown). To increase sensitivity, siRNAs were concentrated by gel purification and subjected to northern blot analyses. siRNAs homologous to both the ura4 and cen probes were clearly detected in wild-type cells, but not in dcr1Δ cells. Detection of these ura4 siRNAs was hairpin dependent, indicating that, as in other systems, secondary siRNAs are generated in fission yeast (Fig 2D).

Sequencing of gel-purified siRNA confirmed the presence of ura4 siRNAs in wild-type cells. In total, 1,276 reads for ura4 were obtained in wild-type cells and only 14 in cells lacking Rdp1 (Fig 3A; supplementary Table S3 online), even though 70% more reads were obtained for rdp1Δ (16.42 compared with 11.63 M). Furthermore, siRNAs homologous to the 5′ and 3′ untranslated regions (UTRs) of the ura4–GFP transcript were prevalent in wild type but absent in rdp1Δ. Similarly, fewer siRNAs corresponding to the 5′ UTR and 3′ UTR of the GFP-HP transcript were detected in rdp1Δ. By contrast, a substantial level of GFP siRNAs was detected in rdp1Δ cells (Fig 3A,B; supplementary Table S3 online).

Figure 3.

Secondary siRNAs are dependent on Rdp1. (A) Detection of ura4⁺ secondary siRNAs by Illumina/Solexa sequencing. Sense (top panel) and antisense (bottom panel) siRNA were aligned to the ura4⁺–GFP target gene. Secondary siRNAs complementary to the ura4 portion of the ura4⁺–GFP target are detected in wild-type (left panels) but not rdp1Δ cells (right panels). (B) Detection of primary and secondary siRNAs complementary to the GFP-HP trigger. Sense (top panel) and antisense (bottom panel) siRNA were aligned to the GFP-HP construct. The siRNAs complementary to 5′ UTR and 3′UTR are detected in wild-type (left panels) but far fewer in rdp1Δ cells (right panels). y-axes represent the log2 of the number of reads that align to each nucleotide of the ura4⁺–GFP or GFP-HP construct. GFP, green fluorescent protein; GFP-HP, GFP hairpin; siRNA, small interfering RNA; UTR, untranslated region; wt, wild type.

It has been shown in vitro that Rdp1 can mediate the synthesis of dsRNA from a single-stranded RNA template without a complementary primer (Motamedi et al, 2004; Sugiyama et al, 2005). The analyses we present provide, to the best of our knowledge, the first evidence of the generation of Rdp1-mediated secondary siRNAs in vivo in fission yeast.

GFP siRNAs induce unstable silencing of target genes

To assess the ability of the GFP-HP to promote gene silencing, we tested whether it affects the expression of different target genes, and also whether it leads to chromatin modification indicative of heterochromatin formation. Similar to wild-type ura4⁺ cells, cells expressing the ura4⁺–GFP fusion gene grow on −URA medium, but not on plates containing 5-FOA. By contrast, wild-type cells that express both the GFP-HP and the ura4⁺–GFP target gene grow on 5-FOA, indicating that the GFP-HP mediates repression of ura4⁺–GFP (Fig 4A, wt). We also combined this chromosomally expressed GFP-HP with cells expressing an ade6⁺–GFP fusion gene. On limiting adenine indicator plates, ade6⁻ cells form red colonies, whereas ade6⁺ cells form white colonies. Red colonies were clearly visible in cells expressing GFP-HP, indicating that the ade6⁺–GFP gene can be silenced (Fig 4B, arrowheads). Thus, the GFP-HP seems to mediate RNAi-induced silencing of target genes.

Figure 4.

GFP-HP induces trans-silencing of target genes. (A) Growth assay demonstrating that the ura4⁺–GFP fusion gene is silenced by expression of GFP-HP in wild-type but not the indicated mutant strains. Silencing is indicated by growth on counter-selective 5-FOA. (B) Colony colour assay to monitor silencing of ade6⁺–GFP by the GFP-HP on indicator plates. Expression of GFP-HP induces unstable silencing of the ade6⁺–GFP reporter gene. Silencing is indicated by the appearance of red colonies; 19% of colonies were red. (C) qRT–PCR of ura4⁺–GFP relative to act1⁺ indicates that ura4⁺–GFP levels are reduced in cells expressing GFP-HP when grown in the presence of counter-selective 5-FOA, which selects for cells with silent ura4⁺. 5-FOA, 5-fluorootic acid; GFP, green fluorescent protein; GFP-HP, GFP hairpin; qRT–PCR, quantitative reverse transcriptase PCR; siRNA, small interfering RNA; wt, wild type.

To further characterize how the GFP-HP mediates repression of ura4⁺–GFP, levels were assessed by quantitative reverse transcriptase PCR (RT–PCR). Surprisingly, transcript levels were not significantly reduced when grown in non-selective medium compared with control cells (data not shown). However, when the silenced population was selected by growing cells in counter-selective 5-FOA medium, a reduction in ura4–GFP transcript levels was apparent (Fig 4C). Therefore, although GFP siRNAs were produced at a high level and are loaded into Ago1 (Fig 2B), it seems that target GFP transcripts were not efficiently reduced. Consistent with this observation, GFP-HP only silences ade6⁺–GFP in 19% of colonies (Fig 4B), and re-plating assays of cells from red colonies resulted in a variegated population indicating that silencing is unstable (data not shown). This suggests that classical post-transcriptionally mediated RNAi knockdown of transcripts is not prevalent in fission yeast. S. pombe Ago1 might not efficiently slice engaged transcripts and perhaps, in association with Chp1 and Tas3 in RITS, Ago1 is more tailored to affect transcription of the template by inducing heterochromatin formation. In agreement with this, we observed that silencing of ura4⁺–GFP by the GFP-HP was not only dependent on the RNAi factors Dcr1, Ago1, Arb1 and Rdp1, but also on the H3K9 methyltransferase Clr4, on the chromo-domain proteins Chp1, Swi6 and Chp2, and on the histone deacetylases Sir2 and Clr3 (Fig 4A; supplementary Fig S1 online).

A GFP-HP induces heterochromatin formation

Loss of GFP-HP-mediated silencing in clr4Δ and swi6Δ cells suggests that heterochromatin might be formed on the ura4⁺–GFP target. This was confirmed by chromatin immunoprecipitation (ChIP) assays that indicated that H3K9me2 and Swi6 were present on the ura4⁺–GFP fusion gene in cells expressing GFP-HP (Fig 5A). The GFP-HP itself is expressed from the nmt1 promoter at the ars1 locus; it might also be a target for the GFP siRNAs generated from its hairpin RNA. The ChIP analyses indicated that higher levels of H3K9me2 could be detected on the GFP-HP locus than on the ura4⁺–GFP target gene, and this was also dependent on Dcr1 and Clr4 (Fig 5B). H3K9me2 at this locus recruits Swi6 (supplementary Fig S2 online) and, therefore, probably allows the formation of intact heterochromatin. This high level of H3K9me2 might be related to the production of sense and antisense GFP RNA at the same locus, allowing the formation of dsRNA and its processing in cis. Previously, it was found that U-HP RNAs induced heterochromatin formation on a ura4⁺ target gene when antisense transcripts traverse the target gene (Iida et al, 2008). However, antisense transcripts across the GFP portion of ura4⁺–GFP could not be detected (Fig 5C). This suggests that antisense transcription is not an absolute requirement for RNAi-directed chromatin modification and silencing of a target gene by exogenous siRNAs.

Figure 5.

GFP-HP induces chromatin modifications. (A) H3K9me2 and Swi6 ChIP analyses. The enrichment of ura4⁺–GFP and cen-otr compared with act1⁺ was calculated by quantitative PCR and subsequently the ratio of enrichment in cells expressing GFP-HP relative to cells without the hairpin was plotted. The position of the ura4⁺-specific primers (arrowheads) for ura4⁺–GFP is shown. (B) H3K9me2 ChIP analyses at the GFP-HP locus. Quantitative PCR allowed the quantification of H3K9me2 enrichment on ura4⁺–GFP, GFP-HP and cen-otr relative to act1⁺. The diagram shows the position of the GFP-HP-specific primers used (arrowheads). ura4–GFP primers were as in (A). (C) Strand-specific RT–PCR assay for transcripts from ura4⁺–GFP. Diagram indicating the primer pairs used in RT–PCR to detect transcripts across the middle (mid) and 3′ end (end) of the ura4⁺–GFP fusion gene. Primers used for cDNA synthesis of antisense (upper arrowheads) or sense (lower arrowheads) transcripts. RT–PCR was carried out with an equal amount of total RNAs from strains with and without the GFP-HP. RT–PCR with act1⁺ primers was performed as a control. as, antisense; cDNA, complementary DNA; ChIP, chromatin immunoprecipitation; GFP, green fluorescent protein; GFP-HP, GFP hairpin; RT–PCR, reverse transcriptase PCR; s, sense; siRNA, small interfering RNA.

The fact that the locus expressing the GFP-HP itself attracts heterochromatin presumably also influences the levels of hairpin RNA and, thus, the results in siRNAs. This might cause the level of siRNAs, and therefore silencing itself, to oscillate. This could explain the clear instability of the silent state observed with the GFP-HP, both with the ura4⁺–GFP and ade6⁺–GFP targets.

Conclusion

These analyses demonstrate that GFP siRNAs generated by the expression of a GFP-HP can act in trans to establish heterochromatin on target genes bearing homology to GFP siRNAs and silence their expression. This silencing does not require other manipulations, such as deletion of eri1⁺ or increased expression of the heterochromatin component Swi6^HP1, which have previously been shown to promote RNAi-mediated silencing in trans (Buhler et al, 2006; Iida et al, 2008). Interestingly, the silencing we report involves only a marginal decrease in the levels of target RNA. In addition, on silencing of ura4⁺–GFP by the GFP-HP, we detected ura4 siRNAs, thus providing evidence of the production of secondary siRNAs in S. pombe.

Our data show that GFP-HP-derived siRNAs are sufficient to induce chromatin modification (H3K9me2 and Swi6 recruitment) on a homologous gene. Furthermore, GFP siRNAs clearly act in cis, inducing heterochromatin formation on the GFP-HP construct itself. Thus, centromeres, telomeres and mating-type loci have no inherent special properties for being selected as targets for the RNAi-directed chromatin modification machinery in fission yeast.

Our analyses indicate that the ability of a hairpin RNA to silence in trans depends on the chromosomal location of target genes, and it suggests that the presence of a patch of heterochromatin can provide a foothold for hairpin-mediated silencing. In fact, the ura4 hairpin used here only induced silencing of the ura4⁺ gene when it was placed in close proximity to heterochromatin and known to have some associated H3K9me2 (Noma et al, 2006; Trewick et al, 2007). Non-coding RNAs and siRNAs generated from this region might also contribute to the hairpin-mediated silencing (Cam et al, 2005; Wilhelm et al, 2008). The GFP-HP allowed heterochromatin formation at the arg3:ura4⁺–GFP locus. It is possible that convergent transcripts from the adjacent cmb1⁺ gene might provide sufficient transient H3K9 methylation (Gullerova & Proudfoot, 2008; Zofall et al, 2009) to promote and enable heterochromatin formation on production of GFP siRNAs. Alternatively, other still undefined features of this region, such as the acetylation status of the surrounding chromatin, might allow heterochromatin formation by GFP-HP. Insertion of the ura–GFP target in other regions, such as ‘gene-poor' domains or regions with distinctive chromatin, will determine which feature promotes or prevents RNAi-directed heterochromatin formation.

Methods

Yeast strains. All S. pombe strains used are listed in supplementary Table S1 online. Standard procedures were used for growth and genetic manipulations (Moreno et al, 1991). Detailed descriptions of constructs and strains are presented in the supplementary information online.

RNA analysis. siRNA was prepared as described in the supplementary information online.

For Solexa/Illumina sequencing, 5′ and 3′ adaptors were ligated to the purified RNA, complementary DNA was then synthesized by reverse transcription and sequenced with the Illumina/Solexa 1G Sequencing System (San Diego, CA, USA).

For northern blotting, RNAs were run on denaturing polyacrylamide gels, electro-transferred onto Hybond-NX membranes (GE Healthcare, Buckinghamshire, UK) and UV cross-linked. Membranes were probed with P³²-labelled oligonucleotides or PCR products (see supplementary information online).

For RT–PCR, RNA prepared with the RNeasy kit (Qiagen, Hilden, Germany) was treated with Turbo DNase (Ambion, Austin, TX, USA) and reverse transcribed using Superscript III RT (Invitrogen, Paisley, UK).

Chromatin immunoprecipitation. The ChIP analysis was carried out as described previously (Pidoux et al, 2004). In brief, cells were fixed in 1% paraformaldehyde for 15 min (H3K9me2) or in 3% paraformaldehyde for 30 min (Swi6) at 25°C. Immunoprecipitation was performed using monoclonal AQH3K9me2 or polyclonal Swi6 antibody and the sample was analysed by quantitative PCR.

Analysis of sequencing data. Individual sequence reads were produced by The GenePool, University of Edinburgh (www.genepool.bio.ed.ac.uk). After trimming the adaptor sequences, the sequence reads were mapped onto the S. pombe reference genome and our GFP constructs by using Short Oligonucleotide Analysis Package 2 (SOAP2). The following SOAP2 parameters were used, thus allowing for no mismatches or insertions and reporting all repeats: ‘-M 0 -r 2 -m 0 -x 0 -v 0'. To compare regions of the genome, the number of mapped sequences was calculated per base and per region using in-house perl scripts (available on request).

Supplementary information is available at EMBO reports online (http://www.emboreports.org).

Supplementary Material

Supplementary Information