A Pre-mRNA–Associating Factor Links Endogenous siRNAs to Chromatin Regulation

In plants and fungi, small RNAs silence gene expression in the nucleus by establishing repressive chromatin states. The role of endogenous small RNAs in metazoan nuclei is largely unknown. Here we show that endogenous small interfering RNAs (endo-siRNAs) direct Histone H3 Lysine 9 methylation (H3K9me) in Caenorhabditis elegans. In addition, we report the identification and characterization of nuclear RNAi defective (nrde)-1 and nrde-4. Endo-siRNA–driven H3K9me requires the nuclear RNAi pathway including the Argonaute (Ago) NRDE-3, the conserved nuclear RNAi factor NRDE-2, as well as NRDE-1 and NRDE-4. Small RNAs direct NRDE-1 to associate with the pre-mRNA and chromatin of genes, which have been targeted by RNAi. NRDE-3 and NRDE-2 are required for the association of NRDE-1 with pre-mRNA and chromatin. NRDE-4 is required for NRDE-1/chromatin association, but not NRDE-1/pre-mRNA association. These data establish that NRDE-1 is a novel pre-mRNA and chromatin-associating factor that links small RNAs to H3K9 methylation. In addition, these results demonstrate that endo-siRNAs direct chromatin modifications via the Nrde pathway in C. elegans.


Introduction
Small regulatory RNAs can silence gene expression in the nucleus by establishing repressive chromatin states. This process, termed Transcriptional Gene Silencing (TGS), was first observed in plants, where small RNAs direct DNA methylation and histone modifications (reviewed in [1]). In addition, the fission yeast, Schizosaccharomyces pombe has been an important model in defining the role of small RNAs in heterochromatin formation. In S. pombe, small RNAs direct the formation of heterochromatin primarily at repetitive DNA elements surrounding centromeres [2,3]. At these repetitive elements, nascent RNAs, transcribed by RNA Polymerase II (RNAP II), serve as platforms for the assembly of RNAi machinery. For instance, the RNA Induced Transcriptional Silencing (RITS) complex, composed of the Argonaute Ago1, the chromodomain protein Chp1, and the glycine and tryptophan (GW)-motif-containing protein Tas3, is guided to nascent transcripts by Argonaute and centromeric siRNAs [4]. The RITS complex recruits chromatin-modifying machinery, such as the histone methyltransferase Clr4, to genomic sites of nuclear RNAi [5,6]. Clr4 catalyzes the methylation of Histone H3 on Lysine 9 (H3K9me) [7]. H3K9me is a conserved molecular mark of heterochromatin [8]. Thus, in plants and S. pombe, small RNAs play a central role in regulating chromatin dynamics. The role of TGS and heterochromatin formation in metazoan silencing processes is less clear [3].
Experimentally provided small RNAs can elicit transcriptional silencing and induce heterochromatic marks in metazoans. In mammalian cells, experimentally provided siRNAs directed against promoter regions can lead to transcriptional silencing and induce heterochromatic marks [9][10][11][12]. Paradoxically, experimentally provided small RNAs can also enhance transcription and decrease H3K9me marks [13,14]. In C. elegans, experimentally provided siRNAs are bound by the Ago NRDE-3 in the cytoplasm, and escorted into the nucleus [15]. NRDE-3/siRNA ribonucleoprotein complexes bind nascent transcripts and recruit the conserved nuclear RNAi factor NRDE-2. The Nrde pathway inhibits RNA Polymerase (RNAP) II during the elongation phase of transcription, and directs the deposition of H3K9me marks at genomic sites that exhibit homology to experimentally introduced siRNAs [16].
How and if endogenously expressed small regulatory RNAs silence gene expression in metazoan nuclei is unclear. Dicer deficient mouse embryonic stem cells express high levels of centromeric repeat RNAs and exhibit altered heterochromatic marks at centromeres [17]. In Drosophila, heterochromatic marks, including H3K9me and HP1, are mislocalized in flies lacking components of the RNAi machinery such as Piwi, Aubergine, and Homeless [18]. In addition, the Drosophila Ago-like protein PIWI binds small RNAs, termed piRNAs, and associates with chromatin [19]. Loss of piwi has variable effects on chromatin states at genomic sites homologous to piRNAs [20][21][22][23][24]. Finally, in C. elegans, animals lacking two RNAi-related factors: the RNAdependent RNA Polymerase EGO-1, or the Ago CSR-1, exhibit large-scale changes in chromosomal H3K9me patterns during germline development [25,26]. Thus, endogenous small regulatory RNAs have been implicated in chromatin regulation in metazo-ans. However, a direct link has yet to be established, and the molecular mechanisms by which this might occur are unknown.
Here we show that the endogenous small RNAs, termed endo-siRNAs, direct H3K9me marks at discrete genomic loci in C. elegans. Small RNA-directed H3K9 methylation requires the Nrde pathway and results in the inhibition of transcription from these loci. In addition, we identify two novel nuclear RNAi factors termed NRDE-1 and NRDE-4, and show that these factors are required for small RNA-directed H3K9 methylation. Finally, we show that small RNAs direct NRDE-1 to associate with pre-mRNA and chromatin of genes, which have been targeted by RNAi. Thus, the Nrde pathway links endogenously expressed small regulatory RNAs to the regulation of transcription and chromatin dynamics in C. elegans.

A genetic screen identifies novel nrde genes
We previously reported a forward genetic screen that identified two genes (termed nrde-2 and nrde-3) required for nuclear RNAi [15,16]. The mechanism(s) by which NRDE-2/3 silence nascent transcripts and inhibit RNAP II transcription are unknown. To understand this mechanism we continued screening for nuclear RNAi factors. .80% of the nrde alleles identified in our original genetic screen were alleles of nrde-3 (Table S1, [15]). To maximize our chances of identifying novel nuclear RNAi factors, we performed our modified screen in animals harboring ectopic copies of nrde-3 (nrde-3::gfp), which was integrated into the genome on chromosome V ( Figure S1). eri-1 encodes an exonuclease that negatively regulates RNAi [27]. Our original screen was conducted in an eri-1(2) genetic background. Our modified screen was conducted in eri-1(+) animals ( Figure S1). Our modified screen identified twenty-three alleles of nrde-2, nineteen alleles of nrde-1, nine alleles of the RNA-dependent RNA Polymerase (RdRP) rrf-1, four alleles of nrde-4, and one additional nrde allele, which complements the known nrde genes, but has not yet been assigned a nrde gene designation (Figure 1a, and Table S1). Here we report the identification and characterization of nrde-1 and nrde-4.

NRDE-1 is a nuclear-localized protein
To determine the molecular identity of nrde-1, we used a single nucleotide polymorphism (SNP)-based mapping approach [32]. We mapped nrde-1 to a 0.86cM interval on Chromosome III that contained 42 genes. The open reading frame (ORF) c14b1.6 lies within this mapping interval. Sequencing of c14b1.6 from three independent nrde-1 alleles revealed three mutations in c14b1.6 ( Figure 1c). Two of these alleles encode premature stop codons, and therefore likely reveal the null phenotype of nrde-1. Expression of a wild-type copy of c14b1.6 was sufficient to rescue the Nrde phenotype associated with nrde-1 (see below). We conclude that c14b1.6 corresponds to nrde-1. Analysis of nrde-1 expressed sequence tags (ESTs) indicated that nrde-1 encodes a protein containing 793 amino acids [33]. Database searches revealed that nrde-1 is conserved in other nematode species, but these searches failed to detect any obvious orthologues of nrde-1 outside nematodes. In addition, these database searches did not identify any obvious protein domains within NRDE-1.

Author Summary
Chromatin consists of DNA and proteins. Chromatin can exist in many different states. The state of chromatin in highly regulated in order to ensure that genes are expressed correctly. RNAs play an important role in the regulation of chromatin. For example, in plants and fungi small RNAs drive the formation of heterochromatin, a repressive chromatin state. Many types of small RNAs have been identified in animal cells, but the functions of these small RNAs are largely unknown. Using the nematode C. elegans as a model system, we identified a small RNA pathway that regulates the state of chromatin. We report the identification of two new factors, termed NRDE-1 and NRDE-4, which act in this nuclear small RNA pathway. NRDE-1 and NRDE-4 link small RNAs to chromatin regulation. Additionally, we show that endogenously expressed small RNAs, termed the endo-siRNAs, direct the post-translational modification of histone proteins, which are core components of chromatin. These results establish a direct connection between small RNAs and chromatin regulation in animals.

NRDE-1 is recruited to pre-mRNAs in response to RNAi
We asked if NRDE-1 was recruited to pre-mRNA following RNAi. We performed NRDE-1 RNA Immuno-Precipitation (RIP) experiments in animals exposed to lin-15b dsRNA. lin-15b RNAi induced a ,30-706 enrichment in un-spliced lin-15b RNA that co-precipitated with FLAG::NRDE-1 (Figure 2d, 2e). The dpy-28 gene encodes a subunit of the C. elegans dosage compensation complex [35]. We tested if dpy-28 dsRNA would induce NRDE-1dpy-28 pre-mRNA association. Following dpy-28 RNAi, NRDE-1 associated with dpy-28 pre-mRNA ( Figure 2d). Finally, dpy-28 or lin-15b RNAi did not result in enrichment of NRDE-1 with lin-15b or dpy-28 pre-mRNA, respectively, indicating that the association of NRDE-1 with pre-mRNA (induced by RNAi) is sequence specific (Figure 2d). We were concerned that NRDE-1 might associate with pre-mRNA targets, in vitro, during sample preparation. To address this issue we pooled extracts from animals exposed to lin-15b dsRNA, and extracts from NRDE-1::GFP expressing animals not exposed to lin-15b, dsRNA and failed to detect an association of NRDE-1 with lin-15b pre-mRNA, indicating that NRDE-1/pre-mRNA interactions likely occurs in vivo (Figure 2e). Taken together, these data show that NRDE-1 associates with pre-mRNAs that have been targeted by RNAi.
NRDE-1 co-precipitating pre-mRNA was enriched for RNA sequences encoded at, or near, the site of RNAi-relative to sequences encoded 59 or 39 to the site of RNAi (Figure 2e). We have previously shown that NRDE factors fail to associate with pre-mRNA sequences encoded 39 to the site of RNAi due to RNAi-mediated inhibition of transcription elongation [16]. We investigated the apparent lack of pre-mRNA sequences encoded 59 to the site of RNAi and found that, while the NRDE factors fail to associate with un-spliced RNA 59 to the site of RNAi, the Nrdes do associate with spliced RNA 59 to the site of RNAi ( Figure S3). Splicing is thought to occur co-transcriptionally [29]. Therefore, the apparent lack of NRDE-1/pre-mRNA association 59 to sites of RNAi may be due to co-transcriptional splicing of nascent transcripts.

NRDE-1 promotes RNAi-directed Histone 3 Lysine 9 methylation
In plants and S. pombe small RNAs direct the methylation of Histone 3 Lysine 9 (H3K9me). Histone methylation results from small RNA-mediated recruitment of histone methyltransferase enzymes to genomic sites exhibiting sequence homology to small RNAs [5]. RNAi also directs H3K9 methylation in C. elegans [16]. nrde-2 is required for RNAi-mediated H3K9 methylation in C. elegans [16]. The mechanism by which the C. elegans Nrde pathway mediates H3K9 methylation is unknown. We conducted H3K9me Chromatin Immuno Precipitation (ChIP) to determine if NRDE-1 was required to link small RNAs to H3K9 methylation. lin-15b RNAi induced a ,306 increase in H3K9me marks at the lin-15 locus (Figure 3). In nrde-1(2) animals, however, lin-15b RNAi had no effect on the methylation status of chromatin at the lin-15b gene ( Figure 3). We conclude that NRDE-1 is required to link small RNAs to H3K9 methylation at a genomic site that has been targeted by RNAi.

RNAi directs NRDE-1 to associate with chromatin
We asked if the NRDE factors themselves might become associated with chromatin in response to RNAi. In order to address this question, we performed NRDE-1/2/3 ChIP exper- Animals of the indicated genotypes were fed bacteria expressing indicated dsRNAs (e.g. lin-15b). eri-1 encodes an exonuclease that is required for the biogenesis of endogenous small interfering RNAs (endo-siRNAs) [36]. eri-1(2) animals have an enhanced RNAi phenotype; they respond more robustly to dsRNA than wild-type animals [27]. Therefore, an eri-1(2) background was used to facilitate phenotypic analysis. The phenotypes (e.g. Multi-vulva) of eri-1(mg366) animals exposed to dsRNA were defined as '+' (,90-100% of animals with phenotype), the phenotypes of eri-1(mg366);rde-1(ne219) were defined as '2' (0% of animals with phenotype). 50-250 animals were scored blind in each trial (n$3  Data are expressed as a ratio of NRDE-1 co-precipitating pre-mRNAs with or without indicated RNAi. Samples exposed to dpy-28 RNAi or lin-15b RNAi were probed with primers targeting iments before or after exposure of animals to dsRNA. In response to lin-15b RNAi, we did not detect any significant increase in the association of NRDE-2 or NRDE-3 with chromatin at the lin-15b gene (Figure 4a). Interestingly, NRDE-1 precipitated ,66 more lin-15b DNA following lin-15b RNAi (Figure 4a). In nrde-2(2) and nrde-3(2) animals, lin-15b RNAi failed to trigger an increase in lin-15b DNA that co-precipitated with NRDE-1 (Figure 4b). We conclude that NRDE-1 is able to IP chromatin of a gene that has been targeted by RNAi, and that the association of NRDE-1 with chromatin requires NRDE-2 and NRDE-3. It is possible that the ability of NRDE-1 to co-precipitate with chromatin may occur as an indirect consequence of NRDE-1/pre-mRNA interactions. To address this issue, we turned our attention to nrde-4.

NRDE-4 is required for RNAi-directed recruitment of NRDE-1 to chromatin
We mapped and cloned nrde-4 ( Figure S4). nrde-4 is predicted to encode a protein containing 788 amino acids [33]. Database searches revealed that nrde-4 is conserved within other nematode species, but not in other species. nrde-4 encodes a predicted bipartite nuclear localization signal (NLS) and no other obvious protein domains ( Figure S4). NRDE-4 is required for silencing nuclear localized RNAs (Table 1), for linking small RNAs to the inhibition of transcription (Figure 1b), and for linking small RNAs to H3K9 methylation ( Figure 3). Interestingly, the recruitment of NRDE-1 (and NRDE-2/3) to pre-mRNA was largely unaffected in animals lacking NRDE-4 (Figure 2c, 2e, Figure S5). NRDE-4 was, however, required for recruitment of NRDE-1 to chromatin in response to RNAi (Figure 4b). These data indicate that NRDE-4 functions downstream of NRDE-1/2/3/pre-mRNA interactions during nuclear RNAi. These data also demonstrate that the ability of NRDE-1 to associate with chromatin is dissociable from the ability of NRDE-1 to associate with pre-mRNA, supporting the idea that NRDE-1 associates with chromatin at genomic sites targeted by RNAi.
Endo-siRNAs direct NRDE-dependent H3K9 methylation C. elegans express at least three types of endogenous small RNAs; the microRNAs, the piRNAs, and the endo-siRNAs. A sub-set of the endo-siRNAs requires ERI-1 for their expression [36,37]. NRDE-3 associates with the ERI-1-dependent endo-siRNAs, but not the other classes of endogenous small RNAs [15,37]. Five lines of evidence cumulatively argue that ERI-1 dependent endo-siRNAs are able to direct the deposition of H3K9me marks in C. elegans. First, in animals that fail to express endo-siRNAs H3K9me marks are depleted at genomic regions exhibiting sequence complementarity to endo-siRNAs. For instance, e01g4.5 siRNAs are amongst the most abundant endo-siRNAs expressed in C. elegans [36]. eri-1(2) animals do not express e01g4.5 endo-siRNAs ( [36,37] and Figure 5a). We conducted H3K9me ChIP and detected a ,66 depletion of H3K9me marks at the e01g4.5 gene in eri-1(2) animals ( Figure 5b). The changes in H3K9me marks were restricted to genomic regions exhibiting homology to endo-siRNAs; surrounding genomic regions, which are not homologous to known small regulatory RNAs, did not exhibited altered H3K9me marks (Figure 5b). Second, in nrde-1/2/3/4 mutant animals we observed a similar localized depletion of H3K9me marks at e01g4.5 (Figure 5c and Figure S6). Third, the e01g4.5 pre-mRNA was over-expressed 2-56 in eri-1 and nrde-1/2/3/4 mutant animals ( [15], Figure S7, and data not shown). Fourth, we performed NRDE-1 RIP and quantified the amount of e01g4.5 pre-mRNA that co-precipitated with NRDE-1. We conducted this experiment in both nrde-2(+) and nrde-2(2) animals as NRDE-2 is required for NRDE-1 recruitment to pre-mRNAs in response to feeding RNAi. We found that NRDE-1 associated with ,56more e01g4.5 pre-mRNA in nrde-2(+) animals than in nrde-2(2) animals ( Figure 5d). These data suggest that NRDE-1 can associate with pre-mRNAs that are homologous to endo-siRNAs, and that this process depends upon components of the Nrde pathway. Five, we detected a subtle and complex, yet reproducible, increase in transcription at the e01g4.5 gene in eri-1 and nrde-1/2/4 mutant animals (Figure 5e and Figure S8). Taken together, these data indicate that e01g4.5 endo-siRNAs are able to direct chromatin modification in C. elegans and that this process requires the Nrde pathway.
Lastly, we investigated the generality of small RNA-mediated chromatin regulation in C. elegans. We queried seven additional genomic sites that exhibit sequence homology to eri-1-dependent endo-siRNAs. At four of these loci H3K9me marks were depleted in eri-1 and nrde-1/2/3/4 animals (Figure 5f). At three of these loci, no significant differences in H3K9me marks were observed. We conclude that Nrde-dependent endogenous small RNA-mediated chromatin modification occurs at multiple loci in C. elegans.

Discussion
Here we report that small RNAs are necessary and sufficient to direct chromatin modification in C. elegans. We show that a class of endogenous small RNAs, termed the endo-siRNAs, direct H3K9 methylation at discrete genomic loci, and that this process requires the Nrde pathway. Finally, we identify two novel nuclear RNAi factors including NRDE-1, which we show is recruited to pre-mRNAs and chromatin by RNAi, and is required to link small RNAs to chromatin regulation.

NRDE-1 and NRDE-4 in nuclear RNAi
What is the role of NRDE-4 in nuclear RNAi? Our preliminary investigation has shown that: NRDE-4 is required to link small RNAs to transcription and chromatin regulation. Interestingly, we find that NRDE-4 is not required for small RNA-directed NRDE-1/pre-mRNA association, but is, required for the recruitment of NRDE-1 to chromatin. Therefore, it seems reasonable to speculate that one role of NRDE-4 during nuclear RNAi may be to load/stabilize NRDE-1 on chromatin, following the recruitment of NRDE-1 to pre-mRNAs by NRDE-2/3 ( Figure 6).
What is the role of NRDE-1 in nuclear RNAi? In response to RNAi, NRDE-1 co-precipitates with both pre-mRNAs and chromatin. We did not detect an association of NRDE-3 or NRDE-2 with chromatin despite the fact that NRDE-2/3 are able, like NRDE-1, to associate with pre-mRNA in response to RNAi (Figure 4a). These data hint that NRDE-1 may possess a chromatin associating property not exhibited by NRDE-2/3. We considered the possibility that NRDE-1 might IP chromatin indirectly via pre-mRNA/RNAP II intermediates. However, we found that in nrde-4(2) animals, NRDE-1 is recruited to pre-mRNAs by RNAi, but does not become associated with chromatin (Figure 2e and Figure 4b). These data demonstrate that the RNA and chromatin associating properties of NRDE-1 can be separated. Additionally, we find that NRDE-1 association with RNA occurs predominantly 59 to the site of RNAi, whereas NRDE-1 association with chromatin occurs predominantly 39 to the site of RNAi (Figure 2e and Figure 4a). Taken together, these data argue that NRDE-1 associates with chromatin in response to RNAi, and that NRDE-1 interacts with pre-mRNAs first and chromatin second during nuclear silencing processes ( Figure 6). The question then becomes; what is the role of NRDE-1 at chromatin?

H3K9me and nuclear RNAi
Here we show that small RNAs promote H3K9 methylation in C. elegans. We show that experimentally introduced small RNAs are sufficient to direct H3K9me marks at genomic sites targeted by RNAi. We also show that small RNAs are necessary to establish H3K9me marks; in animals that fail to express endogenous siRNAs, H3K9me marks are depleted at genomic sites homologous to endo-siRNAs. In S. Pombe, the RNAi machinery directs H3K9 methylation at pericentromeric repeats via recruitment of the H3K9 methyltransferase Clr4 to pre-mRNAs exhibiting homology to pericentromeric siRNAs [5]. Interestingly, fungi lacking H3K9me, due to loss of Clr4, fail to express abundant pericentromeric siRNAs [38]. Thus, H3K9me and the RNAi machinery are thought to comprise a self-reinforcing loop that facilitates heterochromatin formation at pericentromeric regions in S. pombe [39,40]. We find that, in C. elegans, RNAi directs both Figure 3. NRDE-1 is required for RNAi-directed H3K9 methylation. Chromatin Immunoprecipitation (ChIP) with anti-H3K9me3 (Upstate, 07-523) was performed on extracts derived from embryos of animals exposed to +/2 lin-15b RNAi. Co-precipitating H3K9me3 DNA was quantified with qRT-PCR and data are expressed as ratios of samples exposed to lin-15b RNAi or no RNAi (n = 3 +/2 s.d). doi:10.1371/journal.pgen.1002249.g003 H3K9 methylation and the association of NRDE-1 with chromatin. These data hint that C. elegans may employ a similar strategy as S. pombe for establishing heterochromatin; e.g. RNAi promotes H3K9 methylation and H3K9 methylation may help recruit components of the RNAi machinery, such as NRDE-1, to chromatin. In order to test this model, the C. elegans methyltransferase(s) responsible for depositing H3K9me marks in response to RNAi will need to be identified.

Why nuclear RNAi?
In S. pombe small RNAs primarily target repetitive genomic elements. RNAi-directed heterochromatization at pericentromeric repeats permits efficient segregation of chromosomes during meiosis [41]. In plants, small RNAs silence genomic regions enriched in transposons, pericentromeric regions, and rRNA genes [1]. Here we show that ERI-1-dependent endo-siRNAs direct the establishment of heterochromatic marks on chromatin. The biological role(s) of this small RNA-mediated chromatin regulation in C. elegans is unknown. The ERI-1-dependent endo-siRNAs are anti-sense to several hundred cellular mRNAs [36]. In general, these mRNAs appear to be poorly conserved and repetitive, hinting that these mRNAs may represent the products of dead and dying genes [42]. The purpose of nuclear RNAi may be to prevent expression of these dysfunctional genes. Alternatively, these mRNAs may simply serve as templates for the creation of small RNAs, which, in turn, regulate chromatin dynamics.
There are 26 Agos encoded in the worm genome, in addition to nrde-3 [43]. We have detected pleiotropic fertility defects exhibited by nrde-1/2/4(2), but not nrde-3 (2), animals, hinting that other Ago proteins and, perhaps, other types of small RNAs, may engage NRDE-1/2/4 to promote H3K9 methylation during development ( Figure S9). In support of this idea, we find that the recruitment of NRDE-1 to pre-mRNAs and chromatin, in response to RNAi, is not completely abolished in animals harboring null alleles of nrde-3 ( Figure 2e). These data support the idea that other Ago proteins may engage the Nrde pathway to elicit nuclear silencing and chromatin regulation in C. elegans (Figure 2e, Figure 6). The identification of these Ago factors and their small RNA partners will be important for unraveling the cellular connections that exist between endogenous small RNAs and chromatin dynamics in metazoans.
Nuclear run on (NRO) assay NRO was performed as described previously [16]. Hypochlorite-isolated embryos were used for NROs.