RNA interference (RNAi) and microRNA (miRNA) pathways represent RNA silencing mechanisms utilizing short RNA molecules, produced by RNAse III family enzyme Dicer, to guide sequence-specific silencing of gene expression. Factors involved in RNA silencing also participate in the formation and maintenance of heterochromatin. The connection between RNA silencing and chromatin is best established for Schizosaccharomyces pombe and plants where chromatin-associated silencing complexes were extensively characterized (reviewed in [1], [2]). Components of RNA silencing were implicated in transcriptional silencing and heterochromatin also in animals but, unlike in plants and S. pombe, the mechanistic aspects of the connection remain unclear [3]. In mammals, transfected siRNAs were reported to cause heterochromatinization of targeted genes in a process possibly involving Argonaute proteins [4]–[6]. In addition, mammalian Dicer was implicated in the formation of centromeric heterochromatin [7], [8], and in regulation of intergenic transcription at the human and chicken β-globin locus [9], [10]. However, recent data indicate that some of the Dicer effects on epigenetic mechanisms can be indirectly mediated by miRNAs, what warrants caution when interpreting results obtained with cells depleted of Dicer or Argonaute proteins [11], [12].

The uncertainty about small-RNA mediated epigenetic silencing in mammals contrasts with experimentally supported model of heterochromatin formation in S. pombe where Dicer cleaves transcripts from centromeric and other repeats to produce short RNAs that recruit complexes inducing local formation of heterochromatin (reviewed in [2]). Other RNA silencing factors such as Ago1 and Rdp1 operate in cis as stable components of heterochromatin at centromeres, telomeres, mating type locus, and ribosomal DNA (rDNA) repeats [13]. Since mutations of RNAi machinery components lead to increased mitotic recombination of rDNA repeats in S. pombe, the RNAi pathway may also control integrity of rRNA gene arrays [13]. A similar role has been proposed also for quelling, an RNA silencing mechanism in Neurospora crassa [14].

Mammalian ribosomal RNA (rRNA) genes are organized, like in other organisms, in tandemly-repeated arrays. Human rDNA comprise approximately four hundred 43-kb repeats composed of a 13-kb transcribed region and a 30-kb intergenic spacer (IGS). Human rDNA arrays are localized together with satellite repeats on short arms of acrocentric chromosomes [15]. Mouse rDNA has a similar organization [16]. Transcriptional and epigenetic regulation of rDNA has been extensively studied (reviewed in [16], [17]). In established mammalian cell lines, approximately half of rDNA promoters are inactive and methylated [16], [18]. Silencing of rDNA is mediated by the chromatin nucleolar remodeling complex (NoRC), which is guided by 150–300 nucleotide-long noncoding RNAs (pRNAs) that are complementary to the rDNA promoter. Mutations that abrogate RNA binding of TIP5, the large subunit of NoRC, impair association of NoRC with rDNA and fail to promote histone H3 lysine 9 (H3K9) and and histone H4 lysine 20 (H4K20) methylation and HP1 protein recruitment. Knockdown of pRNAs abolishes the nucleolar localization of NoRC, decreases DNA methylation, and enhances rDNA transcription [19], [20]. Connection between pRNAs and RNA silencing pathways is uncertain as analysis of rDNA-derived small RNAs provided no evidence that pRNAs are processed by Dicer [21].

However, several links between RNA silencing and rRNA expression in animals can be found in the literature. Drosha, a mammalian RNase III family enzyme responsible for processing of primary miRNA transcripts to precursor miRNAs, was implicated in pre-rRNA processing [22] though reports contradicting this possibility have also appeared [23]. rRNA fragments are always found among cloned small 20-25-nt-long RNAs although they do not seem to stem from Dicer processing and their physiological role remains unclear [21], [24]. Interestingly, Peng and Karpen have reported decreased H3K9 dimethylation and increased recombination of repetitive DNA, including rDNA, in Drosophila mutants in Su(var)3-9 and Dicer-2 genes, suggesting that RNA silencing pathway is also involved in modulating rDNA chromatin in insects [25].

Here, we report that Dicer is physically associated with rDNA repeats on chromosomes of human and mouse cells. Analysis of rDNA chromatin and rRNA biogenesis suggests that Dicer is not directly regulating expression or processing of rRNA in mammalian cells. Possibly, Dicer association with rDNA chromatin is indicative of its role in maintenance of rDNA tandem repeats. Such a role would be consistent with the fact that Dicer associates with both active and silenced rDNA genes throughout the cell cycle. In addition, our data provide an insight into epigenetic status of rRNA genes in mouse embryonic stem (ES) cells and different tissues.

We have found that mammalian Dicer is physically associated with rDNA chromosomal in mammalian cells. Dicer associates with both transcribed and silenced rRNA genes and it is mostly localized to rDNA region encoding to 45S pre-rRNA. Our data argue against a possibility that the rDNA-associated Dicer has a role in establishing transcriptionally repressive or permissive chromatin, or in rRNA precursor processing. Apart from the analysis of rDNA-associated Dicer, our data provide a comprehensive view of the rDNA chromatin structure in the transcribed region and ISG in mammalian cells.

Analysis of H3K9 and H3K27 modifications confirmed results of a previous study of rDNA in ES cells, which analyzed chromatin structure at the 3′ end of the transcribed rDNA region [31]. Additionally, we characterized chromatin structure of the rRNA promoter and the IGS. We found UBF present at the rRNA promoter but not in the IGS. Interestingly, the rRNA promoter had rather low levels of “active” marks, such as H3K4me2 and H3K9ac when compared to the GAPDH promoter. Repressive histone marks, such as H3K9 methylation and acetylation levels were also low. In contrast to the previous analyses of UBF binding in human HeLa and mouse A9 cells, which suggested that UBF binds throughout rDNA repeat [34], we did not observe enrichment of UBF in the intergenic spacer in mouse ES cells; similarly, Nemeth et al. [35] found that in mouse NIH 3T3 cells the UBF factor is most abundant at the rDNA enhancer/promoter region and not enriched at the intergenic spacer. Nemeth et al. [35] also provided data about occupancy of other transcription factors and on chromatin structure, including some histone modifications, at rDNA genes in NIH 3T3 cells.

It has been repeatedly documented that DNA methylation is associated with transcriptional inactivation of rDNA (for recent review, see [17]). Our data indicate that DNA methylation is not consistently found in mammalian rDNA. In fact, except of transformed cell lines, we only rarely saw clones diagnostic of hypermethylation upon bisulfite sequencing of DNA isolated from primary tissues or cells, including liver and testis, oocytes and blastocysts, and also undifferentiated and differentiated ES cells. Only in brain DNA we observed methylation of CpG at position -143 in approximately 40% of clones. However, even in this case the distribution of methylation in sequenced clones was heterogeneous, which contrasts with the two very distinct patterns of methylation in HEK293 cells (Fig. 3). A survey of the literature and our own data point to extreme diversity of the DNA methylation patterns of rDNA promoters in somatic cells. Published bisulfite sequencing data reveal three patterns of DNA methylation at rRNA promoters: (1) very low (if any) methylation in some clones and strong hypermethylation represented by long stretches of methylated CpGs in other clones [36], [37]; (2) more mosaic distribution of methylated CpGs across individual clones [38]; (3) low, if any, methylation in mouse cells and some primary human cell types (Fig. 7). Considering very different CpG density of the mouse and human rRNA promoters, we propose that CpG methylation at rRNA promoters is an accessory, possibly non-essential mechanism, which stabilizes rDNA silencing but is of variable significance in different cells of different mammalian species. rDNA methylation seems to have a stochastic cumulative nature. It accumulates to larger extend in long-established transformed cell lines such as NIH 3T3 [35], HEK293 or HeLa [36] (and this work), in non-dividing neuronal cells [37], and aged cells [39]. This suggests that the main role of rDNA methylation is long-term stabilization of the most advantageous expression level of rRNA in specific cells.

We report for the first time association of Dicer with chromatin. This association is relatively stable and remains preserved during mitosis as evidenced by staining of metaphase chromosomes. This is similar to several non-histone proteins (for example, the transcription factor UBF [40] or the chromatin remodeling protein ATRX [41]), which remain associated with rDNA chromatin during mitosis. Immunofluorescent staining of interphase cells with anti-Dicer antibodies did not provide conclusive evidence in support of nucleolar localization of Dicer because of a high background staining of whole mount cells (data not shown). In contrast to interphase cells, the signal on mitotic chromosomes is much more concentrated while the background signal is reduced by elimination of the nucleoplasm, therefore allowing for Dicer detection. It is unlikely that the observed staining of mitotic chromosomes is due to non-specific cross-reactivity of Dicer antibodies. Using different human and mouse cells, we detected the same unique staining pattern with two different antibodies raised against distinct Dicer peptides (Fig. 1). Furthermore, similar staining was observed with four different anti-tag (anti-HA, anti-FLAG, anti-Myc, and anti-EGFP) antibodies used to detect Dicer fusions with the respective tags, which were transiently expressed in HEK293 cells (Fig. 2).

Association of Dicer with rDNA presumes that Dicer enters the nucleus. Interestingly, dsRNA-binding domain (dsRBD) was recently shown to play a role in nucleo-cytoplasmic shuttling of ADAR1 [42] and we found that dsRBD of Dicer acts as a nuclear localization signal and that Dicer shuttles between the cytoplasm and the nucleus in human cells (M. Doyle and W. Filipowicz, unpublished results). Shuttling of Dicer and the role of its dsRBD in nuclear localization was also recently reported in S. pombe [43].

While Dicer localization to rDNA is supported by several lines of evidence, its role there remains unclear. It is unlikely that Dicer is involved in processing of pre-rRNA or its fragments into siRNA since siRNA-like small RNAs originating from rRNA and accumulating in mammalian cells were demonstrated to be produced in a Dicer-independent way [24], [44]. However, rRNA-derived small RNAs were found bound to P19 suppressor of RNA silencing known to interact with small siRNA-like duplexes [21]. In addition, small RNAs were shown to induce transcriptional silencing and changes in chromatin structure in mammalian cells [4]–[6]. Could the rDNA associated Dicer be a part of a cis-acting epigenetic system regulating rDNA expression by small RNAs? Several lines of evidence argue against such a role.

Regulation of rDNA chromatin is complex and involves non-coding RNAs. A recent study by Mayer et al. [20] demonstrated that 150–300 nt-long RNA transcribed from the rDNA promoter region is required for the functional integrity of the NoRC complex, which is involved in heterochromatization and silencing of rRNA genes. However, it is unlikely that Dicer plays a role in the NoRC-mediated regulation. In contrast to NoRC, Dicer association extends over an entire rDNA transcribed region. In addition, small RNAs (<25-nt; size expected for the products of Dicer cleavage) were not found among RNAs associated with TIP-5, an RNA binding protein component of NoRC [20].

Furthermore, metabolic labeling experiments performed with Dicer^−/− ES cells showed no apparent defect in pre-rRNA processing in these cells. Although pulse-chase experiments revealed decrease in the level of pre-rRNA synthesis in Dicer^−/− ES cells, this finding is readily explained by a substantially slower growth of these cells when compared to control Dicer^+/− cells ([12], [26]). Furthermore, we previously found that Retinoblastoma-like 2 (Rbl2), a transcriptional repressor, is inhibited in ES cells by miRNAs of the miR-290-295 cluster and, consequently, the Rbl2 level is elevated in Dicer^−/− cells, which are depleted of mature miRNAs [12]. Since Rbl2 overexpression has negative effect on rRNA synthesis [45], it is likely that reduced rRNA expression in Dicer^−/− cells is, at least partially, a consequence of the miRNA loss.

Recently, Peng and Karpen [25] reported that the RNAi machinery is responsible for proper nucleolar architecture in Drosophila. Loss of Dicer-2 in Drosophila led to a pronounced decrease in dimethylation of H3K9 at the rDNA loci in differentiated fly tissues. Dicer-2 mutant flies also showed disrupted organization of nucleoli and decreased stability of rDNA repeats. In contrast, we did not find any decrease in H3K9 methylation in mouse ES cells lacking Dicer. However, it should be stressed that ES cells do not represent an optimal system to study H3K9 methylation of the rRNA promoter as H3K9 repressive marks outside the IGS region are low, if not absent, in rDNA in ES cells (Fig, 6E, D and [31]). Although estimation of size of nucleoli in Dicer^+/− and Dicer^−/− ES cells by light microscopy suggested a small increase in size of nucleoli in Dicer^−/− ES cells (our unpublished results), this observation does not provide any insight into a possible role of Dicer in rDNA chromatin as the aforementioned changes might result from disregulation of miRNA repression or other factors.

Mutations in components of the RNAi machinery, such as Rdp1, Ago1 and Dcr1, exhibit increased mitotic recombination frequency of rDNA repeats also in S. pombe [13]. While the frequency of wild-type colonies showing a reporter gene loss at the tandem rDNA array was 3.3×10⁻³, the recombination frequencies of the RNAi mutant cells were 5- to 10-fold higher [13]. A similar role have been proposed in N. crassa [14]. The lack of evidence supporting involvement of Dicer in transcriptional regulation of rDNA or in pre-rRNA maturation raises a possibility that, by analogy to the situation in S. pombe, N. crassa and Drosophila, the Dicer association with rDNA in mouse and human cells reflects a role of this protein in maintaining the integrity of rDNA repeats also in mammals. While this hypothesis needs to be tested in the future, our data represent the first step towards understanding the role of Dicer association with rDNA in mammalian cells.