RNA Polymerase IV Directs Silencing of Endogenous DNA

See allHide authors and affiliations

Science  01 Apr 2005:
Vol. 308, Issue 5718, pp. 118-120
DOI: 10.1126/science.1106910


Plants encode subunits for a fourth RNA polymerase (Pol IV) in addition to the well-known DNA-dependent RNA polymerases I, II, and III. By mutation of the two largest subunits (NRPD1a and NRPD2), we show that Pol IV silences certain transposons and repetitive DNA in a short interfering RNA pathway involving RNA-dependent RNA polymerase 2 and Dicer-like 3. The existence of this distinct silencing polymerase may explain the paradoxical involvement of an RNA silencing pathway in maintenance of transcriptional silencing.

We previously identified an Arabidopsis sde4 mutant that exhibited partial loss of a transgene-silencing pathway that is otherwise dependent on RNA-dependent RNA polymerase 6 (RDR6) (1). The sde4 mutant plants were also defective for small interfering RNA (siRNA) production and methylation of the SINE retroelement AtSN1 (2) in a pathway that was subsequently associated with RDR2, Dicer-like 3 (DCL3), and other endogenous siRNAs (3). It is likely that this silencing pathway is related to the RNA interference (RNAi)–mediated heterochromatinization in Schizosaccharomyces pombe that is also dependent on siRNA, Dicer, and an RDR (4). Therefore, to investigate the mechanism of both transgene and retroelement silencing we investigated the molecular identity of the SDE4 locus. We describe here how SDE4 encodes the largest subunit of a putative RNA polymerase that is distinct from eukaryotic RNA polymerases I, II, and III (whose subunits in Arabidopsis are designated by the prefix NRPA, NRPB, or NRPC, respectively). In keeping with the convention of naming RNA polymerases, we henceforth refer to this enzyme as RNA polymerase IV (Pol IV) and the largest subunit encoded at SDE4 as NRPD1a. The previously described sde4 mutation is now designated nrpd1a-1.

The screen for mutants defective in transgene silencing used an Arabidopsis line (GxA) in which a green fluorescent protein (GFP) transgene (fig. S1A, designated G) was silenced by a potato virus X (PVX)–GFP transgene (fig. S1A, designated A) (1). Unlike rdr6 mutants, which completely lose GFP silencing in the GxA background, nrpd1a-1 displayed a delayed onset of silencing in the growing points of the plant (Fig. 1A and fig. S1F) (1). In young plants, the leaves initially emerged GFP-fluorescent, but, within a week, silencing appeared in localized areas and then spread throughout the leaf. This delayed onset phenotype persisted until flowering when the young inflorescences were GFP-fluorescent.

Fig. 1.

GFP silencing in nrpd1a-1. (A) Ultraviolet pictures of GxA wild-type (WT) and nrpd1a-1 flowering plants. (B) Northern analysis of GFP mRNA and siRNA in flowers (F), stems (S), and leaves (L) from WT and nrpd1a-1 GxA plants. (C) Southern blot analysis of genomic DNA purified from the same tissues as in (B), digested with the methylation-sensitive enzyme Sau96I, and probed for GFP. Unmethylated G and GxA sgs3 lines serve as controls. DNA fragments larger than 554 nt are caused by DNA methylation.

The patterns of transgene mRNA and siRNA accumulation corresponded to the amount of GFP fluorescence. Thus, the Northern analysis of wild-type and nrpd1a-1 flowers showed that increased accumulation of GFP mRNA (4.5-fold) and PVX-GFP RNA in nrpd1a-1 corresponded to a sixfold reduction of 21- to 24-nucleotide (nt) GFP siRNAs (Fig. 1B) relative to wild type. In fully silenced rosette leaves, where the nrpd1a-mediated loss of silencing was less pronounced than in flowers, GFP mRNA and siRNA amounts were comparable.

RNA-directed DNA methylation (RdDM) is associated with silencing in plants and is likely mediated by an siRNA-guided effector complex. Consistent with RdDM at the GFP locus in GxA plants, GFP DNA methylation is lost in flowers of rdr6 and sgs3 mutants along with GFP siRNAs (1) (Fig. 1C). In nrpd1a-1 flowers, where GFP siRNAs are reduced but not absent, we detected only slight changes in the pattern of GFP DNA methylation (Fig. 1C) that contrast with a more pronounced loss of AtSN1 DNA methylation in nrpd1a-1 plants (2, 5). Thus, the NRPD1a pathway is not the only source of suitable siRNAs for RdDM.

We mapped nrpd1a-1 to a large rearrangement on chromosome 1 that disrupted At1g63020 (fig. S1, B to G) and then used other nrpd1a alleles and complementation with an NRPD1a transgene to confirm that this gene encoded NRPD1a. The sequence of NRPD1a aligned with an A. thaliana NRPD1a homolog (encoded by At2g40030) and two putative proteins from rice in a plant-specific clade for the largest subunit of RNA polymerase (Fig. 2A and fig.S2, A to C). Consistent with NRPD1a being the largest subunit of a multisubunit Pol IV, the rice and A. thaliana genomes encode two proteins that could be the second-largest subunit (At3g18090 and At3g23780) (6) (Fig. 2B and fig. S2D). To test the silencing role of these putative Arabidopsis NRPD2 subunits, we transformed wild-type GxA plants with inverted repeat (IR) constructs that would target RNAi to NRPD1a, the putative NRPD2 genes, or GUS (as a control) (fig. S3, A and B). The transformants with the IR directed against NRPD1a and the putative NRPD2 genes, but not against GUS (fig. S3, C and D), exhibited a delayed onset of silencing like that of nrpd1a-1. Although both NRPD2 genes would be inactivated by the IR construct (fig. S3B), two lines of evidence indicate that the active NRPD2 locus is At3g23780 rather than At3g18090. First, gene-specific reverse transcription polymerase chain reaction (RT-PCR) analysis showed that most of the NRPD2 transcripts come from At3g23780 (fig. S3E), and, second, an insertion mutant in At3g23780 (nrpd2a-1, fig. S3F) but not At3g18090 (nrpd2b-1, fig. S3F) eliminated endogenous siRNA (Fig. 3A).

Fig. 2.

Phylogenetic trees of the largest (A) and second largest (B) RNA polymerase subunit families from plants, worms, and yeast. I to IV on the right of each tree indicates RNA polymerase subfamily. Gray and black boxes indicate Arabidopsis Pol IV subunits.

Fig. 3.

Molecular indicators of silencing. (A) Northern analysis of endogenous small RNAs: lanes 1 to 8, GxA (C24) background; lanes 9 to 13, Col-O background. Lane 1, WT; lane 2, rdr2-3; lanes 3 and 4, two independent rdr2-3 RDR2p::RDR2 T2 lines; lane 5, nrpd1a-1; lanes 6 and 7, two independent nrpd1a-1 NRPD1ap::NRPD1a T2 lines; lane 8, nrpd1a-2; lane 9, WT; lane 10, nrpd1a-3; lane 11, nrpd1a-4; lane12, nrpd2a-1; lane 13, nrpd2b-1. Blots were stripped and reprobed multiple times. (B) Southern analysis of 5S rDNA repeats in the NRPD1a allelic series after digestion with the methylation-sensitive enzyme HpaII. (C) RT-PCR analysis of endogenous AtSN1 transcripts that are elevated in rdr2-2 and nrpd1a mutants.

NRPD1a and NRPD2 have sequence similarities to their NRPA, NRPB, and NRPC homologs in regions corresponding to functionally important features of yeast RNA polymerase II (7). For NRPD1a, regions of identity with NRPB1 include the N-terminal clamp core (20%), the Zn metal binding sites, the active site (41%), the funnel and portions of the cleft domain (beginning, 42%, and end, 24%). The middle of the cleft domain in NRPA1, NRPB1, and NRPC1 is absent in NRPD1a (conserved region G) (fig. S2B). However, the C-terminal clamp core (fig. S2B), which defines the end of the globular domain before the C-terminal domain (CTD) of NRPB1 (7), is present (23% identical). NRPD2 is 34% identical to NRPB2 and has conserved regions corresponding to the binding sites of smaller core subunits (NRPB3/AC40, NRPB10, and NRPB12) (fig. S2D) that are thought to play a role in enzyme assembly (7). There are multiple genes for NRPB3/AC40, and some of them could be Pol IV–specific. However, for NRPB10 and NRPB12 there are only two genes each in Arabidopsis, and they could be shared between Pol IV and other RNA polymerases because they are in other eukaryotes (8).

The most striking difference between NRPB1 and NRPD1a is at the C terminus of NRPD1, where the rice and Arabidopsis proteins share similarity to the C-terminal half of a nuclear-encoded protein (defective chloroplasts and leaves) that regulates rRNA processing in chloroplasts (9) (fig. S2A). This C-terminal extension may coordinate RNA polymerase activity with downstream processing steps in a manner analogous to the CTD of Pol II (8). The combination of this unusual C terminus with conserved features of RNA polymerases suggests that Pol IV is an active RNA polymerase with important differences from RNA polymerases I, II, and III.

To further investigate the mechanism of Pol IV–mediated silencing, we characterized another mutant with a delayed onset of GFP silencing (fig. S4A). It has an inversion on chromosome 4 that disrupts RDR2 (fig. S4B, rdr2-3), and the GFP silencing phenotype could be complemented by transformation with wild-type RDR2 genomic DNA (fig. S4A). In this mutant, as in nrpd2a-1 and each of the nrpd1a mutants, there were lower amounts of endogenous 24-nt siRNAs (AtSN1, 1003, 02, and cluster 2) than in wild type, whereas miR167 amounts were unaffected (Fig. 3A) (3). These siRNAs are all in the 24-nt size class, and the DCL3 Dicer is involved in their biogenesis (3). It is likely therefore that Pol IV, RDR2, and DCL3 act together in a silencing pathway. Pol IV would generate RNA species that are copied into double-stranded RNA (dsRNA) by RDR2. This dsRNA would then be processed into siRNA by DCL3.

The RDR2 and RDP1 mutations also affected long RNAs from the siRNA loci. However, opposite to the effect on siRNA, the long RNAs were more abundant in the silencing mutants. Thus, AtSN1 RNA was more abundant in the nrpd1a and rdr2 mutants than in wild-type plants (Fig. 3C), indicating that they are not Pol IV transcripts. Presumably this RNA results from the RNA polymerase III transcription (10) of AtSN1 that is silenced by the Pol IV–RDR2–DCL3 pathway in wild-type plants. Our inability to detect long Pol IV transcripts from AtSN1 is presumably because they are present at only very low amounts and are masked by more abundant RNAs produced by other RNA polymerases.

The Pol IV–RDR2–DCL3 pathway associated with cluster 2 and 02 siRNAs does not require AGO4 or DNA methylation (3). However, the AtSN1 and 1003 loci are hypermethylated in wild-type plants relative to the nrpd1a mutants (2) (Fig. 3B), and accumulation of their siRNAs is dependent on the Argonaute protein AGO4, the de novo DNA methyltransferases DRM1 and DRM2 (5, 11, 12), as well as Pol IV, RDR2, and DCL3. In these examples, it seems, as has been proposed in S. pombe (13), that maintenance of silencing involves a self-reinforcing silencing pathway in which Pol IV–mediated siRNA production is dependent on the AGO4-mediated DNA methylation and vice versa.

The role of Pol IV as described here could resolve a paradox of chromatin silencing mechanisms that are dependent on RNA. If the silenced loci are transcribed by a single polymerase, such RNA-dependent mechanisms would not be stable because the suppression of transcription from these loci would also lead to loss of silencing RNA. However, a silencing-specific polymerase like Pol IV may be resistant to chromatin or DNA modifications affecting polymerases I to III and so could stably maintain the silenced state. In animals and fungi the Pol IV clade is absent, but the silencing paradox may be resolved if there are forms of the RNA polymerases I to III with silencing-specific subunits that allow transcription of heterochromatin and, consequently, maintenance of the RNA-dependent silencing.

A final general point about silencing mechanisms is illustrated by the GFP transgene silencing and endogenous 24-nt siRNA silencing pathways that are both dependent on RDR proteins (fig. S4). In rosette leaves of the plant, these two silencing pathways are independent of each other because transgene silencing is unaffected by rdr2 (fig. S4) and because endogenous 24nt siRNA persists in rdr6 plants (2). However, in flowers these pathways are interdependent, because the GFP silencing is affected by both rdr2 and rdr6. This potential of silencing pathways to interact, combined with their ability to form feedback and self-reinforcing loops (14), illustrates the potential complexity of endogenous regulatory networks involving siRNAs.

Supporting Online Material

Materials and Methods

Figs. S1 to S4

Tables S1 and S2

References and Notes

References and Notes

View Abstract

Stay Connected to Science

Navigate This Article