Noncoding transcription by RNA Polymerase Pol IVb/Pol V mediates transcriptional silencing of overlapping and adjacent genes

Identification of Pol V-dependent transcripts in intergenic noncoding regions

A heterochromatic knob, or chromomere, on the northern arm of A. thaliana chromosome 4 is a well-characterized interval rich in transposons and other heterochromatic repeats (Fransz et al., 2000; Lippman et al., 2004). Within this domain are intergenic noncoding (IGN) regions at which RNA transcripts have not been detected using tiling DNA microarrays (Lippman et al., 2004). Nonetheless, siRNAs and DNA hypermethylation often map to these regions (Kasschau et al., 2007; Lippman et al., 2005; Lister et al., 2008), suggesting that low abundance transcripts might serve as siRNA precursors. Therefore, we used RT-PCR to search for IGN RNAs present in wild-type plants but missing in Pol IV or Pol V mutants. Of 14 IGN regions examined, six had RNAs that were lost or reduced in Pol V mutants (Fig. 1, Fig. S1). For instance, at IGN5 and IGN6 (intergenic noncoding regions 5 and 6; Fig. 1A, B), transcripts detected in wild type (ecotype Col-0) and nrpd1 mutants are depleted in nrpe1 (nrpd1b-11) or nrpd2 mutants (Fig. 1E, top three rows), indicating that Pol V, but not Pol IV, is required for their production. However, AtSN1-family retrotransposons are derepressed (activated) in both the Pol IV and Pol V mutants (Fig. 1E, fourth row from the top). Actin 2 mRNA abundance is unaffected by the mutations (Fig. 1E).

IGN5 and IGN6 are located in regions rich in transposon-derived elements, siRNA production and DNA hypermethylation (Lister et al., 2008), all characteristic of heterochromatic domains. Pol V-dependent transcripts are also detected at IGN7 and IGN17 (Fig. 1F), which are located in pericentromeric heterochromatic regions (Fig. S1). However, IGN10 and IGN15 are present in gene-rich environments with relatively few transposon-related repeats (Figs. 1C, D), yet also give rise to Pol V-dependent transcripts (Fig. 1F). Collectively, these data suggest that Pol V contributes to IGN transcription in both heterochromatic and euchromatic environments.

Characterization of Pol V-dependent transcripts

To determine if Pol V-dependent RNAs initiate at specific sites, we performed 5′ RACE at IGN5 and IGN6 (Figs. 2A–C). Resulting PCR-amplified RACE products yielded distinct bands upon agarose gel electrophoresis (Fig. 2C) but excising the bands and cloning and sequencing of the cDNAs revealed heterogeneity at the 5′ ends. At IGN5, top-strand clones initiated at two sites, seven nucleotides apart (Fig. 2A, Fig. S2). An IGN5 bottom-strand-specific primer yielded five different 5′ ends spanning a 33 nt interval (Fig. 2A, Fig. S2). At IGN6, clones derived from the gel-purified upper and lower bands collectively revealed four distinct 5′ ends spanning a 94 nt interval (Fig. 2B, Fig. S2). Bottom strand-specific transcripts were not detected at IGN6.

It is noteworthy that the 5′ terminal nucleotides of all RACE products were adenosine or guanosine (Fig. S2) given that transcripts of eukaryotic Pol I, II, III and bacterial RNA polymerase typically begin with purines (Smale and Kadonaga, 2003; Sollner-Webb and Reeder, 1979; Zecherle et al., 1996). To test whether RACE 5′ ends represent transcription start sites or cleavage sites, we exploited the fact that initiating nucleotides have 5′ triphosphate groups (Pol I, Pol III) or 7-methylguanosine caps (Pol II). By contrast, cleaved RNAs have 5′ monophosphate or hydroxyl groups. Terminator^™ exonuclease (Epicentre Biotechnologies) is a 5′->3′ exonuclease that degrades RNAs having 5′ monophosphates but not RNAs that have 5′ triphosphate groups, 5′ hydroxyl groups or 7-methylguanosine caps. Total RNA treated with Terminator endonuclease was subjected to RT-PCR using IGN5-specific primers (Fig. 2D; interval A is depicted in Fig 2A). In agreement with Figure 1, IGN5 transcripts were detected in wild-type (Col-0) plants but were absent in the Pol V mutant (nrpe1/nrpd1b-11). Terminator exonuclease treatment prior to RT-PCR caused a ~70% reduction in the Pol V-dependent IGN5 transcript signal, suggesting that the majority of the transcripts amplified by PCR are 5′-monophosphorylated; however, the remaining transcripts are resistant to the exonuclease (Fig. 2D). Treatment of the RNA with Tobacco Acid Pyrophosphatase, which removes 7-methylguanosine caps or triphosphates and leaves a 5′ monophosphate, rendered the IGN5 transcripts and actin control fully susceptible to Terminator exonuclease digestion. Therefore, IGN5 transcripts that require Tobacco Acid Pyrophosphatase in order to be made Terminator-susceptible are deduced to be triphosphorylated or capped (Fig. 2D), indicative of transcription start sites. It is noteworthy that 5′ RACE requires a 5′ monophosphate for adaptor ligation. RACE products were only obtained upon treating RNA with Tobacco Acid Pyrophosphatase, but not upon treating RNA with T4 Polynucleotide Kinase and ATP (data not shown), which would have converted 5′ hydroxyls to phosphates and allowed their cloning. Collectively, our observations suggest that the 5′ ends detected by RACE are transcription start sites. However, much of the RNA detected by RT-PCR consists of processed RNAs.

To test whether Pol V-dependent transcripts are polyadenylated, total RNA was fractionated using oligo d(T) magnetic beads. IGN5 transcripts were detected in total RNA and polyA-depleted fractions of wild-type Col-0, but were not detected in Poly A-enriched RNA (Fig. 2E), unlike Actin 2 mRNA. AtSN1 transcripts produced in nrpe1 (nrpb1b-11) mutants were present in total and polyA-depleted, but not Poly A-enriched RNA, consistent with Pol III transcription of AtSN1 (see below).

Collectively, the assays of Figure 2 suggest that Pol V-dependent transcripts can be at least ~200 nt in size, can initiate from multiple sites, have triphosphates or 7meG caps at their 5′ ends, and lack poly A tails.

Evidence that Pol V synthesizes IGN transcripts

The largest subunits of Pol IV and Pol V include sequences that are invariant among DNA-dependent RNA polymerases, including a DFDGD at the active site (Metal A site) that coordinates a magnesium ion essential for nucleoside polymerization (Cramer, 2004). We tested the importance of the presumptive NRPE1 Metal A site by analyzing nrpe1 (nrpd1b-11) mutants transformed with a wild-type NRPE1 transgene or a transgene in which the invariant aspartates were changed to alanines (active site mutant; abbreviated ASM) (Fig. 3A). Both transgenes utilized the native NRPE1 promoter, included their full complement of introns and exons, and were similarly expressed, as shown by immunoblot detection of the FLAG epitope tags added to their C-termini (Fig. 3B, bottom row). Moreover, the wild-type and ASM mutant proteins both co-immunoprecipitate NRPD2/NRPE2, the second-largest subunit of both Pol IV and Pol V, suggesting that the ASM mutation does not disrupt Pol V subunit assembly (Haag and Pikaard, unpublished). The wild-type NRPE1 transgene restored Pol V-dependent IGN5 and IGN6 transcripts in the nrpe1 (nrpd1b-11) mutant background but the NRPE1-ASM transgene did not (Fig. 3B), indicating that synthesis of Pol V-dependent transcripts requires the conserved active site.

To determine if NRPE1 physically interacts with loci giving rise to Pol V-dependent transcripts, we performed chromatin immunoprecipitation (ChIP) of FLAG-tagged NRPE1 as well as FLAG-tagged NRPB2, the second-largest subunit of RNA polymerase II (Figure 3C). Subsequent quantitative real-time PCR showed that NRPE1 physically associates with IGN5, whereas NRPB2 does not. A retrotransposon-derived solo LTR (long terminal repeat) shown to be silenced in a Pol V-dependent manner (Huettel et al., 2006) is also occupied by NRPE1. The solo LTR most likely programs Pol II transcription, and Pol II is detected at this locus above background (defined as ChIP signals obtained with Col-0 plants that lack a FLAG-tagged transgene), but at lower levels that at the actin 2 gene locus At3g18780. Collectively, the ChIP data indicate that Pol V is present at loci that give rise to Pol V-dependent RNAs.

We next asked whether Pol V-dependent RNAs could be immunoprecipitated (IPed) in association with NRPE1. Formaldehyde-crosslinked chromatin preparations of non-transgenic Col-0, or nrpe1 (nrpd1b-11) lines expressing FLAG-tagged NRPE1 were IPed using anti-FLAG antibody. Following DNase I treatment, samples were tested by RT-PCR (Fig. 3D). IGN5, IGN6, AtSN1 and solo LTR RNAs were all enriched in IP fractions of NRPE1-FLAG plants compared to non-transgenic Col-0 controls that were also subjected to anti-FLAG IP (Fig. 3D). Background levels of abundant actin mRNA were equivalent in Col-0 and NRPE1-FLAG IP fractions, indicating that the enrichment of the IGN and transposon RNAs in NRPE1-FLAG IP fractions compared to Col-0 reflects specific interaction of these RNAs with Pol V. Because Pol V-dependent transcripts require the presumptive NRPE1 active site, NRPE1 physically associates with loci giving rise to these transcripts and NRPE1 physically associates with the transcripts themselves, we deduce that the Pol V synthesizes the transcripts.

Pol V transcription is necessary in order to silence overlapping and adjacent genes

Transcriptional silencing of AtSN1 retroelements requires both Pol IV and Pol V (see Fig. 1E). AtSN1-family elements are short interspersed nuclear elements (SINEs) that possess A-box and B-box elements (see diagram in Fig. 4A) typical of the internal promoters of Pol III-transcribed genes (Myouga et al., 2001). In wild-type (Col-0) plants, AtSN1 elements are silenced, but in nrpe1 (nrpd1b-11) mutants they are derepressed (Fig. 1E and Fig. 4C, interval A). AtSN1 silencing is restored in nrpe1 mutants by the full-length NRPE1 transgene but not by the active site mutant NRPE1-ASM transgene (Fig. 4C, top row), indicating that Pol V transcription is required for AtSN1 silencing. In the intergenic region, and overlapping the expected Pol III transcription start site (see Fig. 4A and Fig. S3), IGN transcripts corresponding to both DNA strands can be detected by RT-PCR. These transcripts, within intervals B and C, are readily detected in wild-type plants but are absent, or much reduced, in nrpe1 mutants (Fig. 4C, rows 2–5). The interval B and C transcripts are restored in nrpe1 mutants by the wild-type NRPE1 transgene, but not by the NRPE1-ASM transgene. Collectively, the data indicate that AtSN1 transcripts are only generated if Pol V transcripts are absent.

Like AtSN1, a long interspersed nuclear element (LINE), At5g27845, which overlaps the solo LTR (see Fig. 4B), is silenced in a Pol V (Pol IVb)-dependent manner (Huettel et al., 2006). Transcription of this LINE is low in wild-type plants, but increases substantially in the nrpe1 (nrpd1b-11) mutant (Fig. 4D, RT-PCR interval A). Silencing is restored by the wild-type NRPE1 transgene, but not by the NRPE1-ASM transgene (Fig. 4D).

In wild-type plants, transcripts are detected from both strands upstream of the LINE and solo LTR (interval B), including intergenic sequences and overlapping an adjacent transcription unit, At5g27850 (see Fig. 4B). These RNAs in wild-type plants might be Pol V transcripts. However, unlike the intergenic region adjacent to AtSN1, where transcripts disappear in nrpe1 (nrpd1b-11) mutants, suggesting that Pol V is the sole polymerase transcribing the region, transcript abundance in the region adjacent to the solo LTR increases dramatically in nrpe1 or NRPE1-ASM transgenic plants (Fig. 4D). This increased transcription is attributable to RNA polymerase II, as shown by ChIP (Fig. 4F). Whereas Pol II occupancy of the locus is low in wild-type plants, it increases dramatically in the nrpe1 mutant. Transformation of nrpe1 with the wild-type NRPE1 transgene reduces Pol II occupancy of the locus, whereas the NRPE1-ASM mutant is ineffective (Fig. 4F). Taken together, the data indicate that derepression of Pol II transcription in the solo LTR region occurs in the absence of Pol V transcription.

A LINE element located to the right of IGN5 is expressed at low levels in wild type plants but is derepressed in the nrpe1 mutant (Fig. 4E). Silencing is restored by the wild-type NRPE1 transgene but not by the NRPE1-ASM mutant transgene. Collectively, the data of Figure 4 indicate that intergenic Pol V transcription plays a direct role in suppressing transcription from overlapping or adjacent LINE and SINE transposons.

Pol V transcription is necessary for heterochromatin formation at affected loci

We next examined histone modifications and cytosine methylation at Pol V affected loci (Fig. 5). ChIP using an antibody specific for histone H3 lysine 27 mono-methylation (H3K27me1), a heterochromatic mark previously shown to be dependent on Pol V (Pol IVb) (Huettel et al., 2006), resulted in significant enrichment of IGN5, the solo LTR region and AtSN1 relative to the actin gene control (Fig. 5A). Decreased H3K27me1 at the IGN5, solo LTR and AtSN1 loci in nrpe1 (nrpd1b-11) was restored by the NRPE1 transgene but not the NRPE1-ASM transgene (Fig. 5A). ChIP controls in which antibody was omitted yielded negligible background signals (Fig. S4). ChIP using an antibody specific for dimethylated histone H3 lysine 9 (H3K9me2), also a heterochromatic mark, showed association of IGN5 and the solo LTR region that was reduced in nrpe1 and rescued by the wild-type NRPE1 transgene but not the NRPE1-ASM transgene (Fig. 4). Interestingly, Pol V mutations did not significantly affect H3K9me2 at AtSN1, despite their pronounced effect on H3K27me1 at the locus.

Diacetylation of histone H3 on lysines 9 and 14 (abbreviated H3Ac2) is a characteristic of active, euchromatic genes, such as actin (Fig. 5C). At the solo LTR, H3Ac2 levels increased significantly in the nrpe1 (nrpd1b-11) mutant and were restored by the wild-type NRPE1 transgene but not the NRPE1-ASM transgene (Fig. 5C). These results parallel increased Pol II occupancy of the locus in the absence of functional NRPE1 (see Fig. 4F). H3Ac2 levels at IGN5 and AtSN1 were not influenced by NRPE1. Differences in histone hyperacetylation at the loci may reflect the different RNA polymerases transcribing them: IGN5 is transcribed by Pol V and AtSN1 is presumably transcribed by Pol III, whereas Pol II transcribes the solo LTR.

We assayed IGN5, IGN6 and solo LTR DNA methylation status based on McrBC endonuclease sensitivity (Fig. 5D). McrBC specifically cleaves methylated DNA, preventing its subsequent amplification by PCR. In wild-type Col-0, methylcytosine levels are high at IGN5, IGN6 and the solo LTR, such that McrBC digestion reduces their PCR amplification by ~80% (Fig. 5D). At IGN5 and the solo LTR, DNA methylation is significantly reduced in the nrpe1 (nrpd1b-11) mutant and in a null mutant for RNA-DEPENDENT RNA POLYMERASE 2 (RDR2), a protein required for 24 nt siRNA biogenesis (Xie et al., 2004). In the nrpe1 mutant background, IGN5 and solo LTR methylation are restored by the wild-type NRPE1 transgene but not by the NRPE1-ASM transgene. The data indicate that Pol V transcription, like RDR2, is needed for siRNA-directed DNA methylation at these loci.

Unlike IGN5 and the solo LTR, DNA methylation at IGN6 does not require Pol V or RDR2, but does require DDM1 (DECREASE IN DNA METHYLATION 1), a SWI/SNF-family chromatin remodeller that acts primarily in the maintenance, rather than RNA-mediated establishment of cytosine methylation (Jeddeloh et al., 1999). DDM1 also affects maintenance methylation at IGN5, but has no appreciable effect at the solo LTR, which may rely exclusively on RNA-directed DNA methylation.

Loss of DNA methylation at the AtSN1, IGN5 and solo LTR loci in the nrpe1 (nrpd1b-11) mutant was also demonstrated using methylation-sensitive restriction endonucleases (Figs. 5E, F). Methylation of HaeIII or AluI recognition sites blocks the enzymes from cutting the DNA, allowing PCR amplification of the region. However, unmethylated sites are cleaved such that PCR amplification fails. DNA methylation was lost at HaeIII or AluI sites of the AtSN1, IGN5, and solo LTR loci in the nrpe1 mutant and was restored by the wild-type NRPE1 transgene but not by the NRPE1-ASM transgene (Fig. 5E, F). At IGN6, no effect of nrpe1 was observed on methylation of the sole AluI site tested (Fig. 5F). Collectively, the data indicate that Pol V mediates the establishment of heterochromatic histone modifications and DNA methylation changes that correlate with the silencing of Pol II or Pol III-transcribed genes that overlap the Pol V-transcribed regions.

Pol V-dependent transcription does not require small RNA biogenesis

Because Pol V is required for siRNA-dependent DNA methylation, we asked whether mutations in genes required for siRNA biogenesis, RNA-directed gene silencing or DNA methylation affect Pol V transcription (Fig. 6A). At IGN5 and IGN6, Pol V transcripts lost in nrpe1 (nrpd1b-11) and nrpd2 mutants were unaffected by mutation of the four Dicers that process double-stranded RNA precursors into siRNAs, including a quadruple mutant that combines a hypomorphic dcl1 allele with null alleles of dcl2, dcl3 and dcl4. Pol V-dependent transcripts were also unaffected in mutants defective for RNA-dependent RNA polymerases (rdr2, rdr1 and rdr6) implicated in generating siRNA precursors, or in mutants affecting cytosine methylation (drm2, met1, ddm1). However, many of these mutants interfere with AtSN1 silencing, including the dicer quadruple mutant, rdr2, nrpd1a, drm2, and drd1 (Fig. 6A, row 4). Collectively, the results reveal that Pol V transcription occurs independent of small RNA biogenesis, de novo cytosine methylation (drm2) or maintenance cytosine methylation (met1, ddm1). However, Pol V and siRNA biogenesis are both required for AtSN1 silencing.

DRD1 facilitates the association of Pol V with chromatin

As shown in Figure 6A, Pol V transcripts are lost in drd1-6 mutants. DRD1 is a member of the SWI2/SNF2 family of ATP-dependent chromatin remodelers and was identified in a genetic screen that also identified nrpe1 (nrpd1b) and nrpd2 alleles, suggesting that DRD1 and Pol V act in collaboration (Huettel et al., 2007). ChIP of FLAG-tagged NRPE1 in wild-type or drd1 mutant backgrounds was conducted to determine if DRD1 regulates NRPE1 association with chromatin (Fig. 6B). NRPE1-FLAG protein levels were similar in both genetic backgrounds (Fig. 6C). In nrpe1 plants that are wild-type at the DRD1 locus, the NRPE1-FLAG protein physically associates with IGN5, IGN6 and the solo LTR locus (Fig. 6B). However, in the drd1 mutant background, NRPE1 association with these loci is reduced to background levels resembling the actin gene control (Fig. 6B). We conclude that DRD1 mediates Pol V recruitment to chromatin.

Polymerase activity of Pol V

RNA polymerase activity has not yet been demonstrated for Pol IV or Pol V in vitro. However, our study provides in vivo evidence for Pol V polymerase activity by demonstrating the existence of Pol V-dependent transcripts, by showing that these RNAs require the conserved polymerase active site, by showing that Pol V physically associates with DNA loci corresponding to Pol V-dependent transcripts and by showing that Pol V physically associates with the transcripts themselves. The most parsimonious explanation for the results is that Pol V transcribes DNA into RNA, which fits with the crosslinking of Pol V to both DNA and RNA and with the requirement for the putative chromatin remodeler DRD1 in order for Pol V to associate with transcribed loci. DRD1 and Pol V do not appear to physically interact, based on co-IP experiments (Ream, Wierzbiecki and Pikaard, unpublished), suggesting that DRD1 functions upstream of Pol V, presumably by remodelling chromatin to facilitate Pol V recruitment to the DNA. If Pol V were to utilize RNA templates, a prediction is that Pol V-dependent transcript abundance would increase in accord with the abundance of RNAs serving as templates. However, mutations that derepress transposons, including rdr2, drm2, met1 or ddm1, have no effect on Pol V transcript abundance. Likewise, Pol V transcripts do not decrease in mutants for the major RNA-directed RNA polymerases, rdr2 or rdr6, which could potentially generate RNA templates for Pol V.

Detection of multiple Pol V transcript 5′ ends using RACE suggests that Pol V may initiate transcription in a promoter-independent fashion. How sites of Pol V initiation are chosen is unclear. One hypothesis is that specific DNA methylation patterns or histone modifications recruit Pol V. However, Pol V transcripts are detectable in both heterochromatic, transposon-rich regions as well as gene-rich, presumably euchromatic environments. Moreover, mutants affecting siRNA production or DNA methylation have no effect on Pol V transcript abundance. An alternative possibility, which we favor, is that Pol V initiates transcripts throughout the genome, both in silenced and non-silenced regions, and these transcripts are necessary but not sufficient for gene silencing. Instead, we envision that Pol V transcription renders a locus competent for silencing, but silencing only occurs if siRNAs complementary to the locus are also produced (see below).

The role of Pol V transcription in transcriptional gene silencing

Noncoding RNAs (ncRNAs) originating in intergenic regions are prevalent in eukaryotes, including Arabidopsis, but their functional significance is mostly unknown. Our results indicate that Pol V-transcribed ncRNAs play direct roles in silencing overlapping or adjacent genes. At the AtSN1 locus, Pol V transcripts and retrotransposon transcripts presumably generated by Pol III are mutually exclusive, suggesting that Pol V transcription prevents Pol III transcription. Likewise, at the solo LTR locus, Pol II association is low in wild-type plants but increases 35-fold in nrpe1 mutants. Similar increases in transcription of the LINE element adjacent to IGN5 occur in nrpe1 mutants. Collectively, the data indicate that Pol V transcription facilitates the silencing of overlapping genes as a result of repressive chromatin modifications, including H3K9 methylation, H3K27 methylation and cytosine hypermethylation.

Pol V transcription is necessary but not sufficient to silence AtSN1 and solo LTR elements. Other necessary proteins include Pol IV, RDR2, one or more DCL proteins, AGO4, DRD1 and DRM2 (see Fig. 6), which are components of the 24 nt siRNA-directed DNA methylation pathway. Because mutants that disrupt siRNA biogenesis (e.yrpd1, rdr2, dicer) have no effect on the production of Pol V dependent transcripts, our results suggest that Pol V transcription and siRNA production occur independently but collaborate in gene silencing. This hypothesis fits with the observation that Pol V is not required for siRNA production at the majority of the >4000 loci giving rise to 24 nt siRNAs (Mosher et al., 2008), including the AtSN1(Kanno et al., 2005; Pontes et al., 2006) and solo LTR (Huettel et al., 2006) loci we have examined. At other endogenous repeat loci giving rise to siRNAs, all of which require Pol IV, Pol V is apparently required (Mosher et al., 2008). However, this does not necessarily imply that Pol V transcripts serve as siRNA precursors. Instead, Pol V-dependent heterochromatin formation may stimulate Pol IV-dependent production of siRNAs in a positive feedback loop that enforces gene silencing (Li et al., 2006; Pontes et al., 2006).

In our alternative models (Figure 7), we envision that chromatin remodeling by DRD1 is required for Pol V transcription initiation. In parallel, siRNAs produced by the combined actions of Pol IV, RDR2 and DCL3 are incorporated into AGO4. Our favored model is that Pol V transcripts basepair with siRNAs that are associated with AGO4 (Figure 7A), similar to the way Pol II transcripts reading through silenced fission yeast pericentromeric regions are proposed to interact with the siRNA-AGO moiety of the RNA-induced Transcriptional Silencing (RITS) complex (Buhler et al., 2006; Irvine et al., 2006). The interaction of the siRNA with the nascent transcript might then direct the silencing machinery, including the de novo cytosine methyltransferase DRM2 and/or histone modifying activities, to the adjacent DNA. Alternatively, Pol V transcripts may directly bind to AGO4 and stabilize siRNA-DNA interactions (Figure 7B). It is also possible that Pol V transcripts, or the act of transcription itself, influences structural features of heterochromatin that are required by AGO4 for efficient interactions with target loci (Figure 7C). In each of these scenarios, AGO4 recruitment is expected to be co-transcriptional and may involve direct interactions between AGO4 and the C-terminal domain of NRPE1/NRPD1b (El-Shami et al., 2007; Li et al., 2006). A prediction of all of the models is that transcriptional silencing does not occur everywhere that Pol V transcription occurs, but only at sites where Pol V transcription and siRNA production overlap. Testing this hypothesis on a whole genome basis is a goal for future studies.

Plant strains

A. thaliana nrpd1a-3 (nrpd1), nrpd1b-11 (nrpe1) and nrpd2a-2 nrpd2b-1 mutants were described previously (Onodera et al., 2005; Pontes et al., 2006), as were nrdp1b-11 NRPD1b-FLAG (NRPE1-FLAG)(Pontes et al., 2006) and NRPB2-FLAG (Onodera et al., 2008) transgenic lines. NRPE1 mutagenesis and production of transgenic lines expressing Pol IV and Pol V active site mutants will be described elsewhere (Haag and Pikaard, submitted). rdr1-1, rdr2-1, dcl2-1 and dcl3-1 were provided by J. Carrington, sgs2-1 (rdr6) and dcl4-1 were provided by H. Vaucheret, drd1-6 was provided by M. Matzke, met1-1 and ddm2-1 were provided by E. Richards, dcl234 (dcl2-5 dcl3-1 dcl4-2) and dcl1234 (dcl1-9 dcl2-5 dcl3-1 dcl4-2) were provided by T. Blevins, drm2-2 (SAIL_70_E12) was provided by E. Richards.

RNA analysis

RNA was isolated from 2-week old plants using a RNeasy Kit (Qiagen). 5′ RACE was performed using a GeneRacer Kit (Invitrogen) with two nested amplification steps; see Table S1 for primers. 5′ RACE products were gel-purified and cloned into TOPO-TA (Invitrogen). Tobacco Acid Pyrophosphatase (Invitrogen) or Terminator Exonuclease (Epicentre) treatments followed manufacturers’ instructions. Polyadenylated RNA was purified using a FastTrack MAG Kit (Invitrogen). For RT-PCR, 1μg of RNA digested with DNase (Invitrogen) was reverse-transcribed 30 min at 55°C using 60 units SuperScript III Reverse Transcriptase (Invitrogen), 1.5 units Platinum Taq (Invitrogen) and a gene-specific primer. After heat-inactivation of reverse transcriptase, the second primer was added and PCR was performed. Alternatively, the One Step RT-PCR Kit (Qiagen) was used. Table S1 shows primer pairs.

ChIP and RNA-IP

ChIP was performed by adapting existing protocols (Lawrence et al., 2004; Nelson et al., 2006), as was RNA-IP (Gilbert and Svejstrup, 2006; Martianov et al., 2007). Details are provided in the Supplemental material. All ChIP and RNA IP experiments were reproduced at least twice.

Real-time quantitative PCR

DNA was amplified using an Applied Biosystems model 7500 thermocycler with 0.5 units of Platinum Taq (Invitrogen), SYBR Green I (Invitrogen) and Internal Reference Dye (Sigma). Primer pairs are shown in Table S1. Results were analyzed using the comparative C_T method (Livak and Schmittgen, 2001) relative to input or undigested samples.

Antibodies

Anti-FLAG M2 mouse monoclonal and rabbit polyclonal antibodies were purchased from Sigma-Aldrich. Anti-Pol II (anti-NRPB2) was described previously (Onodera et al., 2005). Anti-H3K27me1 antibody #8835 (Peters et al., 2003) was provided by Thomas Jenuwein. Antibody against diacetyl-H3 (K9 and K14) was obtained from Upstate Biologicals (cat. #06599, lot #31994). Anti-H3K9me2 was obtained from Abcam (cat. #ab7312, lot #133588).