|Home | About | Journals | Submit | Contact Us | Français|
In Arabidopsis, RNA-dependent DNA methylation and transcriptional silencing involves three nuclear RNA polymerases that are biochemically undefined: the presumptive DNA-dependent RNA polymerases, Pol IV and Pol V and the putative RNA-dependent RNA polymerase, RDR2. Here, we demonstrate their RNA polymerase activities in vitro. Unlike Pol II, Pols IV and V require an RNA primer, are insensitive to alpha-amanitin and differ in their ability to displace the non-template DNA strand during transcription. Biogenesis of 24 nt small interfering RNAs (siRNAs), which guide cytosine methylation to corresponding sequences, requires both Pol IV and RDR2, which physically associate in vivo. Whereas Pol IV does not require RDR2 for activity, RDR2 is non-functional in the absence of associated Pol IV. These results suggest that the physical and mechanistic coupling of Pol IV and RDR2 results in the channeled synthesis of double-stranded precursors for 24 nt siRNA biogenesis.
Eukaryotes have three essential multisubunit DNA-dependent RNA polymerases, abbreviated as Pols I, II and III (Cramer et al., 2008; Werner and Grohmann, 2011). In plants, two additional multisubunit RNA polymerases, Pol IV and Pol V evolved as specialized forms of Pol II (Huang et al., 2009; Ream et al., 2009) and play important roles in development, transposon taming, transgene silencing and pathogen defense (Haag and Pikaard, 2011).
Pols IV and V are best understood with respect to their roles in RNA-directed DNA methylation (RdDM) in Arabidopsis thaliana (Herr et al., 2005; Kanno et al., 2005; Onodera et al., 2005; Pontes et al., 2006; Pontier et al., 2005)(Figure 1A). Genetic and cytological evidence suggests that Pol IV acts early in the pathway, upstream of RNA-DEPENDENT RNA POLYMERASE 2 (RDR2) (Pontes et al., 2006). RDR2 is thought to transcribe Pol IV transcripts to generate double-stranded RNAs (dsRNAs) that are then cleaved into 24 nt siRNAs by DICER-LIKE 3 (DCL3) (Xie et al., 2004), 3′ end-methylated by HUA-ENHANCER 1 (HEN1) (Li et al., 2005) and loaded into ARGONAUTE 4 (AGO4), or a related Argonaute protein (Havecker et al., 2010; Qi et al., 2006). Independent of 24 nt siRNA biogenesis, Pol V generates RNA transcripts to which AGO4-siRNA complexes bind (Wierzbicki et al., 2009), facilitating recruitment of the de novo DNA methyltransferase, DRM2, and other chromatin modifying activities that repress Pol I, II or III transcription (Haag and Pikaard, 2011; Law and Jacobsen, 2010; Zhang and Zhu, 2011).
Detection of Pol IV or Pol V polymerase activities has proven elusive using standard promoter-independent transcription assays or nuclear run-on assays (Erhard et al., 2009; Huang et al., 2009; Onodera et al., 2005). These negative results have suggested that Pols IV and V might require unconventional templates, or possibly lack RNA polymerase activity, consistent with the divergence, or absence, in Pols IV and V, of amino acids that are invariant in Pols I, II or III (Haag et al., 2009; Herr, 2005; Landick, 2009). However, Pols IV and V retain key amino acids of the magnesium-binding Metal A and Metal B sites that are invariant at the active sites of all multisubunit RNA polymerases (Haag et al., 2009; Herr, 2005; Landick, 2009). Mutagenesis of these sites abolishes Pol IV or Pol V functions in vivo, including siRNA biogenesis, de novo cytosine methylation and transposon silencing (Haag et al., 2009; Lahmy et al., 2009). Moreover, Pol V transcripts detectable in vivo are lost upon mutation of Pol V’s Metal A site (Wierzbicki et al., 2008).
Here, we demonstrate RNA-primed transcription of DNA templates by Pols IV and V and differences in Pol IV, Pol V and Pol II with respect to their sensitivities to the fungal toxin, alpha-amanitin and their abilities to transcribe RNA-RNA templates or displace non-template DNA during transcription. We find that RDR2 activity is Pol IV-dependent, suggesting that RNAs are channeled from Pol IV to RDR2 to generate dsRNAs for subsequent dicing.
We rescued an nrpd1-3 null mutant lacking the Pol IV largest subunit with a FLAG epitope-tagged NRPD1 transgene (NRPD1-FLAG), allowing Pol IV affinity purification using anti-FLAG resin. Trypsin digestion and LC-MS/MS mass spectrometry identified peptides of Pol IV’s twelve core subunits (Ream et al., 2009) as well as ten peptides corresponding to RDR2 (Figure 1B), confirming a recent report (Law et al., 2011).
As an independent test of Pol IV- RDR2 interaction, we rescued an rdr2-1 null mutant with a RDR2 transgene (see Figures S1A–C) that includes the RDR2 promoter, all exons and introns, and a C-terminal HA epitope tag. Following anti-HA immunoprecipitation (IP) and immunoblotting, RDR2-HA is readily detected using anti-RDR2 antisera (Figure 1C, lane 2, row 2), as are the catalytic subunits of Pol IV, NRPD1 and NRPD2 (Figure 1C, lane 2, rows 3 and 5). LC-MS/MS analysis of affinity-purified RDR2-HA identified nine of the twelve Pol IV subunits, including major (3a), and alternative (3b) forms of the third subunit (Tables S1 and S2). No Pol I, II, III or V-specific subunits were detected.
Consistent with the RDR2-HA IP and mass spectrometry results, RDR2 co-IPs with FLAG-tagged NRPD1 (Figure 1C, lane 3) but not with Pol V (NRPE1-FLAG, lane 4), Pol II (NRPB2-FLAG; lane 7), or Pols I or III (lane 2, row 8). NRPD1 does not co-IP with RNA-DEPENDENT RNA POLYMERASE 6 (Figure 1C, lane 6 and lane 3), involved in 21 nt siRNA biogenesis (Figure S1D), indicative of Pol IV’s specificity for RDR2. No association between RDR2 and DCL3 was detected by immunoblot (Figure 1C, lanes 2 and 5) or LC-MS/MS analyses.
To test if Pol IV and RDR2 might associate via RNA, we made use of Pol IV rendered catalytically inactive (Haag et al., 2009) by changing to alanines the three invariant aspartates of the NRPD1 Metal A site (Figure 1D). Whereas nrpd1-3 null mutants are rescued by a wild-type NRPD1-FLAG transgene, NRPD1 bearing the active site mutations (ASM) fails to restore siRNA biogenesis, RNA-directed DNA methylation or transposon silencing (Haag et al., 2009). For example, a soloLTR element silenced in wild-type cells (Figure 1E, lane 1), but derepressed in nrpd1-3 (Pol IV) or nrpe1–11 (Pol V) null mutants (lanes 3 and 4), is re-silenced by NRPD1-FLAG or NRPE1-FLAG transgenes (lanes 5 and 6), but not by NRPD1(ASM)-FLAG (active site mutant) or NRPE1(ASM)-FLAG transgenes (Figure 1E, lanes 7 and 8).
NRPD1-FLAG or NRPD1(ASM)-FLAG associate with equivalent amounts of NRPD2 (the Pol IV second subunit) and RDR2 (Figure 1F, compare lanes 4 and 5), suggesting that active site mutations do not disrupt Pol IV assembly or RDR2 association. Pol IV-RDR2 association is also unaffected by RNase A (Figure 1G, lane 2). Collectively, these results suggest that Pol IV-RDR2 interaction does not require Pol IV transcripts or other RNAs.
In the search for templates for Pols IV or V (e.g. see Fig. S2), we found that Pol IV, like Pol II, will transcribe a tripartite oligonucleotide template that mimics a transcription bubble (Figure 2A). Features of this template include an 8 bp RNA-DNA hybrid, single stranded DNA and RNA upstream of the hybrid region and double-stranded DNA downstream of the DNA-RNA hybrid (Figure 2B). Pols I and II transcribe such tripartite templates, extending the RNA in a DNA-templated manner (Kuhn et al., 2007; Lehmann et al., 2007).
Using the tripartite template, Arabidopsis Pol II and Pol IV-RDR2 complexes catalyze alpha 32P-CTP incorporation into RNA extension products that can be resolved on sequencing gels and visualized by autoradiography (Figure 2C). The RNA of the tripartite template is 16 nt; its DNA-templated extension can yield a full-length product of 32 nt. Consistent with previous studies using yeast Pol II, Arabidopsis Pol II catalyzes the synthesis of RNA products up to 32 nt (lane 5) and is inhibited by 5 μg/ml α-amanitin (lane 6). The Pol IV-RDR2 complex generates abundant 12–16 nt transcripts and longer transcripts up to 32 nt (lane 2). Pol IV-RDR2 transcripts are insensitive to α-amanitin (Figure 2C, lane 3), consistent with the divergence in Pols IV and V of the α-amanitin binding pocket of Pol II (Figure S3) and the expected α-amanitin insensitivity of RNA-dependent RNA polymerases, such as RDR2. Cloning and sequencing of RNA-primed extension products confirmed that all Pol II, IV and V transcripts are DNA-templated (Figure S4).
Catalytically crippled (ASM) and wild-type Pol IV both associate with RDR2 (Figure 1F). Affinity purified Pol IV(ASM)-RDR2 generates 12–16 nt RNA products (Figure 2C, lane 4) as efficiently as wild-type Pol IV-RDR2 (lanes 1 and 2) but most long transcription products are absent. Long RNAs dependent on the Pol IV active site are interpreted to be DNA-templated Pol IV transcripts. Transcripts unaffected by mutating the Pol IV active site are presumably generated by RDR2; these are mostly smaller than the 16 nt RNA oligonucleotide in the reactions (a 5′ end-labeled aliquot of this RNA is present in lane 8), consistent with RDR2 transcribing the 16 nt RNA. A transcript of ~26 nt generated by the Pol IV(ASM)-RDR2 complex (lane 4) is also RDR2-dependent based on subsequent experiments using Pol IV isolated from an rdr2 null mutant background (see below).
We next deconstructed the tripartite template, testing its component oligonucleotides as templates (Figure 2D). Pol IV-RDR2 or Pol II transcription reactions performed using a bipartite template, consisting of the 16 nt RNA hybridized to the 31 nt DNA template, yielded products similar to those obtained using the tripartite template (Figure 2D, lanes 6–8; compare to lanes 2–4), indicating that non-template DNA downstream of the DNA-RNA hybrid is dispensable. In fact, transcription was more robust without the need to displace the non-template DNA oligonucleotide, allowing more full-length transcription by Pols II and IV.
Using only the 31 nt DNA oligonucleotide as the template, a ladder of transcription products were generated by both Pol IV-RDR2 and Pol II (Figure 2D, lanes 9–12). Many of these products were less abundant using the Pol IV active site mutant form of the Pol IV-RDR2 complex, indicating that Pol IV (like Pol II) is able to transcribe single-stranded DNA to some extent. The 12–16 nt RNA products obtained using the tripartite or bipartite templates are absent in reactions containing only the 31 nt DNA template (lanes 9–12). Conversely, transcription reactions using the 16 nt RNA oligonucleotide alone support 12–16 nt RNA production in Pol IV-RDR2 and Pol IV(ASM)-RDR2 reactions (Figure 2D, lanes 14 and 15), but not in Pol II reactions (lane 16), consistent with these being RDR2 transcripts templated by the 16 nt RNA oligonucleotide.
Because Pol IV and RDR2 copurify, we disentangled them by introgressing NRPD1-FLAG or NRPD1(ASM)-FLAG transgenes into an rdr2-1 null mutant background and by introgressing an RDR2-HA transgene into a Pol IV null mutant (nrpd1-3). In the rdr2-1 mutant background, affinity purified NRPD1 or NRPD1(ASM) lack associated RDR2, as expected (Figure 3A, lanes 4 and 5). Likewise, RDR2-HA normally associates with Pol IV (Figure 3B lane 2), but not in an nrpd1-3 null mutant background (Figure 3B, lane 3).
As shown previously, Pol IV-RDR2 generates both long (>16 nt) and short (<16 nt) transcripts using the bipartite template (Figure 3C, lane 2), with most long transcripts dependent on the Pol IV active site (lane 3). Importantly, 12–16 nt transcripts are no longer produced in reactions utilizing Pol IV isolated from an rdr2 mutant (Figure 3C, lane 4), consistent with their synthesis by RDR2. Products of ~16 and 31 nt observed in anti-FLAG and anti-HA IP controls from non-transgenic plants (Figure 3C, lanes 1 and 6, respectively) are due to end-labeling activities that are neither Pol IV nor RDR2-dependent.
Pol IV-RDR2 complex(es) isolated upon IP of RDR2 or NRPD1 have similar activities (Figure 3C, compare lanes 7 and 2). However, RDR2 isolated from the Pol IV null mutant background (nrpd1-3) no longer synthesizes 12–16 nt transcripts (Figure 3C, lane 8). We conclude that RDR2 requires association with Pol IV, or a Pol IV-associated factor, for activity in vitro. In contrast, Pol IV activity is not dependent on RDR2 association.
We tested the activity of Pol V affinity purified from an nrpe1-11 null mutant rescued with wild-type or active site mutant (ASM) forms of FLAG-tagged NRPE1 (see Figure 1D). NRPE1-FLAG complements the nrpe1-11 mutant but NRPE1(ASM)-FLAG does not (Figure 1E, compare lanes 6 and 8) (Haag et al., 2009).
Like Pol IV, no significant Pol V activity was detectable using sheared genomic DNA, chromatin or ssDNA templates (Figure S2). Unlike Pol IV, no Pol V activity was detected using the tripartite template, (data not shown). However, using the bipartite template, which lacks a non-template DNA strand in need of displacement, Pol V generates transcription products that are similar to those of Pol II (Figure 4A compare lanes 6 and 7 to lanes 9 and 10). Pol V transcripts are abolished upon mutation of the Pol V active site (lane 8) but their synthesis is insensitive to alpha-amanitin (lanes 4, 7).
The ability of both Pols IV and V to transcribe bipartite DNA-RNA templates prompted us to test their ability to transcribe bipartite RNA-RNA templates (Figure 4B). Interestingly, Pol IV is able to generate transcripts up to ~27 nt in length (Figure 4B, lanes 3 and 4), but Pol V lacks significant activity using the all-RNA template.
Our results provide biochemical demonstrations of RNA polymerase activity for Pol IV, Pol V and RDR2. Pol IV and RDR2 physically associate and we find that RDR2 activity is dependent on this association, suggesting that Pol IV and RDR2 activities are coupled, thereby channeling RNA intermediates early in the 24 nt siRNA-directed DNA methylation pathway.\
Compared to Pol II, the transcriptional activity of Pol IV and Pol V is relatively weak. One possibility is that cofactors that increase Pol IV and Pol V activity do not copurify with the polymerases. Pols IV and V also have numerous amino acid substitutions or deletions at positions that are invariant, or highly conserved, in Pols I, II and III, which might compromise their activities. Compared to Pol IV, associated RDR2 activity is strong. Low-abundance Pol IV transcripts might be amplified significantly by RDR2 in vivo, particularly if RDR2 can use its own transcripts as templates. If so, Pol IV’s RNA polymerase activity may not need to be particularly robust.
Thus far, we have been unable to detect significant Pol IV or Pol V transcription in vitro in the absence of an RNA primer. A trivial explanation could be that Pols IV and V require unidentified cofactors that are not needed by Pols I, II or III to initiate transcription using sheared genomic DNA or other DNA templates. However, the mislocalization of Pols IV and V, but not Pol II, in nuclei treated with RNase A is consistent with an involvement of RNA as a template or primer (Pontes et al., 2006). Abortive transcripts or persistent RNA-DNA hybrids resulting from Pol II transcription, or small RNAs that invade duplex DNA to generate R loops are potential sources of DNA-RNA hybrid templates.
Pol V’s ability to carry out transcription using the bipartite oligonucleotide template, but not the tripartite template, suggests an inability to disrupt downstream dsDNA during transcription. The potential ramifications of this observation in vivo are intriguing. We’ve shown that DRD1, a putative chromatin remodeler in the SWI2/SNF2 protein family, and DMS3, a protein that shares homology with the hinge domain region of cohesin and condensin ATPases, enable Pol V transcription in vivo (Wierzbicki et al., 2008; Wierzbicki et al., 2009). These proteins interact with RDM1 (Law et al., 2010), a single-stranded DNA binding protein (Gao et al., 2010), suggesting that the complex may play a role in generating, and stabilizing, melted duplex DNA regions to provide Pol V with a single-stranded template. Further development of Pol V in vitro transcription assays should allow tests of this hypothesis.
A. thaliana mutants studied were nrpd1-3, nrpe1-11 and rdr2-1 (Onodera et al., 2005; Pontier et al., 2005; Xie et al., 2004). Transgenic lines were NRPD1-FLAG nrpd1-3, NRPE1-FLAG nrpe1-11, DCL3-FLAG dcl3-1, NRPB2-FLAG nrpb2-1, NRPD1(ASM)-FLAG nrpd1-3 and NRPE1(ASM)-FLAG nrpe1-11 (Haag et al., 2009; Pontes et al., 2006; Ream et al., 2009).
Full-length RDR2 sequences were PCR amplified, captured in pENTR-TOPO (Invitrogen) and recombined into the pEarleyGate301 plant transformation vector (Earley et al., 2006). Cloning details are provided in the supplemental information. A. thaliana transformation was by the floral dip method (Clough and Bent, 1998).
Affinity purified anti-NRPD1, anti-NRPE1, anti-NRP(D/E)2, anti-NRPA2/NRPB2/NRPC2 (anti-Pol I, II, III) and anti-RDR6 were described previously (Hoffer et al., 2011; Onodera et al., 2005; Ream et al., 2009). Anti-FLAG M2-HRP and anti-HA were purchased from Sigma (St. Louis, USA). Antibodies against bacterially expressed 6xHis-RDR2-C (amino acids 786–1133) and 6xHis-DCL3-N (amino acids 1–393) were raised in rabbits and affinity purified. Additional details are provided in the supplemental information.
Frozen leaf tissue (4.0 g) ground in liquid nitrogen using a mortar and pestle was homogenized in extraction buffer, subjected to centrifugation to pellet cell debris, and the supernatant incubated with 25 μL anti-FLAG-M2 or anti-HA resin (Sigma) for >2 hr on a rotating mixer. Resin was recovered and washed twice with extraction buffer supplemented with 0.5% NP-40. For immunoblot experiments, proteins were eluted from the resin by boiling 5 min in two bed volumes of 2x SDS sample buffer. Proteins resolved on 7.5% Tris-glycine SDS-PAGE gels were transferred to nitrocellulose or PVDF and probed with rabbit antibodies (see supplemental information for antibody dilutions) and anti-rabbit-HRP (Amersham) secondary antibody. Immunoblots were visualized using ECL or ECL Plus (GE Healthcare) chemiluminescent detection.
Pol IV or Pol V transcription reactions used the total IP fraction from 4.0 g leaf tissue and were performed according to (Kuhn et al., 2007). Polymerase-bound resin was mixed with transcription reaction buffer containing 2 mM each of ATP, 2 UTP and GTP, 0.08 mM unlabeled CTP, 0.2 miC/mL alpha-32P-CTP and 4 pmol template nucleic acid(s). As appropriate, α-amanitin was added 5 min prior to the reaction buffer. Labeled transcripts were resolved on 15% polyacrylamide, 7 M urea gels, transferred to Whatman 3MM filter paper and dried under vacuum prior to phosphorimaging or film exposure. Additional details are provided in the supplemental information.
To prepare tripartite and bipartite templates, oligonucleotides at a concentration of 10 μM each in T4 Polynucleotide Kinase buffer (New England Biolabs), 50 mM NaCl, were annealed by incubating for 2 min in a 95 °C water bath, then allowing the water bath to cool to room temperature.
J.R.H. and C.S.P. designed the study and wrote the paper. T.S.R. generated the RDR6-FLAG line and DCL3 and RDR6 antibodies. Pol IV samples for LC-MS/MS were prepared by T.S.R. and analyzed by C.D.N., A.D.N., and L.P.-T. Transcript sequences were determined by M.M. J.R.H. performed all other experiments. Portions of this research were supported by the NIH National Center for Research Resources (RR18522) and the W.R. Wiley Environmental Molecular Science Laboratory, a national scientific user facility sponsored by the U.S. Department of Energy, located at PNNL and operated by Battelle Memorial Institute under DOE contract DE-AC05-76RL01830. Pikaard lab research was supported by National Institutes of Health grant GM077590. C.S.P. is an Investigator of the Howard Hughes Medical Institute and the Gordon & Betty Moore Foundation. Opinions are those of the authors and do not necessarily reflect the views of our sponsors. The authors declare that no competing interests exist.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.