Multisubunit RNA Polymerases IV and V: Purveyors of Non-Coding RNA for Plant Gene Silencing

REVIEWS

Multisubunit RNA polymerases IV and V: purveyors of non-coding RNA for plant gene silencing

Jeremy R. Haag and Craig S. Pikaard Abstract | In all eukaryotes, nuclear DNA-dependent RNA polymerases I, II and III synthesize the myriad RNAs that are essential for life. Remarkably, plants have evolved two additional multisubunit RNA polymerases, RNA polymerases IV and V, which orchestrate non-coding RNA-mediated gene silencing processes affecting development, transposon taming, antiviral defence and allelic crosstalk. Biochemical details concerning the templates and products of RNA polymerases IV and V are lacking. However, their subunit compositions reveal that they evolved as specialized forms of RNA polymerase II, which provides the unique opportunity to study the functional diversification of a eukaryotic RNA polymerase family.

Short interspersed nuclear DNA-dependent RNA polymerases are essential enzymes Recent articles have discussed the evolution of Pol IV elements that read the information that is stored in DNA and and Pol V subunits and the divergence of their catalytic (SINES). Retrotransposable convert it into RNA. Unlike bacteria and archaea, the centres compared with those of other RNA polymer- elements in eukaryotic genomes of which encode only one multisubunit RNA ases6–8. In this Review, we focus on the biological roles genomes that are ancestrally related to tRNAs. polymerase, eukaryotes use three essential RNA polymer- of Pol IV and Pol V in gene silencing, their differences ases to decode their nuclear genomes1,2. RNA polymerase I in subunit composition compared with Pol II and the (Pol I) is the most specialized of the three, transcribing questions that remain concerning the templates and hundreds of copies of genes that are essentially iden transcripts of Pol IV and Pol V. tical and whose primary transcripts are processed into the 18S, 5.8S and 28S (25S in yeast) catalytic RNAs of Discovery of Pol IV and Pol V ribosomes. Pol II is the most versatile of the RNA poly- Given the expectation that there would be three, and only merases, transcribing thousands of protein-coding genes, three, nuclear multisubunit RNA polymerases in plants, long non-coding RNAs and small RNAs that modify or as in other eukaryotes, the Arabidopsis thaliana genome process messenger RNAs or ribosomal RNAs. Pol III is sequence contained a surprise. During the annota- also versatile, as it transcribes several classes of smaller tion process, C. S. P. noted that the A. thaliana genome RNAs (typically under 500 nucleotides), such as 5S ribo- possessed sequences for the largest and second-largest somal RNA, transfer RNAs, small regulatory RNAs and subunits of two potential RNA polymerases that were short interspersed nuclear elements (SINEs). distinct from those of Pol I, Pol II and Pol III9. As these The specialized functions of Pol I, Pol II and Pol III two subunits form the active site of RNA polymerases, must have been established in the last common ancestor this observation suggested the existence of atypical RNA of the Eukarya, given their ubiquity and conservation. polymerases. Additional phylogenetic analyses confirmed How they arose and diversified from a single archaeal that the atypical RNA polymerase subunit genes were progenitor enzyme1 is an intriguing — but intractable present only in plants, suggesting that the atypical RNA Department of Biology and 9 Department of Molecular — question owing to a lack of alternative evolutionary polymerases might carry out plant-specific functions . and Cellular Biochemistry, outcomes, such as groups of eukaryotes that are missing These functions began to come to light 5 years later, when Indiana University, one or more of the polymerases. Against this intellectual forward and reverse genetic screens defined two nuclear Bloomington, backdrop, considerable excitement was generated by the activities that were non-essential for viability but required Indiana 47405, USA. first hints, a decade ago, of the existence of novel plant- for RNA-mediated transcriptional silencing and hetero- Correspondence to C.S.P. 10–13 e-mail: specific RNA polymerases, followed by the elucidation of chromatin formation . These activities were initially [email protected] the functions of Pol IV and Pol V and the demonstration only defined by their largest subunits (FIG. 1) and were first doi:10.1038/nrm3152 of their Pol II‑like subunit composition3–5. called Pol IVa and Pol IVb but have since been renamed

/GVCN#DKPFKPIUKVG JGRVCFTGRGCVU 2QN++ 1$')'*'(0 %QPUGPUWU<637636 # $ % & ' ( ) * CC

5*')'*'&/ 2QN+8 &G%.NKMG # $ % & ' ( ) * CC

6$')'*'&9 9))9OQVKHU 2QN8 &G%.NKMG 35TKEJ # $ % & ' ( ) * CC

CCTGRGCVU '..1:*7(6$3$$:*6 '..16(7(6'$$$:*6 1..66(7(61*$7:*6 '..1,(7'6(3$$:*6 *..16(7(6*3$$:*$ '...6(7(3*3$*:*0 '..16(7(/*3$$0*1 '...6'7.6*3$$:*6 '.116(7(6'$$$:*6 1..76(,(6*$*$:*6 2QN+ 1$')'*'(0 # $ % & ' ( ) * CC

1$')'*'(0 2QN+++ # $ % & ' ( ) * CC CC Figure 1 | Domain structures of the largest subunit of the five Arabidopsis thaliana nuclear multisubunit RNA polymerases. The conserved domains A–H are present in the largest subunit of all multisubunit0CVWTG4GXKGYU RNA^ polymerases,/QNGEWNCT%GNN$KQNQI[ and they are important for organizing the active site or mediating interactions with other subunits. The G domains of the RNA polymerase IV (Pol IV) and Pol V largest subunits are diffusely shaded to reflect their substantial sequence divergence relative to the other paralogues. The invariant Asps are highlighted in orange within the Metal A binding sites of the catalytic centre. Heptad (7 amino acid (aa)) repeats in the carboxy‑terminal domain (CTD) of the Pol II largest subunit are denoted by open arrowheads. These repeats, which facilitate interactions with numerous regulatory activities, are absent in the CTDs of the largest subunits of Pol IV and Pol V. Instead, the CTDs of Pol IV and Pol V share a domain (the DeCL-like domain) of unknown function that is similar in sequence to a presumptive ribosomal RNA processing protein, DEFECTIVE CHLOROPLASTS AND LEAVES. The CTD of the Pol V largest subunit, which is larger than that of Pol IV, also includes ten 16 aa repeats (yellow region with closed arrowheads) within a WG/GW-rich region that interacts with ARGONAUTE 4 (AGO4), as well as a QS-rich domain of unknown function at the C terminus. All aa positions are numbered from the amino terminus.

Pol IV and Pol V14–16 to reflect their distinct functions and vast majority of loci18. Moreover, in Pol IV-null mutants, subunit compositions (BOX 1). other proteins of the RdDM pathway are mislocalized, suggesting that they all act downstream of Pol IV21. This RNA-directed DNA methylation observation suggests that proteins of the RdDM pathway Pol IV and Pol V are best understood with regard to require the expression of Pol IV, or the products of Pol IV their roles in RNA-directed DNA methylation (RdDM), transcription, for stability. a process by which 24-nucleotide small-interfering Genetic evidence suggests that Pol IV transcripts are RNAs (siRNAs) direct the de novo cytosine methylation copied into double-stranded RNAs (dsRNAs) by RNA- of complementary DNA sequences (FIG. 2). Thousands of DEPENDENT RNA POLYMERASE 2 (RDR2), one of six retrotransposons and endogenous repeats are silenced A. thaliana RNA-dependent RNA polymerases22,23 (FIG. 2, and controlled by RdDM, as are invading DNA viruses, step 2). Pol IV and RDR2 interact in vivo, as shown by transgenes and some protein-coding genes17–20. Pol IV mass spectrometry and co-immunoprecipitation analyses and Pol V have distinct roles at the beginning and ends (J.R.H., T. Ream and C.S.P., unpublished observation). A of the RdDM pathway, respectively, with Pol IV being putative chromatin remodeller, CLASSY1 (CLSY1; also required for siRNA biogenesis and Pol V transcripts known as CHR38), which is a SWI2/SNF2‑like helicase being required for siRNA targeting of RdDM-affected loci. domain protein, assists in Pol IV and RDR2‑dependent siRNA biogenesis24. The nuclear localization pattern of Initiating RdDM: Pol IV, RDR2 and siRNA biogenesis. CLSY1 resembles that of RDR2, and in clsy1 mutants Although Pol IV primary transcripts have proven elu- RDR2 is substantially mislocalized. Pol IV localization is Retrotransposons sive and remain undefined, Pol IV is thought to initiate also perturbed slightly in clsy1 mutants, suggesting that Transposons that replicate FIG. 2 through RNA intermediates the RdDM pathway ( , step 1) because it physically CLSY1 acts at the interface between Pol IV and RDR2 21 and can induce mutations by colocalizes with loci that undergo RdDM and it is (FIG. 2, step 2), perhaps facilitating transcription or the inserting near or within genes. essential for 24-nucleotide siRNA biogenesis from the transfer of RNA from Pol IV to RDR2 (REF. 24).

Box 1 | RNA polymerase subunit nomenclature RNA polymerase IV (Pol IV) and Pol V subunit nomenclature stems from naming systems that are used in other eukaryotes, such as Saccharomyces cerevisiae, in which Pol I, Pol II, and Pol III are designated RPA, RPB and RPC, respectively. In plants, the prefix N (for nuclear) is added to circumvent nomenclature conflicts with other genes, thus Pol I, Pol II, Pol III, Pol IV and Pol V subunits are denoted by the prefixes NRPA, NRPB, NRPC, NRPD and NRPE, respectively. In the nomenclature system we recently proposed15, which is based on the complete subunit compositions of Pol II, Pol IV and Pol V and unpublished subunit compositions we have determined for Arabidopsis thaliana Pol I and Pol III (T. Ream and C.S.P., unpublished observations), plant RNA polymerase subunits are numbered based on their similarity to the corresponding subunits of Pol II in yeast. Thus, NRPD1 and NRPE1 are the largest subunits of Pol IV and Pol V, respectively, and NRPE5 is the Pol V‑specific homologue of the fifth-largest subunit in yeast (the yeast subunits decrease in size with an increase in number). In cases where there is more than one variant of a particular subunit gene, these variants are designated with small letters; for instance, the two genes encoding potential NRPB9 subunits are designated NRPB9a and NRPB9b.

In the next step of the RdDM pathway (FIG. 2, Several proteins that were identified in genetic screens step 3), dsRNAs are diced into 24-nucleotide duplexes facilitate Pol V transcription (FIG. 2, step 6): DRD1 by DICER-LIKE 3 (DCL3), one of four DICER endo is a SWI2/SNF2‑related putative chromatin remod- nucleases in A. thaliana23,25. In dcl3 mutants, some dicing eler35; DMS3 shares similarity with the hinge domains of precursors still occurs, by virtue of DCL2 and DCL4 (dimerization regions) of structural maintenance of (REFS 22,26), but the 21- and 22-nucleotide siRNAs that chromosomes (SMC) proteins, such as cohesins and they produce are relatively ineffective for RdDM23. condensins36,37; and RDM1 (REQUIRED FOR DNA Following dicing, the RNA methylase HUA- METHYLATION 1) is a single-stranded DNA binding ENHANCER 1 (HEN1) adds a methyl group to the protein with a strong preference for methylated DNA38. 2′ hydroxyl groups of siRNAs at their 3′ ends27, which Initial studies showed a loss of Pol V transcripts and a helps to stabilize these small RNAs28 (FIG. 2, step 4). loss of Pol V association at target loci in drd1 and dms3 Consequently, siRNA levels and RdDM are reduced in mutants14,39. DRD1, DMS3 and RDM1 were subsequently hen1 mutants11,23. Following 3′ end methylation, one shown to form a multifunctional complex, known as the strand of the 24-nucleotide siRNA duplex is loaded into DDR complex40. The DDR complex elutes with Pol V ARGONAUTE 4 (AGO4), or its close relatives AGO6 during gel filtration chromatography, and immuno and AGO9, forming an RNA-induced silencing com- precipitated DDR fractions that were analyzed by mass plex (RISC)29,30 (FIG. 2, step 5). The strand with the more spectrometry include subunits of Pol V40. Collectively, weakly hydrogen bonded 5′ end is preferentially selected this evidence suggests that the DDR complex might as the siRNA31. The size of the diced product, and the recruit Pol V to target loci and/or facilitate Pol V nucleotide at its 5′ end, dictate its loading into compat- transcription through chromatin remodelling. ible AGO proteins32,33, depending on their availability in DMS4 (also known as RDM4), which is a homologue a given cell type or time in development29. of the yeast Pol II‑associated protein Iwr1, is a potential Collectively, Pol IV, RDR2, DCL3 and HEN1 are cru- regulator of polymerase abundance or activity41,42 (FIG. 2, cial for siRNA biogenesis and stability, arming AGO4 step 6). Iwr1 binds Pol II in the cytoplasm and directs with the small RNA guidance system that is needed to Pol II import into the nucleus, possibly acting as a sensor seek out homologous target sites in the genome. In the for the fully assembled polymerase43. Consistent with the absence of compatible siRNAs AGO4 is unstable; thus, yeast data, DMS4 interacts with both Pol II and Pol V in Pol IV (nrpd1), rdr2 or dcl3 mutants, AGO4 levels in vivo41. Yeast two-hybrid analyses indicate that DMS4 are substantially reduced34. Whether AGO4 is actively can interact with the Pol IV and Pol V largest subunits’ targeted for degradation in the absence of an associated carboxy‑terminal domains (CTDs), within their DeCL- siRNA is unknown. like domains42 (FIGS 1,2,3) (the DeCl-like domain has homology to the DEFECTIVE CHLOROPLAST AND Generating the target for RdDM: the role of Pol V. LEAVES protein, which is needed for chloroplast rRNA Unlike Pol IV, Pol V is not required for 24‑nucleotide biogenesis44). However, NRPB1, which forms the largest siRNA biogenesis at most loci; however, it is critical for subunit of Pol II, lacks this domain (FIG. 1), indicating that RdDM12,13,17. So, what does Pol V do? A breakthrough the DeCL-like domain cannot be the sole determinant of in our understanding came from the identification of DMS4–polymerase interactions. Indeed, recent evidence Pol V‑dependent non-coding RNAs produced at multi- shows that Iwr1 binds to Pol II in the active site cleft that ple loci that are subject to RdDM14 (FIG. 2, step 6). These is formed by the two largest subunits43. Based on the yeast Pol V‑dependent RNAs span the promoter regions of studies, it is likely that DMS4 function in plants has co- silenced loci, are dependent on the Pol V active site and evolved with the Pol II, Pol IV and Pol V family in order can be chemically crosslinked to Pol V, indicating that to facilitate the nuclear import of all three enzymes. they are direct Pol V transcripts6,14. Importantly, Pol V transcripts are generated even if 24‑nucleotide siRNA RISC recruitment to Pol V transcribed target sites. biogenesis is abolished, indicating that Pol IV and Pol V AGO4 can be crosslinked to Pol V transcripts, pre- carry out their functions independently, in parallel sumably because of base-pairing of siRNAs with the pathways that converge in order for RdDM to occur14. Pol V transcripts39 (FIG. 2, step 7). AGO4 also interacts

2QN8 4GOQXCNQHCEVKXGJKUVQPG VTCPUETKRV ′ OCTMU HQTGZCORNG +&0 ! JKUVQPGFGCEGV[NCVKQPCPF *-FGOGVJ[NCVKQP &/5 #)1 4&/ &4/ 4&/ -6( 2QN+8 &4& &/5 2QN8 %6& &&4EQORNGZ 'UVCDNKUJOGPVQHTGRTGUUKXG UU40# JKUVQPGOCTMU HQTGZCORNG *-OGVJ[NCVKQPCPF %.5; 4&4 *-OGVJ[NCVKQP UK40# /GVJ[NCVGF FWRNGZ UK40#FWRNGZ FU40# #)1s4+5% &%. *'0 #)1 /GVJ[NCVKQP Figure 2 | Model for the RNA-directed DNA methylation pathway in Arabidopsis thaliana. RNA polymerase IV (Pol IV) initiates the RNA-directed DNA methylation (RdDM) pathway (step 1), generating0CVWTG4G transcriptsXKGYU^/QNGEWNCT%GNN$KQNQI[ that are then copied into double-stranded RNA (dsRNA) by RNA-DEPENDENT RNA POLYMERASE 2 (RDR2) (step 2). The putative chromatin remodeller and/or helicase CLASSY 1 (CLSY1) assists in one or more of these steps. DICER-LIKE 3 (DCL3) cleaves the dsRNA into 24-nucleotide small interfering RNA (siRNA) duplexes (step 3) that are methylated at their 3′ ends by HUA-ENHANCER 1 (HEN1) (step 4). A single strand of the siRNA duplex associates with ARGONAUTE 4 (AGO4) to form an RNA-induced silencing complex (RISC)–AGO4 complex (step 5). Independently of siRNA biogenesis, Pol V transcription is assisted by the DDR complex (DRD1 (DEFECTIVE IN RNA-DIRECTED DNA METHYLATION 1), DMS3 (DEFECTIVE IN MERISTEM SILENCING 3) and RDM1 (REQUIRED FOR DNA METHYLATION 1)) and DMS4 (step 6). AGO4 binds Pol V transcripts through base-pairing with the siRNA and is stabilized by AGO4 interaction with the NRPE1 (the largest subunit of Pol V) carboxy-terminal domain (CTD) and KTF1 (KOW DOMAIN-CONTAINING TRANSCRIPTION FACTOR 1), which also binds RNA (step 7). IDN2 may stabilize Pol V transcript–siRNA pairing (step 8). The RDM1 protein of the DDR complex binds AGO4 and the de novo cytosine methyltransferase DOMAINS REARRANGED METHYLTRANSFERASE 2 (DRM2), bringing them to Pol V transcribed regions and resulting in DNA methylation (step 9). Histone modifications resulting from the RdDM pathway include the removal of activating acetylation and methylation marks (the deacetylation of multiple Lys of several core histone proteins and the demethylation of histone H3 Lys 4 (H3K4)) and the establishment of alternative, repressive histone methylation marks (such as the methylation of H3K9 and H3K27), thereby facilitating transcriptional silencing (step 10). The dashed arrow indicates that Pol IV and RDR2 physically interact in vivo, as determined by mass spectrometry, co-immunoprecipitation and enzymatic assays (T. Ream, J.R.H. and C.S.P., unpublished observations). ssRNA, single-stranded RNA.

with a WG/GW-rich‌ domain that is present in the and secondary siRNA biogenesis; this might explain the CTD‑terminal domain of the largest subunit of Pol V34,45 reduced levels of siRNAs that are detected at many loci (FIGS 1,3), indicating that a physical interaction between when Pol V is mutated. It is possible that RNA cleavage Pol V and/or its transcripts with AGO4 can occur. AGO4 and siRNA amplification is not needed at the loci that also associates with chromatin at Pol V transcribed loci are not affected when AGO4 RNase activity is abolished. and this association is lost if Pol V transcription is abol- Alternatively, AGO6 or AGO9 may substitute for AGO4 ished through mutation of its active site39. Collectively, at such loci (REFS 29,48). these observations suggest that AGO4–RISC complexes At least two proteins are thought to assist in RISC are recruited to target loci through binding to Pol V tran- recruitment to target sites. One is IDN2 (also known scripts and the Pol V largest subunit, thereby bringing the as RDM12), which is a protein that was identified in RISC complex into the proximity of the chromatin to be two independent genetic screens as a component of modified39 (FIG. 2, step 7). the RdDM pathway37,49. IDN2 binds dsRNAs that have Once recruited to sites of Pol V transcription, AGO4 5′ overhangs37, but siRNA levels in idn2 mutants are unaf- may have a role in siRNA amplification as well as in fected, suggesting a role for IDN2 downstream of siRNA the recruitment of chromatin modifiers. AGO proteins biogenesis, possibly in the stabilization of siRNAs that are modular, including MID and PAZ domains that are are base-paired with Pol V transcripts50 (FIG. 2, step 8). involved in small RNA binding and PIWI domains that A second protein that is thought to assist AGO4– account for the ‘slicing’ of long RNAs that are base-paired RISC targeting is KOW DOMAIN-CONTAINING with AGO-associated siRNA46,47. An Asp-Asp-His triad TRANSCRIPTION FACTOR 1 (KTF1; also known as is key to slicer activity and is present in AGO4 (REF. 30). RDM3 and SPT5‑LIKE)16,51,52. KTF1 is similar to the Interestingly, mutagenesis of the AGO4 catalytic triad yeast transcription elongation factor Spt5, and was found has differential effects on siRNA accumulation and DNA by mass spectrometry to be associated with immuno methylation at different loci30. One possibility is that precipitated cauliflower Pol V16 (FIG. 3). However, loss AGO4‑mediated slicing of Pol V transcripts generates of KTF1 does not cause a decrease in Pol V transcript new substrates for RDR2, facilitating dsRNA production levels, as would be expected if KTF1 were a positive

2QN++ de novo methylation is DOMAINS REARRANGED METHYLTRANSFERASE 2 (DRM2) (FIG. 2, step 9). A &4/ related protein, DRM1, is expressed at very low levels #)1 and is thought to have a minor role50. It is not yet clear how DRM2 is recruited to its sites of action, but the DDR complex may be key, based on the finding that the RDM1 subunit of the DDR complex interacts with 4&/ both AGO4 and DRM2 and also binds methylated single- stranded DNA38. As the DDR complex also facilitates #)1 Pol V transcription, this suggests that DDR might inte- grate the activities of Pol V, AGO4 and DRM2 in order to bring about de novo cytosine methylation at Pol V transcribed loci. &/5 Chromatin modifications associated with RdDM. In addition to cytosine methylation, several chromatin mod- 2QN8 ifying enzymes are known to have roles in silencing the loci that are subjected to RdDM50 (FIG. 2, step 10). Among these are histone deacetylase 6 (HDAC6)55,56, a broad- &4& specificity enzyme that can remove the acetyl groups that are added by at least seven different A. thaliana histone acetyltransferases in vitro57. UBIQUITIN PROTEASE 26 (UBP26; also known as SUP32) is required for histone -6( &/5 H2B deubiquitylation, which, in turn, is required for the establishment of repressive histone H3 Lys 9 (H3K9) +PXKXQEQ+2 dimethylation58. The Jumonji C domain-containing +PXKVTQEQ+2 #)1 /CUUURGEVTQOGVT[ protein JMJ14 demethylates H3K4, which is a modifica- ;GCUVVYQJ[DTKF tion that is typical of actively transcribed loci, thereby Figure 3 | Interactions among proteins of the RNA-directed DNA methylation paving the way for repressive histone modifications to be 59–61 pathway. The use of several experimental techniques0CVWT hasG4G revealedXKGYU^/QNGEWNCT%GNN$KQNQI that many of the [ added . Collectively, the removal of active chromatin proteins in the RNA-directed DNA methylation (RdDM) pathway interact with each other, marks (histone acetylation and H3K4 trimethylation) thus defining complexes or means of coordination among key activities, particularly those and the establishment of repressive chromatin marks associated with RNA polymerase V (Pol V). The key indicates the experimental method (H3K9 methylation and H3K27 methylation) contrib- that was used to identify the interaction, and the arrowheads point from the bait or query utes to transcriptional silencing at sites that are subject protein towards the proteins that were co-purified with it. Lines with arrowheads at both to Pol IV- and Pol V‑mediated RdDM. ends indicate protein–protein interactions that were confirmed using each partner as the bait protein. The line with no arrowhead indicates an interaction that is supported by A role for Pol II in RdDM. In the fission yeast Schizo multiple lines of evidence but has not yet been demonstrated directly using either protein as bait. The interaction network used to draw the figure was generated using Osprey, saccharomyces pombe, Pol II carries out the functions version 1.2. AGO, ARGONAUTE; co-IP, co-immunoprecipitation; DMS3, DEFECTIVE IN that are attributed to Pol IV and Pol V, generating MERISTEM SILENCING 3; DRD1, DEFECTIVE IN RNA-DIRECTED DNA METHYLATION 1; both siRNA precursors and scaffold transcripts for DRM2, DOMAINS REARRANGED METHYLTRANSFERASE 2; KTF1, KOW DOMAIN- siRNA binding at loci that are subject to RNA-directed CONTAINING TRANSCRIPTION FACTOR 1; RDM1, REQUIRED FOR DNA METHYLATION 1. silencing62. In plants, Pol IV and Pol V have apparently taken over most of the RNA silencing duties of Pol II. However, recent findings indicate that Pol II still has a regulator of Pol V transcription. Instead, Pol V tran- role in RdDM in plants. Specifically, a mutation in the script abundance increases slightly in ktf1 mutants, as it gene encoding NRPB2 (the Pol II second-largest sub does in ago4 mutants39,52. Moreover, KTF1 binds RNA unit) that causes developmental defects also causes subtle in vitro52 and has a WG/GW-rich domain, which medi- RdDM defects, including decreased siRNA accumulation ates its physical interaction with AGO4, similarly to the and decreased silencing at some loci63. Pol II requires WG/GW-rich region in the CTD of the largest subunit DRD1 for transcription at some of these loci, interacts of Pol V51,52. Collectively, these observations suggest that with AGO4, DMS4 and RDM1 and influences Pol IV and KTF1 assists in AGO4 recruitment to Pol V transcripts, Pol V recruitment38,41,63 (FIG. 3), indicating that Pol II can thereby facilitating AGO4 slicing of Pol V transcripts and effect multiple steps of the RdDM process in ways that AGO4‑mediated recruitment of the silencing machinery. are not yet understood.

De novo DNA methyltransferase recruitment. siRNA- Pol IV and RdDM proteins in paramutation directed cytosine methylation occurs in all sequence When brought together in the same nucleus, so-called contexts (CG, CHG and CHH, where H represents a ‘paramutagenic’ silent alleles can transform active, nucleotide other than G) and accounts for approximately paramutable alleles into the silent, paramutagenic state, 30% of all DNA methylation in A. thaliana53,54. The cyto- which is stable and heritable through meiosis. This sine methyltransferase that is primarily responsible for remarkable phenomenon of paramutation occurs in

plants and mammals and has been studied extensively to shoot, as well as from shoot to root75. It has been in maize64,65. Genetic screens have shown that the pro- found, through the analysis by high-throughput small teins that are required for maize paramutation include RNA sequencing of siRNA spread from shoot to root, orthologues or paralogues of A. thaliana proteins that that NRPD1 and DICER function (DCL2, DCL3 and are involved in RdDM. For instance, REQUIRED TO DCL4) are required for silencing signal production76. MAINTAIN REPRESSION 6 (RMR6) is the maize The mobile 24‑nucleotide siRNAs requiring NRPD1 orthologue of NRPD1 (the largest subunit of A. thali and DICER activity can direct long-distance RdDM ana Pol IV66). MEDIATOR OF PARAMUTATION 1 at transgenic or endogenous targets75,76. These results (MOP1) is the maize orthologue of A. thaliana RDR2 run counter to an earlier report demonstrating that (REF. 67). RMR1 is a maize SWI2/SNF2‑like protein that Pol IV and DCL2, DCL3 and DCL4 were dispensable is related to A. thaliana CLSY1 and DRD1 (REF. 68), and for generating the silencing signals in the root that ulti- MOP2 (also known as RMR7) is one of three maize mately cause green fluorescent protein (GFP) reporter homologues of NRPD2 (also known as NRPE2), which silencing in the shoot. However, Pol IV and DICER is the second-largest subunit of both A. thaliana Pol IV activities were required for perception or amplification and Pol V69,70. Whether Pol V is also involved in para- of the silencing signal or signals in shoots77. Moreover, mutation is not yet clear, but the evidence clearly points siRNAs that were detected in shoot tissue, where GFP to RNA as the conduit for inter-allelic communication silencing occurred, neatly corresponded to sequences 3′ and the transcriptional silencing of paramutable alleles. to the trigger sequences that were expressed in the root and not to the trigger sequences themselves77. The dif- Spreading of RNA silencing ferent conclusions obtained from different studies might RNA silencing signals can spread locally within a tissue be reconciled by the possibility that long RNAs as well or organ as well as move long-distance between organs, mobile siRNAs have roles in long-distance silencing, and presumably as an antiviral defence mechanism. Short- that target genes in different contexts might be sensitive range spreading occurs by signals moving through inter- to different classes of mobile RNAs. cellular connections (plasmodesmata) over a range of 10–15 cells, whereas long-range spreading involves sig- Pol V and heterochromatin organization nals that are transported in the phloem of the vascular Centromeric repeats and other heterochromatic system71. Genetic screens to identify mobile silencing sig- sequences coalesce into chromocenters, which are inten- nalling components have made clever use of transgenes sively stained by DNA-binding dyes. An early observa- expressing inverted repeats of sequences of a gene whose tion was that chromocenter interactions are perturbed in knockdown produces a visual phenotype. When tran- nrpd2 mutants11, which lack both Pol IV and Pol V activ- scribed, the inverted repeats form dsRNAs that are diced ity. Subsequent studies of nrpd1 and nrpe1 mutants show into siRNAs, triggering the silencing of the homologous that loss of Pol V, but not Pol IV, causes chromocenter ‘reporter’ gene that is responsible for the phenotype being disruption78. Mutations in DRD1, which assists Pol V monitored. By expressing the inverted repeat in cells of transcription as part of the DDR complex, or in MET1, leaf veins (specifically, phloem companion cells), silencing the major maintenance cytosine methyltransferase, simi- of the endogenous reporter gene in adjacent cells of the larly cause chromocenter disruption. This chromocenter leaf can be monitored24,72,73. Both 21- and 24‑nucleotide disruption is coincident with the derepression of peri- siRNAs are produced during the cell–cell silencing pro- centromeric transcription units. By contrast, rdr2, dcl3, cess and Pol IV, RDR2 (REFS 24,73) and CLSY1 (REF. 24) are ago4 or drm2 mutations have no effect on chromocenter all involved. However, these three genes are dispensable distruption or pericentromeric transcription78. Pol V is for the accumulation of the predominant 21-nucleotide also required for the condensation of a 5S rRNA gene Inverted repeats siRNAs from the inverted repeat triggers, suggesting that locus on chromosome 4, independent of the RdDM Nucleotide sequences that 24‑nucleotide siRNAs arise as secondary siRNAs during pathway79. Collectively, these observations suggest that contain regions of self-comple- mentarity, such that they are spreading. Delivery of 21-nucleotide and longer RNA Pol V transcripts might have roles as structural compo- capable of folding back on duplexes, but not single-stranded RNA, directly into cells nents of heterochromatin or as components of chromatin themselves to generate by particle bombardment is sufficient to induce silencing modification pathways that are independent of Pol IV or double-stranded RNA that extends several cells beyond the bombardment site74. 24‑nucleotide siRNAs. hairpin structures. Twenty four nucleotide siRNA duplexes are also capable 74 Companion cells of short-distance spreading and transgene silencing , Pol IV and Pol V in development Specialized parenchyma cells suggesting that all siRNAs are mobile to some extent. The transition from vegetative growth to flowering in that are located in the phloem The silencing that is associated with short-range siRNA plants involves the reprogramming of stem cells in the of flowering plants and carry mobility is presumably due to a combination of post- shoot and is influenced by endogenous and environ- out the loading and unloading 80 of molecules into associated transcriptional silencing (mRNA slicing or inhibition of mental signals, including day length and temperature . phloem sieve-tubes. translation) and RdDM. A. thaliana RdDM mutants are only slightly delayed Long-distance spreading of RNA silencing signals in flowering under long-day, short-night conditions, Chromocenters has been studied by grafting roots and shoots of plants but the delay in flowering is exacerbated when plants Sites of heterochromatin from different genetic or transgenic backgrounds and are grown under short-day and long-night condi- association in the nucleus, 12,15,81–83 typically involving assessing whether silencing signals can traverse the graft tions . FLOWERING LOCUS C (FLC), a repressor pericentromeric repeats junction. Deep sequencing of small RNAs suggests that of flowering, is regulated during development by Pol IV-, and other repeats. all siRNA size classes are capable of moving from root RDR2-, DCL3- and AGO4-dependent chromatin

Regulating convergently transcribed genes &0#GPVT[ The A. thaliana genome, like that of other eukaryotes, has hundreds of convergently transcribed gene pairs 88 with the potential to generate dsRNAs and siRNAs . The regulatory potential of this arrangement has been 89 demonstrated for the salt stress response , bacterial &0#GZKV 90 91 defence and seed development . In the cases of the gene pairs that are involved in the salt stress and bacte- 40#GZKV rial defence responses, one gene is constitutively active and the antisense transcript is induced as a result of 5WDWPKVUQH2QN++2QN+8CPF2QN8VJCVCTGGPEQFGFD[VJGUCOGIGPG the stress. The resulting siRNAs post-transcriptionally 5WDWPKVUQH2QN++2QN+8CPF2QN8VJCVCTGWPKSWGVQGCEJGP\[OG 5WDWPKVUEQOOQPVQ2QN++CPF2QN+8DWVWPKSWGKP2QN8 silence the constitutive locus, facilitating an adaptive 89,90 5WDWPKVUEQOOQPVQ2QN+8CPF2QN8DWVWPKSWGKP2QN++ response . In the case of the gene pair that affects 5WDWPKVUEQOOQPVQ2QN++CPF2QN8DWVPQVQDUGTXGFKP2QN+8 fertilization and seed development, both transcripts are Figure 4 | Summary of Arabidopsis thaliana Pol II, Pol IV and Pol V subunit sperm cell-specific, giving rise to siRNAs that inhibit the compositions. The cartoon depicts the relative positions0CVWTG4G ofXKGYU RNA^ polymerase/QNGEWNCT%GNN$KQNQI[ subunits, expression of an E3 ubiquitin ligase that interferes with based on yeast RNA polymerase II (Pol II) and the relationships between Arabidopsis proper fertilization and seed development91. Genetic thaliana Pol II, Pol IV and Pol V in terms of their subunit compositions. RNA polymerase evidence indicates that Pol IV is required in each of the subunits are identified by their numbers and are colour-coded as indicated. Subunits three examples above, with Pol V involvement being three and nine are shown in two colours to represent their alternative forms. variable, but precisely how Pol IV (or Pol V) contributes to these phenomena is unclear.

modifications83,84. FLOWERING WAGENINGEN Pol II, Pol IV and Pol V: a diversified family (FWA) is another repressor of flowering in A. thaliana. The subunit compositions for A. thaliana Pol II, Pol IV The FWA promoter region has tandem repeats that are and Pol V, which have been defined by mass spectro targeted by siRNAs and RdDM. When wild-type plants metry15, reveal their striking similarities, with half of are transformed with a FWA transgene, RdDM compo- their 12 subunits encoded by the same genes (FIG. 4; nents are required for its de novo methylation and silenc- BOX 2). The remaining subunits of Pol IV and/or Pol V ing. Mutants of the RdDM pathway, including mutants are encoded by paralogues of Pol II subunit genes that of nrpd1, rdr2, dcl3, ago4, nrpe1 and drm2, release FWA underwent duplication at different times in plant evolu- transgene silencing and cause late flowering12,81,85. tion8,15. Multiple subunits of cauliflower Pol V have also In maize, rmr6, mop2 and mop1 mutants, which are been identified by mass spectrometry16, the gene encod- impaired for paramutation, also display developmental ing the fourth subunit of Pol IV and Pol V was identi- defects, including aberrant leaf development and polar- fied in a genetic screen56 and a reverse genetics approach ity, late flowering and problems with sex determina- identified the fifth subunit of Pol V92. All of these studies tion66,67. Collectively, the observations indicate that Pol support the conclusion that Pol IV and Pol V evolved as IV, Pol V and the RdDM pathway have important roles specialized forms of Pol II. in plant development in both A. thaliana and maize, pre- Many of the intron–exon junctions of the Pol IV sumably as a result of RdDM or other RNA-mediated and Pol V largest subunit genes are identical to those in phenomena. their Pol II paralogue, but not the corresponding Pol I or Pol III paralogues93. Based on phylogenetic analyses of Imprinted Pol IV‑dependent siRNA loci the largest subunits, Luo and Hall93 proposed that Pol IV Pol IV expression and siRNA abundance peak in evolved first through duplication of the Pol II largest sub- A. thaliana flowers before fertilization, with more than unit and Pol V evolved later, through duplication of the 100,000 Pol IV‑dependent siRNAs (p4‑siRNAs) gener Pol IV largest subunit. This hypothesis is supported and ated specifically from maternal chromosomes86. The extended by the recent determinations of the complete basis for the female gametophyte-specific expression subunit compositions of Pol II, Pol IV and Pol V, which of Pol IV‑derived siRNA during early embryogenesis is show that Pol IV is intermediate in subunit composition unclear, as siRNAs from both parents are later expressed between Pol II and Pol V15. in vegetative tissues of mature plants86. One hypothesis The second largest subunits of Pols IV and Pol V is that the maternal-specific expression of p4‑siRNAs are encoded by the same gene, NRPD2 (also known as may have a role in distinguishing self from non-self, or NRPE2), which is distinct from the Pol II NRPB2 gene. in modification of the incoming paternal genome. An In A. thaliana, a gene closely related to NRPD2 was intriguing parallel is that in pollen, transposable ele- duplicated after the divergence of dicots and monocots, ments in the vegetative nucleus are reactivated owing but it is a non-functional pseudogene. Interestingly, as to a loss of repressive cytosine methylation, resulting mentioned previously, there are three NRPD2‑like genes in the production of transposon-specific siRNAs that in maize, underscoring the fact that Pol IV and Pol V are thought to programme siRNA-mediated epigenetic are still evolving in different plant lineages8. In fact, the modifications in the sperm cells prior to fertilization87. Pol IV and Pol V largest subunits, NRPD1 and NRPE1, How these waves of maternal or paternal mobile siRNAs have estimated amino acid substitution rates that are instruct the gametes or early embryo is not yet clear. 20‑fold greater than the Pol II largest subunit, and

Box 2 | Functions of Pol II subunits and implications for Pol IV and Pol V NRPD4 reveals that the protein does not always colo- calize with Pol IV and Pol V catalytic subunits, suggest- The two largest subunits, subunits 1 and 2, form the active site, DNA entry channel and ing that the fourth–seventh subunit subcomplexes of RNA exit channel. The RNA polymerase II (Pol II), Pol IV and Pol V largest subunits have Pol IV and/or Pol V may also be able to dissociate from distinct carboxy‑terminal domains (CTDs) (FIG. 1). The CTD of the Pol II largest subunit the polymerase holoenzyme, as in yeast, possibly having mediates its interactions with numerous regulatory proteins, and the long CTD of the 56 Pol V largest subunit is likely to serve as an interaction site for multiple regulatory roles as a chaperone of Pol IV and/or Pol V transcripts . proteins, including ARGONAUTE 4 (AGO4). In yeast, the Rpb3, Rpb10, Rpb11 and The subunits that are unique to Pol II, Pol IV and Rpb12 subunits help to assemble and stabilize Pol II. Arabidopsis thaliana Pol II, Pol IV Pol V must be responsible for the unique functions of and Pol V make use of the same genes for these four subunits, but Pol V can use an the three polymerases. Future studies aimed at uncover alternative form of the third subunit (NRPE3b) that is not appreciably used by Pol II or ing the functions of these subunits are likely to yield Pol IV. Whether this results in a Pol V isoform with unique properties is unclear. important insights. Yeast Rpb4, Rpb5, Rpb6, Rpb7, Rpb8 and Rpb9 influence the functionality of Pol II. Rpb4 and Rpb7 are positioned near the RNA exit pore and interact to form a dissociable Pol IV and Pol V as enzymes subcomplex in yeast. Rpb7 has an RNA binding domain and mutations in Rpb7 Pol IV and Pol V enzymatic activity has not yet been disrupt small interfering RNA (siRNA)-dependent heterochromatin formation in demonstrated in vitro, despite attempts by several Schizosaccharomyces pombe. A. thaliana Pol II, Pol IV and Pol V each have unique 16,66,11 fourth–seventh subunit subcomplexes, which are likely to confer distinctive properties groups . Many of the amino acids that are invariant on the three enzymes15. in the catalytic subunits of Pol I, Pol II and Pol III are In yeast, Rpb5 and Rpb9 are positioned near where DNA enters the polymerase, different in Pol IV and Pol V. Interestingly, the sequences with Rpb5 contacting the DNA2. Rpb9 is involved in transcription start site selection, of the Pol IV and Pol V catalytic subunits that have processivity and transcript cleavage, which is mediated by the elongation factor TFIIS. diverged most are in the vicinity of their active sites6,7. The fifth subunits of A. thaliana Pol I, Pol II, Pol III and Pol IV are encoded by the same Nonetheless, the core sequences of the Metal A and gene, which is the yeast Rpb5 orthologue94,95. However, Pol V uses a unique fifth subunit, Metal B sites (which coordinate two magnesium ions NRPE5, that may impart unique template recognition properties on the enzyme15,16,92. that are essential for positioning the incoming nucleo- A. thaliana has two highly similar Rpb9‑like genes. It is unclear whether they are tide triphosphate relative to the 3′ end of the growing redundant with respect to Pol II, Pol IV and Pol V15. RNA chain within the active site) are retained in Pol IV The RNA exit pore separates Rpb6 and Rpb8 in yeast Pol II2. A. thaliana Pol II, Pol IV and Pol V use the same genes for these subunits. and Pol V, as in all other multisubunit RNA polymerases. Clustered site-directed mutagenesis of these essential amino acids abolishes all known biological functions of Pol IV and Pol V6, as do single mutations92. Moreover, NRPD2 has a substitution rate that is ten times greater Pol V‑dependent transcripts that are detected in vivo than its Pol II counterpart93. have either 5′ triphosphates or 7‑methylguanosine caps, A. thaliana has two highly similar NRPB3 genes, but not 5′ phosphates or 5′ hydroxyl groups, consistent one used by Pol II, Pol IV and Pol V and the other used with being initiated, rather than processed, RNAs14. only in a subset of Pol V complexes15,16. Whether the use Collectively, the evidence supports the hypothesis that of this alternative third subunit results in functionally Pol IV and Pol V are DNA-dependent RNA polymerases, distinct forms of Pol V is unclear. despite the lack of biochemical proof in vitro. The fourth subunits of Pol IV and Pol V are encoded by the same gene, NRPD4 (also known as NRPE4), Conclusions and perspective which is distinct from the Pol II NRPB4 gene15,56. Pol IV and Pol V have diverse roles in non-coding RNA- Subunits six, eight, ten and twelve are common to mediated silencing and heterochromatin formation. Pol I, Pol II and Pol III, from yeast to humans. These Most of what we know about these activities is based subunits are also common to Pol II, Pol IV and Pol V on genetic analyses, as development of biochemical in A. thaliana, as are subunits nine and eleven. For assays has been problematic. Numerous questions are most of these subunits, there are duplicate genes whose in need of answers. What are the templates for Pol IV redundancy is assumed but not proven. and Pol V? Are these templates DNA, RNA, chromatin The fifth subunits of Pol I, Pol II, Pol III and Pol IV that bears specific modifications, or some combination are encoded by the same gene, from yeast to humans94,95. thereof? What recruits Pol IV or Pol V to their sites of However, Pol V makes use of a distinct fifth subunit, action? Do Pol IV and Pol V recognize conventional NRPE5 (REFS 15,16,92), which is likely to impart different promoters, assisted by transcription factors, in a man- template specificities and/or interactions with regulatory ner similar to Pol II? How do changes in amino acids proteins based on the roles of the Pol II form of this sub that are otherwise invariant in other RNA polymerases unit 15.The yeast Rpb4 and Rpb7 subunits of Pol II form a affect the templates that are recognized by Pol IV and subcomplex that has roles in transcription initiation and Pol V, or the processivity or fidelity of the enzymes? How RNA 3′ end processing96. This Rpb4–Rpb7 subcomplex is do the subunits that differ between Pol II, Pol IV and able to dissociate from Pol II and chaperone mRNA to the Pol V account for their different functions? What is the cytoplasm to stimulate mRNA decay97–100. The A. thaliana full, functional significance of the unique NRPD1 and fourth–seventh subunit subcomplexes of Pol II, Pol IV NRPE1 CTDs? Are there tissue- or development-specific and Pol V are unique in all three polymerases owing isoforms of Pol IV and/or Pol V that make use of alterna- to the use of distinct seventh subunits and the use of a tive subunit variants? These and other questions abound, fourth subunit in Pol IV and Pol V that is different from and answers are few, presenting important challenges for the NRPB4 subunit of Pol II15,56. Immunolocalization of studies in the years to come.

1. Werner, F. & Grohmann, D. Evolution of multisubunit 16. Huang, L. et al. An atypical RNA polymerase involved methylation in Arabidopsis thaliana. Nature Struct. RNA polymerases in the three domains of life. in RNA silencing shares small subunits with RNA Mol. Biol. 16, 1325–1327 (2009). Nature Rev. Microbiol. 9, 85–98 (2011). polymerase II. Nature Struct. Mol. Biol. 16, 91–93 38. Gao, Z. et al. An RNA polymerase II‑ and This review provides a useful comparison of (2009). AGO4‑associated protein acts in RNA-directed DNA multisubunit RNA polymerase subunit 17. Mosher, R. A., Schwach, F., Studholme, D. & methylation. Nature 465, 106–109 (2010). compositions and the striking similarities between Baulcombe, D. C. PolIVb influences RNA-directed DNA 39. Wierzbicki, A. T., Ream, T. S., Haag, J. R. & Pikaard, archaeal and eukaryotic RNA polymerases. methylation independently of its role in siRNA C. S. RNA polymerase V transcription guides 2. Cramer, P. et al. Structure of eukaryotic RNA biogenesis. Proc. Natl Acad. Sci. USA 105, ARGONAUTE4 to chromatin. Nature Genet. 41, polymerases. Annu. Rev. Biophys. 37, 337–352 3145–3150 (2008). 630–634 (2009). (2008). In this study, the authors conducted an insightful This study showed that AGO4 can be crosslinked to 3. Lahmy, S., Bies-Etheve, N. & Lagrange, T. Plant-specific genome-wide analysis of Pol IV and Pol V transcripts, and associates with adjacent multisubunit RNA polymerase in gene silencing. Pol V‑dependent loci and the roles of each enzyme chromatin, providing the basis for current models Epigenetics 5, 4–8 (2010). in siRNA biogenesis. for how AGO4 and the silencing machinery is 4. Pikaard, C. S., Haag, J. R., Ream, T. & Wierzbicki, A. T. 18. Zhang, X., Henderson, I. R., Lu, C., Green, P. J. & recruited to its sites of action. Roles of RNA polymerase IV in gene silencing. Jacobsen, S. E. Role of RNA polymerase IV in plant 40. Law, J. A. et al. A protein complex required for Trends Plant Sci. 13, 390–397 (2008). small RNA metabolism. Proc. Natl Acad. Sci. USA polymerase V transcripts and RNA-directed DNA 5. Herr, A. J. Pathways through the small RNA world of 104, 4536–4541 (2007). methylation in Arabidopsis. Curr. Biol. 20, 951–956 plants. FEBS Lett. 579, 5879–5888 (2005). This paper provided the first genome-wide study to (2010). 6. Haag, J. R., Pontes, O. & Pikaard, C. S. Metal A and show that Pol IV and Pol V are required for siRNA 41. He, X. J. et al. A conserved transcriptional regulator is Metal B sites of nuclear RNA polymerases Pol IV and biogenesis at thousands of loci in the A. thaliana required for RNA-directed DNA methylation and plant Pol V are required for siRNA-dependent DNA genome. development. Genes Dev. 23, 2717–2722 (2009). methylation and gene silencing. PLoS ONE 4, e4110 19. Huettel, B. et al. Endogenous targets of RNA-directed 42. Kanno, T. et al. RNA-directed DNA methylation and (2009). DNA methylation and Pol IV in Arabidopsis. EMBO J. plant development require an IWR1‑type transcription 7. Landick, R. Functional divergence in the growing 25, 2828–2836 (2006). factor. EMBO Rep. 11, 65–71 (2010). family of RNA polymerases. Structure 17, 323–325 20. Raja, P., Wolf, J. N. & Bisaro, D. M. RNA silencing 43. Czeko, E., Seizl, M., Augsberger, C., Mielke, T. & (2009). directed against geminiviruses: post-transcriptional Cramer, P. Iwr1 directs RNA polymerase II nuclear This minireview presents a mechanistic discussion, and epigenetic components. Biochim. Biophys. Acta import. Mol. Cell 42, 261–266 (2011). and illuminating figures, highlighting the surprising 1799, 337–351 (2010). 44. Bellaoui, M., Keddie, J. S. & Gruissem, W. DCL is a amino acid divergence in the catalytic centres of 21. Pontes, O. et al. The Arabidopsis chromatin-modifying plant-specific protein required for plastid ribosomal Pol IV and Pol V relative to other RNA polymerases. nuclear siRNA pathway involves a nucleolar RNA RNA processing and embryo development. Plant Mol. 8. Tucker, S. L., Reece, J., Ream, T. S. & Pikaard, C. S. processing center. Cell 126, 79–92 (2006). Biol. 53, 531–543 (2003). Evolutionary history of plant multisubunit RNA This paper used protein immunolocalization in 45. El-Shami, M. et al. Reiterated WG/GW motifs form polymerases IV and V: subunit origins via genome- combination with genetic analyses to determine the functionally and evolutionarily conserved wide and segmental gene duplications, order of events in the RdDM pathway. This cell ARGONAUTE-binding platforms in RNAi-related retrotransposition, and lineage-specific biological approach allowed the activities that had components. Genes Dev. 21, 2539–2544 (2007). subfunctionalization. Cold Spring Harb. Symp. Quant. been identified genetically to be fitted into the 46. Song, J. J., Smith, S. K., Hannon, G. J. & Joshua-Tor, L. Biol. 75, 285–297 (2011). pathway. Crystal structure of Argonaute and its implications for 9. Arabidopsis Genome Initiative. Analysis of the genome 22. Kasschau, K. D. et al. Genome-wide profiling and RISC slicer activity. Science 305, 1434–1437 (2004). sequence of the flowering plant Arabidopsis thaliana. analysis of Arabidopsis siRNAs. PLoS Biol. 5, e57 47. Rivas, F. V. et al. Purified Argonaute2 and an siRNA Nature 408, 796–815 (2000). (2007). form recombinant human RISC. Nature Struct. Mol. 10. Herr, A. J., Jensen, M. B., Dalmay, T. & 23. Xie, Z. et al. Genetic and functional diversification of Biol. 12, 340–349 (2005). Baulcombe, D. C. RNA polymerase IV directs silencing small RNA pathways in plants. PLoS Biol. 2, 642–652 48. Zheng, X., Zhu, J., Kapoor, A. & Zhu, J. K. Role of of endogenous DNA. Science 308, 118–120 (2005). (2004). Arabidopsis AGO6 in siRNA accumulation, DNA This paper revealed that a well-studied mutation 24. Smith, L. M. et al. An SNF2 protein associated with methylation and transcriptional gene silencing. that is required for transgene silencing and RdDM, nuclear RNA silencing and the spread of a silencing EMBO J. 26, 1691–1701 (2007). silencing defective 4, mapped to the NRPD1 gene, signal between cells in Arabidopsis. Plant Cell 19, 49. Zheng, Z. et al. An SGS3‑like protein functions in revealing the function of Pol IV in siRNA 1507–1521 (2007). RNA-directed DNA methylation and transcriptional biogenesis. 25. Qi, Y., Denli, A. M. & Hannon, G. J. Biochemical gene silencing in Arabidopsis. Plant J. 62, 92–99 11. Onodera, Y. et al. Plant nuclear RNA polymerase IV specialization within Arabidopsis RNA silencing (2010). mediates siRNA and DNA methylation-dependent pathways. Mol. Cell 19, 421–428 (2005). 50. Law, J. A. & Jacobsen, S. E. Establishing, maintaining heterochromatin formation. Cell 120, 613–622 26. Henderson, I. R. et al. Dissecting Arabidopsis thaliana and modifying DNA methylation patterns in plants (2005). DICER function in small RNA processing, gene and animals. Nature Rev. Genet. 11, 204–220 Using a reverse genetics approach to define the silencing and DNA methylation patterning. (2010). function of Pol IV, this paper showed that Pol IV is a Nature Genet. 38, 721–725 (2006). This excellent Review article provides an overview nuclear enzyme that is required for 27. Yu, B. et al. Methylation as a crucial step in plant of how DNA methylation is established, maintained heterochromatin formation and RdDM of nuclear microRNA biogenesis. Science 307, 932–935 (2005). and interpreted. repeats. 28. Li, J., Yang, Z., Yu, B., Liu, J. & Chen, X. Methylation 51. Bies-Etheve, N. et al. RNA-directed DNA methylation 12. Pontier, D. et al. Reinforcement of silencing at protects miRNAs and siRNAs from a 3′-end uridylation requires an AGO4‑interacting member of the SPT5 transposons and highly repeated sequences requires activity in Arabidopsis. Curr. Biol. 15, 1501–1507 elongation factor family. EMBO Rep. 10, 649–654 the concerted action of two distinct RNA polymerases (2005). (2009). IV in Arabidopsis. Genes Dev. 19, 2030–2040 29. Havecker, E. R. et al. The Arabidopsis RNA-directed 52. He, X. J. et al. An effector of RNA-directed DNA (2005). DNA methylation Argonautes functionally diverge methylation in Arabidopsis is an ARGONAUTE 4‑ and Using a reverse genetics approach, these authors based on their expression and interaction with target RNA-binding protein. Cell 137, 498–508 (2009). identified a role for the second atypical largest loci. Plant Cell 22, 321–334 (2010). 53. Cokus, S. J. et al. Shotgun bisulphite sequencing of the subunit as a defining feature of Pol V (originally 30. Qi, Y. et al. Distinct catalytic and non-catalytic roles of Arabidopsis genome reveals DNA methylation known as Pol IVb) and showed that Pol IV and Pol V ARGONAUTE4 in RNA-directed DNA methylation. patterning. Nature 452, 215–219 (2008). have distinct roles in RdDM. Nature 443, 1008–1012 (2006). 54. Lister, R. et al. Highly integrated single-base resolution 13. Kanno, T. et al. Atypical RNA polymerase subunits This paper examined the need for AGO4 RNase maps of the epigenome in Arabidopsis. Cell 133, required for RNA-directed DNA methylation. Nature activity in RdDM, showing that the activity is 523–536 (2008). Genet. 37, 761–765 (2005). differentially required at different loci. 55. Aufsatz, W., Mette, M. F., Van Der Winden, J., In this study, a genetic screen for mutants defective 31. Schwarz, D. S. et al. Asymmetry in the assembly of the Matzke, M. & Matzke, A. J. HDA6, a putative histone for RdDM identified numerous alleles of the Pol V RNAi enzyme complex. Cell 115, 199–208 (2003). deacetylase needed to enhance DNA methylation catalytic subunits and showed their involvement in 32. Mi, S. et al. Sorting of small RNAs into Arabidopsis induced by double-stranded RNA. EMBO J. 21, RdDM downstream of siRNA biogenesis, thus Argonaute complexes is directed by the 5′ terminal 6832–6841 (2002). defining Pol V as an activity distinct from Pol IV. nucleotide. Cell 133, 116–127 (2008). 56. He, X. J. et al. NRPD4, a protein related to the RPB4 14. Wierzbicki, A. T., Haag, J. R. & Pikaard, C. S. 33. Montgomery, T. A. et al. Specificity of ARGONAUTE7– subunit of RNA polymerase II, is a component of RNA Noncoding transcription by RNA polymerase Pol IVb/ miR390 interaction and dual functionality in TAS3 polymerases IV and V and is required for RNA-directed Pol V mediates transcriptional silencing of overlapping trans-acting siRNA formation. Cell 133, 128–141 DNA methylation. Genes Dev. 23, 318–330 (2009). and adjacent genes. Cell 135, 635–648 (2008). (2008). 57. Earley, K. W. et al. Mechanisms of HDA6‑mediated This paper identified Pol V transcripts in vivo and 34. Li, C. F. et al. An ARGONAUTE4‑containing nuclear rRNA gene silencing: suppression of intergenic Pol II showed that Pol V transcription traversing the processing center colocalized with Cajal bodies in transcription and differential effects on maintenance promoter regions of other genes is required for Arabidopsis thaliana. Cell 126, 93–106 (2006). versus siRNA-directed cytosine methylation. siRNA-mediated transcriptional silencing. 35. Kanno, T. et al. Involvement of putative SNF2 Genes Dev. 24, 1119–1132 (2010). 15. Ream, T. S. et al. Subunit compositions of the RNA- chromatin remodeling protein DRD1 in RNA-directed 58. Sridhar, V. V. et al. Control of DNA methylation and silencing enzymes Pol IV and Pol V reveal their origins DNA methylation. Curr. Biol. 14, 801–805 (2004). heterochromatic silencing by histone H2B as specialized forms of RNA polymerase II. Mol. Cell 36. Kanno, T. et al. A structural‑maintenance‑of- deubiquitination. Nature 447, 735–738 (2007). 33, 192–203 (2009). chromosomes hinge domain-containing protein is 59. Searle, I. R., Pontes, O., Melnyk, C. W., Smith, L. M. & This paper revealed the subunit compositions of required for RNA-directed DNA methylation. Baulcombe, D. C. JMJ14, a JmjC domain protein, is A. thaliana Pol II, Pol IV and Pol V, showing that Nature Genet. 40, 670–675 (2008). required for RNA silencing and cell‑to‑cell movement each has twelve subunits, some of which are shared 37. Ausin, I., Mockler, T. C., Chory, J. & Jacobsen, S. E. of an RNA silencing signal in Arabidopsis. Genes Dev. and some of which are unique. IDN1 and IDN2 are required for de novo DNA 24, 986–991 (2010).

60. Deleris, A. et al. Involvement of a Jumonji‑C domain- 76. Molnar, A. et al. Small silencing RNAs in plants are 91. Ron, M., Alandete Saez, M., Eshed Williams, L., containing histone demethylase in DRM2‑mediated mobile and direct epigenetic modification in recipient Fletcher, J. C. & McCormick, S. Proper regulation of a maintenance of DNA methylation. EMBO Rep. 11, cells. Science 328, 872–875 (2010). sperm-specific cis‑nat‑siRNA is essential for double 950–955 (2010). 77. Brosnan, C. A. et al. Nuclear gene silencing directs fertilization in Arabidopsis. Genes Dev. 24, 61. Lu, F., Cui, X., Zhang, S., Liu, C. & Cao, X. JMJ14 is reception of long-distance mRNA silencing in 1010–1021 (2010). an H3K4 demethylase regulating flowering time in Arabidopsis. Proc. Natl Acad. Sci. USA 104, 92. Lahmy, S. et al. PolV(PolIVb) function in RNA-directed Arabidopsis. Cell Res. 20, 387–390 (2010). 14741–14746 (2007). DNA methylation requires the conserved active site 62. Kato, H. et al. RNA polymerase II is required for 78. Pontes, O., Costa-Nunes, P., Vithayathil, P. & and an additional plant-specific subunit. Proc. Natl RNAi-dependent heterochromatin assembly. Pikaard, C. S. RNA polymerase V functions in Acad. Sci. USA 106, 941–946 (2009). Science 309, 467–469 (2005). Arabidopsis interphase heterochromatin organization 93. Luo, J. & Hall, B. D. A multistep process gave rise to 63. Zheng, B. et al. Intergenic transcription by RNA independently of the 24‑nt siRNA-directed DNA RNA polymerase IV of land plants. J. Mol. Evol. 64, polymerase II coordinates Pol IV and Pol V in methylation pathway. Mol. Plant 2, 700–710 101–112 (2007). siRNA-directed transcriptional gene silencing in (2009). This interesting paper details the stepwise Arabidopsis. Genes Dev. 23, 2850–2860 (2009). 79. Douet, J., Tutois, S. & Tourmente, S. A Pol V‑mediated evolution of Pol IV and Pol V catalytic subunits 64. Arteaga-Vazquez, M. A. & Chandler, V. L. silencing, independent of RNA-directed DNA from genes encoding Pol II subunits, beginning in Paramutation in maize: RNA mediated trans- methylation, applies to 5S rDNA. PLoS Genet. 5, green algae. generational gene silencing. Curr. Opin. Genet. Dev. e1000690 (2009). 94. Larkin, R. M., Hagen, G. & Guilfoyle, T. J. Arabidopsis 20, 156–163 (2010). 80. Amasino, R. Seasonal and developmental timing of thaliana RNA polymerase II subunits related to yeast 65. Hollick, J. B. Paramutation and development. flowering. Plant J. 61, 1001–1013 (2010). and human RPB5. Gene 231, 41–47 (1999). Annu. Rev. Cell Dev. Biol. 26, 557–579 (2010). 81. Chan, S. W. et al. RNA silencing genes control de novo 95. Saez-Vasquez, J. & Pikaard, C. S. Extensive 66. Erhard, K. F., Jr. et al. RNA polymerase IV functions in DNA methylation. Science 303, 1336 (2004). purification of a putative RNA polymerase I paramutation in Zea mays. Science 323, 1201–1205 82. Liu, F. et al. The Arabidopsis RNA-binding protein FCA holoenzyme from plants that accurately initiates rRNA (2009). requires a lysine-specific demethylase 1 homolog to gene transcription in vitro. Proc. Natl Acad. Sci. USA In this study, the authors used a genetic screen to downregulate FLC. Mol. Cell 28, 398–407 (2007). 94, 11869–11874 (1997). identify the maize Pol IV largest subunit as a critical 83. Liu, J., He, Y., Amasino, R. & Chen, X. siRNAs 96. Orlicky, S. M., Tran, P. T., Sayre, M. H. & activity that is required for paramutation. targeting an intronic transposon in the regulation of Edwards, A. M. Dissociable Rpb4–Rpb7 subassembly 67. Alleman, M. et al. An RNA-dependent RNA natural flowering behavior in Arabidopsis. Genes Dev. of RNA polymerase II binds to single-strand nucleic polymerase is required for paramutation in maize. 18, 2873–2878 (2004). acid and mediates a post-recruitment step in Nature 442, 295–298 (2006). 84. Swiezewski, S. et al. Small RNA-mediated chromatin transcription initiation. J. Biol. Chem. 276, This study provided the first indication that maize silencing directed to the 3′ region of the Arabidopsis 10097–10102 (2001). orthologues of A. thaliana RdDM-pathway proteins gene encoding the developmental regulator, FLC. 97. Selitrennik, M., Duek, L., Lotan, R. & Choder, M. are required for paramutation by identifying the Proc. Natl Acad. Sci. USA 104, 3633–3638 (2007). Nucleocytoplasmic shuttling of the Rpb4p and maize orthologue of A. thaliana RDR2 as a key 85. Soppe, W. J. et al. The late flowering phenotype of Rpb7p subunits of Saccharomyces cerevisiae RNA protein for paramutation. fwa mutants is caused by gain‑of‑function epigenetic polymerase II by two pathways. Eukaryot. Cell 5, 68. Hale, C. J., Erhard, K. F., Jr, Lisch, D. & Hollick, J. B. alleles of a homeodomain gene. Mol. Cell 6, 791–802 2092–2103 (2006). Production and processing of siRNA precursor (2000). 98. Lotan, R. et al. The RNA polymerase II subunit Rpb4p transcripts from the highly repetitive maize genome. 86. Mosher, R. A. et al. Uniparental expression of PolIV- mediates decay of a specific class of mRNAs. PLoS Genet. 5, e1000598 (2009). dependent siRNAs in developing endosperm of Genes Dev. 19, 3004–3016 (2005). 69. Sidorenko, L. et al. A dominant mutation in mediator Arabidopsis. Nature 460, 283–286 (2009). 99. Lotan, R., Goler-Baron, V., Duek, L., Haimovich, G. & of paramutation2, one of three second-largest This study revealed that the abundant siRNAs Choder, M. The Rpb7p subunit of yeast RNA subunits of a plant-specific RNA polymerase, disrupts that are produced in developing seeds are entirely polymerase II plays roles in the two major cytoplasmic multiple siRNA silencing processes. PLoS Genet. 5, derived from the maternal chromosomes owing to mRNA decay mechanisms. J. Cell Biol. 178, e1000725 (2009). a previously unknown imprinting mechanism. 1133–1143 (2007). 70. Stonaker, J. L., Lim, J. P., Erhard, K. F. Jr & 87. Slotkin, R. K. et al. Epigenetic reprogramming and 100. Goler-Baron, V. et al. Transcription in the nucleus and Hollick, J. B. Diversity of Pol IV function is defined by small RNA silencing of transposable elements in mRNA decay in the cytoplasm are coupled processes. mutations at the maize rmr7 locus. PLoS Genet. 5, pollen. Cell 136, 461–472 (2009). Genes Dev. 22, 2022–2027 (2008). e1000706 (2009). This intriguing paper shows that the vegetative 71. Molnar, A., Melnyk, C. & Baulcombe, D. C. Silencing cell of pollen derepresses transposons in order to Acknowledgements signals in plants: a long journey for small RNAs. generate siRNAs that move to the adjacent sperm Our studies of RNA polymerases IV and V are supported by Genome Biol. 12, 215 (2011). cells to programme gene silencing. grants from the US National Institutes of Health 72. Dunoyer, P., Himber, C. & Voinnet, O. DICER-LIKE 4 88. Wang, X. J., Gaasterland, T. & Chua, N. H. (GM077590) and the Monsanto Company. All opinions is required for RNA interference and produces the Genome-wide prediction and identification of cis- expressed are those of the authors and do not necessarily 21‑nucleotide small interfering RNA component of the natural antisense transcripts in Arabidopsis thaliana. reflect the views of our sponsors. We apologize to our many plant cell‑to‑cell silencing signal. Nature Genet. 37, Genome Biol. 6, R30 (2005). colleagues whose work was not cited owing to limitations on 1356–1360 (2005). 89. Borsani, O., Zhu, J., Verslues, P. E., Sunkar, R. & the numbers of references. 73. Dunoyer, P., Himber, C., Ruiz-Ferrer, V., Alioua, A. & Zhu, J. K. Endogenous siRNAs derived from a pair Voinnet, O. Intra- and intercellular RNA interference of natural cis-antisense transcripts regulate salt Competing interests statement in Arabidopsis thaliana requires components of the tolerance in Arabidopsis. Cell 123, 1279–1291 The authors declare no competing financial interests. microRNA and heterochromatic silencing pathways. (2005). Nature Genet. 39, 848–856 (2007). The authors report how siRNA-mediated regulation 74. Dunoyer, P. et al. Small RNA duplexes function as of convergently transcribed gene pairs can mediate FURTHER INFORMATION mobile silencing signals between plant cells. Science an adaptive response. Craig S. Pikaard’s homepage: 328, 912–916 (2010). 90. Katiyar-Agarwal, S., Gao, S., Vivian-Smith, A. & Jin, H. http://sites.bio.indiana.edu/~pikaardlab 75. Dunoyer, P. et al. An endogenous, systemic RNAi A novel class of bacteria-induced small RNAs in ALL LINKS ARE ACTIVE IN THE ONLINE PDF pathway in plants. EMBO J. 29, 1699–1712 (2010). Arabidopsis. Genes Dev. 21, 3123–3134 (2007).