Downloaded from genesdev.cshlp.org on October 4, 2021 - Published by Cold Spring Harbor Laboratory Press PERSPECTIVE Seeing the forest for the trees: a wide perspective on RNA-directed DNA methylation Huiming Zhang1 and Jian-Kang Zhu1,2,3 1Department of Horticulture and Landscape Architecture, Purdue University, West Lafayette, Indiana 47907, USA; 2Shanghai Center for Plant Stress Biology, Shanghai Institutes of Biological Sciences, Chinese Academy of Sciences, Shanghai 200032, China In this issue of Genes & Development, Wierzbicki and remaining subunits of Pol IV and/or Pol V are different but colleagues (pp. 1825–1836) examine the current model have apparently evolved from their paralogs in Pol II of RNA-directed DNA methylation (RdDM) by deter- (Ream et al. 2009). NRPD1 and NRPE1 are the unique and mining genome-wide distributions of RNA polymerase V largest subunits of Pol IV and Pol V, respectively. Muta- (Pol V) occupancy, siRNAs, and DNA methylation. Their tion in NRPD1 or NRPD/E2, which is the second largest data support the key role of base-pairing between Pol V subunit shared by Pol IV and Pol V, resulted in >90% transcripts and siRNAs in targeting de novo DNA meth- reduction of 24-nt siRNA accumulation (Zhang et al. ylation. Importantly, the study also reveals unexpected 2007; Mosher et al. 2008), thereby demonstrating a pre- complexity and provides a global view of the RdDM dominant role of Pol IV in generating 24-nt siRNAs. pathway. Although Pol IV transcripts remain to be identified, Pol IV has been thought to initiate siRNA production by generating an aberrant ssRNA, which is subsequently copied by RDR2 (RNA-dependent RNA polymerase 2) to DNA methylation at the fifth position of cytosine regu- produce dsRNAs that are cleaved by DCL3 (dicer-like 3) lates many critical biological processes, such as gene and then loaded onto AGO4 (argonaute 4) or its closely imprinting, silencing of transposable elements, and related argonaute proteins (Law and Jacobsen 2010; Haag X-chromosome inactivation. In Arabidopsis, the DNA and Pikaard 2011; Zhang and Zhu 2011). These argo- methyltransferase DRM2 (domains rearranged methyl- naute-bound 24-nt siRNAs can then serve as sequence- transferase 2) catalyzes de novo methylation in all cyto- specific guides for methylation by pairing with comple- sine contexts, including CG, CHG, and CHH (H repre- mentary DNA or nascent scaffold RNA. AGO4 can be sents either A, T, or G) (Cao and Jacobsen 2002). The cross-linked to scaffold RNAs, supporting the model of targeting of DRM2 for DNA methylation can be achieved siRNA–scaffold RNA pairing (Wierzbicki et al. 2009). by the RNA-directed DNA methylation (RdDM) pathway Production of scaffold RNA is independent of siRNA that consists of three phases: biogenesis of 24-nucleotide biogenesis, as shown in Arabidopsis mutants defective (nt) siRNAs, production of scaffold RNAs, and recruit- in NRPD1, RDR2, or DCL3 (Wierzbicki et al. 2008). ment of DRM2 assisted by complementary pairing be- Wierzbicki et al. (2008) previously demonstrated that Pol tween 24-nt siRNAs and nascent scaffold RNAs. V is necessary for scaffold RNA production. In addition to generating scaffold RNAs, Pol V can reinforce siRNA RNA polymerases in the RdDM pathway production at some RdDM target loci (Zhang et al. 2007; Mosher et al. 2008) and can also interact with AGO4 In the RdDM pathway, transcription of the noncoding through an argonaute-binding motif in the C-terminal tail RNAs involves three DNA-dependent RNA polymerases: of NPRE1 (El-Shami et al. 2007), thereby strengthening Pol II, Pol IV, and Pol V. Arabidopsis Pol IV and Pol V are the recruitment of the silencing complex guided by plant-specific RNA polymerases, each of which contains AGO4. Direct interaction between AGO4 and the meth- 12 subunits (Ream et al. 2009). While six of these 12 yltransferase DRM2 has not been shown. Nevertheless, subunits are identical among Pol IV, Pol V, and Pol II, the cytological analyses revealed colocalization of AGO4 and DRM2, together with other RdDM components in dis- tinct nuclear foci (Li et al. 2006, 2008; Pontes et al. 2006; [Keywords: DNA-dependent RNA polymerase; gene silencing; DNA methylation; epigenetics; short interfering RNA; RNA-directed DNA He et al. 2009; Gao et al. 2010). Additionally, AGO4 and methylation] DRM2 both coimmunoprecipitate with RDM1 (RNA- 3Corresponding author E-mail [email protected] directed DNA methylation 1), which is an ssDNA-binding Article is online at http://www.genesdev.org/cgi/doi/10.1101/gad.200410.112. protein with preference to methylated DNA (Gao et al. GENES & DEVELOPMENT 26:1769–1773 Ó 2012 by Cold Spring Harbor Laboratory Press ISSN 0890-9369/12; www.genesdev.org 1769 Downloaded from genesdev.cshlp.org on October 4, 2021 - Published by Cold Spring Harbor Laboratory Press Zhang and Zhu 2010). Therefore, Pol V and its transcriptional prod- the RdDM model genome-wide, Wierzbicki et al. (2012) ucts seemed to be essential for targeting de novo DNA also performed parallel analyses of the DNA methylome methylation. and siRNA transcriptome. Among the identified Pol V- Transcription for noncoding RNAs during RdDM can occupied loci, a majority (56%) encompasses regions also be Pol II-dependent. Pol II transcription of inverted where complementary 24-nt siRNAs and the asymmet- DNA repeats can generate dsRNAs that serve as DCL3 ric CHH methylation can be found. Such observations substrates, thus contributing to 24-nt siRNA production support the RdDM model in which the pairing between (Zhang et al. 2007; Pikaard et al. 2008). At some inter- siRNA and scaffold RNA recruits the silencing complex, genic low-copy-number repeat loci, Pol II was shown to given that methylation in CHH context must be main- be indispensable for scaffold RNA generation (Zheng tained by de novo methylation and is therefore a hall- et al. 2009). Like Pol V, Pol II also possesses the GW/ mark of RdDM activity. In contrast, only 1% of Pol V WG repeats that bind to AGO4 (Zheng et al. 2009). target sites show CHH methylation without the 24-nt Although the argonaute-binding motif appears to be siRNA. weaker in Pol II than Pol V, Pol II clearly interacts with Locus-specific reduction of DNA methylation has been AGO4 and RDM1 (Zheng et al. 2009; Gao et al. 2010). In well documented in Arabidopsis mutants with defective a weak loss-of-function allele of the NRPB2 (the second RdDM. Wierzbicki et al. (2012) observed an unexpected largest subunit of Pol II) mutant, occupancy of Pol IV and shift, instead of a global loss, of the CHH methylation Pol V at some heterochromatic loci is reduced (Zheng mark in Pol IV and Pol V mutants compared with wild- et al. 2009), indicating that these three RNA polymerases type plants. In the mutants, ;50% of CHH cytosine may cooperate during RdDM. Meanwhile, NRPE1 colo- became unmethylated, and this was accompanied by the calizes with RDM1, DRM2, and AGO4 in the perinucleo- ectopic occurrence of a similar number of CHH methyl- lar bodies but shows no clear colocalization patterns in ations at new positions. Intriguingly, CHH methylation nucleoplasmic foci (Li et al. 2006, 2008; Pontes et al. was reduced in the chromosome arms, while new CHH 2006; He et al. 2009; Gao et al. 2010). In contrast, methylation occurred in the pericentromeric regions. colocalization between NRPB1 and RDM1 or AGO4 These results indicate that both Pol IV and Pol V function can only be observed in discrete nucleoplasmic foci (Gao in targeting DNA methylation to certain loci but are not et al. 2010). Therefore, it appears that Pol V and Pol II can required for DNA methyltransferase activities, and in the mediate RdDM in a spatially separated manner. It is absence of these polymerases, DNA methylation is unclear whether the activities in the perinucleolar bodies redirected to other loci. and nucleoplasmic foci may be parallel or sequential (Zhang and Zhu 2011). Discovered and to be discovered The discoveries made by Wierzbicki et al. (2012) have Genome-wide examination of the RdDM model allowed a genome-wide evaluation of the role of non- Although a key role of Pol V in RdDM has been estab- coding RNA pairing in mediating de novo DNA methyl- lished, only a few scaffold RNAs had been detected before ation. In addition to providing support for the current the study by Wierzbicki et al. (2012). In addition, Pol V RdDM model, their results also provide novel informa- also mediates RdDM-independent silencing at pericen- tion that will cause researchers to evaluate de novo DNA tromeric repeats (Douet et al. 2009; Pontes et al. 2009). methylation from a broader perspective (Fig. 1). It was unknown how broadly Pol V is needed across In nrpd1, nrpe1, and nrpd/e2 mutants, a significant the whole genome. By using ChIP-seq (chromatin immu- portion (slightly <50%) of cytosine positions in the CHH noprecipitation [ChIP] followed by deep sequencing), context remain to be methylated (Wierzbicki et al. 2012). Wierzbicki et al. (2012) compared genome-wide NRPE1 One possible explanation is that RNA-directed de novo occupancy in wild-type Arabidopsis and the nrpe1-11-null DNA methylation at these positions is at least in part mutant and thereby revealed potential Pol V-transcribed independent of Pol IV and Pol V. Generation of 24-nt sites throughout the whole genome. NRPE1 occupies siRNAs and scaffold RNAs at these loci may also be >1000 genomic regions, including loci that had been accomplished by Pol II, which is known to be involved in known to be transcribed by Pol V. These Pol V-occupied RdDM in addition to having a canonical role in mRNA regions range from several hundred to a few thousands of transcription (Zheng et al.