Talk:PLOS/RNA-directed DNA methylation (RdDM)

Review by R. Keith Slotkin

Review of version as of 18:07, 13 October 2019: http://compbiolwiki.plos.org/w/index.php?title=RNA-directed_DNA_methylation_(RdDM)&oldid=8176

This is a very high quality effort from the Authors. This reviewer thanks them for writing this Wiki page for the community. The comments here are meant to clarify, expand and improve what has already been written.

Overall, I commend the authors on this article / Wiki page. However, I think the written text is considerably better compared to the figures. These figures will be used a lot by students that google for RdDM figures, so I urge the authors to improve these figures. Good examples are the Nature Reviews Genetics articles on RdDM. Points for the RdDM Pathway figure: Pol IV and RDR2 work as a complex, so please put them together. And the double-stranded RNA they make is very short. Also, AGO4 (and likely 6) interact with Pol V, so those two proteins should also be touching at the chromatin.
In the Abstract, I think it is important to say that small RNA-directed modification of chromatin is found in plants, fungi and animals. Many of the best studied fungi (fission yeast) and animals (Drosophila) don’t methylate their DNA, so the Histone H3 Lysine #9 is the target of methylation. In fungi this is performed by plant-like siRNAs, while in animals this is performed by piRNAs, which are not found in plants.
In the abstract, please do not refer to DNA methylation as an “epigenetic” mark. It is a chromatin mark that can lead to transcriptional silencing, and may lead to epigenetic inheritance of a transcriptionally silenced state.
In the last sentence of the abstract, I would rewrite it to say: “Overall, RdDM is an important pathway in plants that regulates a number of processes by establishing and reinforcing specific DNA methylation patterns, which may lead to trans-generational epigenetic transcriptional silencing.
In the History and Discovery section, before introducing Pol IV, Pol V and other factors, I would add a sentence or two about the discovery to two types of siRNAs (21-22 vs 24 nt) and the discovery that the larger size class associated with RdDM. This was an important discovery that predated and primed the discovery of piRNA.
In the third paragraph of the section “Repression of Transposable Elements”, I would frame the function of RdDM something like this: RdDM primarily targets small TEs and TE fragments that are located near genes, and therefore these TEs are located in euchromatic regions of the genome. RdDM is necessary to constantly reinforce DNA methylation and transcriptional silencing of the TEs in this transcriptionally-permissive environment. RdDM is necessary to constantly redraw the boundary between heterochromatin (the TE) and the euchromatic neighboring genic region. RdDM constantly acts at this boundary to create this distinction, which is akin to drawing a line in the sand. The line will disappear if not constantly redrawn, which is the role of RdDM. Without RdDM, the genic euchromatin pushes into the TE, resulting in the expression of the TE. With too much RdDM, the heterochromatic marks push into the gene, and can result in its silencing.
I think it is worth mentioning that since HP1 has a distinct function in plants, RdDM is one of two known ways to spread H3K9me2 heterochromatin in plant genomes (the other being the loop between CMT3 and KYP).
Please mention that RdDM targets all cytosine contexts equally, which is different from maintenance DNA methylation. Because maintenance methylation acts at a low level at CHH context DNA methylation (Arabidopsis) or not at all at CHH (maize), RdDM is typically assayed by bisulfite sequencing focusing on the CHH context.
In the “Development and Reproduction” section, perhaps describe the FWA example, as this was the classic example used to identify many of the factors, and effects flowering time.
One KEY function of RdDM that needs to be elaborated: Repression of introduced transgenes! A lot of interest in RdDM comes from the desire to avoid transgene silencing to efficiently produce improved plants.
Most of the field refers to RNAP IV as ‘Pol IV’. The authors may want to make this change. Same with ‘Pol V’.
In the section “Silencing of target loci by canonical RdDM”, Existing DNA methylation does not recruit Pol IV. Instead, H3K9me2 recruits Pol IV through SHH1. The authors should add H3K9 methylation to this section so it is clear to the reader that the DNA methylation of RdDM leads to histone methylation.
This sentence is not accurate: “The RdDM pathway is just one of several pathways involved in establishing and maintaining DNA methylation patterns in the plant.” RdDM is the ONLY pathway known to establish DNA methylation in plants. Other pathways can maintain.

Kslotkin (talk) 11:35, 6 November 2019 (PST)

Review by Todd Blevins

Review of version as of 18:07, 13 October 2019: http://compbiolwiki.plos.org/w/index.php?title=RNA-directed_DNA_methylation_(RdDM)&oldid=8176

This topic page provides a very accessible overview of RNA-directed DNA methylation (RdDM) in plants. I congratulate the authors on collecting and curating the abundant literature on this subject. The community will benefit from this Wiki, a natural entry point for new students and scientists from other fields. To help revise this resource, I have several suggestions (points 1-6) and one vital correction (point 7).

I join the previous reviewer, Keith Slotkin, in recommending upgraded figures. In seeking to be schematic, the authors have omitted key protein-DNA, protein-protein and protein-RNA interactions that explain how RdDM works. Here is a non-exhaustive list of improvements that could be made to Figure 2:
1. One cannot understand RdDM outside of its chromatin context. Figure 2 should thus convey some basic notions about histone modifications. I recommend showing SHH1 reading repressive chromatin marks (unmethylated H3K4 + dimethylated H3K9) to recruit Pol IV to the silenced TE.
2. These RNA polymerases require nucleic acid templates. Pol IV should thus be shown transcribing chromosomal DNA to produce short primary transcripts, not floating below the DNA. Similarly, Pol II and Pol V act on chromosomal DNA. RDR2 could be shown transcribing a Pol IV transcript (see next item).
3. The Pol IV-RDR2 complex couples synthesis of primary transcripts (Pol IV products) to the synthesis of short dsRNAs (RDR2 products). These dsRNA products average ~30 bases in vivo, which is supported by Pol IV-RDR2 activities in vitro (Blevins et al. 2015; Zhai et al. 2015; Singh et al. 2019).
4. Dicers (DCL2, DCL3, DCL4) cleave dsRNA substrates into duplex siRNAs, not a pool of single-stranded siRNAs, as currently depicted in Figure 2.
5. AGO4-siRNA effectors are thought to find nascent RNAs in chromatin, tethering AGO4 to these Pol V transcripts and recruiting DRM2 to the DNA. The RNA scaffold would not be floating free, but should rather link AGO4 to Pol V in chromatin.
6. Using a line segment to join AGO1, AGO4 or AGO6 to each siRNA guide is an odd convention. I recommend showing each guide directly bound to its AGO.
Having a "History and discovery of RdDM" section is a good idea, but many key discoveries are omitted from this short account. The notion of siRNAs as putative guides for RdDM, particularly the 24 nt siRNA class, comes from a series of visionary papers including but not limited to: Mette el al. (2000) EMBO J, Hamilton et al. (2002) EMBO J, Zilberman et al. (2003) Science, and Xie et al. (2004) PLoS Biol. The intriguing sde4 allele from the silencing defective screen (Dalmay et al. 2000 Cell) would later be understood as an nrpd1 null mutant (pol IV largest subunit mutant).
The canonical RdDM pathway is indeed well-characterized, as the authors state. Because Pol IV and Pol V were critical to unravelling this mechanism, I recommend citing the four papers that reported discovering these plant-specific multisubunit RNA polymerases: Herr et al. (2005) Science, Onodera et al. (2005) Cell, Kanno et al. (2005) Nature Genet., and Pontier et al. (2005) Genes Dev.
From the above, the authors will note my preference for the Pol IV / Pol V rather than RNAP IV / RNAP V nomenclature. Wikipedia uses both forms, depending on the article and context, but Pol IV / Pol V are the consensus in the RdDM research community.
In the "Biological functions of RdDM", subsections "TEs" and "Stress response": the early work on EVADE and ONSEN retrotransposons should be highlighted: Mirouze et al. (2009) Nature; Ito et al. (2011) Nature. Moreover, there is a rich literature of follow-up experiments that continues to chart the interaction between RdDM pathways and these LTR/Copia elements in Arabidopsis. For example, Marí-Ordóñez et al. (2013) reconstructed how de novo silencing of EVADE elements could transition from 21 nt siRNA-mediated silencing into the canonical Pol IV-dependent, 24 nt siRNA mechanism. The authors could include this study in the TE subsection, rather than only in "Non-canonical RdDM".
In the abstract, "The fullest complement of RdDM pathway components characterized to date is found within angiosperms ... Other groups of plants, such as gymnosperms and ferns, possess a subset of conserved RdDM components ..." Consider rephrasing or deleting this bit about gymnosperms and ferns. We are just beginning to understand the diversity of RdDM factors encoded in conifer and fern genomes. Genomic coverage in these gigantic reference assemblies is spotty, and some past surveys (circa 2015), seriously underestimated the conservation of core RdDM components. Absence of evidence is not evidence of absence.
The evolutionary analysis depicted in Figure 4 contains an error. RDR2 is conserved in Ferns, Lycophytes and Bryophytes, not only in seed plants. In Physcomitrella patens, the function of PpRDR2 was thoroughly tested in Coruh et al. 2015 Plant Cell. As expected, 24 nt siRNAs that accumulate in WT moss are absent in rdr2 mutants.

Tblevins (talk) 15 December 2019

Review by Craig Pikaard

This is a well-written and ambitious article that covers a lot of ground and will be an excellent online resource at Wikipedia. Because this can be anticipated to be such a useful resource to newcomers to the field of RNA-directed DNA methylation, my major suggestion is that the authors should include more of the essential primary papers in the field instead of falling back, in many cases, to referring to review articles, none of which are comprehensive. There are also a few places where additional discussion is needed or where the statements are inaccurate and need to be modified. But overall, this is an excellent review article.

Specific comments

1. In the abstract in the first sentence, it would be best to say that RNA-directed DNA methylation is a biological process in which noncoding RNA molecules direct the addition of DNA methylation….. This is because RdDM requires both small siRNAs and longer non-coding RNAs transcribed by Pol V.

2. Also in the abstract, when discussing pathways that are similar in other organisms, fission yeast should definitely be included, as this is probably the closest analogous system that like RdDM requires RNA-dependent RNA polymerase activity.

3. On page 2, in the history and discovery section, third to last sentence, the Baulcombe lab’s discovery of siRNAs (Hamilton and Baulcombe) should be cited.

4. The second to last sentence of the history section should also be modified to include the fact that reverse genetic experiments played a major role in the discovery of Pols IV and V, not just genetic screens, as currently stated. In fact, the first evidence for the existence of these novel polymerases came from bioinformatic evidence of Pikaard and reported in the Nature paper that describe the Arabidopsis genome sequence. There is a discussion of the discovery, and naming, of Pols IV and V in a review article by Haag and Pikaard in 2011 in a Nature Reviews journal. The Pikaard lab, using reverse genetics, and the Baulcombe lab, using forward genetics, then independently co-discovered Pol IV’s involvement in DNA methylation, and they agreed on the name Pol IV. The Lagrange lab, using reverse genetics, and the Matzke lab, using forward genetics, next independently discovered the function of what they initially called Pol IVb, involving the second unusual large catalytic subunit that Pikaard had observed bioinformatically. Pol IVb was later renamed Pol V by Pikaard and colleagues (Ream et al, 2009) when they determined the complete subunit compositions of Pol IV, Pol V and Pol II, making it clear that the enzymes are structurally, as well as functionally, distinct and needing different names so that there subunits could be identified with a systematic nomenclature (NRPD--- or NRPE -----). The references to these important papers in the field should be included as Pols IV and V are arguably the key enzymes of the RdDM pathway, being responsible for the RNAs that drive the process.

5. On page 5, DCL2 and DCL3 should be defined when first mentioned. Alternatively, the previous mention of Dicer-Like activity could include (DCL) to introduce the nomenclature. There are other examples too where the authors should be check to make sure that abbreviations are explained.

6. The biotechnology application section is in the wrong place, in my opinion, and should be moved to near the end of the article, as it is logically an extension of how the basic science understanding of the pathway can be exploited.

7. At the top of page 7 where Pols IV and V are discussed, the original papers in which the enzymes were first named should be cited (Onodera et al,2005; Herr et al,2005; Ream et al., 2009; Wierzbicki et al, 2008)

8. On page 8, In the section that discusses production of sRNAs, there is a new 2019 paper by Singh et al in Molecular Cell that demonstrates that Pol IV, RDR2 and DCL3 are all that is needed for siRNA synthesis, and they show biochemically how the enzymes work. This would be good to cite here. The model of Figure 2 is also misleading by showing Pol IV and RDR2 as separate entities. In fact, they are physically associated and work in a coupled reaction. The authors should look at the review by Wendte et al in 2017 for a more up-to-date model.

9. On page 8, in the section on production of sRNAs involved in noncanonical RdDM , the first sentence is a bit misleading in that RDR2 is needed in addition to Pol IV for canonical 24 nt siRNAs, so there are two enzymatic sources of the siRNA precursors, not one. On the other hand, the Singh et al paper mentioned in point #8 above shows that Pol IV and RDR2 act in a coupled reaction, so the Pol IV-RDR2 physical complex might be considered as a single entity that does several reactions.

10. In the Discussions of the DDR complex, the involvement of two of the three components of that complex in Pol V transcription was shown by Wierzbicki et al (2208, 2009), prior to the Jacobsen lab showing that they are together in the same complex. This should be cited appropriately.

11. Page 12, near the top, in the discussion of how Pol IV is recruited, Blevins et al (2014) showed that Histone Deacetylase 6 is also required for Pol IV recruitment and for the epigenetic inheritance of the Pol IV recruitment signal, due to its involvement in maintenance methylation by MET1. It is likely the inherited DNA methylation state that gives rise to the histone modification state that is recognized by SHH1. This should be included and cited again in second paragraph of page 13, and in the first sentence of the 4th paragraph of page 13.

12. On page 12, in the factors affecting RNAP V and DRM2 targeting, the authors incorrectly attribute the hypothesis that DDR unwinds DNA to facilitate Pol V transcription to reference 64. In fact, this hypothesis comes from a 2013 Pikaard review article published in the Cold Spring Harbor Symposiums in Quantitative Biology series.

13. Page 13, third paragraph, the author neglected to note that DDM1 facilitates CG methylation by MET1. The original papers from the Richards lab on the discovery of DDM1 should be cited.

14. In the discussion of ROS1, ROS1 expression has also bee shown to be regulated by condensin (Wang et al, 2017), not just DNA methylation.

15. Page 14, the Ream et al paper (2009) is what firmly established the evolutionary relatedness of Pols IV and V to Pol II. The authors should again cite original papers, wherever possible, rather than reviews.

16. In that same section, it is stated that Pols IV and V have “around 12 subunits”. No, they have (at least) 12 subunits, as shown by Ream et al.

17. In discussing the evolution of Pol IV and Pol V subunits, papers by Luo and Hall and Tucker et al are not cited but preceded the paper from Mosher and colleagues that is cited.

18. In the figure legend to Figure 3, the reference to the subunit nomenclature should be cited: again, this is Ream et al, 2009.

19. The final paragraph, about miRNAs is really not relevant to RdDM and I suggest that the authors delete it.

20. One thing that is missing is mention of the discovery/identification of the RNA transcripts that made by Pol IV and Pol V. The Pol V transcripts were first identified in 2008 by Wierzbicki et al. The Pol IV and RDR2-dependent transcripts were not identified until 2015, by the Chen, Jacobsen and Pikaard labs.

In summary, this is an excellent article and I hope these suggestions will help improve it and allow readers to find more of the original papers that have established current thinking I the field.

I focus on science content and in bringing academics/researchers/experts to become contributors to the movement.

I help PLOS run their ongoing 'topic pages' initiative where academics write review-style articles that can be then copied into Wikipedia. http://topicpageswiki.plos.org/

I also am chair of the WikiJournal User Group, which has massively expanded on this concept to engage experts in writing and peer-reviewing content that is then integrated into several wikimedia projects. https://en.wikiversity.org/wiki/WikiJournal_User_Group https://commons.wikimedia.org/wiki/File:Sister_Project_application_for_Wikimedia_Journals_(combined_%2B_addendum_1).pdf

I create high-quality diagrams for Wikipedia's science pages (especially on under-illustrated topics). https://en.wikipedia.org/wiki/User:Evolution_and_evolvability#GALLERY

I ran the Australian wing of the Wiki Science image competition. http://www.wikisciencecompetition.org/participate/

Response to reviews

We thank each of the three reviewers for their detailed and eminently constructive comments - the revised manuscript has definitely benefited from the feedback. Below, we detail our response to each point individually, with our responses in italics.

Review by R. Keith Slotkin

Review of version as of 18:07, 13 October 2019: http://compbiolwiki.plos.org/w/index.php?title=RNA-directed_DNA_methylation_(RdDM)&oldid=8176 This is a very high quality effort from the Authors. This reviewer thanks them for writing this Wiki page for the community. The comments here are meant to clarify, expand and improve what has already been written.

1. Overall, I commend the authors on this article / Wiki page. However, I think the written text is considerably better compared to the figures. These figures will be used a lot by students that google for RdDM figures, so I urge the authors to improve these figures. Good examples are the Nature Reviews Genetics articles on RdDM. Points for the RdDM Pathway figure: Pol IV and RDR2 work as a complex, so please put them together. And the double-stranded RNA they make is very short. Also, AGO4 (and likely 6) interact with Pol V, so those two proteins should also be touching at the chromatin.

We’ve made significant changes to the RdDM pathway figure, and believe we have addressed all of the specific suggestions made here, along with other improvements. We have also made some revisions to the figure outlining RdDM functions (Figure 1).

2. In the Abstract, I think it is important to say that small RNA-directed modification of chromatin is found in plants, fungi and animals. Many of the best studied fungi (fission yeast) and animals (Drosophila) don’t methylate their DNA, so the Histone H3 Lysine #9 is the target of methylation. In fungi this is performed by plant-like siRNAs, while in animals this is performed by piRNAs, which are not found in plants.

We now mention RNA directed chromatin modification within fungi and animals in the abstract, and have also added additional information about similar pathways in fission yeast within the “relationships with sRNA silencing pathways in other kingdoms” section.

3. In the abstract, please do not refer to DNA methylation as an “epigenetic” mark. It is a chromatin mark that can lead to transcriptional silencing, and may lead to epigenetic inheritance of a transcriptionally silenced state.

We have rephrased this sentence.

4. In the last sentence of the abstract, I would rewrite it to say: “Overall, RdDM is an important pathway in plants that regulates a number of processes by establishing and reinforcing specific DNA methylation patterns, which may lead to trans-generational epigenetic transcriptional silencing.

Thank you for this suggestion - we have replaced the final sentence with this one, with minor modifications. We also added a sentence earlier in the abstract for context indicating that DNA methylation patterns are often heritable in plants, since this is not true in mammals.

5. In the History and Discovery section, before introducing Pol IV, Pol V and other factors, I would add a sentence or two about the discovery to two types of siRNAs (21-22 vs 24 nt) and the discovery that the larger size class associated with RdDM. This was an important discovery that predated and primed the discovery of piRNA.

We have added a sentence discussing these findings in the revised and relocated History and Discovery section.

6. In the third paragraph of the section “Repression of Transposable Elements”, I would frame the function of RdDM something like this: RdDM primarily targets small TEs and TE fragments that are located near genes, and therefore these TEs are located in euchromatic regions of the genome. RdDM is necessary to constantly reinforce DNA methylation and transcriptional silencing of the TEs in this transcriptionally-permissive environment. RdDM is necessary to constantly redraw the boundary between heterochromatin (the TE) and the euchromatic neighboring genic region. RdDM constantly acts at this boundary to create this distinction, which is akin to drawing a line in the sand. The line will disappear if not constantly redrawn, which is the role of RdDM. Without RdDM, the genic euchromatin pushes into the TE, resulting in the expression of the TE. With too much RdDM, the heterochromatic marks push into the gene, and can result in its silencing.

We have made some major revisions to the section on TE silencing (now renamed “Transposable element silencing and genome stability”) to address this and other reviewer comments, and believe we now frame the role of RdDM in TE silencing as suggested.

7. I think it is worth mentioning that since HP1 has a distinct function in plants, RdDM is one of two known ways to spread H3K9me2 heterochromatin in plant genomes (the other being the loop between CMT3 and KYP).

We have added a paragraph on HP1 to the section now titled “Interaction between RdDM and other chromatin modifying pathways” (formerly “RdDM and other DNA methylation pathways - localization and interactions”) to address this point.

8. Please mention that RdDM targets all cytosine contexts equally, which is different from maintenance DNA methylation. Because maintenance methylation acts at a low level at CHH context DNA methylation (Arabidopsis) or not at all at CHH (maize), RdDM is typically assayed by bisulfite sequencing focusing on the CHH context.

In the “overview of RdDM mechanism” section, we have added a new section (“RdDM and DNA methylation context”) detailing why cytosine context matters for DNA methylation in plants, and which methyltransferases can add methylation to which contexts. We’ve also included a new figure to illustrate this (Figure 2).

9. In the “Development and Reproduction” section, perhaps describe the FWA example, as this was the classic example used to identify many of the factors, and effects flowering time.

We agree that this is a good example, and have added it to the section.

10. One KEY function of RdDM that needs to be elaborated: Repression of introduced transgenes! A lot of interest in RdDM comes from the desire to avoid transgene silencing to efficiently produce improved plants.

We agree that this is a critical point to address, and have included more discussion of this topic at two points in the article - first, at the beginning of the history and discovery section, and second, in the biological functions section, as a new subsection entitled “transgene silencing.”

11. Most of the field refers to RNAP IV as ‘Pol IV’. The authors may want to make this change. Same with ‘Pol V’.

We agree, and now use Pol IV and Pol V throughout the text.

12. In the section “Silencing of target loci by canonical RdDM”, Existing DNA methylation does not recruit Pol IV. Instead, H3K9me2 recruits Pol IV through SHH1. The authors should add H3K9 methylation to this section so it is clear to the reader that the DNA methylation of RdDM leads to histone methylation.

We have removed this over-simplified statement, and our revised manuscript includes a section dedicated to interactions between RdDM and histone modifications (“Interactions between RdDM and other chromatin modifying pathways”), where the role of histone modifications in recruiting RdDM is discussed in more detail. This interaction between Pol IV, SHH1, and histone modifications was also added to figure 3.

13. This sentence is not accurate: “The RdDM pathway is just one of several pathways involved in establishing and maintaining DNA methylation patterns in the plant.” RdDM is the ONLY pathway known to establish DNA methylation in plants. Other pathways can maintain.

We agree that our phrasing of this point was ambiguous. We have removed this statement from the article. Instead, in a new section on DNA methylation context (“RdDM and DNA methylation context”) where we briefly discuss the other maintenance methylation pathways, we also mention that RdDM is the only pathway capable of adding de novo methylation.

Review by Todd Blevins

Review of version as of 18:07, 13 October 2019: http://compbiolwiki.plos.org/w/index.php?title=RNA-directed_DNA_methylation_(RdDM)&oldid=8176 This topic page provides a very accessible overview of RNA-directed DNA methylation (RdDM) in plants. I congratulate the authors on collecting and curating the abundant literature on this subject. The community will benefit from this Wiki, a natural entry point for new students and scientists from other fields. To help revise this resource, I have several suggestions (points 1-6) and one vital correction (point 7).

1. I join the previous reviewer, Keith Slotkin, in recommending upgraded figures. In seeking to be schematic, the authors have omitted key protein-DNA, protein-protein and protein-RNA interactions that explain how RdDM works. Here is a non-exhaustive list of improvements that could be made to Figure 2:

1. One cannot understand RdDM outside of its chromatin context. Figure 2 should thus convey some basic notions about histone modifications. I recommend showing SHH1 reading repressive chromatin marks (unmethylated H3K4 + dimethylated H3K9) to recruit Pol IV to the silenced TE.

2. These RNA polymerases require nucleic acid templates. Pol IV should thus be shown transcribing chromosomal DNA to produce short primary transcripts, not floating below the DNA. Similarly, Pol II and Pol V act on chromosomal DNA. RDR2 could be shown transcribing a Pol IV transcript (see next item).

3. The Pol IV-RDR2 complex couples synthesis of primary transcripts (Pol IV products) to the synthesis of short dsRNAs (RDR2 products). These dsRNA products average ~30 bases in vivo, which is supported by Pol IV-RDR2 activities in vitro (Blevins et al. 2015; Zhai et al. 2015; Singh et al. 2019).

4. Dicers (DCL2, DCL3, DCL4) cleave dsRNA substrates into duplex siRNAs, not a pool of single-stranded siRNAs, as currently depicted in Figure 2.

5. AGO4-siRNA effectors are thought to find nascent RNAs in chromatin, tethering AGO4 to these Pol V transcripts and recruiting DRM2 to the DNA. The RNA scaffold would not be floating free, but should rather link AGO4 to Pol V in chromatin.

6. Using a line segment to join AGO1, AGO4 or AGO6 to each siRNA guide is an odd convention. I recommend showing each guide directly bound to its AGO.

:Our new version of this figure has incorporated all of these thoughtful and helpful suggestions.

2. Having a "History and discovery of RdDM" section is a good idea, but many key discoveries are omitted from this short account. The notion of siRNAs as putative guides for RdDM, particularly the 24 nt siRNA class, comes from a series of visionary papers including but not limited to: Mette el al. (2000) EMBO J, Hamilton et al. (2002) EMBO J, Zilberman et al. (2003) Science, and Xie et al. (2004) PLoS Biol. The intriguing sde4 allele from the silencing defective screen (Dalmay et al. 2000 Cell) would later be understood as an nrpd1 null mutant (pol IV largest subunit mutant).

This is a great point - we’ve incorporated these key studies into our expanded History and Discovery section.

3. The canonical RdDM pathway is indeed well-characterized, as the authors state. Because Pol IV and Pol V were critical to unravelling this mechanism, I recommend citing the four papers that reported discovering these plant-specific multisubunit RNA polymerases: Herr et al. (2005) Science, Onodera et al. (2005) Cell, Kanno et al. (2005) Nature Genet., and Pontier et al. (2005) Genes Dev.

We have added a paragraph to the history and discovery section specifically focused on Pol IV/V, and have added these citations.

4. From the above, the authors will note my preference for the Pol IV / Pol V rather than RNAP IV / RNAP V nomenclature. Wikipedia uses both forms, depending on the article and context, but Pol IV / Pol V are the consensus in the RdDM research community.

We have replaced RNAP IV/V with Pol IV/V throughout.

5. In the "Biological functions of RdDM", subsections "TEs" and "Stress response": the early work on EVADE and ONSEN retrotransposons should be highlighted: Mirouze et al. (2009) Nature; Ito et al. (2011) Nature. Moreover, there is a rich literature of follow-up experiments that continues to chart the interaction between RdDM pathways and these LTR/Copia elements in Arabidopsis. For example, Marí-Ordóñez et al. (2013) reconstructed how de novo silencing of EVADE elements could transition from 21 nt siRNA-mediated silencing into the canonical Pol IV-dependent, 24 nt siRNA mechanism. The authors could include this study in the TE subsection, rather than only in "Non-canonical RdDM".

We have added a new paragraph discussing these three studies and how TEs have helped uncover RdDM mechanisms to the section on transposon regulation by RdDM.

6. In the abstract, "The fullest complement of RdDM pathway components characterized to date is found within angiosperms ... Other groups of plants, such as gymnosperms and ferns, possess a subset of conserved RdDM components ..." Consider rephrasing or deleting this bit about gymnosperms and ferns. We are just beginning to understand the diversity of RdDM factors encoded in conifer and fern genomes. Genomic coverage in these gigantic reference assemblies is spotty, and some past surveys (circa 2015), seriously underestimated the conservation of core RdDM components. Absence of evidence is not evidence of absence.

Thank you for raising this point, we agree and have rephrased the sentence to avoid suggesting that conservation of RdDM components between angiosperms and other species is limited, since this is not well established.

7. The evolutionary analysis depicted in Figure 4 contains an error. RDR2 is conserved in Ferns, Lycophytes and Bryophytes, not only in seed plants. In Physcomitrella patens, the function of PpRDR2 was thoroughly tested in Coruh et al. 2015 Plant Cell. As expected, 24 nt siRNAs that accumulate in WT moss are absent in rdr2 mutants.

Thanks for calling this omission to our attention - we have modified the figure accordingly.

Review by Craig Pikaard

This is a well-written and ambitious article that covers a lot of ground and will be an excellent online resource at Wikipedia. Because this can be anticipated to be such a useful resource to newcomers to the field of RNA-directed DNA methylation, my major suggestion is that the authors should include more of the essential primary papers in the field instead of falling back, in many cases, to referring to review articles, none of which are comprehensive. There are also a few places where additional discussion is needed or where the statements are inaccurate and need to be modified. But overall, this is an excellent review article.

Specific comments 1. In the abstract in the first sentence, it would be best to say that RNA-directed DNA methylation is a biological process in which noncoding RNA molecules direct the addition of DNA methylation... This is because RdDM requires both small siRNAs and longer non-coding RNAs transcribed by Pol V.

This is a good point, and we have edited the abstract as suggested.

2. Also in the abstract, when discussing pathways that are similar in other organisms, fission yeast should definitely be included, as this is probably the closest analogous system that like RdDM requires RNA-dependent RNA polymerase activity.

This is a great suggestion. We now mention fungi in the abstract, and have added a description of TGS/CTGS in S. pombe within the “Relationships with sRNA silencing pathways in other kingdoms” section.

3. On page 2, in the history and discovery section, third to last sentence, the Baulcombe lab’s discovery of siRNAs (Hamilton and Baulcombe) should be cited..

We have added this citation to our expanded history and discovery section.

4. The second to last sentence of the history section should also be modified to include the fact that reverse genetic experiments played a major role in the discovery of Pols IV and V, not just genetic screens, as currently stated. In fact, the first evidence for the existence of these novel polymerases came from bioinformatic evidence of Pikaard and reported in the Nature paper that describe the Arabidopsis genome sequence. There is a discussion of the discovery, and naming, of Pols IV and V in a review article by Haag and Pikaard in 2011 in a Nature Reviews journal. The Pikaard lab, using reverse genetics, and the Baulcombe lab, using forward genetics, then independently co-discovered Pol IV’s involvement in DNA methylation, and they agreed on the name Pol IV. The Lagrange lab, using reverse genetics, and the Matzke lab, using forward genetics, next independently discovered the function of what they initially called Pol IVb, involving the second unusual large catalytic subunit that Pikaard had observed bioinformatically. Pol IVb was later renamed Pol V by Pikaard and colleagues (Ream et al, 2009) when they determined the complete subunit compositions of Pol IV, Pol V and Pol II, making it clear that the enzymes are structurally, as well as functionally, distinct and needing different names so that there subunits could be identified with a systematic nomenclature (NRPD--- or NRPE -----). The references to these important papers in the field should be included as Pols IV and V are arguably the key enzymes of the RdDM pathway, being responsible for the RNAs that drive the process.

This is a good point. We have expanded the history and discovery section and now include the role of non-screen-based experiments in the characterization of Pol IV and Pol V, and have added additional references to provide a more complete perspective.

5. On page 5, DCL2 and DCL3 should be defined when first mentioned. Alternatively, the previous mention of Dicer-Like activity could include (DCL) to introduce the nomenclature. There are other examples too where the authors should be check to make sure that abbreviations are explained.

We’ve gone through and defined abbreviations at the first mention throughout the revised manuscript.

6. The biotechnology application section is in the wrong place, in my opinion, and should be moved to near the end of the article, as it is logically an extension of how the basic science understanding of the pathway can be exploited.

We’ve moved this section to the closing of the article, and agree that this change enables us to add specifics to this section that refer to details introduced in the remainder of the text.

7. At the top of page 7 where Pols IV and V are discussed, the original papers in which the enzymes were first named should be cited (Onodera et al,2005; Herr et al,2005; Ream et al., 2009; Wierzbicki et al, 2008)

We’ve added references to these papers in appropriate points in the text (mainly in the relevant portion of the history and discovery section).

8. On page 8, In the section that discusses production of sRNAs, there is a new 2019 paper by Singh et al in Molecular Cell that demonstrates that Pol IV, RDR2 and DCL3 are all that is needed for siRNA synthesis, and they show biochemically how the enzymes work. This would be good to cite here. The model of Figure 2 is also misleading by showing Pol IV and RDR2 as separate entities. In fact, they are physically associated and work in a coupled reaction. The authors should look at the review by Wendte et al in 2017 for a more up-to-date model.

We have added this citation and have updated our figures and text throughout the article to reflect that Pol IV and RDR2 are physically associated.

9. On page 8, in the section on production of sRNAs involved in noncanonical RdDM , the first sentence is a bit misleading in that RDR2 is needed in addition to Pol IV for canonical 24 nt siRNAs, so there are two enzymatic sources of the siRNA precursors, not one. On the other hand, the Singh et al paper mentioned in point #8 above shows that Pol IV and RDR2 act in a coupled reaction, so the Pol IV-RDR2 physical complex might be considered as a single entity that does several reactions.

We’ve tried to clarify this point in the text.

10. In the Discussions of the DDR complex, the involvement of two of the three components of that complex in Pol V transcription was shown by Wierzbicki et al (2208, 2009), prior to the Jacobsen lab showing that they are together in the same complex. This should be cited appropriately.

We have added the suggested citations to the section discussing the DDR complex.

11. Page 12, near the top, in the discussion of how Pol IV is recruited, Blevins et al (2014) showed that Histone Deacetylase 6 is also required for Pol IV recruitment and for the epigenetic inheritance of the Pol IV recruitment signal, due to its involvement in maintenance methylation by MET1. It is likely the inherited DNA methylation state that gives rise to the histone modification state that is recognized by SHH1. This should be included and cited again in second paragraph of page 13, and in the first sentence of the 4th paragraph of page 13.

HDA6’s role is now fully detailed within Table 1, along with the suggested citation; we also briefly mention the role of HDA6 and MET1 in the new “Interactions between RdDM and other chromatin modifying pathways” section.

12. On page 12, in the factors affecting RNAP V and DRM2 targeting, the authors incorrectly attribute the hypothesis that DDR unwinds DNA to facilitate Pol V transcription to reference 64. In fact, this hypothesis comes from a 2013 Pikaard review article published in the Cold Spring Harbor Symposiums in Quantitative Biology series.

Thank you for pointing this out. We have updated our attribution accordingly.

13. Page 13, third paragraph, the author neglected to note that DDM1 facilitates CG methylation by MET1. The original papers from the Richards lab on the discovery of DDM1 should be cited.

In the renamed section ‘Interactions between RdDM and other chromatin modifying pathways’, we now note DDM1’s connection to methylation and MET1. Several of the Richards lab papers are now cited in this section.

14. In the discussion of ROS1, ROS1 expression has also been shown to be regulated by condensin (Wang et al, 2017), not just DNA methylation.

We have added a statement mentioning other mechanisms involved in ROS1 regulation beyond DNA methylation.

15. Page 14, the Ream et al paper (2009) is what firmly established the evolutionary relatedness of Pols IV and V to Pol II. The authors should again cite original papers, wherever possible, rather than reviews.

We have added the suggested original reference for this point.

16. In that same section, it is stated that Pols IV and V have “around 12 subunits”. No, they have (at least) 12 subunits, as shown by Ream et al.

We have clarified this statement, now stating “at least 12” subunits per polymerase, and linked this to the suggested reference.

17. In discussing the evolution of Pol IV and Pol V subunits, papers by Luo and Hall and Tucker et al are not cited but preceded the paper from Mosher and colleagues that is cited.

The suggested references have been integrated into the Pol IV/V evolution section.

18. In the figure legend to Figure 3, the reference to the subunit nomenclature should be cited: again, this is Ream et al, 2009.

The suggested reference has been added to the figure legend.

19. The final paragraph, about miRNAs is really not relevant to RdDM and I suggest that the authors delete it.

We agree that this paragraph was tangential to the article’s focus. We removed some unnecessary detail regarding miRNAs and merged the limited remainder with the piRNA paragraph.

20. One thing that is missing is mention of the discovery/identification of the RNA transcripts that made by Pol IV and Pol V. The Pol V transcripts were first identified in 2008 by Wierzbicki et al. The Pol IV and RDR2-dependent transcripts were not identified until 2015, by the Chen, Jacobsen and Pikaard labs.

Information about the nature of the Pol IV and Pol V transcripts is found in the “Canonical RdDM” section of the article, complete with the references noted here.

In summary, this is an excellent article and I hope these suggestions will help improve it and allow readers to find more of the original papers that have established current thinking in the field.

Summary of changes

Add topic