- Mar 2019
MIWI2 targets RNA Transcribed from piRNA-dependent Regions to drive DNA Methylation in Mouse Prospermatogonia
Toshiaki Watanabe, Xiekui Cui, Zhongyu Yuan, Hongying Qi, and Haifan Lin.
Review timeline: Submission date: 17th October 2016<br> Editorial Decision: 21st November 2016<br> Revision received: 2nd May 2018<br> Editorial Decision: 25th June 2018<br> Revision received: 6th July 2018<br> Accepted: 9th July 2018
Editor: Anne Nielsen.
(Note: With the exception of the correction of typographical or spelling errors that could be a source of ambiguity, letters and reports are not edited. The original formatting of letters and referee reports may not be reflected in this compilation.)
1st Editorial Decision 21st November 2016 Thank you for submitting your manuscript for consideration by the EMBO Journal. It has now been seen by three referees whose comments are shown below.
As you will see from the reports, our three referees all express interest in the findings reported in your manuscript, although they also raise a number of concerns that you will have to address before they can support publication of the manuscript here. In particular, refs #2 and #3 both point out that the current data does not unequivocally demonstrate that the RNAs isolated via the MIWI2 IP are bona fide target RNAs. I would therefore encourage you to include additional experimental data to strengthen this part of the study (especially since this will also go a long way to address the overall concern from ref # that the current manuscript is in parts confirmatory of earlier studies). Finally, the referees all point to a number of minor issues and controls that will also need to be addressed in full.
Given the overall positive recommendations from our three referees, I would like to invite you to submit a revised version of the manuscript, addressing the comments of all three reviewers. I should add that it is EMBO Journal policy to allow only a single round of revision, and acceptance of your manuscript will therefore depend on the completeness of your responses in this revised version.
Comments for Watanabe and Lin Piwi-interacting RNAs (piRNAs) arise from genomic regions called piRNA clusters as a long transcript that is then processed into ~26-30 nt piRNAs in the mouse male germline. After cytoplasmic loading into a PIWI protein (MIWI2), the complex translocates into the nucleus where it is proposed to target genomic regions for silencing by recruitment of silent histone or DNA methylation marks. How this interaction between nuclear piRNA-guided MIWI2 and the target genomic locus takes place is currently unknown.
- The authors repeatedly state ("perhaps some MILI molecules") that MILI is also nuclear. This is not supported by any data in this study or in the literature (apart from an extrapolation of analysis of whole genome bisulfite data).
- Previous studies examined the role of active piRNA pathway in establishing DNA methylation on genomic loci in mouse male germ cells (Molaro et al. 2014; Pezic et al. 2014; Manakov et al. 2015; Nagamori et al. 2015). The authors repeated the study, but examined an earlier time-point (0-2 days), instead of the 10-day germ cells (as done before). They reach similar conclusions to identify hypomethylated regions (HMR) that depend on the piRNA pathway.
- Figure1-2: A link between presence of DNA methylation on such genomic loci and active transcription of the loci has been presented (Molaro et al. 2014). The authors find a similar correlation between active transcription of the HMRs and DNA methylation. The authors additionally profile the chromatin status in microscopic, embryonic testes to describe that these HMRs have chromatin status consistent with actively transcribed regions.
- Are the HMR-L1 transcripts polyadenylated? The authors suggest that they are exported to be cleaved by MILI and processed into piRNAs.
- Figure 3: They purify EGFP-MIWI2 and show that there is a clear enrichment of transcripts arising from these HMRs (HMR-L1). This was always expected in the field, but now demonstrated.
- Figure 4 and Figure 5: Perhaps the most important data in this manuscript. These figures demonstrate using different mouse lines, a direct relationship between piRNA guides and target DNA methylation. When the binding site on the target locus is removed or when the piRNA locus is deleted (promoter) or guide-generating sequences are removed, there is an impact on DNA methylation of the target. I am wondering how repetitive is this target site and the guide piRNA? The level of DNA methylation changes is dramatic. So are there no other repeat piRNAs from RMER4B that can compensate for loss of these guides form the chromosome 7 cluster? The authors should show the read-counts of RMER4B repeat piRNAs in wt and cluster deletion mice. Reads with perfect match and with varying mismatches should be plotted.
- In the discussion, the authors mention that DNA methylation that is lost is regained in the next generation (12.5 dpc embryos). I think it is an interesting observation and should be put in the manuscript. Presently it is unpublished data.
- Did the authors check DNA methylation in post-natal germ cells of the cluster deletion mice? It would be interesting to know how dynamic DNA methylation is on the target site within the same mouse.
Overall, I am supportive of this study which provides many direct links between piRNA and DNA methylation of target sites in the mammalian germline.
Piwi-interacting RNAs (piRNAs) play a critical role in transcriptional silencing of transposons in mice. MIWI2 is the a predominantly nuclear Piwi protein implicated in DNA methylation of piRNA genomic targets. This study examines the relationship between MIWI2, piRNAs and expression of their genomic targets. Experiments show expression of piRNA genomic target loci at the time of DNA methylation and MIWI2 interaction with RNAs transcribed from those. Finally, authors test for the requirement of piRNA or its genomic target sequence for de novo DNA methylation. Based on these data, the authors propose that MIWI2-piRNA effector complex guides de novo DNA methylation via piRNA:RNA base pairing.
Authors build their case gradually by:
- identifying hypomethylated regions (HMRs) in Miwi2 KO "gonocytes" (see #3 below on the term usage);
- demonstrating active transcriptional state of HMRs;
- demonstrating the enrichment of piRNA target RNAs in MIWI2 immunoprecipitates;
- providing evidence of the production of piRNAs targeting HMR RNAs as well as piRNA production from same RNAs;
- finely dissecting piRNA-target RNA requirements for de novo methylation of the imprinted Rasgrf1 locus.
Overall, this is an interesting study coming from researches with a proven track of breakthrough studies in the piRNA field. In principle, there are no major issues with experimental designs, primary data, data interpretation, conclusions and claims. There are few issues that should be addressed to improve the manuscript:
The least convincing aspect in the paper is that of MIWI2 interaction with piRNA target RNAs. Here is why: A. MIWI2 is not just nuclear, it also localizes to a cytoplasmic germinal granule, a derivative of the P-body (Aravin, PLOS Gen 2009). Thus, it is not immediately clear if the detected RNA associations are reflective of nuclear (as the authors want to believe) and/or cytoplasmic interactions. In this regard, it would be useful to examine MIWI2 associated "nascent" RNAs for the presence of introns. If no intron-containing HMR RNAs are enriched in MIWI2 IPs, an artificial intron-containing RNA should be generated by modifying an associated RNA. Subsequent sequence analysis of intron splicing should clarify nuclear vs cytoplasmic origin of the RNA. B. The MIWI2 IP data are presented as those implying direct interaction of MIWI2-piRNA effector complexes with their target RNAs. But, as we know, IPed MIWI2 complexes are quite heterogeneous and contain dozens of proteins many of which should bind RNA (Vagin, GD, 2009). Therefore, authors must a) acknowledge this shortcoming of their approach and b) re-write this section accordingly. C. A critical experiment is missing - MIWI2 IP and RNA-seq (long for RNAs and short for piRNAs) from Mili KO p0-p2. This will show which of MIWI2 associated RNAs are present in this complex in a piRNA-dependent manner.
A key message in this manuscript is that of direct piRNA: nascent RNA pairing as the guide of DNA methylation. And yet, the presented data are insufficient to unequivocally make that conclusion. Importantly, they do not rule out a possibility of piRNA:DNA interactions a la what is observed in the case of Myc regulation and few other cases. While this seems a remote possibility at a first approximation, it cannot be excluded simply because we do not find it attractive or because previous studies from other species support the RNA:RNA model. Authors state in Discussion that this claim is based on three key findings: A. RNA expression; B. MIWI2 interaction with RNA and C. Rasgfr1 locus regulation. A. With respect to A. authors state on page 6 that "... positive correlation between the extent of demethylation in Mili -/- spermatogonia and RNA expression level at the stage of piRNA-mediated DNA methylation among individual L1 sequences, supporting the view that piRNA mediated epigenetic regulation requires RNA expression". In fact, this results only shows that in the absence of a functionally active piRNA pathway its target loci are derepressed. The authors even themselves recognize the shortcoming of this argument since they acknowledge ERV RNAs are not upregulated despite demethylation in the Mili KO. That brings us to the crux of the argument that expression of a target locus is not sufficient to claim RNA:RNA interactions. It could be simply a reflection of open/derepressed chromatin of the locus. B. As already addressed above, MIWI2 IP plus RNA-seq data are not sufficiently stringent to claim piRNA-mediated association of MIWI2 with nascent RNAs. C. The section of the paper describing analysis of DNA methylation at Rasgrf1 DMR is very strong. It clearly demonstrates the significance of a specific piRNA and its target sequences for DNA methylation. But do these data unequivocally tell us if this target is RNA or DNA? They don't. Not only they don't tell us that it is RNA but in actuality they support the idea of RNA:DNA interaction. Specifically, deletion of the promoter and abrogation of pit-RNA expression has almost no effect on DNA methylation! The authors quickly dismiss the lack of effect on DNA methylation as non significant since they have a much stronger data using a specific deletion in the piRNA sequence in the genome. But this latter experiment only shows that complementarity is important but does not distinguish the two alternative modes of interaction of piRNA with its targets (RNA vs DNA). The authors are strongly encouraged to consider their existing data in light of this alternative model. This could be done in writing and wouldn't require any additional wet lab efforts.
Authors should use the correct nomenclature of male germ cells. "Gonocyte" is a misnomer. Please see McCarrey JR, BOR, 2013 for further explanation.
"piRNA-dependently methylated regions" is a real mouthful. Why not "piRNA-dependent regions"? It will be pretty clear from the context of the paper what authors mean by that.
The manuscript by Watanabe et al. addresses the issue of DNA methylation that is primed by piRNAs in mice. In particular, the MIWI2 protein is involved in that, but MIWI2 can only be loaded with piRNAs though activity of another piwi protein, MILI. There have been a number of papers describing the impact of piRNAs and MIWI2 on DNA methylation, but in these cases spermatocytes were used. In the current study, gonocytes from just born pups are used. The authors argue these represent a better point of view since they are developmentally closer to the time-window when MIWI2 is actually expressed (around E14.5). Shortly after MIWI2 is expressed the PGCs stop cycling until just after birth, which is when the current study analyses DNA methylation.
Conceptually, one can argue that the study by Watanabe and Lin does not reveal many novel features. Basically, their results confirm the previous studies. On the other hand, it is important that the MIWI2 effect on DNA methylation is studied as close to its actual expression window as possible. In that light, the current study does have its value. In addition, the results described on the imprinted locus are quite elegant and interesting, and the description of chromatin profiles of the piRNA-affected loci is useful as well. It would have been great if the authors would have shown, as they claim, that MIWI2 acts (as we expect, but is still not proven) on nascent RNA. However, the current manuscript does not prove that, and hence my overall feeling is that the novelty brought by the manuscript may not be at the level for EMBO Journal.
Below I bring forward more specific issues that require attention by the authors.
-coverage BSseq: the authors document a 17-fold average coverage. However, this does not reflect true average but is heavily influenced by a small number of heavily covered regions: >60% is covered between only 1-10 times. (table 1).
-A scatter plot of methylation frequencies across the genome, comparing wild-type and mili mutant gonocytes would be good to have. What is shown now is only the HMRs. A complete view would be more informative.
-A direct comparison between spermatocyte and gonocyte HMRs would also be very useful, as this would directly visualize the added value of the gonocyte data.
-page 5: "..contrary to our expectation..."; why is this so? I personally prefer a less opinionated way of writing. Just write what you find. If the gonocyte data is the same as the spermatocyte data, so be it.
-Please present a systematic comparison of HMR+L1 and HMR-L1 loci across the manuscript, using the same visualization. The manuscript focuses on the -L1 HMRs, but I do not see why the +L1 loci would be less interesting. Both contain repeat elements known to be affected by piRNAs anyway.
-page 6: 'Hence, it appears that more HMR_-L1s actually express RNAs". Why 'more'. More than what? Also, the last sentence is an overstatement: based on mere presence of RNA one cannot draw conclusions on active transcription. RNA could also be a remains of earlier events.
-page 6 'presence of L1 promoter regions': this is purely based on annotated promoter activity. Why not also show the chromatin profiles for the +L1 loci (as they do for the -L1 loci); this would allow the authors to indeed make statements like this.
-For the RIP experiments a MIWI2 IP from a mili mutant would be a better control than oct4 I think. Why has the oct4 been chosen as control?
-Regarding the RIP results: the RNAs coming down with MIWI2 might well be in the process of being turned into piRNAs, and may not be the nascent transcripts that authors take them to be. How do the 5' ends of these RNAs align with mature piRNAs from that stage? Importantly, the authors recognize the fact that their claim to seeing nascent RNA in MIWI2 IPs is weak, as they write on page 8/9: "these results are nevertheless consistent with the notion that RNAs transcribed from HMRs are targeted by piRNAs." True, but they are equally consistent with the notion that they may represent primary piRNA precursors that have not seen targeting by any Piwi protein. Based on these experiments the authors simply cannot make statements on the nature of the transcripts detected in the RIP.
-Figure 3: Since HMRs have many repeats (L1's, or LRTs in case of HMR-_L1s) the finding that they produce piRNAs is rather trivial and does not bring much novelty.
-The effects of RMER4B on Rasgrf1 are intriguing. This part of the manuscript brings novelty. Still, it is a pitty that the authors do not follow the methylation status longer during development, to see when it gets back to normal (as it seems to based on a comment in the discussion). At the same time: the fact that methylation does get back to normal in the new (elegant!) system does limit the impact of piRNAs on imprinting of course. This should be more clearly indicated.
Minor issues: -HMRs are only hypo-methylated in mili mutants, so they are technically not HMRs in wild-type animals.
-Figure 5D: why is only region E analysed? It is anyway difficult to compare panels D/E versus A. That's a pitty.
-5C: how does the piRNA coverage of the chr 7 cluster in the 4b deletion look like?
-Figure S5 is not cited
-Figure S6: what is exactly presented on the Y-axis? Enrichment against what? And how can we assess coverage? On how many data points/observations is the enrichment based?
-please clearly indicate what stages of development are analyzed in each of the panels of figure 4 and 5.
-Figure 1: The authors write that 55k genes are expressed in gonocytes. Is that realistic? I assume they consider splice isoforms and the like as different genes, because to my knowledge there are not that many genes in the mouse genome. Please clarify.
-Figure 3: enrichment should not be calculated against Ensemble mRNAs but against gonocyte-expressed mRNAs.
-Figure S1: how many CpGs have >10 fold coverage in both replicates (and also in both genotypes?)?
-The effect of the RMER4B deletion in the chr7 piRNA cluster is still significantly weaker in region E compared to the Rasgfr1-RMER4B deletion mutant, while the authors write "...very similar to those observed in mili and rasgrf1 mice..." (page 12, just before discussion). Please rephrase more carefully.