Binding of guide piRNA triggers methylation of the unstructured N-terminal region of Aub leading to assembly of the piRNA amplification complex

PIWI proteins use guide piRNAs to repress selfish genomic elements, protecting the genomic integrity of gametes and ensuring the fertility of animal species. Efficient transposon repression depends on amplification of piRNA guides in the ping-pong cycle, which in Drosophila entails tight cooperation between two PIWI proteins, Aub and Ago3. Here we show that post-translational modification, symmetric dimethylarginine (sDMA), of Aub is essential for piRNA biogenesis, transposon silencing and fertility. Methylation is triggered by loading of a piRNA guide into Aub, which exposes its unstructured N-terminal region to the PRMT5 methylosome complex. Thus, sDMA modification is a signal that Aub is loaded with piRNA guide. Amplification of piRNA in the ping-pong cycle requires assembly of a tertiary complex scaffolded by Krimper, which simultaneously binds the N-terminal regions of Aub and Ago3. To promote generation of new piRNA, Krimper uses its two Tudor domains to bind Aub and Ago3 in opposite modification and piRNA-loading states. Our results reveal that post-translational modifications in unstructured regions of PIWI proteins and their binding by Tudor domains that are capable of discriminating between modification states is essential for piRNA biogenesis and silencing.

methylosome complex in Fig 5 in piRNA-free S2 cells, given that piRNA-loading is essential for sDMA of Aub. 3. I am wondering if loading of piRNA precursor could also induce the conformational change and sDMA methylation of Aub. The PAZmut mutant, which is deficient in mature piRNA loading but could bind precursors (EMBO Reports. 2020), could be a good one to test this. 4. In Fig 3a and 4a, the author should include the IB panels to show whether the mutations alter Aub expression in flies. Minor points 1. Overall, the labelling in western blot panels should be improved. 2. In Fig 5d, the labeling of lanes 3 and 4 should be be "md" rather than "rd"?
Reviewer #3 (Remarks to the Author): In this manuscript, Huang and Hu et al. show that the two Tudor domains of Krimper (Tud1 and Tud2) bind unmodified Ago3 and sDMA-modified Aub, respectively, and that the sDMA modification of Aub is regulated by loading of a piRNA guide, thereby dynamically assembling the piRNA amplification complex. In general, the experiments are well executed and the data support the conclusion. The following points should be addressed before publication.
Major points: 1. The small RNA seq data in Fig. 3f should be more thoroughly analyzed. 1) According to the result in Fig 3g, the authors should not simply focus on the top 20 most abundant TE families but should appropriately categorize the piRNA clusters in their analysis. 2) Rather than just focusing on sense & 10A & non-1U piRNAs, the data for antisense piRNAs should also be presented and the 1U/10A bias should be more rigorously analyzed.
3) The ping-pong pair factions of some transposons seem to be well rescued by mdAub. How do the authors explain this observation? What is the common feature(s) among them (e.g., more dependent on homotypic ping-pong)?
Minor points: 1. The authors have previously shown that Aub and Ago3 interact with both Tud1 and Tud2 (2015 Mol Cell, 59, 564-575). The apparent discrepancy should be more carefully explained. Specifically, does the Aub peptide fragment containing R26 bind to Tud1? 2. The authors should specify the tools used in their bioinformatic analyses and also how small RNA seq data were normalized. 3. In Fig 3i upper panel, the wtAub data should be presented on the right of the mdAub data. 4. In Fig 4g, h, sDMA modification is detected in the absence of 26-nt ssRNA. Does this mean that FLAG-Aub is partially loaded with endogenous RNA fragments in S2 cells? 5. In Fig. 5d: "rd" should read "md". 6. Exd Data Fig. 5a requires molecular markers. 7. The order of Figs and Exd Data Figs should be amended so that they are sequentially cited in the text. 8. The authors should discriminate PIWI (subfamily name) and Piwi (Drosophila protein name) throughout the text. 9. Page 3, line 7: "Argonautes bind nucleic acids guides" should read "Argonautes bind nucleic acid guides". 10. Page 3, line 6 from the bottom: "produces symmetrically methylated arginine (sDMA) residues" should read "produces symmetrically dimethylated arginine (sDMA) residues". 11. Page 5, line 1: "Krimper that lacks the N-terminal self-interaction region (residues 1-301aa)" should probably read "residues 1-300 aa" (in Fig. 1c, ΔN-Krimp is shown as 301-746). 12. Page 10, line 2 from the bottom: the underlines in "piRNA binding" make the abbreviation "pbAub", not "pdAub".
13. The discussion part is reasonable but rather lengthy.
We are very grateful to all three reviewers for their valuable criticism and suggestions, which we believe have greatly contributed to improving the manuscript. We performed new experiments and analyses as well as implemented changes in the text of the paper to address the comments. Below, we provide our point-by-point response (shown in blue) to the comments of each reviewer.

Reviewer #1
In this manuscript, Huang X et al reported that Krimper uses its N-terminal eTudor domain (eTud1) to bind Ago3 in a methylation-independent manner and methylation of the target arginine residues will disrupt binding, and use the eTud2 domain to recognize the methylated N-terminus of Aub and this binding depends on sDMA of the target arginine resideus. They further show that sDMA modification of Aub is required for piRNA biogenesis through the ping-pong cycle but is dispensable for loading of piRNA into Aub. RNA binding promotes Aub arginine methylation, and triggers conformational change in Aub, exposing its N terminus to the methyltransferase Csul/Vls complex. This is a very interesting piece of work, which represents some exciting advance in the field.
Overall, this paper is well written, and the data are of solid quality. However, I have some concerns, which should be addressed prior to publication.
1.What is the structural explanation why eTud1 cannot bind to unmethylated Aub-1 peptide now that it also contains a similar GRGRAR motif found in Ago3-2? By the way, in Fig. 1e, the Aub peptide is labelled as Aub-1, but as aub in Although the overall topology of eTud1 is similar to the other extended Tudor domains, eTud1 has its own unique folding. It forms both a hydrophilic and a hydrophobic concave region in the cleft between the Tudor and SN-like subdomains of eTud1 (Fig. 2h). The AGO3-2 peptide forms an a-helical structure which fits into the narrow binding cleft of eTud1. The N-terminal AUB (NH2-NPVIARGRGRGRK-COOH), sequence with an additional "GR" motif appears to be more hydrophilic than AGO3-2 and may not readily fit into the narrow hydrophobic concave cleft of eTud1. We have added this comparison in the text of the revised manuscript.
We thanked the reviewer for pointing out the typo and we corrected it as suggested (Fig. 1f).
2.Why does eTud1 not bind Ago3-1 at all? Does it mean that more than 1 arginine residue in the motif is required in binding?
As stated in the Q1, unlike Ago3-2, Ago3-1 with only one "GR" motif appears to be unable to form strong interactions with eTud1. We have added this comparison in the main text.
3.eTub2 binds tighter to methylated Ago3-2 than to methylated Aub, is there any biological significance, such as methylated Ago3-2 can compete out Aub to release Aub 5. Although the manuscript is well written, but there are still many grammar errors, which need to be fixed as well.
We thanke the reviewer for pointing out grammar mistakes and we corrected them as suggested.

Reviewer #2 (Remarks to the Author):
The symmetric dimethylarginine (sDMA) modification is widely present in unstructured N-terminal region of Piwi proteins, which is known critical for Piwi binding to Tudor domain-containing proteins. Nevertheless, the regulatory function of such modification in piRNA pathway has not been fully understood. We thank the reviewer for this suggestion. As suggested, we have performed in vitro assays to compare slicer activities of the wild-type and mdAub proteins.
The results show that mdAub exhibits a normal slicer activity similar to wildtype Aub. We added this result to revised manuscript (Extended Data Fig. 4d experiments from other groups demonstrate that Aub can bind any 5'phosphorylated RNA of proper length and use it as a guide in vitro or in cell extract. Therefore, it is plausible that Aub finds opportunities to bind cellular RNA present in S2 cells, such as tRNA fragments known to be bound by PIWI proteins in various systems. To explore if Aub is loaded with guide RNA in S2 cells, we labeled nucleic acids associated with purified protein and found that a small amount of RNAs of various sizes is indeed bound by Aub (data not shown). Thus, Aub binding of RNA in S2 cells might promote its methylation even in the absence of genuine piRNA. This conclusion is further supported by the observation that sDMA modification is suppressed in the Aub mutant deficient in RNA binding (pdAub) if expressed in S2 cells (Fig. 6b). However, it should be noted that pdAub expressed in germ or S2 cells still has miniscule but detectable levels of sDMA modification (Fig. 5a, 6b, compare pdAub with mdAub), indicating that the correlation between piRNA binding and methylation is not absolute. We also observed that insertion of a heterologous sequence between the N-terminus and the rest of Aub renders it a good substarte for sDMA modification even if it does not bind RNA (Fig 6a). Furthermore, we generated new Aub PAZ mutant lacking the 3'-end piRNA binding residues and found that it failed to be modified, further confirming the role of guide binding in promoting sDMA modification (Fig. 5i). Overall, the results from several experimental systems together indicate that RNA loading strongly promotes sDMA modification, however, while binding of genuine piRNAs induce modification in germ cells, other RNAs can substitute for genuine piRNAs leading to modification in heterologous systems.
3. I am wondering if loading of piRNA precursor could also induce the conformational change and sDMA methylation of Aub. The PAZmut mutant, which is deficient in mature piRNA loading but could bind precursors (EMBO Reports. 2020), could be a good one to test this.
We are thankful for suggesting this experiment that allowed us to further test requirements for sDMA modification. We expressed Aub PAZmut mutant, which is deficient in 3' end binding but could bind 5' end of precursors and tested its methylation level upon loading with synthetic 26nt ssRNA. We found that unlike the wild-type protein, Aub PAZmut has miniscule level of sDMA modification (Fig 5i), indicating that complete binding of the guide RNAboth its 5' and 3'endis required to induce modification. We added this result to the Results and Discussion sections of the revised manuscript.
4. In Fig 3a and 4a, the author should include the IB panels to show whether the mutations alter Aub expression in flies.
We have added panels to show the amount of wild-type, mdAub and pdAub Aub in ovarian lysates to the new Extended Data Fig. 4a. mdAub is expressed at a level similar to the wild-type protein, while expression of pdAub, which lacks piRNA-binding capacity is diminished.
Minor points 1. Overall, the labelling in western blot panels should be improved.
We have re-labelled western blot panels to clarify source of material, antibodies used for IP and Western. Fig 5d, the labeling of lanes 3 and 4 should be be "md" rather than "rd"?

In
Thank you for noticing, we have corrected this mistake.

Reviewer #3 (Remarks to the Author):
In this manuscript, Huang and Hu et al. show that the two Tudor domains of Krimper (Tud1 and Tud2) bind unmodified Ago3 and sDMA-modified Aub, respectively, and that the sDMA modification of Aub is regulated by loading of a piRNA guide, thereby dynamically assembling the piRNA amplification complex. In general, the experiments are well executed and the data support the conclusion. The following points should be addressed before publication.
Major points: 1. The small RNA seq data in Fig. 3f should be more thoroughly analyzed. 1) According to the result in Fig 3g,  3) While performing piRNA analysis for the revision, we have found a mistake in processing of data for Fig. 3f (Fig. 4b in Fig. 1b).
On the other hand, there is no discrepancy between our new findings and previous results regarding interaction between Ago3 and Krimp. Previously, we showed that in S2 cells Ago3 co-IPs with both N+Tud1 and Tud2 fragments (though strongest interaction was observed for the fragement containing both Tud1+Tud2 domains). The new findings also demonstrate that sDMA-modified, but not unmodified Ago3 peptides bind Tud2 in vitro (Extended Data Fig.1a).
However, as was shown previously, in germ cells Krimp complexes contain unmodified Ago3, which according to our data can only bind to Tud1 domain.
We discuss these results and possible reasons for the difference between the in vitro and in vivo results in the manuscript.
2. The authors should specify the tools used in their bioinformatic analyses and also how small RNA seq data were normalized.
We have added the information about bioinformatic tools and RNA-seq data normalization in the Material and Method section.
3. In Fig 3i upper panel, the wtAub data should be presented on the right of the mdAub data.
We have fixed the presentation of data on Fig. 3i (new Fig. 4g) as suggested.
4. In Fig 4g, h, sDMA modification is detected in the absence of 26-nt ssRNA.
Does this mean that FLAG-Aub is partially loaded with endogenous RNA fragments in S2 cells?
The similar question was also raised by reviewer #2 (point 2), so we provide a combined answer: There are two non-mutually exclusive explanantions for these results: (a) a fraction of Aub molecules expressed in S2 cells might be loaded with cellular RNA (b) a fraction of Aub molecules might be modified by the methylosome complex even if they lack RNA guides due to random expose of their N-termini to the methylosome complex. Our results (Extended Data Fig. 4d) and experiments from other groups demonstrate that Aub can bind any 5'phosphorylated RNA of proper length and use it as a guide in vitro or in cell extract. Therefore, it is plausible that Aub finds opportunities to bind cellular RNA present in S2 cells, such as tRNA fragments known to be bound by PIWI proteins in various systems. To explore if Aub is loaded with guide RNA in S2 cells, we labeled nucleic acids associated with purified protein and found that a small amount of RNAs of various sizes is indeed bound by Aub (data not shown). Thus, Aub binding of RNA in S2 cells might promote its methylation even in the absence of genuine piRNA. This conclusion is further supported by the observation that sDMA modification is suppressed in the Aub mutant deficient in RNA binding (pdAub) if expressed in S2 cells (Fig. 6b). However, it should be noted that pdAub expressed in germ or S2 cells still has miniscule but detectable levels of sDMA modification (Fig. 5a, 6b, compare pdAub with mdAub), indicating that the correlation between piRNA binding and methylation is not absolute. We also observed that insertion of a heterologous sequence between the N-terminus and the rest of Aub renders it a good substarte for sDMA modification even if it does not bind RNA (Fig 6a). Furthermore, we generated new Aub PAZ mutant lacking the 3'-end piRNA binding residues and found that it failed to be modified, further confirming the role of guide binding in promoting sDMA modification (Fig. 5i). Overall, the results from several experimental systems together indicate that RNA loading strongly promotes sDMA modification, however, while binding of genuine piRNAs induce modification in germ cells, other RNAs can substitute for genuine piRNAs leading to modification in heterologous systems.
Thank you, we have corrected the error.
We have modified the figure to show molecular markers (updated Extended Fig.   6a).
7. The order of Figs and Exd Data Figs should be amended so that they are sequentially cited in the text.
We have corrected the figures to align panel order with the text.
8. The authors should discriminate PIWI (subfamily name) and Piwi (Drosophila protein name) throughout the text.
We have chaged the text throughout as suggested to discriminate subfamily and protein names. 9. Page 3, line 7: "Argonautes bind nucleic acids guides" should read "Argonautes bind nucleic acid guides".
Thank you for noticing, we have corrected the text as suggested.
Thank you for noticing, we have corrected the text as suggested.
Thank you for noticing, we have corrected the text as suggested.
Thank you for pointing this out, we have corrected it as suggested.
13. The discussion part is reasonable but rather lengthy.
We are very grateful to all three reviewers for their valuable criticism and suggestions, which we believe have greatly contributed to improving the manuscript. Below, we provide our point-by-point response (shown in blue) to the comments of each reviewer.
Reviewer #2 (Remarks to the Author): Most of concerns have been adequately addressed by additional data in revised version of manuscript.
Minor point: The labelling in western blot panels, including molecular markers, should be improved.
Molecular weight for all western blots were added as suggested.
Reviewer #3 (Remarks to the Author): The authors have adequately addressed most of my previous concerns. The following minor points should be fixed before publication. 1) The main text does not contain any reference to Fig 4c. I assume that "Unlike the wildtype protein, expression of mdAub fails to rescue piRNA generation throughout the genome indicating that sDMA is crucial for piRNA biogenesis (Fig 4b and 4d)." in page 10 is the corresponding part.
We apologized for this mistake. Figure references were added in the text. "Almost all regions that generate piRNAs with notable exception of uni-strand piRNA clusters such as flam and 20A show dramatic loss of piRNA in the aub HN/QC mutant (Fig. 4c). Unlike the wild-type protein, expression of mdAub fails to rescue piRNA generation throughout the genome indicating that sDMA is crucial for piRNA biogenesis (Fig. 4b, 4c and 4d)." 2) The legend for Fig 4c says "Genomic windows with more than 5 RPM in libraries from heterozygous ovaries and more than 80% reduction in aub mutant were selected for further analysis." If this statement is correct, the authors have specifically selected genomic windows with >80% reduction of piRNAs, and then claimed that piRNAs from those regions are greatly reduced (which is too natural to be claimed). Isn't it much more important that piRNA production from those regions cannot be rescued by mdAub?
We found that mdAub couldn't rescue piRNA derived from genomic region affected by aub mutant using genome wide analysis. We modified our figure legends for Fig. 4c to clarify the result as below " Aub sDMA modification is required for generation of piRNAs from piRNA clusters. Reads from 5 kb genomic windows were normalized to total miRNAs reads count. Heatmap shows piRNA fold change compared to control (aub heterozygotes).
wtAub, but not mdAub rescues piRNA expression from genomic regions affected by aub mutation. Genomic windows with more than 5 RPM in control and more than 80% reduction in aub mutant are shown; uni-strand clusters flam and 20A that are not affected in aub mutants are shown for comparison." 3) In Fig. 4d, 42AB and 38C show log2FC=~0 for the wtAub rescue, but those regions are quite green in Fig. 4c. Why?
Thanks for pointing out the discrepancy.
There are two reasons.
1. Overall, there are slightly more piRNAs compared to miRNAs in wtAub rescue compared to control (~1.5 fold in Fig. 4a) that lead to slightly elevated levels of piRNA from all regions on Fig. 4c. The same trend is presented on revised Fig. 4d: piRNAs in wtAub rescue are generally have log2(wtAub/control) values slightly above 0. The total normalized uniquely genome mapped piRNAs reads count in wtAub rescue is around 1.5fold compared to control.
2. The color bar in the Fig. 4c was mislabeled. Should be -5(Red) to 5(Green). And the bars indicating region of cluster (42AB, 38C, 80EF) weren't correctly labeled. We revised them in the new Fig. 4c.