miR2118-dependent U-rich phasiRNA production in rice anther wall development

Reproduction-specific small RNAs are vital regulators of germline development in animals and plants. MicroRNA2118 (miR2118) is conserved in plants and induces the production of phased small interfering RNAs (phasiRNAs). To reveal the biological functions of miR2118, we describe here rice mutants with large deletions of the miR2118 cluster. Our results demonstrate that the loss of miR2118 causes severe male and female sterility in rice, associated with marked morphological and developmental abnormalities in somatic anther wall cells. Small RNA profiling reveals that miR2118-dependent 21-nucleotide (nt) phasiRNAs in the anther wall are U-rich, distinct from the phasiRNAs in germ cells. Furthermore, the miR2118-dependent biogenesis of 21-nt phasiRNAs may involve the Argonaute proteins OsAGO1b/OsAGO1d, which are abundant in anther wall cell layers. Our study highlights the site-specific differences of phasiRNAs between somatic anther wall and germ cells, and demonstrates the significance of miR2118/U-phasiRNA functions in anther wall development and rice reproduction.

The manuscript describes the role of miR2118-dependent regulatory module in rice anther wall development by gene editing. The author integrated small RNA profiling, proteome, and 3Dhistochemical analyses of miR2118 mutants to reveal the novel miR2118 functions and the biogenesis of miR2118-dependent U-phasiRNA/AGO1 subfamilies in the anther wall development. Overall, the experiments were well-designed and results were interesting. The data looked robust and well presented. Comments/queries that need to be addressed: 1. Authors produced miR2118 mutants by genome editing and obtained three independent mutant alleles, mir2118-1, mir2118-2, and mir2118-3. Whether the authors conducted experiments such as southern blot to confirm these three mutants were actually independent mutant alleles? In Fig  1b, only mi-1 and mi-2 were presented, how about mi-3? Authors should add it to Fig 1b. 2. In manuscript, authors described that there was 16,230bp deleted in ir2118-1, mir2118-2, and mir2118-3(mi-1, mi-2, and mi-3) by whole-genome sequencing. So what is the PAM sequences to achieve such large fragment deletion? And what about the vector information? 3. Authors said microRNA2118 (miR2118) conserved among plants, is expressed at reproductive stages, causing the production of secondary phased small interfering RNAs (phasiRNAs) and it plays an important role in rice anther wall development. Does miR2118 has the similar function in other plants? 4. Authors found that there was no significant difference in plant height between WT and mi-1, while mi-2 showed a dwarf phenotype under SD condition. And is there any difference between WT and mutants under LD condition? Why did authors choose LD and SD conditions to investigate the role of miR2118 in rice anther wall development? 5. Authors classified rice anther development into six stages and selected Stage 2 to analyze the overall structure of anthers. And moreover, 0.5 mm anthers at Stage 2 were also used to perform the transcriptome analysis of small RNAs. So how about the different/differential expressions of small RNAs of the other five stages anthers between WT and miR2118 mutants? I suggested authors supply these data on the manuscript. 6. Expressions of encoding protein genes play important roles in rice anther development. Therefore, I wonder the differentially expressed genes between WT and miR2118 mutants. These data can be added to the manuscript if possible. 7. What's the role of AGO proteins in rice anther wall development? It should be clarified in the manuscript. 8. In present study, the inner mechanism of miR2118 in rice anther wall development is not so adequate and further research is suggested to be done to clarify the specific function of miR2118 in rice wall development.

Responses to reviewer's comments
First of all, we would like to thank the reviewers for their constructive and valuable comments with regard to our manuscript. We have indicated the major changes made in the revised manuscript, and provided further point-by-point responses to the reviewers' comments below.

Additional results and major changes
1. Long-deleted mir2118 mutants are renamed as mir2118-1-1, mir2118-1-2, and mir2118-1-3 in the main text (the reasons are as follows). In this manuscript, Araki et al. report an analysis of a rice mir2118 mutant generated using CRISPR-Cas9. The mutant appears to be a partial mir2118 mutant, as at least 4 (known) copies of miR2118 are not knocked out, and mature miRNAs derived from these loci are detected in anther and pistil. The mutant exhibits defects in both male and female reproductive development, causing male & female sterility. The developmental defects in anthers appear to be thicker tapetum and delayed death of middle layer. The authors also identified miR2118-triggered phasiRNAs in the somatic cells of anther and observed an enrichment of 5'-U in these phasiRNAs, distinct from the meiocyte-enriched phasiRNAs that are enriched for 5'-C. The analyses provide insights into the role of a subset of the miR2118 family in male (& female) reproductive development.
However, the study is lacking some of the key data in support of the conclusions.
Below are a number of specific comments/suggestions that may help improve the manuscript.
Major issues: 1. There is no quantitative data (e.g., segregation ratios) provided for the genetic analysis of the mutant alleles.
3. The deletion of a 16-kb genomic region using two guide RNAs is remarkable. However, it seems that the deleted region includes at least an intact protein-coding gene and a portion of another protein-coding gene, but the authors did not rule out (or discuss) these genes as potential causal genes of the mutant phenotypes.
We agree with the reviewer regarding this point. Therefore, in the revised manuscript, we further analyzed putative genes located in the deleted region. According to the annotation in RAP-DB 4. On a related note, the authors state that "the phenotypic variations of the two alleles might originate from the gRNA sequences used for the deletion of the loci that have high homology with the miR2118 recognition sites in over 700 reproductive lincRNAs", implying editing at 21-PHAS loci; if this is the case, whole-genome sequencing of the two mutants, which is done is this study, may identify mutations in the 21-PHAS loci.
However, this possibility is not discussed.
We thank the reviewer for the comment. We further analyzed genomic variations present in mi1-1, mi1-2 and mi1-3, using the whole-genome sequencing data. Variant searches against the Nipponbare genome, however, did not detect genomic variants of deletion or/and insertion within 21PHAS loci (body) in mi1-1, mi1-2, and mi1-3, while six and two variants were detected within 1 kb upstream or downstream of 21PHAS in mi1-2, and mi1-3, respectively (Supplementary Figure 1f). Accordingly, we revised the paragraph in the main text to incorporate the results (Page 7, line 19~ ). 6. There does not seem to be a clear set of hypotheses about the roles of AGO1b/AGO1d. The dysregulation of the AGO genes in the mir2118 mutant may simply suggest that they are regulated, either direct or indirectly, by miR2118 or the miR2118-triggered phasiRNAs, instead of being directly involved in miR2118/phasiRNA biogenesis as stated in the manuscript. The latter is in fact not well supported by the available data.

Some important information is
We agree with the reviewer, and in the revised manuscript, we deemphasized the descriptions about the involvement of the AGOs in miR2118/phasiRNA biogenesis. Understanding the functions of AGO1b/1d are great interests in the related research fields, and would be the subject of future studies. Actually, we are currently conducting projects to understand the role of the AGOs during the rice reproduction. In the discussion part of this revision, we further added the expression analysis of AGO1b/d in mir2118 mutants (Sup Figure 9) and the following sentences to discuss the possibilities of the observed reduction of the AGO proteins in mir2118 (Page14 line10-17): "There are two possibilities to explain the reduction of AGO1b, AGO1d, and MEL1 proteins in mir2118 mutants: (1) the reduction of miR2118 or phasiRNAs may cause the destabilization of AGO1b/AGO1d/MEL1 proteins in the absence of miR2118/phasiRNAs; or (2) defects of anther wall development in the mir2118 may reduce the anther wall cell mass, which then reduces the relative amounts of AGO1 proteins. Reduced AGO1b/d proteins with no difference of the mRNA levels in mir2118 mutants ( Supplementary Fig. 9) may support the first hypothesis, in which the proteins are associated with the anther wall U-rich phasiRNAs or U-miR2118 family members." 7. The writing of the manuscript needs polishing, perhaps by a native speaker. There are many issues in the writing that make it taxing to follow. For example, on page 7, I don't agree that the results indicate that mir2118 causes defects in the elongation of OUTER anther locules, as the change in the shape of the outer locules is likely a consequence of curled inner locules -as described in the manuscript.
The revised manuscript was further edited by a professional editing service (Editage). We also revised the pointed sentences about the shape of anther locules in the main text (Page8 line14-16).
Minor issues: 8. Supplementary Figure 1: labels of the y-axes are missing, so it's not clear what type of data is shown.
We thank the reviewer for the comment. We added labels indicating the "fertility Rates  Supplementary Figure 4), the experiment had done without the fertilizer, which did not result in the optimal growth of the plants. We repeated the experiment for this revision by growing the mutant plants with a fertilizer (HypoNex; https://www.hyponex.co.jp/en/) under SD and LD conditions. In the experiment, no significant difference in plant height between WT and mi1-1 or mi1-2 were observed (Supplementary Figure 4). Currently the cause of the developmental phenotype observed in the previous experiment is not clear, except the absence of the fertilizer. We now added a plant picture and data for plant height of mi1-1 and mi1-2 under LD and SD conditions ( Supplementary Fig.4), and more detailed description in the main text. 11. Page 5: The authors state that "MEL1 is the only AGO protein known to bind phasiRNAs during reproduction" -it would be better to clarify that this is only in rice, as there are known AGO proteins loading reproductive phasiRNAs in other plant species (e.g., maize).
We agree that this is an important point. We changed the sentence "MEL1 is the AGO protein known to bind to phasiRNAs during reproduction in rice" (Page5 line11-12).

Reviewer #2 (Remarks to the Author):
The manuscript describes the role of miR2118-dependent regulatory module in rice anther wall development by gene editing. The author integrated small RNA profiling, proteome, and 3D-histochemical analyses of miR2118 mutants to reveal the novel miR2118 functions and the biogenesis of miR2118-dependent U-phasiRNA/AGO1 subfamilies in the anther wall development. Overall, the experiments were well-designed and results were interesting. The data looked robust and well presented.
2. In manuscript, authors described that there was 16,230bp deleted in ir2118-1, mir2118-2, and mir2118-3(mi-1, mi-2, and mi-3) by whole-genome sequencing. So what is the PAM sequences to achieve such large fragment deletion? And what about the vector information?
In the response to the comment, the PAM sequence of the guide RNA used for the genome-editing was shown in Figure 1a, and also described in the main text. miR2118b and miR2118n that are flanking the deletion sites have the identical sequence to the guide RNA except one C/U mismatch in the PAM sequence (Figure 1a In monocots (Maize, Brachypodium), miR2118 mainly targets long non-coding RNAs, which are specifically expressed at reproductive stages. In contrast, in dicot, coding genes targeted by miR2118 include the NBS-LRR genes involved in pathogen defense and transcription factors. These results suggest that the target RNAs of miR2118 are diverged between monocots (reproductive noncoding RNAs) and dicots (coding genes).
In 2019, it is reported that tomato miR2118 (dicot) is involved in the pathogen defense via 21-nt phasiRNAs production (Canto-Pastor et al., 2019). However, the molecular functions of miR2118 during the reproductive stages are not fully understood in monocots (Page 4 line23-28).
In this study, we showed the function of miR2118 in the rice reproduction. Numerous reproductive long non-coding RNAs, targeted by miR2118, have been reported in monocots other than rice, suggesting the importance of miR2118 as a key regulator of the reproductive process in other species. 4. Authors found that there was no significant difference in plant height between WT and mi-1, while mi-2 showed a dwarf phenotype under SD condition. And is there any difference between WT and mutants under LD condition? Why did authors choose LD and SD conditions to investigate the role of miR2118 in rice anther wall development?
In terms of the plant height in Sup. Fig. 2 in the previous manuscript (now in Supplementary Figure 4), the experiment had done without the fertilizer, which did not result in the optimal growth of the plants. We repeated the experiment for this revision by growing the mutant plants with a fertilizer (HypoNex; https://www.hyponex.co.jp/en/) under SD and LD conditions. In the experiment, no significant difference in plant height between WT and mi1-1 or mi1-2 were observed (Supplementary Figure 4).  Fig. 9) may support the first hypothesis, in which the proteins are associated with the anther wall U-rich phasiRNAs or U-miR2118 family members." 8. In present study, the inner mechanism of miR2118 in rice anther wall development is not so adequate and further research is suggested to be done to clarify the specific function of miR2118 in rice wall development.
We agree with the reviewer's comment. The underlying miR2118-mediated molecular mechanisms in the rice anther wall development remain unclear. In this revision, one paragraph is added to discuss the possibilities of miR2118 function: (1) dual functions in the specific anther wall cell layers, (2) secondary effects of the abnormal differentiation process of outer layers during pre-meiosis. Furthermore, we refer the publications of OCL4 and MSP1, which regulate outer cell layer in anther, and these mutants show the similar phenotype of mir2118. We discussed an interaction between the soma and germ and/or a non-cell autonomous regulation via miR2118/secondary siRNAs (Page15 line4-18).
We are currently conducting research projects to understand the role of miR2118 and the AGOs during the rice reproduction by focusing on the AGO1b/1d and the interacting miR2118 and/or U-phasiRNAs, the results of which may be published hopefully in the near future.
Overall, the authors adequately addressed the major concerns I raised for the first version of the manuscript. Key new data have been added, and the writing has been greatly improved. I am glad to see that the authors discussed the protein-coding genes that are knockout along with the miR2118 cluster. However, in Supplementary Figure 3, it seems that Os04g0435300 does show a preferential expression in the 0.5-mm anthers than in other tissues/organs -rather than ubiquitously expressed, as the authors suggested. Moreover, a ubiquitously expressed gene may still have specialized functions in distinct tissues. I am not asking for more experiments to rule out the protein-coding genes that are knocked out in the current study, but the authors need to be objective about the data and explicitly discuss the limitation of the approach. Below are a few other (minor) comments: 1. It's unclear why the authors used a "-1" in the middle of the mutant names; it does not help with distinguishing the three alleles. 2. Supplemental Figure 6c: a statistical test should be included for the differences. 3. Supplemental Figure 10a: Many of the error bars overlap and therefore cause confusion in the interpretation of the data. It would be great if the authors could figure out a way to fix this. (My apologies that I did not point this out in the first round of review.) 4. I'm glad to see that the authors added discussion about PMS1T, the known 21-PHAS locus that has previously been shown to influence male fertility of rice; suggesting the miR2118 cluster analyzed in this work functions in an independent pathway than PMS1T. However, why didn't the author show any data for LDMAR, another male fertility-associated lncRNA locus that the authors mentioned in their response to the other reviewer? 5. Line 12: should the first gene ID be Os04g0435300 (see Supplementary Figure 3) instead of Os04g043500?
Reviewer #2 (Remarks to the Author): Almost all the queries that need to be addressed in the revised manuscript entitled "miR2118dependent U-rich phasiRNA production in rice anther wall development" have been modified carefully. The authors didn't perform experiments to confirm the inner mechanism of miR2118 in rice anther wall development but added much efficient text supplement to prove the idea.
We would like to thank the reviewers for their valuable comments on our revision. We have made changes in this revision of the manuscript in responses to the reviewer's comments as below.
Reviewer #1 (Remarks to the Author): Overall, the authors adequately addressed the major concerns I raised for the first version of the manuscript. Key new data have been added, and the writing has been greatly improved. I am glad to see that the authors discussed the protein-coding genes that are knockout along with the miR2118 cluster. However, in Supplementary Figure 3, it seems that Os04g0435300 does show a preferential expression in the 0.5-mm anthers than in other tissues/organs -rather than ubiquitously expressed, as the authors suggested. Moreover, a ubiquitously expressed gene may still have specialized functions in distinct tissues. I am not asking for more experiments to rule out the protein-coding genes that are knocked out in the current study, but the authors need to be objective about the data and explicitly discuss the limitation of the approach. Below are a few other (minor) comments: As commented, we could not completely rule out the possibility of these genes' effects on the anther phenotypes. Therefore, we removed the word "ubiquitously" and added the sentence in the mir2118 mutant part of the result as follows (Page 7, line 18-20). "~ while we could not exclude the possibilities that the loss of Os04g0435300 and Os04g0435475/LOC_04g35550 had effects on the observed phenotypes in mir2118."