Export of piRNA precursors by EJC triggers assembly of cytoplasmic Yb-body in Drosophila

PIWI-interacting RNAs (piRNAs) are effectors of transposable element (TE) silencing in the reproductive apparatus. In Drosophila ovarian somatic cells, piRNAs arise from longer single-stranded RNA precursors that are processed in the cytoplasm presumably within the Yb-bodies. piRNA precursors encoded by the flamenco (flam) piRNA cluster accumulate in a single focus away from their sites of transcription. In this study, we identify the exportin complex containing Nxf1 and Nxt1 as required for flam precursor nuclear export. Together with components of the exon junction complex (EJC), it is necessary for the efficient transfer of flam precursors away from their site of transcription. Indeed, depletion of these components greatly affects flam intra-nuclear transit. Moreover, we show that Yb-body assembly is dependent on the nucleo-cytoplasmic export of flam transcripts. These results suggest that somatic piRNA precursors are thus required for the assembly of the cytoplasmic transposon silencing machinery.

Reviewer #1 (Remarks to the Author) 5 6 The manuscript by Dennis et al. presents a set of data linking the EJC 7 complex and its associated exportins Nxt1 and Nxf1 to the nuclear 8 transfer and export of piRNA precursor transcripts originating from the 9 flamenco locus in Drosophila follicle cells. The authors show that nuclear 10 transport of the flam transcript is compromised when EJC components are 11 knocked down, while depletion of Nxt1 and Nxf1 affects flam RNA export 12 to the cytoplasm. In addition, they show that assembly of cytoplasmic  bodies, the piRNA processing machineries in OSCs, is strongly affected in 14 export-deficient mutants, but not in EJC mutants. Based on this data, the 15 authors conclude that flam transcripts are directly required for Yb-body 16 assembly, and that nuclear transport of piRNA precursors is essential for 17 the production of functional piRNAs. 18 The manuscript is compactly written and fairly easy to follow. The results 19 are interesting and of clear importance to the field. The finding that piRNA 20 precursors hijack the general RNA splicing and transport machineries for 21 their biogenesis, is also generally interesting, and consistent with their 22 essential function in genome defense. The data is mostly well presented 23 and clearly explained, however the following points should be addressed: transport may affect the results. Eg. is the processing or expression of 29 other piRNA factor proteins altered in these cell lines? 30 3. Why does the Nxf1-SKD have a much weaker effect than Nxt1-SKD? Is 31 the reduction of piRNAs ( Fig 2D,E), the loss of DNA-RNA co-localization 32 (Fig 2F), and the reduction of Yb-body assembly ( Fig 5B)  10. In Fig 2E (and Fig 4B) are the Y-axes identical for the 3 graphs? 55 Should be noted in the legend. 56 11. On page 9, second paragraph, the third sentence starting with "As a 57 consequence..." is confusing. Does this refer to the RnpS1 knock-down 58 cells? 59 12. In Fig 4D and Fig 6D,  conclude that each mutant tested reflects a direct role for the gene 76 product in intra-nuclear transfer, nuclear export, or Yb-body localization 77 of piRNA precursors. In truth, genetic and molecular data can rarely 78 distinguish direct from indirect effects. The data presented here cannot 79 distinguish between direct and indirect roles for exportins or EJC 80 components in piRNA biogenesis. I recommend that the manuscript be 81 rejected.

83
Major Concerns 84 (1) In Figure 1, the authors report that the <i>flamenco</i> transcript 85 localizes to three distinct cellular compartments: a body inside the 86 nucleus, a body coincident with the nuclear periphery, and a body in the 87 cytoplasm. In contrast, Murota et al. (2014) reported detecting only 88 cytoplasmic "Flam Bodies." Given the discrepancy, it is unclear whether 89 the observed localization patterns are an artifact of confocal imaging. 90 Clearly additional markers for the three compartments are needed. 91 92 (2) No consistent metric is used for <i>flam</i> RNA localization: some 93 figures report three classes (inside nucleus, peripheral, or in cytoplasm) 94 while others divide the <i>flam</i> RNA between just two classes or 95 present data for only a single category. It is impossible to make 96 thoughtful comparisons among the mutants with such data. 97 98 (3) Small and often contradictory changes in the localization of Armi and 99 <i>flam</i> RNA, or in the level of mature piRNA are presented as 100 meaningful, or are ignored without justification. Much of the apparent 101 changes between mutants may simply reflect experimental variation.

103
Minor Points 104 (1) Abstract: There is no evidence that piRNA processing occurs in Yb 105 bodies; all current data simply identifies the subcellular regions containing 106 the highest local concentrations of piRNA precursors and processing 107 components, not where the actual endonucleolytic and exonucleolytic 108 steps or PIWI-protein loading occur. 109 110 (2) Introduction: piRNA precursors in mammals do not come from TE 111 clusters, even those piRNAs that target transposons. The entire 112 Introduction describes piRNA biology in flies, but is written as if it applies 113 more generally to all animals (metazoan).  (5) Results: It is unusual to refer to one's own data as "accurate," but 134 even more unusual to refer to it as "quantitative" and then immediately 135 refer to "a few cells" rather than an actual percentage or quantity.

137
Reviewer #3 (Remarks to the Author) 138 139 General comments. These interesting studies regarding Drosophila 140 ovarian somatic cell piRNAs report two overall conclusions: 141 (1) intranuclear movement of pre-piRNA from the site of transcription to 142 the DotCOM licences future pi-RNAs for function in TE silencing. The 143 authors present convincing imaging (DNA/RNA FISH and DNA/RNA/FISH-144 IF) documenting that deficiencies in the Nxf1/Nxt1 mRNA nuclear export 145 machinery result in failure of flam pre-piRNAs to transit from their site of 146 transcription to intranuclear foci (DotCOM) as well as a failure to be 147 exported to the cytoplasm. In contrast, knockdowns of EJC components, 148 while resulting in fewer DotCOM foci, do not affect pre-piRNA splicing, 149 nuclear export, or cytoplasmic processing of the pre-piRNAs into mature 150 piRNAs. Thus, gathering of pre-piRNAs into DotCom foci does not affect 151 their nuclear export or cytoplasmic processing. However, EJC deficiencies 152 result in defects in TE silencing. Although the data regarding the affects of 153 the knockdowns on pre-piRNA intranuclear movement and nuclear export 154 are convincing, this reviewer is not convinced that the authors can 155 conclude that defects in flam pre-piRNA intranuclear movement causes 156 defects in TE silencing since, as the authors admit, disruption of EJC 157 components could indirectly affect TE silencing.

158
(2) Yb-body assembly requires pre-piRNA nuclear export and occurs at 159 the site of pre-piRNA nuclear exit. The data document that Yb-bodies fail 160 to efficiently form in the absence of flam transcription (Fig. 6) as well as 161 when nuclear exit of pre-piRNA is aberrant (as happens upon Nxf1/Nxt1 162 knockdowns). The data that Yb-bodies locate at the cytoplasmic face of 163 nuclei, in foci close to the nuclear site of flam transcription/exit are 164 convincing and interesting; however, the mechanism by which Yb-body 165 proteins locate to pre-piRNA foci exiting the nucleus remains unknown. 166 Specific criticisms: 167 (1) Figure  2A, 2C, S1B, and S2 each have misspellings (Y-axis). The color scheme 172 for Fig. S1B may not be described.

173
(2) Level of knockdowns -Nxt1 and Nxf1 knockdowns have quantitatively 174 different consequences; is this due to different knockdown efficiencies? 175 The levels of knockdown for all constructs should be provided.

176
(3 )Number of nuclei viewed -The data presented in the imaging studies 177 are "pretty"; however, for most panels only ~2 nuclei can be viewed.

178
Although for most studies, quantitative results are provided, it would be 179 preferential to be able to view additional images as supplementary figures 180 181 They also fit with results provided by Hayashi & al who reported that "splicing of most 190 introns is EJC-independent and no piRNA pathway mRNAs other than piwi seem mispliced 191 in Tsu-or Acn-SKD." 192 These two earlier studies are mentioned and a new paragraph has been added page 8/9: 193 " we analyzed by RT-qPCR the impact of Mago-and RnpS1-SKD on the expression of major 194 genes involved in the somatic piRNA pathway like piwi, armitage (armi), maelstrom (mael) 195 or yb (Supplementary Figure 5). We found that the expression of none of these genes is 196 affected in these mutants except piwi whose expression decreases in Rnps1 depleted line as 197 earlier reported ((36, 37) see discussion).". It must be noted that, in the study reported here, 198 mRNA depletion may not be detected when production is mostly provided by germ cells. 199 3-We now discuss the difference observed between Nxf1 and Nxt1-SKD lines. It is mainly 200 stated page 13 in the following paragraph : 201 "..It is possible that the redundancy which has been reported between export components 202 binding nuclear pore complex may compensate depletion of Nxf1 (41,42). Alternatively, in 203 our SKD mutant, a residual Nxf1 function could be sufficient to allow the export of enough 204 flam precursors that are processed to an almost WT amount of piRNAs (37 and 41 rpkm 205 respectively) ( Figure 2E). Nxt1 plays an essential role in the export of piRNA precursors in 206 follicle cells. Several studies have reported the ability of Nxt1 to mediate nuclear mRNA 207 export by controlling the interaction of Nxf1 with components of the nuclear pore (29, 43)…" 208 We have added a statistical analysis to all our results. In the new version, we now discuss conservation of EJC/exportins in other organisms and 214 their potential role: "…it can be anticipated that components of EJC and exportin pathways 215 which are well conserved throughout eukaryotes fully participate in the defense system 216 protecting genomes from TE invasions…" (page 15).

217
Concerning the role of long piRNA precursors in initiating the assembly of their processing 218 machineries, we added a discussion highlighting the similarities and differences observed 219 between piRNA precursors in the somatic lineage and the germline "Unlike somatic cells, 220 germline has evolved a mechanism that inhibits splicing of the nascent piRNA transcript (39 12 -We agree that in Fig 4D and Fig 6D, the WT bars should be similar to each other. We 250 performed new experiments to increase the number of counts but a difference remains. Since 251 WT bars are always similar between several DNA/RNA FISH experiments (See Figure 2F  252 and 4D), it may be argued that counting can hardly be compared between DNA/RNA ( Fig.  253 2F and 4D) and Immuno DNA FISH experiments (Fig 6D). This is potentially due to 254 differences in the detection of signals varying with the type of probe.

256
Reviewer 2 257 The two major flaws reported by reviewer 2 do not exist anymore.

258
-We added all the statistics in each Figure  -We are aware that genetic and molecular approaches as used here cannot discriminate 262 between direct and indirect roles of the genes identified. This was not clearly stated in the 263 former version. So, we added a paragraph and discuss this point in detail page 13: "At 264 present, we do not know whether interaction between the flam precursor and these factors is 265 direct or indirect. EJC components may each bind flam precursor directly and independently. 266 Alternatively, some may bind directly whereas others may only associate indirectly through 267 interactions within the complex. It is also possible that these interactions occur with other yet 268 unknown factors. Moreover, we cannot exclude that through a more general effect on mRNA 269 processing and transport, EJC mutants may indirectly affect flam export and the piRNA performed on ovaries, to analyze more closely the sub-cellular localization of flam 279 transcripts. We bring enough data to do statistics which fit with each other. We do not 280 believe that the observed localization patterns originate from an artifact of confocal imaging. 281 If they were, we should not see an increase in the proportion of cells having a nuclear 282 accumulation of flam transcripts in Exportin-SKD as we see in Figure 2A and B. 283 Nevertheless, to answer this comment, we added a film showing an intranuclear flam dot with 284 no ambiguity. 285 286 2 -We agree with this comment. We corrected the manuscript to present and discuss data in 287 light with the three categories of localization (nuclear, trans-membrane and cytoplasmic). 288 This has been modified for WT and depleted lines (Figure 2A and 3E). Statistics are added 289 for all of them so that comparisons can be made. 290 291 3-In the revised version, the statistical analysis performed for each experiment shows clearly 292 that we did not present, or ignored or over-interpreted differences which should not have 293 been. Error bar are now shown on each histogram and p values are given.

295
Minor points: 296 1-We agree that no evidence exists showing that piRNA processing occurs in Yb bodies. We 297 corrected the abstract to take this comment into account. 298 2-We focused the introduction on what is known in Drosophila and indicated it clearly. 299 3-We indeed did not refer to the recent papers reporting that primary piRNA biogenesis is 300 different in the somatic cells and in the germline. We believe that introducing the mechanism 301 of piRNA processing taking place in the germline, including phasing, is not appropriate in 302 this article. 303 To respond to this comment, we first modified the introduction to only focus on the somatic 304 piRNA clusters. Second, we added a full paragraph on somatic and germline differences in 305 the discussion page 15, in which the traffic of flam transcripts in the follicle cells is compared 306 to the one of piRNA precursors in the germline. To this purpose, we report the differences in 307 expression of piRNA clusters in both lineages: uni-strand vs. dual strand piRNA clusters, 308 canonical vs. non-canonical transcription, spliced vs. unspliced piRNA precursors. Moreover, 309 we also refer to Zhang et al (2014) who proposed an extrinsic control of nuage assembly and 310 function exerted by nuclear components, and discuss it in link with our data (page 16). 311 4-Reference to UAP56 has been completely modified. It is now presented and its function 312 more clearly reported in the introduction page 4, in the results page 8 and in the discussion 313 page 14 and 16. 314 5-We agree with this comment. We corrected the first paragraph of the results accordingly.

315
"Accurate quantitative analysis" has been removed.

317
Reviewer 3  2-To answer this comment, we added a paragraph page 16. In the germline, an extrinsic 329 control of nuage assembly and function exerted by nuclear components has also been 330 reported (38). The mechanism remains unknown but the authors proposed that Rhino and 331 UAP56 are connected to the nuage via the nuclear pore complex that could trigger release of 332 cluster transcripts in the nuage. A nuclear envelope-spanning machinery would then be 333 required for piRNA biogenesis. To answer to reviewer 3, we now report this study and 334 compare germline data to data reported here from the somatic follicle cells. We highlight 335 similarities and differences. This leads us to suggest that in the follicle cells, Nxt1 could play 336 a key role by stimulating interactions of exportins with components of the nuclear pore. 337 Specific criticisms: 338 1-We apologize for having posted misaligned figures. We verified that figures are aligned in 339 this version. We also corrected misspellings in figures 2A, 2C, S1B, S2 and described the 340 color scheme in figures when required. 341 2-Quantitative differences between Nxf1-and Nxt1-SKD lines are now discussed page 342 13:"… . It is possible that the redundancy which has been reported between export 343 components binding nuclear pore complex may compensate depletion of Nxf1 (41,42). 344 Alternatively, in our SKD mutant, a residual Nxf1 function could be sufficient to allow the 345 export of enough flam precursors that are processed to an almost WT amount of piRNAs (37 346 and 41 rpkm respectively) ( Figure 2E). Nxt1 plays an essential role in the export of piRNA We believe that this revised version of the manuscript answers each comment made by the 364 reviewers. We hope that it will make our manuscript suitable for publication. 365 366