flam piRNA precursors channel from the nucleus to the cytoplasm in a temporally regulated manner along Drosophila oogenesis

Background PIWI-interacting RNAs (piRNAs) are the effectors of transposable element silencing in the reproductive apparatus. In Drosophila ovarian somatic cells, piRNAs arise from long RNA precursors presumably processed within cytoplasmic Yb-bodies. Results Here we show that the nucleo-cytoplasmic traffic of piRNA precursors encoded by the flamenco locus is subjected to a spatio-temporal regulation. Precursor RNAs first gather in a single nuclear focus, Dot COM, close to the nuclear periphery, and transit through the membrane before being delivered to the cytoplasmic Yb-bodies. Early in oogenesis, flamenco transcripts are rapidly transferred to the cytoplasm making their initial nuclear gathering in Dot COM too transient to be visualized. As oogenesis proceeds, the cytoplasmic delivery steadily decreases concomitantly with the decrease in the protein levels of Armi and Yb, two components of the Yb-bodies. Both events lead to a reduction of Yb-body assembly in late stages of oogenesis, which likely results in a drop in piRNA production. Conclusion Our findings show a spatio-temporal regulation of the piRNA biogenesis in the follicle cells of Drosophila ovaries, that involves coordinated control of both piRNA precursors and components of the piRNA processing machinery. This newly unveiled regulation establishes another level of complexity in the production of piRNAs and suggests a stage-dependent involvement of the piRNA biogenesis in the mechanism of transposable elements silencing along oogenesis. Electronic supplementary material The online version of this article (10.1186/s13100-019-0170-7) contains supplementary material, which is available to authorized users.


Background
Eukaryotic genomes are composed of a variable proportion of transposable elements (TEs) accumulated throughout evolution. These sequences are silenced by the host to protect itself and its progeny against potentially deleterious mutations and genome invasion. In the gonads, where it is essential to ensure the maintenance of genome integrity for the next generation, the piRNA pathway is responsible for TE silencing in both somatic and germinal tissues [1][2][3][4]. This process involves small guide piRNAs of [23][24][25][26][27][28][29] nucleotides (nts) that originate from discrete genomic regions called piRNA clusters.
142 piRNA clusters have been identified in Drosophila melanogaster [2], mostly located in pericentromeric and telomeric regions. These clusters vary considerably in size from a few kilobases (kb) to several hundred kb. They are enriched in full length or truncated TEs that are often nested within one another [2,5]. In ovarian somatic support cells surrounding the germline, piRNAs are mainly produced from two piRNA clusters: traffic jam [6] and flamenco (flam) [2,7]. The flam cluster is located at the pericentromeric region of the X-chromosome, spans over more than 200 kb and is strongly enriched in both ancient and recent retrotransposons mostly inserted in an orientation antisense to the TE transcription [5]. flam is transcribed from a polymerase II promoter as a long single-stranded RNA that is a substrate for piRNA biogenesis. Flam transcripts undergo alternative splicing to generate diverse piRNA precursors that all share the first exon at their 5′ end [8]. These transcripts are then processed into 23-29 nt piRNAs, presumably in cytoplasmic Yb-bodies [9,10]. Mature piRNAs associated with Piwi protein form a piRNA-induced silencing complex (piRISC) that is delivered to the nucleus to target nascent TE mRNAs and initiate transcriptional gene silencing [11,12].
Two studies have reported that flam precursor transcripts, together with transcripts from other piRNA clusters, concentrate in 1 to 2 foci in ovarian follicle cells [13,14]. Analysis of the subcellular localization of these sites of accumulation at different stages of development have yielded varying findings. The first study focused on follicle cells in late stages of oogenesis and showed that flam RNA precursors accumulate in a single nuclear substructure, named Dot COM, that faces the cytoplasmic Yb-body to which the flam precursors are channeled by nucleo-cytoplasmic transfer [13]. A subsequent report [14] visualized flam transcripts concentrated within the cytoplasm in 1 to 2 substructures named flam bodies, located close to Yb-bodies. The authors worked mainly with a cultured Drosophila ovarian somatic stem cell line (OSS cells) derived from a somatic stem cell population of the germarium [15] that expresses a functional piRNA pathway [6].
We show here that flam transcripts are channeled to the cytoplasm in a temporally regulated manner. In early stages of oogenesis, they are mainly detected within the cytoplasm. As oogenesis proceeds, flam transcripts accumulate in a focus detected within the nucleus or within the nuclear membrane as though their delivery to the cytoplasm was impeded. Combined with a drop in Armi and Yb protein levels after stage 8, this affects the assembly of Yb-bodies which then are very small and absent from most cells. These findings emphasize the temporal regulation of Yb-body assembly, which requires both cytoplasmic flam delivery and Armi and Yb proteins. In their absence, Yb-bodies fail to assemble correctly, which potentially can cause a decrease in the production of piRNAs.

Results and discussion
The cytoplasmic transfer of flam precursor transcripts decreases during ovariole development In a previous study [16], we performed a global quantitative analysis of the subcellular localization of flam transcripts within ovarian somatic follicle cells, independently of the developmental stages of oogenesis. Depending on the follicle cell observed, the focus where flam precursors gathered was visualized either completely within the nucleus close to the nuclear periphery, or stretching across the nuclear membrane, or within the cytoplasm close to the nuclear membrane. To investigate whether this differential localization is somehow related to the stage of the developing egg chambers, we examined flam RNA precursors in follicle cells within egg chambers in stages 3 to 10 of oogenesis. To do so, we performed an immuno-RNA FISH experiment on WT Drosophila ovaries using an anti-lamin antibody to label the nuclear membrane, and a specific flam RNA probe described in [13]. The same experiment was performed in parallel with OSS cells. The localization of flam foci was quantified as follows. When positioned close to the inner side or outer side of the lamin signal, foci were considered as nuclear within Dot COM or cytoplasmic within the Yb-bodies respectively. When co-localized with the lamin signal (totally or partially), the flam focus was considered as stretching across the nuclear membrane.
Firstly we investigated egg chambers from stage 3 to 10 of oogenesis and quantified the proportion of cells in which a flam signal could be visualized (Fig. 1a). We found no significant difference between stages 3, 4 and 7-10 where flam foci are detected in more than 60% of cells. This proportion appears slightly lower around 50% at stages 5 and 6 ( Fig. 1a).
Secondly, focusing on the follicle cells that harbored a flam signal, we investigated its subcellular localization. We found that flam transcripts that accumulated mostly in a single focus, vary drastically in localization from early to late stages of oogenesis. In early stages, in a majority of follicle cells, 1 to 2 flam foci are observed within the cytoplasm (Fig. 1b, d, e). OSS cells, which derive from a somatic stem cell population of the germarium, and follicle cells from early stage 3, display a similar percentage of flam foci within the cytoplasm (around 62%) ( Fig. 1b & c). This similarity may be explained by the intrinsic nature of these two cell types; both are highly dividing cells required to rapidly encapsulate the 16-cell germinal cyst. The percentage of cells with cytoplasmic flam foci decreases progressively to 10% at stage 10 of oogenesis. Concomitantly, the proportion of follicle cells in which a flam focus is detected within Dot COM increases, from 4% in stage 3 to 32% in stage 10 ( Fig. 1b-g). The switch from a cytoplasmic to a nuclear position is progressive, suggesting that the export of piRNA precursors to the cytoplasm steadily decreases as oogenesis proceeds. This stage-by-stage analysis further emphasizes that an increasing fraction of flam RNAs co-localizes with the lamin staining. Interestingly, in late stages of oogenesis, the flam transcripts are even more likely to be detected spanning the nuclear membrane (57%) than within the nucleus (33%) and the cytoplasm (10%) (Fig. 1b). These data indicate that flam precursors can enter the nuclear membrane throughout oogenesis even in late stages but seem to lose progressively the ability to be delivered to the cytoplasm. This suggests that fewer piRNA precursors reach the Ybbodies, thereby giving rise to fewer piRNAs in late stages of oogenenesis.
At early stages of Drosophila oogenesis, flam precursors rapidly transit to the cytoplasm To better understand the discrepancy between flam localization in early and late stages of oogenesis, we first considered the early stages of oogenesis and posed the question of why no nuclear accumulation is observed. The nuclear gathering in Dot COM could be specific to later stages of oogenesis and not occur in early stages.
Alternatively, the nuclear accumulation could be too transient before export to be detected in our immuno-RNA FISH experiments. To distinguish between these two hypotheses, we examined the localization of flam transcripts in follicle cells of Nxt1-SKD flies in early stages of oogenesis (Nxt1 somatic knockdown using tj-Gal4 driver) when the export of flam precursors is impeded. Nxt1 protein is a RNA export factor that is required, together with its partner Nxf1, for the export of flam precursors [16]. We observed a clear increase in the proportion of follicle cells with a nuclear accumulation of flam transcripts in early stages 3 and 5 in Nxt1-SKD compared to WT ISO1A (Fig. 2 a & b), which suggests that nuclear accumulation of flam precursors can occur in early stages of oogenesis.  Table S1. c-g Subcellular localization of flam foci in OSS cells (c) or ovarian follicle cells of stage 3 (d), stage 5 (e), stage 8 (f) or stage 10 (g). flam transcripts visualized as 1 to 2 foci per cell are labelled in red using flam 508 RNA probe and nuclear membrane is labelled in green with anti-lamin antibody These findings indicate that, at early stages, in their traffic to the cytoplasm, flam transcripts are shuttled to the cytoplasm very rapidly. Export of the piRNA precursors is thus more efficient in early than in late stages of oogenesis.
At late stages of Drosophila oogenesis, flam precursors fail to be delivered to the cytoplasm The second issue we addressed was to understand why the delivery of flam precursors from the membrane to the cytoplasm is steadily affected as oogenesis proceeds. We have previously shown that the export of flam precursor transcripts to the cytoplasm is the signal for the Yb-bodies to be assembled [5]. If, as suggested above, flam precursors do not reach the cytoplasm in late stages, then the proportion of Yb-bodies is expected to decrease.
In immunostaining experiments using Armi and Yb antibodies, we indeed observed that the size of the Yb-bodies (Fig. 3a) detected in follicle cells decreases during oogenesis together with the percentage of cells that have a discernable Yb-body (Fig. 3b). At stage 10 less than 10% of cells have Armi and Yb foci compared to 80% in stage 3 or 90% in OSS cells (Fig. 3b and Additional file 1: Figure S1A).
However, the decrease in Yb-body assembly in late stages of oogenesis cannot be attributed solely to the reduction of flam piRNA precursors exported to the cytoplasm. In the germarium, the follicle cells encapsulate the germline cyst and carry out a mitotic division program from stage 2 to stage 6 of oogenesis [17]. At stage 6, they cease to undergo mitotic cycles and start endocycles. This transition is regulated by the Notch pathway [17]. It has been demonstrated that Armi protein level decreases at this switch from mitosis to endoreplication owing to Notch signaling pathway activation [18]. Immuno-histochemical experiments performed with Armi and Yb-antibodies on whole ovaries showed that not only is Armi protein level reduced in late stages of oogenesis, but also Yb staining. We observed a clear drop during oogenesis with a high level of Armi and Yb proteins only until stage 8 and a weak staining afterwards (Fig. 3b & Additional file 1: Figure S1) indicating that both proteins are presumably regulated at the transition from mitosis to endoreplication.
To determine whether the reduced export of flam transcripts at late stage of oogenesis is related to the decrease in Armi and Yb protein levels after stage 8, we analyzed whether a reduction of Armi or Yb protein level in early stages of oogenesis would have the same impact on RNA precursor export. To do so we quantified the sub-cellular localization of flam transcripts in follicle cells of stage 5 depleted for Armi or Yb using RNAi or mutant lines. The localization of flam transcripts was found similar between WT and Armi-and Yb-depleted lines, with a majority of follicle cells having transcripts located within the cytoplasm. However, quantifications showed a low but significant decrease in the proportion of cells exhibiting cytoplasmic flam foci ( Fig. 4a & b). A reduced accumulation of flam transcripts in the cytoplasm may result from an increased instability of the transcripts due to the absence of Armi and Yb in the cytoplasm. Alternatively, the mild decrease in the proportion of cells with cytoplasmic flam  Table S1 foci is expected if the export of flam transcripts is impeded. It has been suggested that armi is expressed in the follicle cells of the trans-heterozygous mutant armi 1 / armi 72.1 [18] and we cannot exclude an incomplete depletion of Armi and Yb proteins in early stages in the RNAi lines used. Nevertheless, these mutants reveal a possible involvement of Armi and Yb proteins in the delivery of piRNA precursors to the cytoplasm. It has previously been shown that Armi and Yb proteins bind piRNA precursors [19][20][21][22]. It has also been reported that artificially tethering Armi or Yb protein to a heterologous RNA channels this RNA to the cytoplasmic processing machinery [23,24]. Armi ATP-dependent helicase activity has also been observed in vitro [24] and been implicated in Armi specific binding to piRNAs precursors and piRNA production [21,22,25], with the hypothesis that Armi binds and unpacks piRNA precursors while being exported to the cytoplasm. Whether this binding happens in the nucleus or at the nuclear membrane remains to be elucidated. In late stages of oogenesis in Nxt1-SKD flies, we observed that Armi but not Yb protein can occasionally be visualized within the nucleus as a dotty pattern (Additional file 2: Figure S2). Of the 3074 follicle cells examined from 20 stage 10 egg chambers, Armi was detected in 7.9% of the nuclei, along with a great variability among the follicles (0 to 18%). This raises the possibility that Armi protein shuttles between the cytoplasm and nucleus which is not detectable in WT conditions. However, we cannot exclude that this Armi pattern of staining is a consequence of a  Table S1 detrimental lack of the Nxt1 export protein in the ovarian follicle cells.
On the basis of these findings, we propose that two interconnected events occur during oogenesis that lead to a drop in Yb-body assembly in late stages. On the one hand, following Notch-signaling pathway activation in stage 6-7, the levels of Armi and Yb proteins are drastically reduced, which not only decreases the assembly of the Yb-bodies but also impedes the export of piRNA precursors (Fig. 4). On the other hand, the export of piRNA precursors is progressively reduced along oogenesis which further hampers the assembly of the remaining Armi and Yb into Yb-bodies.
Reduction of the assembly of Yb-bodies and of the export of flam piRNA precursors is thought to cause a decrease in the production of piRNAs in late stages of oogenesis. However, several studies indicate that the piRNA silencing is active throughout oogenesis. For example, if we consider the two retrotransposons, ZAM and gypsy, their enhancers are active at the posterior pole of the follicle cells throughout oogenesis including in late stages [26,27]. Nevertheless, their transcripts are never detected in WT conditions owing to the silencing exerted by the piRNA pathway, which indicates that the silencing exerted on ZAM and gypsy is active even in late stages. One possible explanation of TE silencing in late stages of oogenesis when piRNA production seems impeded could be that the few and highly diminished Yb-bodies present in late stages give rise to a piRNA population that is sufficient to silence TEs. Alternatively, and not exclusively, piRNAs produced anteriorly in early stages are stable enough to remain active in late stages. Finally, it cannot be excluded that a transcriptional silencing exerted in early stages can still prevent transcription of the TEs in late stages with no need for piRNAs. The differential quantification of mature piRNAs produced in early versus late stages of oogenesis will provide a far better view of piRNAs produced during oogenesis.

Conclusions
From transcription to processing, multiple steps and numerous protein factors are required to drive piRNA precursors into the cytoplasmic structure where functional piRNAs are produced. In the follicle cells of Drosophila  Table S1 ovaries, RNAs transcribed from piRNA clusters are first spliced, presumably concomitantly to transcription [8], before being transferred throughout the nucleoplasm in an Exon Junction complex, UAP56 protein and Nxt1-Nxf1-dependent manner, to be assembled in a single nuclear focus [13,16]. Then, the transfer to the cytoplasm requires the export complex Nxt1-Nxf1 and our present study suggests that Armi and Yb proteins may somehow be implicated in this process (Fig. 4). Armi protein is thought to be able to bind several types of RNAs being exported -including mRNAs. In the cytoplasm, its specific interaction with piRNA precursors is regulated by its ATPase hydrolysis activity and Yb protein [21,25]. Our present study suggests that Armi could play an earlier role within the nucleus before RNA export. When in the cytoplasm, piRNA precursors are delivered to the Yb-bodies and mitochondria where Yb and Armi play distinct and collaborative roles to ensure the production of Zuc-dependent phased piRNAs [21,25].
Our study shows that a spatially and temporally controlled piRNA biogenesis exists with two critical periods taking place before and after the switch from a mitotic cycle to an endo-replication cycle. It can be anticipated that the major pool of somatic piRNAs is produced during the first period when piRNA precursors exit from the nucleus and Yb-bodies are correctly assembled.
Overall, these data show that in addition to factors specifically required for piRNA biogenesis, the spatiotemporal regulation of the whole system takes place at another level of complexity, the understanding of which will certainly help to interpret delayed and unexpected regulations.

Drosophila stocks
The used fly strains were: ISO1A from the collection of the GReD; armi Cell culture OSS cells were grown in prepared from Shields and Sang M3 Insect Medium (Sigma) supplemented with 0.6 mg/ ml glutathione, 10 mU/ml insulin, 10% fetal bovine serum and 10% fly extract.

RNA fish
The DNA fragment to prepare the specific flam 508 probe to detect flam transcripts was PCR amplified from the ISO1A line using primers 5′-ATTCTCCTTTCTCAG-GATGC-3′ and 5′-GCATTGCTACCTTACGTTTC-3′ and cloned into pGEMT easy vector.
Riboprobe was synthesized by digestion of pGEMT easy plasmids with NcoI or SpeI enzyme, followed by in vitro transcription using Sp6 or T7 polymerase and digoxygenin labeled UTP (Roche), DNAse I treatment and purification.

Immunofluorescence
Ovaries from 2-to 4-days-old flies were dissected in PBT fixed in 4% formaldehyde/PBT at RT for 20 min, rinsed three times with PBT, incubated 1 h in PBS-0.3% Triton, rinsed three times with PBT and incubated 1 h in TNB 0,3% Triton prior staining with anti-lamin, anti-Armitage or anti-Yb antibody. Secondary antibodies coupled to Cy3, Cy-5 or Alexa-488 were used.
Three-dimensional images were acquired on Leica SP5 and Leica SP8 confocal microscopes using a 40X objective, acquiring a stack of at least 10 slices with an interval of 0,5 nm for each follicle to ensure signal detection within the entire volume of the follicle cells. Only cells acquired in their totality were considered for quantifications. Manual scoring was performed for the presence/absence of flam RNA signal in follicle cells (Fig. 1a) and for the localization of flam RNA probe signal compared to lamin signal (Figs. 1b, 2b and 4b). The presence/absence of Armi and Yb foci in each follicle cell was assessed manually (Fig. 3b).