Yb body assembly on the flamenco piRNA precursor transcripts reduces genic piRNA production.

In Drosophila ovarian somatic cells, PIWI-interacting small RNAs (piRNAs) against transposable elements are mainly produced from the ∼180-kb flamenco (flam) locus. flam transcripts are gathered into foci, located close to the nuclear envelope, and processed into piRNAs in the cytoplasmic Yb bodies. The mechanism of Yb body formation remains unknown. Using RNA fluorescence in situ hybridization, we found that in the follicle cells of ovaries the 5'-ends of flam transcripts are usually located in close proximity to the nuclear envelope and outside of Yb bodies, whereas their extended downstream regions mostly overlap with Yb bodies. In flamKG mutant ovaries, flam transcripts containing the first and, partially, second exons but lacking downstream regions are gathered into foci at the nuclear envelope, but Yb bodies are not assembled. Strikingly, piRNAs from the protein-coding gene transcripts accumulate at higher levels in flamKG ovaries indicating that piRNA biogenesis may occur without Yb bodies. We propose that normally in follicle cells, flam downstream transcript regions function not only as a substrate for generation of piRNAs but also as a scaffold for Yb body assembly, which competitively decreases piRNA production from the protein-coding gene transcripts. By contrast, in ovarian somatic cap and escort cells Yb body assembly does not require flam transcription.

INTRODUCTION

The PIWI-interacting RNA (piRNA) pathway controls expression of transposable elements (TEs) in the somatic and germ cells of Drosophila ovaries. PIWI clade proteins guided by small (24–30 nt) piRNAs recognize complementary TE transcripts resulting in their degradation or transcriptional repression (for a review, see Czech and Hannon, 2016). In the somatic cells of ovaries, piRNA machinery operates via the single nuclear Piwi protein, whereas the germ cells possess also two cytoplasmic proteins of this family—­Aubergine (Aub) and Argonaute3 (Ago3), which are involved in the process of posttranscriptional silencing coupled with the piRNA amplification, the ping-pong cycle. Most piRNAs against TEs are transcribed as the long RNA precursors from the piRNA clusters, genome regions containing multiple damaged TE copies (Brennecke ). In the germ cells, piRNAs are mainly generated from the convergently transcribed dual-strand clusters, whereas in the somatic cells, piRNA clusters, exemplified by the flam locus, are unidirectionally transcribed (Klattenhoff ; Malone ; Pane ; Zhang ). Although both dual- and unistrand piRNA clusters are located predominantly in the heterochromatin, different mechanisms direct their transcripts for processing into piRNAs (for a review, see Huang ).

The flam piRNA cluster generates long, capped and alternatively spliced transcripts, with the short first exon represented in all spliced variants (Goriaux ; Mohn ). flam transcripts are assembled into one or two foci, designated DotCOM/Flam bodies (herein flam foci), located in close proximity to the nuclear envelope and adjacent to the cytoplasmic Yb bodies (Dennis , 2016; Murota ), where their processing into piRNAs occurs (Olivieri ; Saito ; Qi . The components of the exon junction complex (EJC) as well as the Nxf1/Nxt1 exportin complex were shown to participate in the transfer of flam transcripts away from their site of transcription to the nuclear flam foci and for their further export to the Yb body (Dennis ). Importantly, Yb body assembly appears to be dependent on the nucleocytoplasmic export of flam transcripts (Dennis ). Yb bodies are composed of Yb, Armitage (Armi), Vreteno (Vret), Shutdown (Shu), Sister of Yb (SoYb), and some other proteins required for piRNA biogenesis (Haase ; Olivieri , 2012; Saito ; Handler ; Zamparini ). One or two Yb bodies per cell are usually located nearby to mitochondria (Szakmary ; Saito ), which carries on its surface the Zucchini (Zuc) nuclease, responsible for cleavage of transcripts into piRNAs (Pane ; Ipsaro ; Nishimasu ). Mature Piwi–piRNA complexes then move into the nucleus where they induce transcriptional silencing of TEs.

In the somatic cells of ovaries, generation of piRNAs is thought to be defined by the presence of specific nucleotide motifs in piRNA precursors that facilitate binding of the piRNA biogenesis machinery and further recruitment of nucleases for primary piRNA processing (Homolka ; Ishizu ). Artificial recruitment of Yb or Armi proteins to transcripts was shown to be sufficient for their processing into piRNAs (Pandey ), and the cis-regulatory elements targeting the downstream regions of RNAs for piRNA production were recently identified in flam first and second exons as well as in the 3′-untranslated region (3′-UTR) of traffic jam (tj) gene (Homolka ; Ishizu ). These sequences appear to be bound by Yb protein according to the CLIP assay (Ishizu ). However, similar cis-regulatory elements were revealed in the luciferase (Homolka ) as well as in the CG9257 (Ishizu ) genes indicating their widespread presence in the genome. In agreement with this, a number of gene transcripts in ovarian somatic cells (OSCs) including tj were shown to generate piRNAs (Robine ; Saito ). It is currently unclear whether these cis-­elements are sufficient for the delivery of transcripts into the Yb body and/or for Yb body assembly. Thus, the mechanisms that define selective piRNA production from the piRNA clusters in somatic cells remain underexplored.

Here we demonstrate that in the follicle cells of ovaries, the 5′-ends of flam transcripts are accumulated in close proximity to the nuclear envelope, whereas extended downstream transcript regions overlap with the Yb bodies. The 5′-ends of flam transcripts appear to be sufficient for their aggregation in the flam foci, but not for Yb body assembly. Intriguingly, in flam mutant ovaries lacking Yb bodies, piRNA production from 3′-UTRs of the genic mRNAs was increased. These results indicate that Yb body assembly is not necessary for piRNA biogenesis itself, but is essential for the preferential production of piRNAs against TEs compared with genic transcripts.

RESULTS

The first flam exon and the downstream flam transcript regions are differently positioned relative to Yb bodies and nuclear envelope

Previously, it was shown that in both the follicle cells of ovaries and OSC culture flam transcripts were assembled into one or two DotCOM/Flam bodies juxtaposed with the Yb bodies (Dennis , 2016; Murota ). However, there was controversy concerning the localization of these bodies in OSCs and follicle cells of ovaries. In OSCs, Flam bodies were found in the cytoplasm (Murota ), whereas in follicle cells of ovaries both nuclear and cytoplasmic flam foci were observed (Dennis ). We performed RNA FISH in OSCs with the flam fourth exon probe (#508; Figure 1A; Dennis ) coupled with immuno­staining of the nuclear envelope (with anti-lamin Dm0 antibody) and Yb bodies (with anti-Yb antibody) and observed FISH signals in both nucleus and cytoplasm (Supplemental Figure S1), similarly to the results previously obtained by Dennis in the follicle cells of ovaries. Moreover, in a fraction of OSCs we detected the simultaneous presence of flam signals at both sides of the nuclear envelope, with the cytoplasmic portion overlapping with the Yb bodies (Supplemental Figure S1). The doubling of signal likely reflects an intermediate stage of translocation through the nuclear pore of multiple flam transcripts from the point of their assembly within the nucleus to the newly formed Yb body. Clearly, this indicates that flam transcripts in both the follicle cells and OSCs are first gathered in the foci within the nucleus and after that are transported to the cytoplasmic Yb body, the center of their processing into piRNAs.

FIGURE 1:

The fourth and distal exons of flam transcripts are mostly localized in Yb bodies, whereas the flam first exon is not. (A) Scheme of the flam locus with the designation of exon/intron structure (according to Goriaux and FlyBase r6), sites of TE insertions in flam and flam mutations (black triangles), RNA FISH probes (black rectangles), and primer pairs used for RT-qPCR (arrowheads). (B) RNA FISH (red) with flam fourth exon probe (left panels), distal exon probe (middle panels), or first exon probe (right panels) coupled with immunostaining of Yb bodies with anti-Yb antibody (green; top panels) and nuclear envelope with anti-lamin Dm0 antibody (blue; enlarged view of representative nuclei; bottom panels) in the follicle cells of stage 12 egg chambers. Full Z-series projections of confocal image stacks are shown in the top panels, and single optical sections are shown in the bottom panels. Scale bars are 5 μm in the top panels and 1.5 μm in the bottom panels. (C) Violin plots showing the distribution of distances between centers of Yb bodies and flam RNA FISH signals for the first (red), second (magenta), fourth (blue), or distal (dark blue) exon probes in the follicle cells of early egg chambers. Distances were measured with IMARIS software for 40–50 cells in two biological replicates for each probe. (D) Violin plots showing the distribution of distances from the centers of flam RNA FISH signals and Yb bodies to the nuclear lamina. Distances for 40–50 cells in two biological replicates (for first, second, fourth, and distal exons in WT ovaries), in one replicate (for first exon in flam/Df ovaries), or in eight replicates (for Yb bodies) were measured with IMARIS software. In C and D M-W U-test was used for the pairwise comparison of distributions between the distances of the first exon-Yb, the second exon-Yb, and the fourth exon-Yb (C), or between the distances of the first, second, fourth, and distal exons and Yb bodies to the nuclear lamina (D).

Further, we analyzed the distribution of flam transcripts in the ovaries using several RNA FISH probes to different flam regions. In the follicle cells of germaria and early egg chambers, FISH with the flam fourth exon probe revealed the frequent overlap of FISH signals with Yb bodies (Figure 1B, left panels). The signals of #681 probe (Dennis ), corresponding to the distal flam exon (Figure 1A), also usually overlapped with Yb bodies (Figure 1B, middle panels). However, in some follicle cells distal flam exon signals were not revealed, which may be either because they are not included in all flam splice variants or because they are more rapidly cleaved into piRNAs. Strikingly, when we employed the probe for the first flam exon, the overlap of RNA FISH signals with the Yb bodies was drastically reduced (Figure 1B, right panels). The probe for the extended part of the second exon was intermediately overlapped with the Yb bodies. Accordingly, median distances between the centers of Yb bodies and RNA FISH signals were ∼0.2 µm for the distal exon, ∼0.3 µm for the fourth exon, ∼0.4 µm for the second exon, and ∼0.5 µm for the first exon (Figure 1C). These findings indicate that the 5′-end of flam RNA (the first exon) is mainly located out of the Yb body, whereas the long downstream region of flam transcript is associated with the Yb body. Moreover, the median three-dimensional (3D) distances to the nuclear lamina were notably different between the first and the downstream flam exons (Figure 1D). Whereas the first flam exon was mostly positioned at the nuclear lamina, the downstream transcript regions protruded further into the cytoplasm (at 0.3 µm median distances), and the second exon occupied an intermediate position (at 0.1 µm median distance). Thus, the downstream flam transcripts are more localized in the Yb body than the 5′-end of flam RNA.

Interestingly, we failed to detect flam transcripts simultaneously containing both first and third exons by the conventional RT-PCR, whereas these transcripts were revealed by more sensitive RT-PCR with an exon-specific primer (Supplemental Figure S2). Therefore, the continuous flam transcripts with both first and third exons seem to be very short-lived, whereas first and third exons mostly exist as the separate RNAs.

The presence of the first and a part of the second flam exon is sufficient for the formation of flam foci but not for the Yb body assembly

It was shown that flam and flam mutations are caused by artificial transposon insertions in the promoter region or in the second exon of the flam locus, respectively (Figure 1A), resulting in a more than 10-fold decrease of flam transcript level (Brennecke ; Mével-Ninio ). In the flam line, the TE insertion disrupts the binding site for the Cubitus Interruptus transcription factor located upstream of TSS, thus impairing flam transcription (Goriaux ). However, in the case of flam mutation the mechanism of flam transcription disturbance was less clear. To address this and further questions we analyzed mutant ovaries bearing flam or flam chromosomes combined with the X chromosome deletion covering the whole flam locus (Df), which allowed us to exclude the influence of genetic background. In addition, flam/Df and flam/Df ovaries exhibited more normal morphology than the homozygous mutants. We hypothesized that in flam mutant the termination of flam transcription may occur at the TE sequence, resulting in the retention of the first exon and part of the second exon but not of the downstream transcript regions. Indeed, by RT-qPCR analysis we revealed the preservation of the flam first exon transcripts and the lack of the third exon transcripts in flam/Df ovaries, whereas both transcript regions were absent in flam/Df ovaries (Figure 2A).

FIGURE 2:

flam first exon transcripts are retained and form perinuclear foci in flam/Df ovaries. (A) RT-qPCR analysis of flam transcription in flam and flam mutant ovaries using primer pairs for the first or third exons. Fold change of flam transcription in the ovaries with indicated genotypes relative to flam/flam ovaries, normalized on the Adh transcript levels, is indicated along the Y-axis (mean ± SD for two replicates). The flam third exon transcripts were barely detected in flam and flam mutants, but the flam first exon transcripts are retained in flam mutant. (B) RNA FISH (red) with flam first exon probe coupled with immunostaining of Yb bodies with anti-Yb antibody (green) and nuclear envelope with anti-lamin Dm0 antibody (blue) in the follicle cells of the stage 2 egg chamber of flam/+ (left panel), flam/Df (central panel), and enlarged view of flam (right panel). Scale bars are 5 μm in the left and central panels and 0.7 μm in the right panel.

Next, we assayed by RNA FISH whether the transcripts bearing only the first and part of the second flam exon in the flam/Df ovaries are assembled into foci. Although less efficient than in wild type (WT), the assembly of flam foci was detected using the first exon probe in flam/Df but not in flam/Df ovaries (Figure 2B and Supplemental Figure S3). At the same time, flam fourth exon FISH signals were not revealed in either flam/Df, or in flam/Df ovaries (Supplemental Figure S3). Furthermore, median 3D distances from flam first exon signals to the nuclear lamina in flam/Df ovaries were similar to that seen in WT ovaries (Figure 1D), suggesting the sufficiency of this region for the aggregation of transcripts into foci near the nuclear envelope.

Recently, Dennis have found that the complete absence of flam transcripts in flam ovaries or the blocking of their nuclear export in the nxf1 or nxt1 mutants resulted in Yb body disintegration. Additionally, it was shown that the first and second exons of flam transcripts contain the cis-regulatory signals required for Yb protein binding and for the production of piRNAs from downstream sequences (Homolka ; Ishizu ). We therefore investigated whether the presence of only the first and part of the second flam exon transcripts in flam ovaries is sufficient for the assembly and/or maintenance of the Yb bodies. Immunostaining with anti-Yb antibody revealed the disintegration of Yb bodies in the follicle cells of flam and flam mutants (Figure 3, A and B). We note that in many cases Yb bodies were almost completely dispersed in the cytoplasm of flam as well as flam ovaries, resulting in the underestimation of quantitative effects of mutations in Figure 3B. Moreover, even when incompletely disintegrated, Yb body remnants were not juxtaposed with RNA FISH signals of the flam first exon (Figure 2B). Another component of Yb body, the Armi protein, became dispersed as well in flam mutants (Supplemental Figure S4A). Therefore, whereas the transcripts corresponding to the first flam exon are assembled into flam foci, the presence of flam downstream sequence(s) and/or the longer transcripts are necessary for the formation and maintenance of the nearby Yb bodies. Of note, the intact Yb bodies were observed in piwi mutant follicle cells (Supplemental Figure S4A; Olivieri ), ruling out the possibility that the disintegration of Yb bodies in flam mutants is an indirect effect of the loss of piRNAs or of the activation of TEs.

FIGURE 3:

Disintegration of Yb bodies in flam mutants. (A) Immunostaining of follicle cells from fourth stage egg chambers in flam/+ (WT), flam/flam, and flam/Df ovaries with anti-Yb (green) and anti-lamin Dm0 (red) antibodies. Scale bars are 7 μm. (B) Box plots showing a distribution of volumes of Yb bodies from egg chambers of stages 3–5 measured with IMARIS software in flam mutants in comparison with the corresponding controls. The number of counted signals is indicated for each genotype. M-W U-test was used for the pairwise comparison of distributions between mutant and control ovaries.

piRNA production from the gene transcripts is enhanced in flam ovaries

Interestingly, in spite of the lack of Yb bodies, the total amount of Yb protein remained nearly unchanged in flam mutants (Supplemental Figure S4B). Moreover, we observed some smaller aggregates of colocalized Yb and Armi proteins in the cytoplasm of follicle cells of flam mutants (Supplemental Figure S4C) that potentially may be involved in piRNA biogenesis. It is known that loading of piRNAs into Piwi is required for its nuclear import (Olivieri ; Saito ). In follicle cells of flam mutants, we revealed the unaltered nuclear localization of Piwi (Supplemental Figure S5) indicating the robust piRNA generation in these cells.

Next, to examine piRNA production in the follicle cells of flam mutant ovaries, we performed deep sequencing of small RNA libraries obtained from flam/Df and flam+ ovaries. In flam mutant, piRNA production from the flam locus was greatly reduced along its entire length (Figure 4A), including the first exon and part of the second exon that are presented in the flam line (Supplemental Figure S6). This observation suggests that the 5′-end of flam transcripts is likely not exported to the cytoplasm for piRNA processing in flam mutant. Then we evaluated the effect of flam mutation on the piRNAs generated from the known genic piRNA clusters operating in somatic cells (Robine ; Wen ). The abundance of piRNAs derived from the 3′-UTR of tj mRNA and from other genic piRNA clusters from somatic cells was significantly increased in flam mutant compared with control ovaries (Figure 4, A–C, and Supplemental Data S1). Of note, the levels of piRNAs generated from other piRNA clusters which are predominantly germline-­specific (Malone ; Mohn ), as well as the levels of microRNAs abundant in OSCs, were not noticeably altered (Figure 4A). Our analysis of the publicly available small RNA libraries (Malone ; which, however, were sequenced drastically less deeply than our libraries) confirmed the enhanced generation of genic piRNAs in flam mutant (Supplemental Data S1). Thus, in follicle cells of ovaries, piRNA production from the protein-coding gene transcripts becomes even more effective when Yb body assembly is impaired, indicating that this process can be mediated by piRNA machinery components dispersed in the cytoplasm. Collectively, these data imply that the assembly of Yb bodies in follicle cells makes piRNA processing preferable toward the flam transcripts relative to the protein-coding gene transcripts.

FIGURE 4:

piRNA production from the protein-coding gene transcripts is enhanced in flam ovaries. (A–C) piRNA density (in RPM) for uniquely mapped reads along the entire flam (A), 3′-UTR of tj (B), and 3′-UTR of brat (C) piRNA clusters. Sense and antisense piRNAs are indicated in red and blue, respectively. (D) Scatter plot showing an abundance of piRNAs (in log2 RPM) for somatic genic piRNA clusters (red dots), predominantly germline piRNA clusters (green dots), and for the top 10 microRNAs most abundant in OSCs (purple dots) in flam/+ (control) or flam/Df small RNA-seq libraries. The median fold change for somatic genic piRNA clusters and for predominantly germline piRNA clusters upon flam mutation is 1.6 and 1.1, respectively. The difference in piRNA production between “germline” and genic clusters is highly significant (P < 0.0001, M-W U-test).

Yb body formation depends on the flam transcripts in the follicle cells but not in the cap and escort cells

As we and others (Dennis ) have demonstrated the disintegration of Yb bodies in flam mutant ovaries, the question arises whether the morphology of flam foci and Yb bodies is changed during normal ovarian development. The Drosophila adult ovary is composed of 12–16 ovarioles that contain a germarium and usually six to seven progressively developing egg chambers of sequentially larger sizes. The anterior end of the germarium contains germ stem cells and their daughter cystoblast cells surrounded by a somatic niche that includes terminal filament cells, cap cells, and escort cells. After mitotic divisions, germline cysts become encapsulated by a monolayer of somatic follicle cells leading to the formation of the first stage egg chamber in the germarium region 3. The egg chamber then separates from the germarium and progresses through 14 stages of maturation (Spradling ).

RNA FISH in WT ovaries with flam fourth or first (unpublished data) exon probes coupled with immunostaining of Yb bodies and the ovarian follicle cell marker Tj revealed that flam foci are most prominent in the germaria regions 2b and 3 (first stage egg chamber) as well as in stage 2 to 3 egg chambers, whereas at the later stages of oogenesis flam signals become weaker (Figure 5, A and B, and Supplemental Figure S7). Importantly, this flam foci reduction in the late stage follicle cells correlated with the decrease of Yb body size during the progress of oogenesis (Figure 5, A and C), thus accentuating the link between flam foci and Yb bodies in this cell type.

FIGURE 5:

In ovarian follicle cells flam foci as well as the Yb bodies become progressively smaller during egg chamber maturation. (A) RNA FISH with flam fourth exon probe (red) in WT (flam/+) ovaries coupled with the immunostaining of Yb bodies by anti-Yb antibody (green) and anti-Tj antibody (violet), the nuclear marker of OSCs. Egg chambers at different stages in the same ovariole are shown. The apical region of a germarium is populated by cap and escort cells (outlined by a white dotted line), and the first stage follicle cells are located distally (outlined by a red dotted line). flam foci are most prominent in the first stage follicle cells of germaria (top panel) as well as in stage 2–3 egg chambers, whereas at the later stages of oogenesis (middle and bottom panels) flam as well as Yb body signals become weaker. Scale bars are 10 μm in the top panel and 7 μm in the middle and bottom panels. (B, C) Box plots showing a distribution of volumes of the flam fourth exon (B) or the Yb body (C) signals at different stages of egg chamber maturation in WT (flam/+) ovaries. Volumes of flam foci or Yb bodies from the indicated stages were measured with IMARIS software. The number of counted signals is indicated for each stage. M-W U-test was used for the pairwise comparison of distributions. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001; ns, not significant.

However, in the cap and escort cells, located at the anterior end of germaria, flam fourth or first exon RNA FISH signals were weaker than in the follicle cells of the first egg chamber, while the Yb bodies stained more strongly (Figure 6, A–D). Many cap and escort cells contained large Yb bodies and completely lacked flam foci. Furthermore, immunostaining of flam/Df and flam/Df ovaries with anti-Yb antibody revealed only a slight decrease of Yb body sizes in the cap and escort cells in both mutants (Figure 6, A–D). Therefore, even upon the complete absence of flam transcripts, the Yb bodies do not disintegrate, thus indicating a different mechanism of Yb body formation/maintenance in cap and escort cells.

FIGURE 6:

In cap and escort cells the assembly of Yb bodies does not depend on flam transcription. (A) RNA FISH with flam first (top panel) or fourth (bottom panel) exon probes (red) coupled with Yb (green) and Tj (violet) immunostaining. In cap and escort cells (outlined by a white dotted line in germarium) of WT (flam/+) ovaries (left panel) flam RNA FISH signals are weaker than in the distally located first egg chamber follicle cells, while the Yb bodies are more strongly stained. In cap and escort cells of flam mutants (middle and right panels) Yb bodies are assembled. Scale bars are 10 μm. (B) Germaria of PZ1444 (WT, left panel) or flam; PZ1444 (right panel) ovaries (PZ1444 is a lacZ enhancer trap line that is expressed in the cap and escort cells; Margolis and Spradling, 1995; Xie and Spradling, 2000) immunostained with anti-Yb (green) and anti–β-galactosidase (red) antibodies, counterstained with DAPI (blue). Yb bodies are larger in the cap and escort cells (their nuclei are marked by red) than in the follicle cells (outlined by a red dotted line). In flam/Df ovaries Yb bodies are disintegrated in follicle cells, but not in the cap and escort cells. Scale bars are 10 μm. (C, D) Box plots showing a distribution of volumes of flam fourth exon (C) or Yb body (D) signals in the cap and escort cells (EC) or in the first stage egg chamber follicle cells (1 st FC) of WT (flam/+) or flam/Df ovaries. The number of counted signals is indicated for each cell type. M-W U-test was used for the pairwise comparison of distributions.

DISCUSSION

It was recently shown that in ovarian follicle cells flam piRNA precursor transcripts were assembled into foci, located close to the nuclear envelope and juxtaposed with Yb bodies, their cytoplasmic processing centers (Dennis ; Murota ). Our data (Supplemental Figure S1) and the results of Dennis demonstrate that both in OSCs and in ovarian follicle cells flam transcripts are first gathered into foci within the nucleus and then are exported to the cytoplasm where they, with the exception of the 5′-end, become mainly localized in Yb bodies (Figure 1).

We found that the degree of colocalization of exported to the cytoplasm flam transcripts with the Yb body was progressively increased for the downstream flam regions (Figure 1C), likely reflecting the location of 5′-termini of flam transcripts outside the Yb body, and the rest of the transcript parts—­inside the Yb body. Moreover, the absence of flam transcripts in the flam and flam mutants (Figure 6A and Supplemental Figure S3; Dennis ) or their decline at the later stages of oogenesis (Figure 5, A and B) is accompanied by the disintegration of Yb bodies (Figures 3, A and B, and 5, A and C). Collectively, these results strongly support the model that flam transcripts serve not only as a substrate for piRNA production but also as a scaffold for Yb body assembly from Yb and other piRNA machinery proteins. This finding indicates the functional similarity of flam transcripts with architectural lncRNAs, which are responsible for the assembly of different protein bodies, such as paraspeckles, omega speckles, and others (for a review, see Ip and Nakagawa, 2012; Chujo and Hirose, 2017).

The 5′-end of flam transcripts producing few piRNAs (Murota ) and only weakly colocalizing with the Yb bodies (Figure 1C) likely performs additional, but currently unknown, functions. First, the first and the second flam exons almost obligatorily present in all flam transcripts (Goriaux ). Second, flam 5′-end contains the short cis-elements that bind Yb protein and are responsible for the targeting of downstream transcripts for processing into piRNAs (Homolka ; Ishizu ). Third, our results in flam mutant ovaries indicate that flam transcripts containing only the first and, partially, the second flam exon can still be gathered into foci near the nuclear envelope (Figure 1D). However, flam 5′-regions are not sufficient (or may not even be required) for Yb body formation (Figure 3, A and B), likely because the long flam transcripts carrying multiple sites for Yb protein binding (Murota ) are necessary to assemble enough Yb molecules. Moreover, the peculiarities of behavior of the 5′-end of the flam transcripts imply that this region may be either cut off from the rest of the flam transcripts thus remaining in the nucleus, or shielded by an unknown complex from piRNA processing after export from the nucleus. The low abundance of continuous flam transcripts carrying both first and third exons in WT ovaries (Supplemental Figure S2) supports the former possibility.

What might be the biological function of Yb body assembly? Compartmentalization is an important feature of biological systems that increases the efficiency and specificity of different cellular activities. We propose that the compartmentalization of piRNA biogenesis machinery in the Yb body reduces the concentration of the dissolved Yb body components and thus diminishes the piRNA production from protein-coding gene transcripts, which may by chance possess cis-elements for Yb binding. Indeed, even higher levels of piRNAs from tj and other genic transcripts were generated in somatic cells of flam mutant ovaries (Figure 4 and Supplemental Data S1). Therefore, if flam transcripts are absent (as in flam ovaries) or too short (as in flam ovaries), the dispersed in the cytoplasm Yb body components can be gathered on the subset of protein-coding gene transcripts and induce their processing into piRNAs more efficiently than in WT ovaries.

In the Drosophila ovarian germline, transcription of the dual-strand piRNA clusters is defined by the Rhino-Deadlock-Cutoff (RDC) complex (Mohn ; Zhang ). Transcripts of germline clusters recruit the TREX complex (including RNA helicase UAP56) for their transport to the cytoplasmic piRNA biogenesis centers localized in the nuage (Zhang ; Hur ). In the ovaries, mutant for genes of RDC factors or UAP56, the cluster transcripts are not exported from the nucleus, and nuage components become dispersed in the cytoplasm (Klattenhoff ; Pane ; Zhang ). Strikingly, upon these conditions, genic piRNAs become more abundant (Zhang ). Therefore, as in the case of flam, the transcripts from dual-strand clusters are required for the assembly of nuage structure and this compartmentalization can reduce nonselective piRNA production.

We also show that the assembly of Yb bodies on the flam transcripts operates only in the follicle cells of ovaries. Although the flam locus is required for the normal functioning of somatic escort cells (Upadhyay ), heavily stained Yb bodies may be assembled in the cap and escort cells in the absence of flam transcripts (Figure 6, A–D). This might indicate that in these cell types the transcripts from other piRNA clusters provide the RNA backbone for Yb body formation, or that other protein constituents of the Yb body are abundantly expressed thus aggregating the Yb bodies without flam RNA. Interestingly, when visualized by electron microscopy, Yb bodies appear to be located adjacent to an unknown RNA-rich structure in the cap and escort cells, whereas in the follicle cells such neighborhood was not reported (Szakmary ).

In conclusion, our results uncover new features of the compartmentalization of piRNA biogenesis machinery in the follicle cells of ovaries, necessary for the preferential formation of piRNAs against TEs.

MATERIALS AND METHODS

Fly stocks

Drosophila melanogaster stocks were maintained under standard conditions at 25°C. w (#13912; herein flam), y (#16453; herein flam), and Df(1)Exel6255, w (#7723; herein Df) mutants were obtained from the Bloomington Drosophila Stock Center. For the analysis of flam mutations, we obtained transheterozygotes by crossing Df females with flam or flam males. For the analysis of piwi mutation, piwi and piwi stocks (Lin and Spradling, 1997) were crossed. To visualize escort cells, we combined flam mutations with a PZ1444 lacZ enhancer trap line that is expressed in cap and escort cells (Margolis and Spradling, 1995; Xie and Spradling, 2000). Fly stock bearing Armi-GFP transgene (Olivieri ) was kindly provided by J. Brennecke (Institute of Molecular Biotechnology of the Austrian Academy of Sciences).

OSC maintenance

OSCs (Niki ), kindly provided by M. Siomi (The University of Tokyo), were cultured in Shields and Sang M3 insect medium (Sigma-Aldrich) supplemented with 10% heat-inactivated fetal bovine serum (Life Technologies), 10% fly extract (http://biology.st-andrews.ac.uk/sites/flycell/flyextract.html), 10 µg/ml insulin (Sigma-Aldrich), 0.6 mg/ml glutathione (Sigma-Aldrich), 50 U/ml penicillin, and 50 µg/ml streptomycin.

RNA isolation and RT-qPCR analysis

Total RNA was isolated from manually dissected ovaries using Trizol reagent (Invitrogen) and purified from genomic DNA by DNA-free kit (Ambion). Total RNA (1 μg) was used for the reverse transcription (RT) reaction with random primer and Superscript II reverse transcriptase (Invitrogen). The resulting cDNAs in at least three biological replicates were analyzed by RT-qPCR performed in a MJ Mini instrument (Bio-Rad) using SYBR Green chemistry (Applied Biosystems). For detection of flam transcripts containing first and third joint exons, either 1 μg of total RNA with random primer or 15 μg of total RNA with the specific for the third exon reverse primer was used for the RT reactions. Then, in the case of third exon-specific primer reaction, cDNAs were treated with RNase H (Invitrogen; 10 U, 15 min) and purified by Wizard Gel and PCR Clean-Up System (Promega). Primer pairs for RT-qPCR are listed in Supplemental Table S1.

Immunostaining

Immunostaining was performed basically as described in Shpiz with some modifications. OSCs in the growth phase were collected and rinsed two times in phosphate-buffered saline (PBS). Ovaries from 1- to 5-d-old females were manually isolated in PBT (PBS containing 0.01% Tween-20) at 4°C and then rinsed in PBS. Cells or ovaries were fixed in 4% formaldehyde (on PBT) for 25 min at room temperature. Fixation was stopped by incubation with 0.25 M glycine (Sigma-Aldrich) for 5 min. Then cells or ovaries were washed in PBS three times for 10 min at room temperature, permeabilized with PBTX (PBS with 0.1% Tween-20, 0.3% Triton X-100) for 10 min, blocked with PBTX containing 3% normal goat serum (NGS; Invitrogen) at room temperature for 1 h (cells) or for 3 h (ovaries), incubated with primary antibody in PBTX containing 3% NGS for 3 h (cells) or for 7 h (ovaries) at room temperature, or overnight at 4°C, washed in PBTX three times for 10 min at room temperature, incubated with secondary antibodies (1:1000) in PBTX containing 3% NGS for 3 h (cells) or for 7 h (ovaries) at room temperature, or overnight at 4°C, and then washed in PBTX three times for 10 min at room temperature in a dark chamber. Coverslips were mounted with a drop of SlowFade Gold Antifade reagent (Invitrogen) containing DAPI. As the primary, rabbit polyclonal anti-lamin Dm0 (1:500; Osouda ), mouse monoclonal anti-Yb (1:100; Murota ), anti-Armi (1:50; Saito ), anti-Piwi (1:500; Saito ), chicken anti–β-galactosidase (1:500; Abcam; ab9361), and guinea pig anti-Tj (1:5000; Li ) antibodies were used. As the secondary, Alexa Fluor 546–conjugated goat anti-rabbit immunoglobulin G (IgG) (Invitrogen) or Alexa Fluor 488–conjugated goat anti-mouse IgG (Invitrogen) antibodies were used.

RNA FISH coupled with immunostaining

RNA FISH was performed basically as described in Shpiz with some modifications. DNA fragments corresponding to flam RNA FISH probes were PCR-­amplified on the Drosophila genomic DNA using primer pairs (Supplemental Table S1) carrying the T7 RNA polymerase promoter sequence at their 5′-ends. Digoxigenin (DIG)-11-UTP (Roche) labeling of transcripts generated from these fragments by T7 RNA polymerase was carried out according to manufacturer’s instructions. RNA probes were dissolved in 100 µl of hybridization buffer (HB; 50% formamide, 4× SSC [1× SSC: 150 mM NaCl, 15 mM Na3 citrate, pH 7.0], 100 mM Na3PO4, pH 7.0) at a concentration of ∼10 µg/ml.

OSCs in the growth phase or ovaries, manually isolated from 1- to 5-d-old females, were rinsed twice with PBS and then fixed with 4% formaldehyde (on PBT) for 25 min. Fixation was stopped by incubation with 0.25 M glycine (Sigma-Aldrich) for 5 min. Cells or ovaries were washed in PBT three times for 10 min, permeabilized in PBTX for 10 min, blocked with PBTX containing 3% NGS (Invitrogen) for 1 h at room temperature, and washed in PBTX three times for 10 min. Cells or ovaries were prepared for hybridization by sequential incubation for 15 min at 42°C each in 20, 50, 80, and 100% HB on PBT. RNA FISH probes (in HB) were heated at 80°C for 5 min and then cooled on ice. Herring sperm DNA (Promega) and yeast tRNA (Invitrogen) were then added to the probes (both to 0.5 µg/µl), and RNA hybridization was performed overnight at 42°C. Cells or ovaries were washed sequentially for 15 min at 42°C in 50% formamide, 2× SSC; 40% formamide, 2× SSC; 30, 20, and 10% formamide on PBT. Further, cells or ovaries were washed several times in PBT at room temperature, blocked in PBT containing 3% NGS at room temperature for 2 h, incubated with primary antibody in PBT containing 3% NGS for 7 h at room temperature, or overnight at 4°C, and then washed in PBT three times for 10 min. RNA FISH signals in ovaries were subjected to a tyramide amplification procedure according to manufacturer’s instructions (Perkin Elmer). Cells or ovaries were incubated with the secondary antibody (1:1000) in PBT containing 3% NGS for 7 h at room temperature, or overnight at 4°C, and then washed in PBT three times for 10 min in a dark chamber. Coverslips were mounted with a drop of SlowFade Gold Antifade reagent (Invitrogen) containing DAPI. As the primary, rabbit polyclonal anti-lamin Dm0 (1:500; Osouda ), mouse monoclonal anti-Yb (1:100; Murota ), and sheep polyclonal anti–DIG-FITC (1:200; Roche) antibodies were used. As the secondary, Alexa Fluor 546–conjugated goat anti-rabbit IgG (Invitrogen), Alexa Fluor 633–conjugated goat anti-mouse IgG (Invitrogen), and Alexa Fluor 488–conjugated goat anti-Fluorescein/Oregon Green IgG (Invitrogen) antibodies were used.

Western blot analysis

Ovarian lysates were fractionated by SDS–PAGE (10% acrylamide gel) and transferred to a polyvinylidene difluoride membrane (Immobilon-P; Millipore). Blots were developed using alkaline phosphatase–conjugated secondary antibody (Sigma) and the Immun-Star AP detection system (Bio-Rad). The following primary antibodies were used: mouse monoclonal anti-Yb (1:500; Murota ) and mouse monoclonal anti-beta actin (1:3000; Abcam; ab8224).

Measuring distances between RNA FISH signals, Yb bodies, and nuclear lamina

Three-dimensional image stacks were recorded with a confocal LSM 510 Meta laser scanning microscope (Zeiss). Optical sections with 0.5- to 1.0-μm intervals along the Z-axis were captured. Images were processed and analyzed using IMARIS 7.4.2 software (Bitplane AG) with blind experimental setup. Images were thresholded to eliminate hybridization and immunostaining background, and nuclei containing Yb body foci obviously located in the cytoplasm were taken for further analysis. The distances between the signals and the nuclear envelope were counted as was previously described (Shevelyov ). Briefly, the nuclear rim colored by anti-lamin Dm0 antibody was manually outlined in each section by the middle to automatically reconstruct the surface of the nuclei. One “measurement point” was positioned in the optical section with the brightest FISH or Yb signal, at its visually determined center, and another one was placed on the reconstructed nuclear surface at the point of its earliest intersection with the progressively growing sphere of the first “measurement point.” The distance between two “measurement points” (the shortest distance between the RNA FISH signal or Yb signal and the nuclear rim as well as the distances between the centers of the RNA FISH and Yb signals) was measured for each nucleus (Supplemental Figure S8). Data were obtained in two independent experiments for 40–50 nuclei per experiment (Supplemental Table S2).

Measuring of Yb body volumes in WT and mutant ovaries

Confocal images containing several hundreds of flam RNA FISH signals or immunostained Yb bodies from two to three independent experiments were processed and analyzed with the same set of parameters for WT or for flam mutant ovaries using IMARIS 7.4.2 software (Bitplane AG). The surfaces of flam signals or Yb bodies were automatically reconstructed. For background subtraction, the images were thresholded to the 40% (for Yb) or 25% (for flam) of maximal intensity of the channel because at this value the Yb or flam fluorescence was not detected outside egg chambers. The volumes of flam signals or Yb bodies exceeding the background threshold (>0.09 µm3 for flam signals or >0.01 µm3 and <10 µm3 for Yb signals) were retrieved for further analysis.

Small RNA cloning

Small RNA cloning was performed as described in Muerdter . In brief, total RNA was isolated from flam/Df or flam/+ ovaries by Trizol reagent (Invitrogen), resolved on a denaturing 12% polyacrylamide gel, and small RNAs corresponding to 19–29 nt fraction were isolated. Libraries were barcoded according to Illumina TrueSeq Small RNA sample prep kit instructions and sequenced using the Illumina NovoSec-6000 instrument. Small RNA sequencing data are deposited in the Gene Expression Omnibus (GEO) under GSE126016 accession number.

piRNA mapping

piRNA data for the flam/+ and flam/flam ovaries were retrieved from the GEO NCBI database (GSM379054, GSM379055; Malone ). Adapter sequences were removed using the “FlexBar” tool (Dodt ). Sequenced reads of small RNA libraries obtained in our work were trimmed for adapter sequences, masked for low-complexity or low-quality sequences and filtered by size ranging from 23 to 29 nt (to obtain piRNA fraction) using “Cutadapt” (Martin, 2011). For the analysis of microRNA abundance filtering by size was omitted. Processed reads were then uniquely mapped to Release 5.41 of Drosophila genome using “Bowtie2” (Langmead and Salzberg, 2012). “R” with the additional packages including “GenomicRanges” and “GenomicAlignments” (Lawrence ) was used to calculate reads per million (RPM) values for the various genomic regions: for the 3′-UTRs of genic piRNA clusters (their coordinates were obtained from Robine ), for the flam locus, for the collection of the predominantly germline piRNA clusters (including 42AB cluster; their coordinates were retrieved from Rosenkranz, 2016), as well as for the microRNAs (coordinates of pre-microRNAs were retrieved from FlyBase dmel_r5.57 annotation). For the analysis of data from Malone , only genes and microRNAs with read counts more than 10 were used. The top 10 expressed microRNAs in OSCs were identified by mapping reads from the OSC small RNA libraries (Sienski ; GSM1914769) to Drosophila genome and by counting reads mapped to microRNA genes using STAR (Dobin ). The scatter plots of RPM values were generated using “R” with the “ggplot2” package (Wickham, 2016).

Statistical analysis

For P value estimation, the Mann-Whitney (M-W) U-test was used for comparison of two sample distributions.

Click here for additional data file.

Click here for additional data file.