Drosophila small ovary gene is required for transposon silencing and heterochromatin organization, and ensures germline stem cell maintenance and differentiation

The eukaryotic genome is organized into structurally distinct and functionally specialized chromatin domains, called euchromatin and heterochromatin (Heitz, 1928). The euchromatic domain contains actively transcribed genes, whereas the heterochromatin is mainly associated with a repressive transcriptional state (Wang et al., 2016). The heterochromatin is enriched in repetitive elements and transposons that occupy the centromeric and telomeric regions of the chromosomes (Brutlag et al., 1978; Peacock et al., 1978). Another domain of the heterochromatin is formed at regulatory regions of genes that have to be transcriptionally repressed at specific stages of development. The heterochromatin is epigenetically defined by a combination of specific covalent modifications of histone molecules. The formation of the heterochromatin is accompanied by trimethylation of Histone 3 at Lysine 9 (H3K9me3), which recruits Heterochromatin protein 1a [HP1a, encoded by Su(var)205] initiating the formation of the repressive chromatin environment (Bannister et al., 2001; Rea et al., 2000). Heterochromatic domains are organized around HP1a into phase-separated liquid compartments that physically compact chromatin and recruit additional repressive components (Larson et al., 2017; Strom et al., 2017). Kinetic analysis of HP1-chromatin binding revealed a complex interaction between heterochromatin components; however, the precise molecular mechanisms required for formation and maintenance of heterochromatin domains are not completely understood (Bryan et al., 2017).

In eukaryotes, a heterochromatin-dependent, small non-coding RNA-based defence system has been evolved against transposon-induced mutagenesis (Tóth et al., 2016). Central components of this pathway are the Piwi-interacting RNAs (piRNAs). Long precursors of the piRNAs are transcribed from both uni-strand and dual-strand piRNA clusters containing transposon sequences (Brennecke et al., 2007). Following piRNA biogenesis in the cytoplasm, mature short piRNAs associate with members of the Piwi class of the Argonaute protein superfamily (Piwi, Aub and Ago3 in Drosophila) and form RISC complexes (Huang et al., 2017). In the germ cells, Aub- and Ago3-RISC complexes mediate the post-transcriptional silencing of the transposons by inducing the degradation of transposon transcripts in the cytoplasm. The Piwi-RISC complex, however, migrates into the nucleus and inhibits transposon transcription. In the somatic cells of the ovary, exclusively the Piwi-RISC-mediated transcriptional silencing inhibits transposon activity (Le Thomas et al., 2013; Malone et al., 2009).

Two steps of the piRNA pathway have been shown to depend on heterochromatin function (Sato and Siomi, 2018). First, long precursors of the piRNAs are transcribed from piRNA clusters located mainly at the heterochromatic regions of the genome. Disruption of heterochromatin formation by eggless/dSetdb1 (egg) or Su(var)205 mutations impedes the transcription of the clusters, which results in derepression of transposons (Rangan et al., 2011; Teo et al., 2018). The second heterochromatin-dependent step of the piRNA pathway is the transcriptional silencing of the transposon transcription. At the transposon loci, Piwi-RISC inhibits transposon transcription by inducing the formation of a repressive heterochromatic environment on the transposon loci (Le Thomas et al., 2013). Transcriptional silencing of the transposons includes the deposition of repressive H3K9me3 modification mark and the recruitment of HP1a to the chromatin of the transposon locus.

Drosophila oogenesis provides an excellent model for understanding mechanisms of heterochromatin formation and its function in gene expression regulation and transposon silencing. In the ovary, repeated divisions of germline stem cells (GSCs) ensures continuous production of germ cells (Eliazer and Buszczak, 2011). GSCs reside stem cell niches, which are located in the germaria at the anterior tip of the ovary (Lin and Spradling, 1993). The GSC niches are composed of three somatic cell types, terminal filament cells, cap cells and escort cells (ECs), which provide physical and signalling milieu required for GSC self-renewal and differentiation (Fig. 1A) (Chen et al., 2011). Mitotic division of the GSC reproduces the GSC and generates a committed progenitor cell, the so-called cystoblast. The cystoblast has a limited division capacity and generates 16 interconnected cyst cells. One of the cyst cells differentiates into an oocyte whereas the remaining 15 cyst cells become supportive nurse cells. The developing germ cells are surrounded by an epithelial monolayer of somatic follicle cells forming an egg chamber.

In a previous RNAi-based screen for genes regulating germ cell behaviour, we have identified several essential chromatin regulators, such as Su(var)205 and Su(var)2-10, to be involved in germ cell development (Jankovics et al., 2014). In the same screen, we identified the annotated CG14438 gene, which has been shown to be involved in transposon silencing and to co-immunoprecipitate with HP1a (Alekseyenko et al., 2014; Czech et al., 2013; Muerdter et al., 2013). To gain a better understanding of chromatin regulation during germ cell development, we analysed the function of CG14438 in Drosophila oogenesis. Here, we show that CG14438 is identical to small ovary (sov) gene and it is a novel chromatin regulator that promotes heterochromatin formation by stabilizing the association of HP1a with the chromatin. Our results suggest that Sov suppresses transposon activity by regulating the transcription of the dual-stranded piRNA clusters and by the transcriptional silencing of the transposons. In the stem cell niche, sov function is required both in the somatic and in the germ cells to ensure GSC maintenance and differentiation.

CG14438 encodes a single large protein of 3313 amino acids (Fig. S1A). The N-terminal half of the protein is highly unstructured and contains a putative intrinsically disordered RGG/RG domain mediating degenerate specificity in RNA binding (Ozdilek et al., 2017; Thandapani et al., 2013). The C-terminal half of the protein contains 21 zinc-finger domains and a PxVxL pentamer motif, which is a canonical HP1-binding domain (Smothers and Henikoff, 2000).

We generated a null allele of CG14438, which removes the entire coding region of the gene (Fig. S1B). Animals homozygous for the novel deletion allele, which we called CG14438^del1, died at the third larval stage, indicating that CG14438 is an essential gene. Complementation analysis between CG14438^del1 and alleles of genes mapping to the same genomic region revealed that CG14438^del1 does not complement alleles of sov (Wayne et al., 1995).

To confirm that CG14438 corresponds to sov, we performed a series of rescue experiments. The Dp(1;3)DC486 duplication, which covers a 92.5 kb genomic region around the CG14438/sov locus, rescued all sov phenotypes, indicating that the sov gene is localized in this genomic region. Identical rescue was observed with two overlapping genomic transgenes [Dp(1;2)FF026056 and Dp(1;3)FF184439] (Fig. S1B). Sequencing of the sov² allele revealed a point mutation in the CG14438-coding region generating a premature stop codon after amino acid 3151 that results in a truncated mutant protein lacking 152 C-terminal amino acids (Fig. S1A). Taken together, the complementation analysis, the rescue experiments and the sequencing of the sov mutant allele indicate that CG14438 and sov are identical.

sov is an essential gene but its hypomorphic allelic combinations result in similar ovarian morphology to that of CG14438 RNAi (Jankovics et al., 2014; Wayne et al., 1995). To gain insights into the function of sov in germ cell development, we made use of the hypomorphic allelic combinations sov²/sov^ML150 and sov²/sov^del1. Females were sterile and exhibited rudimentary ovaries. In most of the germaria, no germ cells were found, indicating a role for sov in germ cell maintenance (Fig. 1C,F). In addition to the agametic germaria, ∼30% of the mutant germaria exhibited germ cell tumours (Fig. 1D,F). Each tumour cell contained a single spectrosome, a hallmark of the GSC or cystoblast-like undifferentiated germ cells, indicating that sov is required not only for GSC maintenance but also for GSC differentiation. The GSC maintenance and differentiation defects were rescued by the Dp(1;3)FF184439 transgene expressing the sov gene from its genomic context (Fig. 1E,F).

To narrow the temporal and spatial requirement of sov in germ cell development, tissue-specific RNAi and clonal analysis were performed. Silencing of sov with the germline-specific nosGal4 driver caused female sterility. Young females possessed normal-looking ovaries (Fig. 2B,D) and laid eggs that did not hatch. However, depletion of sov in the germ line resulted in a progressive loss of GSCs. In 4-week old sovRNAi females, most of the niches lost the GSCs and contained no germ cells, indicating that sov is required cell-autonomously in the germ line for GSC maintenance (Fig. 2C,D).

A cell-autonomous requirement for sov in germ cell maintenance was confirmed by analysis of sov mutant germline clones. Homozygous mutant sov^del1 germ cells were induced by FLP/FRT-mediated recombination and identified by the lack of the GFP marker gene (Fig. S2A,B). To analyse sov function at the larval stage, sov mutant germ line clones were induced in L1 larvae and the phenotype was analysed in adults. In 3-day-old females, GSCs were found in the sov mutants similar to the wild-type clones, indicating that sov mutant larval germ cells can populate the niche and can develop into normal GSCs (Fig. S2C). This indicates that sov function is dispensable in the germ line between L1 and the adult stage. To analyse sov function specifically in adult GSCs, clones were induced in the germ cells of young females. Both in the control and in the mutant niches, GFP-minus GSCs appeared in the first week after clone induction (ACI) (Fig. S2D). Wild-type control clones were maintained even later than 4 weeks ACI; however, the number of niches carrying sov mutant GSCs decreased (Fig. S2D).

Taken together, our data show that sov is required for GSC maintenance intrinsically in the germline at the adult stage. Remarkably, loss of sov in the GSCs located in niches composed of wild-type somatic cells did not induce tumour formation, indicating that the differentiation defect observed in sov mutants is not germ-cell-autonomous.

Formation of tumours composed of undifferentiated GSC-like cells and GSC loss could be a consequence of defects in the somatic cells of the niche. Depending on their position in the germarium, ECs have two distinct functions (Wang and Page-McCaw, 2018). Anterior ECs promote GSC self-renewal and maintain GSCs in stem cell state. Posterior ECs, however, promote GSC differentiation. Thus, loss of ECs in sov mutants results in a dual phenotype: GSC loss if the anterior ECs are lost, or a GSC tumour, if the posterior ECs die. In sov mutant niches, the number of the ECs was reduced, indicating that sov is required for EC survival (Fig. 3B,E). Consistent with the EC loss, we observed accumulation of activated Caspase3 in the sov mutant niches (Fig. S3). A similar reduction in EC numbers was observed when sov was silenced specifically in the somatic cells by the c587Gal4 driver, indicating that the requirement of sov for EC survival is cell autonomous (Fig. 3C,D,E).

Next, we investigated how somatic sov function affects germline behaviour. Therefore, the c587Gal4 line was used to induce sov silencing in the ECs and the germ cells were analysed. Silencing of sov in the ECs induced GSC loss and formation of GSC-like tumours in the niches (Fig. 3C,D,F). This indicates that sov is required in the ECs in a non-cell-autonomous manner for GSC maintenance and differentiation.

This sov RNAi phenotype in GSCs could be a consequence of earlier defects induced in the larval ancestors of the ECs. To induce adult-specific sov silencing in the ECs, the temperature-sensitive Gal80^ts mutant was used. At the permissive temperature (18°C), sov is not silenced in c587Gal4;sovRNAi;Gal80^ts flies and no abnormalities are detectable in the niche (Fig. 3F). However, shifting the adult females to the restrictive temperature (29°C) allowed Gal4-driven sov silencing in the ECs that, in turn, resulted in germline defects similar to those of sov mutants. Two weeks after RNAi induction, germ cell tumours were formed and GSCs were lost (Fig. 3F).

Homozygous EC clones were induced in L1 larvae or in 1-day-old adult females, and the numbers of germaria containing homozygous sov mutant and wild-type EC clones were determined in 4-day-old females (Fig. S4A,B). The frequency of germaria containing sov mutant ECs was reduced compared with wild-type clone frequency (Fig. S4C). These germaria contained fewer sov mutant ECs, confirming the results obtained by analysis of mutant allelic combinations and the RNAi data on the requirement of sov in EC survival (Fig. S4D).

In summary, sov is required for EC survival in the adult niches, which ensures GSC differentiation and maintenance in a non-cell-autonomous manner.

In the niche, a complex regulatory network controls GSC differentiation. To explore further the function of sov in GSC differentiation, we analysed the activity of the signalling pathways involved in communication between different cell types of the niche.

In GSCs, the repression of bam expression prevents differentiation of the stem cells (McKearin and Ohlstein, 1995). In the wild type, bam expression is initiated in the GSC daughter cells that lose physical contact with the cap cells and adopt cystoblast fate. In sov mutant germaria, no bam expression was detected when monitored with the reporter line bam-GFP (Fig. S5A,B). Forced expression of bam from the heat shock-inducible hs-bam transgene was sufficient to induce differentiation of the sov mutant germ cells, as indicated by the formation of fusome containing cysts and an almost complete lack of GSC-like tumours in the sov²/sov^ML150; hs-bam females (Fig. S5C-E). Based on these data, we conclude that sov acts upstream of bam in the GSC differentiation process.

In the niche, the primary factor that represses bam expression in the GSCs is Dpp, the Drosophila TGFβ homolog (Song et al., 2004). In the wild type, Dpp activity is restricted to the GSCs and can be monitored by the nuclear translocation of pMad. Sov mutant tumour cells located outside the GSC niche accumulate pMad in their nuclei, indicating that expanded Dpp activity is responsible for the maintenance of bam repression in the sov mutant germ cells (Fig. S5F,G).

Restriction of Dpp activity to the GSCs can be adjusted by controlled diffusion of the secreted Dpp ligand. In wild-type germaria, ECs extend long cellular protrusions that enwrap differentiating GSC daughter cells, separating them from the Dpp signal (Fig. S5H). Sov mutant ECs, however, fail to extend protrusions, indicating that sov controls the accessibility of the secreted Dpp (Fig. S5I). Niche abnormalities caused by the lack of EC protrusions resemble the phenotypes that have been observed in niches with impaired Wnt4 function (Mottier-Pavie et al., 2016; Upadhyay et al., 2016). To test the involvement of sov in Wnt4- mediated niche regulation, Wnt4 expression was analysed in sov mutants. RT-qPCR revealed that Wnt4 mRNA levels were reduced in sov mutant niches, indicating that sov is required for Wnt4 expression (Fig. S5J). Forced expression of Wnt4 in sov RNAi ECs by two different transgenic constructs, however, did not rescue the germ cell differentiation defects caused by sov silencing, indicating that sov regulates GSC development not exclusively by promoting Wnt4 expression (Fig. S5K). Taken together, sov promotes protrusion formation in the ECs that separates the GSC daughter cell from the Dpp source and enables its Bam-driven differentiation.

Remarkably, GSC and EC defects of sov mutants resemble that of egg, HP1a or piwi, which are essential regulators of heterochromatin formation and transposon repression in Drosophila (Jin et al., 2013; Ma et al., 2014; Rangan et al., 2011; Wang et al., 2011). Thus, we hypothesized that Sov regulates niche integrity by suppressing transposon activity via heterochromatin regulation.

Transposons can be classified into three groups depending on whether they are predominantly expressed in the germline cells (e.g. HET-A and Burdock), in somatic cells (e.g. gypsy) or in both cell types (Malone et al., 2009). Therefore, we used transposon sensors, RT-qPCR and RNA-seq to analyse the expression of specific transposons in ovaries silenced for sov specifically in the germ cells or in the soma.

Transposon sensors contain transposon-derived sequences that are targeted by the piRNAs, resulting in repression of the LacZ reporter gene. First, we used sensors for the germline-dominant HET-A and the Burdock transposons (Dönertas et al., 2013). The MTD-Gal4 driver was used to drive the germline-specific expression of three independent sov RNAi lines. In all MTDGal4>sovRNAi ovaries, a robust β-galactosidase (β-Gal) expression was detected from both transposon sensors, suggesting derepression of transposons (Fig. 4A-D). Consistent with these results, a strong accumulation of the endogenous germline-dominant HET-A and Burdock transposon mRNA levels was detected in nosGal4>sovRNAi ovaries using RT-qPCR (Fig. 4I). To determine the germ line-specific effect of sov on the steady-state RNA levels of all transposon classes, polyA-RNAs were sequenced from nosGal4>sovRNAi and nosGal4>wRNAi control ovaries (RNA-seq). In the nosGal4>sovRNAi ovaries, a robust upregulation of transcript levels for all germ-line dominant transposons was detected, indicating that sov is required for repression of transposon activity in the germline (Fig. 4J).

To analyse the effect of sov on transposon silencing in the soma, we made use of the gypsy-LacZ transposon sensor, which reflects the expression of the soma-dominant gypsy transposon (Dönertas et al., 2013). Somatic silencing of sov with the 3-2 and KK103679 RNAi lines resulted in severe morphological abnormalities of the ovaries, which made the analysis of the gypsy-LacZ reporter difficult. Nevertheless, we detected a strong β-Gal accumulation by Tj-Gal4-driven expression of the silencing constructs, indicating transposon derepression by sov silencing (Fig. 4E,F). Somatic silencing of sov with the weak HMC04875 RNAi line resulted in the formation of normal egg chambers with normal-looking somatic cells and the oogenesis was completed. Despite the modest phenotypic consequences of sov silencing by this RNAi construct, a weak β-Gal expression was detected in the somatic cells (Fig. 4G,H). Results obtained through analysis of the gypsy-LacZ transposon sensor were confirmed by measuring endogenous gypsy mRNA levels in the sov-silenced ovaries. RT-qPCR on c587Gal4>sovRNAi ovaries revealed a robust accumulation of the endogenous gypsy mRNAs, indicating that sov functions in the repression of the somatic transposon activity (Fig. 4I).

In summary, derepression of the reporter expression from the sensors and upregulation of the endogenous transposons by impaired sov function revealed an essential role for sov in transposon repression, both in the germline and in the soma.

Derepression of transposons results in accumulation of double-stranded breaks (DSBs) on the chromosomes, which are recognized by the histone variant γH2Av. In the wild type, γH2Av accumulates in nuclei of germ cells in the 16 cell cysts undergoing meiotic recombination in region 2 of the germaria. As meiosis is accomplished, meiotic DSBs are repaired and γH2Av levels are reduced in the wild-type germ cells (Fig. S6A,C). In contrast, an overall high level of γH2Av was detected in the nuclei of sov RNAi germ cells prior and after meiosis (Fig. S6B,D). Accumulation of γH2Av was initiated in the GSCs, indicating the formation of DSBs, presumably caused by transposon mobilization.

Transposon-induced DNA damage has been shown to activate the Chk2-dependent checkpoint (Chen et al., 2007; Klattenhoff et al., 2007). Activation of the Chk2 and ATR kinases induces microtubule organization defects in the oocytes, which lead to Gurken (grk), oskar RNA (osk) and Vasa delocalization and, thus, abnormal axis specification (Abdu et al., 2002; Ghabrial and Schüpbach, 1999). In sov RNAi germ cells, no Vasa was detected at the posterior pole of the oocytes, indicating a disrupted AP polarity (Fig. S6E,F). Prior to Vasa, osk mRNA is localized to the posterior pole of the wild-type oocytes where it is translated. However, neither osk mRNA (Fig. S6G,H) nor Osk protein (Fig. S6E,F) accumulation was detected at the posterior of sov-depleted oocytes. In addition, no anterodorsal localization of Grk was detected in sov RNAi oocytes (Fig. S6E,F). Consistent with the lack of Grk localization, all eggs laid by nosGal4>sovRNAi females exhibited severe ventralized eggshell-patterning defects with fused or absent dorsal appendages (Fig. S6I,J). Analysis of the RNA-seq data revealed an increased expression of the apoptotic genes hid and reaper in the sov RNAi germ cells, suggesting that loss of Sov induces apoptosis (Fig. S6K). Taken together, our data indicate that loss of sov function in the germ line results in transposon derepression, which, in turn, leads to DNA damage, abnormal axis specification and apoptosis.

To identify the mechanisms by which Sov regulates transposon repression, we first determined the subcellular localization of GFP-tagged Sov expressed from a genomic fosmid construct. The tagged Sov variant completely rescued the sterile and the lethal phenotypes associated with sov mutations. In sov^del1; Sov:GFP females, we detected ubiquitous Sov expression in the somatic and germ cells of the ovary. Sov localized in the nuclei and accumulated at nuclear foci (Fig. 5A). At these foci, Sov partially colocalized with HP1a and Centrosome identifier (Cid), suggesting a direct role for Sov in chromatin regulation (Fig. 5B,C).

To identify Sov interacting proteins, we affinity purified Sov from the sov^del1; Sov:GFP ovaries on six independent occasions and performed mass spectrometry (LC-MS/MS) analysis of the samples. Consistent with the colocalization in the nuclei, Sov co-immunoprecipitated with HP1a, indicating that sov affects transposon repression as a chromatin regulator (Fig. 5D). This role of Sov was confirmed in a pericentromeric position effect variegation (PEV) assay. For the assay, the w^m4h allele was used, in which the white gene responsible for the red eye colour is translocated to the border between the pericentromeric heterochromatin and the euchromatin. In the individual facets of the compound fly eye, stochastic spreading of the heterochromatin permits or inhibits expression of the white gene, leading to a variegating eye colour (Elgin and Reuter, 2013).

In heterozygous sov^ML150 and sov^del1 flies, we detected increased eye pigment production, indicating a dominant suppressor effect of sov on PEV (Fig. 5E). Dominant PEV-suppression of the w^m4h allele suggests that Sov promotes heterochromatin formation at the pericentromeric regions of the genome. The positive effect of sov on heterochromatin formation is gene-dose dependent, as demonstrated by the enhancement of PEV with increasing gene copies of sov (Fig. 5E). Furthermore, PEV analysis revealed a dominant genetic interaction between HP1a [encoded by the Su(var)205 gene] and sov. Increasing the copies of the wild-type sov gene in Su(var)205⁵ heterozygous flies caused decreased eye pigment production, indicating that increased amounts of Sov can compensate the heterochromatin formation defects caused by decreased HP1a levels (Fig. 5F). Haplo-suppression and triplo-enhancement of PEV by sov, and its genetic and physical interaction with HP1a indicates that Sov functions as a structural component of the heterochromatin.

Heterochromatin is epigenetically defined by repressive chromatin modifications, such as trimethylation of histone 3 at lysine 9 (H3K9me3), which is mediated by the histone methyl transferase egg and is recognized by HP1a. To test the effect of sov on the deposition of this repressive chromatin modification, we visualized H3K9me3 by immunostaining of sov^del1 homozygous germline clones. In the sov-null mutant cells, no difference in H3K9me3 immunostaining was detectable when compared with the neighbouring sov heterozygous cells, indicating that Sov functions downstream of egg in heterochromatin formation (Fig. S7A). Similar to the wild type, HP1a was recruited to the heterochromatic foci and formed elongated structures in sov homozygous germline clones and in sov RNAi germ cells (Fig. S7B-D). To dissect further the epistatic relationship between HP1a and sov, we analysed Sov localization in HP1a-silenced germ cells. Wild-type Sov localization was found in HP1a RNAi cells, indicating that Sov localization does not depend on HP1a (Fig. S7E,F).

To test the involvement of Sov in HP1a-mediated gene silencing, we applied a LacI/LacO transcriptional reporter assay (Sienski et al., 2015). The NLS-GFP reporter construct containing LacO repeats was expressed in the ovary (Fig. 6A,D). We expressed LacI tagged HP1a in the germline (HP1a:LacI), which was artificially recruited to the reporter via LacI-LacO interaction. Tethering HP1a:LacI to the reporter DNA suppressed GFP expression in the germline (Fig. 6B,D). We silenced sov in the germline and observed no derepression of the reporter expression, suggesting that sov is not required for HP1a-mediated repression of the reporter locus when HP1a is artificially tethered to the DNA, but rather promotes the recruitment of HP1a to the chromatin (Fig. 6C,D).

Formation of heterochromatic domains have been shown to be driven by liquid phase separation via weak hydrophobic interaction between HP1a molecules and other heterochromatin components (Larson et al., 2017; Strom et al., 2017). In the mature heterochromatin, HP1a is mobile in the liquid compartment, whereas chromatin bound HP1a is immobile. To study the function of sov in the regulation of the dynamic properties of heterochromatin, we measured the immobile fraction of HP1a at the heterochromatin foci. We expressed HP1a:GFP in the germ cells and used fluorescent recovery after photobleaching (FRAP) to analyse HP1a dynamics (Movie 1). Silencing of sov resulted in an increase of the mobile fraction of HP1a:GFP (Fig. 6E-G). Increase of HP1a mobility indicates that sov stabilizes the heterochromatin domain by the enhancement of HP1a association with the chromatin polymer.

To determine the global effect of sov on steady-state mRNA levels, we compared mRNA-seq data of nosGal4>sovRNAi and nosGal4>wRNAi control ovaries. Of the 6811 euchromatic genes expressed in the ovary, transcription was increased by more than twofold at 161 genes (2.4%), whereas 146 genes had a more than twofold decrease in mRNA levels (2.1%). Of the 203 expressed genes mapping to the heterochromatic regions of the genome, 26 were downregulated more than twofold (15.1%), whereas no heterochromatic gene was upregulated (Fig. 7A). Over-representation of the heterochromatic genes in the downregulated gene set indicates that Sov preferentially promotes transcription in the heterochromatic genome regions.

Long piRNA precursors are transcribed from piRNA clusters that require a heterochromatic context for transcription (Brennecke et al., 2007; Rangan et al., 2011). We hypothesized that sov is involved in transposon repression by regulating piRNA cluster transcription. In the soma, piRNA clusters are uni-strand clusters, i.e. only the plus strand of the piRNA clusters is transcribed. In the germ cells, piRNA precursors are generated from both uni-strand and dual-strand clusters (Brennecke et al., 2007). We carried out strand-specific RT-qPCR for piRNA precursors derived from the germline-specific dual-strand 42AB cluster, from the germline and soma-expressed uni-strand 20A cluster, and from the soma-specific flamenco (flam) cluster. In sov RNAi cells, no decrease in the transcript levels was detected from the transcribed strands of cl20A cluster and flam; however, transcription of the dual-strand cl42AB was severely affected (Fig. 7B). Although the plus-strand-specific transcript levels were not reduced, we detected decreased piRNA precursor production from the minus strand of the cl42AB. The effect of sov on transcription of the minus strand of the dual-strand piRNA cluster resembles that of egg, suggesting that sov affects chromatin regulation of the cluster (Rangan et al., 2011).

To investigate the chromatin structure at the dual-strand clusters, we analysed the localization Rhino, an HP1 homologue associated exclusively with the dual-strand piRNA clusters. Silencing of sov induced formation of nuclear aggregates of Rhino, suggesting that Sov is required for the formation of the proper chromatin structure at the dual-strand piRNA clusters (Fig. 7D). Silencing of egg in the germline resulted in a similar abnormality in Rhino localization, supporting the hypothesis that Sov affects dual-strand cluster transcription by regulating heterochromatin structure at this cluster in cooperation with egg (Fig. 7E).

Next, we analysed whether sov silencing in the germ line affects expression of the genes involved in the piRNA pathway. Analysis of the RNA-seq data of germline-silenced sov ovaries revealed a significant downregulation of Ago3, a member of the Ago family required for piRNA biogenesis (Fig. S8A). As mature piRNAs enable Piwi to enter into the nucleus, we analysed Piwi localization in the sov^del1 mutant and in the sov RNAi cells. We detected a nuclear accumulation of Piwi, indicating that the nuclear translocation of Piwi is not affected by the reduction in Ago3 levels (Fig. S9A-F). Nevertheless, reduction of Ago3 expression in Sov-depleted germ cells uncovers the involvement of sov in an additional regulatory level of the piRNA-pathway.

Heterochromatin represents a functionally and structurally separate nuclear domain. The heterochromatin-associated DNA includes at least one-third of the Drosophila genome and is mainly composed of repetitive sequences, such as transposon fragments and satellite repeats (Hoskins et al., 2007; Smith et al., 2007). Besides its DNA content, the heterochromatin is rather epigenetically defined by a pattern of various histone modifications and by the binding of additional chromatin-associated proteins (Kharchenko et al., 2011; Riddle et al., 2011; Saha et al., 2018). Although there is a diversity of combinatorial patterns of epigenetic marks in the heterochromatin, the majority of the heterochromatin regions lack the ‘activating’ chromatin marks and are enriched in the ‘repressive’ signatures, such as H3K9me3 and HP1a (Riddle et al., 2011). In addition to the DNA and the specific protein composition, heterochromatin has a higher level of organization. HP1a is capable of demixing from solution and forming liquid droplets that organize the assembly of the heterochromatin into a membrane-less nuclear domain via liquid-liquid phase separation (Larson et al., 2017; Strom et al., 2017). Many proteins driving the formation of phase-separated organelles contain extended intrinsically disordered regions and are enriched in RGG/RG domains (Boeynaems et al., 2018; Mitrea and Kriwacki, 2016). Sov shows a remarkable structural similarity to these proteins as it has an unusually long disordered N terminus that contains an RGG/RG repeat domain. It is tempting to speculate that Sov functions, in concert with its binding partner HP1a, in the establishment of the unique physical properties of heterochromatin. Indeed, the integrity of the mature heterochromatin domain was shown to require formation of the immobile HP1a compartment through the interaction of HP1a with non-histone binding partners (Strom et al., 2017). We propose that Sov is one of these non-histone partners of HP1a. We show that Sov affects the dynamics of HP1a between the liquid and static compartments in the heterochromatin domain, suggesting that the HP1a-mediated phase separation driven heterochromatin formation depends, at least in part, on Sov. The RGG/RG domain found in Sov is a widespread RNA binding domain displaying degenerate RNA-binding specificity (Ozdilek et al., 2017). The HP1a complex containing Sov is enriched in RNAs, raising the possibility that the putative RNA binding of Sov is involved in proper heterochromatin function (Alekseyenko et al., 2014). Targeting of heterochromatin formation at particular genomic regions involves diverse mechanisms, such as function of satellite DNA-specific binding proteins or the RNAi machinery (Elgin and Reuter, 2013). In the LacI/LacO-based tethering assay, we show that HP1a-mediated repression is independent of Sov, whereas Sov enhances the recruitment and binding of HP1 to the chromatin. This function of Sov could be performed by the highly structured C terminus, which contains a large number of tandem Zn fingers. Through these domains, Sov may stabilize the binding of HP1a complexes to specific DNA sequences and target Sov/HP1a-mediated processes at specific regions of the genome.

Heterochromatin is usually associated with transcriptional repression. Many protein-coding genes residing in the repressive domains, however, require the heterochromatic environment for transcription (Clegg et al., 1998; Lu et al., 2000; Schulze et al., 2005; Wakimoto and Hearn, 1990). Disruption of heterochromatin by impaired HP1a function results in reduced expression of these genes (Lundberg et al., 2013). Consistent with this observation, Sov preferentially promotes the transcription of heterochromatin-resident genes by positively regulating HP1a function.

We demonstrate that Sov is involved in the regulation of transposon repression at multiple levels. In the germline, it is required for proper transcription of the piRNA clusters and promotes the expression of the piRNA-pathway component Ago3. In addition, it may mediate transcriptional gene silencing, both in the germline and in the somatic cells.

Efficient transcription of the piRNA clusters requires a heterochromatic environment. Despite the obvious differences in their regulation, both uni-strand and dual-strand clusters were shown to be enriched in the repressive H3k9me3 histone mark (Mohn et al., 2014). Impairment of heterochromatin formation leads to reduction of piRNA cluster transcription (Rangan et al., 2011; Teo et al., 2018). The transcription of the uni-strand clusters critically depends on egg, however; in particular, the minus strand of the dual-stranded cl42AB was sensitive to egg depletion (Rangan et al., 2011). Similar to egg, depletion of sov results in reduction of the transcription of the minus strand of the cl42AB, whereas the transcription of the plus strand remains unaffected. The effect of egg and sov depletion on cluster transcription, however, is not identical. In contrast to egg, the function of sov is restricted to the dual-stranded cluster (Rangan et al., 2011). It is possible that sov specifically regulates the formation of the Rhino-dependent specialized heterochromatin that enables dual-stranded transcription. Further studies are needed to explore how this separation in sov function is regulated.

Another level of transposon regulation, where Sov is involved, is the indirect control of piRNA biogenesis. Like egg, sov is required for Ago3 transcription, raising the possibility that the transposon derepression defects found in sov-depleted germ cells may result not only from reduced cluster transcription, but also from defective piRNA biogenesis (Kang et al., 2018). However, the effect of sov on transposon repression and niche regulation differs from that of Ago3. Unlike egg or sov mutants, no GSC loss was detected in the Ago3 mutants, showing that GSC self-renewal was not affected (Rojas-Ríos et al., 2017). Consistent with the lack of GSC self-renewal defect in Ago3 mutants, piRNAs are produced in the absence of Ago3 through an Aub-dependent homotypic amplification mechanism, and these piRNAs could be used in the regulation of GSCs (Wang et al., 2015a). piRNAs enable Piwi to move from the cytoplasm to the nucleus, where it suppresses transposon transcription (Klenov et al., 2011; Le Thomas et al., 2013; Rozhkov et al., 2013; Sienski et al., 2015; Wang and Elgin, 2011; Yashiro et al., 2018). In sov mutant and sov RNAi cells, we detect nuclear Piwi accumulation, suggesting that piRNA biogenesis was not completely abolished. Loss of Ago3 results in the derepression of a subset of transposons (Wang et al., 2015a). Sov depletion, however, causes a stronger defect in transposon regulation, i.e. the derepression of all transposon classes, indicating that sov has a more general effect on transposon silencing than the sole enhancement of Ago3 transcription.

In the sov RNAi ECs, we find a derepression of the gypsy transposon. piRNAs required for gypsy silencing are generated from the flam cluster. Similar to HP1a, Sov is dispensable for flam transcription (Penke et al., 2016). As there is no Ago3-dependent piRNA amplification in the soma, Sov must have an additional effect on transposon silencing. An attractive explanation for gypsy derepression in the Sov-depleted ECs could be that sov is involved in the transcriptional transposon silencing mechanism. This step of the silencing pathway requires heterochromatinization of the transposon locus. Tethering of the Piwi-RISC to the transposon induces the recruitment of the effector complex, composed of Asterix, Panx, Mael and Egg, to the transposon loci (Sienski et al., 2012, 2015; Yu et al., 2015). Egg initiates the deposition of the repressive H3K9me3 mark and inhibits transcription by recruiting HP1a. In a parallel pathway, Piwi also recruits Histone1 to the transposons, which organizes the chromatin into a higher order repressive state (Iwasaki et al., 2016). Sov physically interacts with HP1a and stabilizes the heterochromatin domain through the enhancement of an association of HP1a with the chromatin. We propose that Sov supports the recruitment of HP1a to the transposon locus by binding to HP1a, which in turn stabilizes the association of H3K9me3-bound HP1a with the chromatin. A similar Sov-mediated mechanism might work also in the germline; however, this requires further investigation.

We show that Sov is involved in germ cell development by regulating GSC survival, GSC differentiation and EC survival. The complex loss-of-function phenotype of sov closely resembles that of egg mutations, supporting our hypothesis that sov contributes to the formation and maintenance of the HP1a-Egg-mediated repressive chromatin environment (Wang et al., 2011).

Germline-specific RNAi revealed that Sov is required in the GSCs to control their self-renewal in a cell-autonomous manner. Loss of function of the components of the heterochromatin machinery and of the piRNA pathway induce apoptotic GSC loss accompanied with transposon derepression, DNA damage and checkpoint activation (Chen et al., 2007; Klattenhoff et al., 2007; Ma et al., 2014, 2017). In egg mutants, upregulation of the apoptotic genes hid and reaper, and activation of Caspase 3 cleavage was found (Clough et al., 2007; Kang et al., 2018). Similarly, we observed an accumulation of DNA damage, checkpoint activation and apoptosis in sov-depleted cells. We propose that sov ensures GSC maintenance through suppression of transposon activity-induced genome damage and, by so doing, supresses the apoptosis of the germ cells.

Cell-type specific RNAi experiments showed that Sov function is required in ECs for controlling both the maintenance and the differentiation of germ cells in a non-cell-autonomous manner. When anterior-most ECs are lost, GSCs are gradually eliminated from the niche, whereas loss of posterior ECs leads to differentiation defects and to the formation of GSC tumours (Wang and Page-McCaw, 2018). Indeed, somatic sov RNAi resulted in loss of both the anterior and the posterior ECs and, as a consequence, a combination of the GSC loss and germ cell tumour phenotypes was observed.

The sov-dependent EC loss may be induced by the improper function of the signalling network operating in the somatic niche. We show that sov has an effect on the Wnt and dpp signalling pathways. Wnt signalling was shown to promote survival of ECs and inhibit the expansion of dpp activity in the niche (Wang et al., 2011, 2015b). Interestingly, transposon derepression results in decreased Wnt4 expression (Upadhyay et al., 2016). We show that sov is required for Wnt4 expression in the ECs; however, the differentiation defect induced by sov depletion is not rescued by simultaneous Wnt overexpression, indicating additional diverse roles of sov in ensuring EC survival.

In ECs, increasing transposon activity by knocking down egg, hp1a, piwi or flam results in cell death (Ma et al., 2014; Upadhyay et al., 2016; Wang et al., 2011). Because in sov-depleted ECs, we detected robust transposon derepression accompanied by caspase activation, we propose that Sov supresses EC death by downregulating transposons.

All animals were raised at 25°C, unless otherwise indicated. The sov^del1 allele was generated by FRT-mediated recombination between PBac{WH}f04480 and P{XP}d07849 transposon insertions (Ryder et al., 2004). The sov² and sov^ML150 alleles were obtained from the Bloomington Drosophila Stock Center. To silence sov, we generated the UAS-sovRNAi[3-2] transgenic construct expressing shRNA targeting the last exon of sov. In this study, the UAS-sovRNAi[3-2] construct was used for sov silencing unless otherwise indicated. The Dp(1;3)FF184439 and Dp(1;2)FF026056 duplications were generated by inserting the FF184439 and the FF026056 fosmid constructs into the attP2 and attP40 docking sites, respectively (Ejsmont et al., 2011). The FlyFos018439-CG14438-SGFP-V5-preTEV-BLRP-3XFLAG fosmid was obtained from the Drosophila TransgeneOme Project and was inserted into the attP40 site. For the sake of simplicity, we refer to this transgenic line as Sov:GFP thereafter.

The strains used in this study include: sov² (Bl#4611), sov^ML150 (Bl#4591), bam-GFP (Chen and McKearin, 2003), nos-Gal4 (Van Doren et al., 1998), c587Gal4 (Manseau et al., 1997; Song et al., 2004), hs-bam (Ohlstein and McKearin, 1997), gypsy-LacZ (Dönertas et al., 2013), Burdock-LacZ (Dönertas et al., 2013), Het-A-LacZ (Dönertas et al., 2013), Vasa:AID:GFP (Bl# 76126) (Bence et al., 2017), HP1a:RFP (BL#30562) (Wen et al., 2008), HP1a:GFP (BL#30561) (Wen et al., 2008), FRT19A,His2Av:GFP,hsFLP/FRT19A; His2Av-mRFP (B#30563), FRT19A (Bl#1709), Su(var)205⁵ (Bl#6234), w^m4h, UAS-HP1aRNAi-TRiP.GL00531 (Bl#36792), UAS-eggRNAi-TRiP.HMS00443 (Bl#32445), UAS-wRNAi-TRiP.HMS0001 (Bl#33623), UAS-sovRNAi-TRiP.HMC04875 (Bl#57558), UAS-sovRNAi-KK103679 (VDRC#v100109), 8XLacO-nos>GFP (Sienski et al., 2015), UASP-LacI:HP1a (Sienski et al., 2015), UAS-Wnt4.ORF.3XHA (FlyORF#F001112), UAS-dWnt4(r13) (Sato et al., 2006) and oskMS2/MS2GFP (Forrest and Gavis, 2003; Zimyanin et al., 2008).

To sequence the sov² mutant allele, genomic DNA was isolated from sov²/sov^del1 mutant females and the sov-coding sequence, including the introns, was PCR amplified and sequenced between the start and the stop codons. Comparison of the sov² sequence with the reference R6.15 Drosophila genomic sequence revealed 20 missense mutations, a 27 bp in-frame deletion, a 54 bp in-frame insertion and a frame-shift mutation in the sov²-coding region. To unambiguously identify the mutation responsible for the sov mutant phenotype, we sequenced the sov-coding sequence of the wisp^12-3147 allele and used it as a reference. The wisp^12-3147 and the sov² alleles were isolated in the same genetic screen and have the same paternal chromosome (Mohler, 1977). Analysis of the sequences revealed that all mutations are shared between sov² and wisp^12-3147, except for the frame-shift mutation.

The negatively marked control and sov mutant EC and germline clones were generated using the FLP/FRT-mediated recombination as described previously (Song et al., 2002; Xie and Spradling, 1998). Briefly, heterozygous sov mutant and wild-type control cells ubiquitously express both Histone2Av:RFP and Histone2Av:GFP. Upon mitotic recombination, cells become homozygous for the sov mutant allele and concomitantly lose Histone2Av:GFP expression. Thus, homozygous sov mutant clones express only Histone2Av:RFP and can be identified by their lack of Histone2Av:GFP expression. To generate sov mutant clones, FRT19A,His2Av:GFP,hsFLP/FRT19A,sov^del1; His2Av-mRFP/+ females were heat shocked either 24 h after egg laying or at the adult stage. Young females were heat shocked on three consecutive days for 1 h at 37°C, and the mutant phenotypes were examined 6 days after clone induction.

To silence sov in the germ line, the nos-Gal4 driver was used. To silence sov in the ECs specifically at the adult stage, c587Gal4/UAS-sovRNAi[3-2];TubGal80^ts females were cultured at 18°C until adulthood and then shifted to 29°C. For the transposon sensor assay, sov was silenced using the germ line-specific MTDGal4 driver. To express Bam in sov mutants, sov²/sov^ML150; hs-bam flies were heat-shocked at 37°C twice for 1 h, separated by a 2-h recovery period at 25°C. Ovaries were analysed 24 h after the first heat-shock.

Female flies were aged at 25°C for 10 days prior to imaging. To measure eye pigmentation, females were frozen in liquid nitrogen. Heads were separated from bodies by brief vortexing. Samples of 10 females were homogenized in 0.5 ml of 0.01 M HCl in ethanol; the homogenate was placed at 4°C overnight, warmed at 50°C for 5 min, clarified by centrifugation and the OD at 480 nm of the supernatant was measured.

β-Gal staining of transposon sensor lines was performed as described previously (Dönertas et al., 2013). Immunostainings were performed as described earlier (Jankovics et al., 2014). The list of primary antibodies used is summarized in Table S1. DAPI was used to label the nuclei, actin was visualized by phalloidin staining. Specimens were examined with Leica TCS SP5 confocal microscope. To determine the number of escort cells, niches were co-immunostained with Traffic jam (Tj) and Fas3 antibodies. Terminal filament cells were double-negative, follicle cells were labelled by both antibodies, whereas ECs and cap cells were labelled exclusively by Tj. Then, cap cells and escort cells were distinguished based on their location, shapes of nuclei and based on the weaker intensity of Tj labelling in the cap cells (Wang and Page-McCaw, 2018).

For live imaging of Sov:GFP, HP1a:GFP, HP1a:RFP and Cid:RFP, samples were prepared as described previously (Forrest and Gavis, 2003) and examined with a VisiScope spinning disc confocal microscope. Fluorescence recovery experiments were performed on stage 10 egg chambers expressing HP1a:GFP. FRAP experiments were performed with a Leica SP5 confocal microscope. The 405 nm laser line was used to photobleach a 4 μm² region of the heterochromatin domain. Recovery was recorded for 1 min at one frame every 1.5 s. Fluorescence recovery curves were analysed using the easyFrap software as described previously (Bancaud et al., 2010; Koulouras et al., 2018). Statistical tests were performed with GraphPad Prism.

For the protein interactome analysis, 2-day-old Sov:GFP and w¹¹¹⁸ females were used. Ovaries were dissected in PBS. Ovaries (150-200 µl) were harvested, frozen in liquid nitrogen and ground with TissueLyser at 30 Hz. Total proteins from Drosophila tissue were extracted using the manufacturer's Lysis buffer supplemented with 1 mM DTT, 1 mM PMSF, 1×Sigma protease inhibitor cocktail, 3 mM pNPP and 1 μM MG132. Total protein extracts (8 mg/immunoprecipitate) were immunopurified using anti-GFP (three replicates) or anti-FLAG (three replicates) antibody coupled 50 nm magnetic beads (MACS Technology, Miltenyi) digested in column with trypsin, and analysed in a single run using a mass spectrometer (Hubner et al., 2010)

The resulting peptide mixture was desalted prior to LC-MS/MS analysis (Omix C18 100 μl tips, Varian) and the purified peptide mixture was analysed by LC-MS/MS using a nanoflow RP-HPLC on-line coupled to a linear ion trap-Orbitrap (Orbitrap-Elite, ThermoFisher Scientific) mass spectrometer operating in positive ion mode. Data acquisition was carried out in data-dependent fashion, the 10 most abundant, multiply charged ions were selected from each MS survey for MS/MS analysis (MS spectra were acquired in the Orbitrap; CID spectra were acquired in the linear ion trap).

Raw data were converted into peak lists using the in-house PAVA software (Guan et al., 2011) and searched against the Swissprot database using the Protein Prospector search engine (v5.15.1) with the following parameters: enzyme, trypsin with maximum 1 missed cleavage; mass accuracies, 5 ppm for precursor ions and 0.6 Da for fragment ions (both monoisotopic); fixed modification, carbamidomethylation of Cys residues; variable modifications, acetylation of protein N-termini; Met oxidation, cyclization of N-terminal Gln residues allowing a maximum of two variable modifications per peptide. Acceptance criteria were: minimum scores, 22 and 15; maximum E values, 0.01 and 0.05 for protein and peptide identifications, respectively. Another database search was also performed using the same search and acceptance parameters except that the Uniprot.random.concat database (downloaded 2015/4/16) was searched with Drosophila melanogaster species restriction (52,524 proteins), including additional proteins identified from the previous Swissprot search (protein score>50). False discovery rate was estimated using peptide identifications representing randomized proteins (2×number of random IDs/total peptide IDs)=2× number of random IDs divided by peptide IDs.

Spectral counting was used to estimate relative abundance of individual proteins in the Sov:GFP and control samples: peptide counts of the individual proteins were normalized to the total number of peptide identifications in each sample, then these normalized peptide counts were compared in the two samples. Enrichment between SovGFP and control experiments was calculated using edgeR (Li and Andrade, 2017). Counts representing Sov were omitted from the analysis.

In general, the total RNA was prepared from whole ovaries using ReliaPrep RNA Tissue Miniprep System (Promega, Z6111). cDNAs were synthesized using oligodT primers (First Strand cDNA Synthesis kit, ThermoScientific, K1612). Strand-specific RT-qPCR was performed as described previously (Klattenhoff et al., 2009). For strand-specific RT-qPCR of the 42AB and 20A piRNA clusters, ovaries of 2-day-old nosGal4>sovRNAi females were used. For strand-specific RT-qPCR of the flam cluster, c587Gal4>sovRNAi; Tub>Gal80^ts females were cultured on 18°C and shifted to 29°C at the adult stage. RT-qPCR was performed 10 days after the temperature shift. qPCR reactions were performed using Maxima SYBR Green/ROX (ThermoScientific, K0221). For each reaction, tree technical replicates were performed on three biological replicates. Rp49 transcript was used as internal control. Data ware analysed using the Rotor-Gene Q Series software. Primer sequences are provided in Table S2.

For RNA-seq, ovaries of nosGal4>sovRNAi3-2 and nosGal4>wRNAi females were dissected. Total RNA was prepared using RNeasy Mini Kit (Qiagen, 74104). RNA samples were quantified and their quality determined by capillary gel electrophoresis with a 2100 Bioanalyzer (Agilent) instrument using Agilent RNA 6000 nano kit. Poly(A) RNAs were selected using the NEBNext Poly(A) mRNA Magnetic Isolation Module then strand-specific RNA-seq libraries were prepared using the NEBNext Ultra RNA Library Prep Kit for Illumina with NEBNext Multiplex Oligos for Illumina following the recommendations of the manufacturer (New England Biolabs). Sequencing libraries were validated and quantified using an Agilent DNA 1000 kit in a 2100 Bioanalyzer instrument. After pooling, paired-end sequencing was carried out with an Illumina MiSeq instrument using MiSeq Reagent kit V3-150.

Fastq files were generated with MiSeq Control Software then quality trimmed and filtered with Trimmomatic v0.33. For transcriptome analysis sequence flies were aligned to the Drosophila melanogaster reference genome r6.13 with Tophat v2.0.9. After alignment files were deduplicated with SAMtools software and differential expression analysis was performed with CuffDiff v2.1.1 using the corresponding transcript annotation file. For transposon expression, analysis-trimmed fastq files were aligned to the USCS dm6 reference genome with Bowtie v2.1 then differential transposon expression was determined by CuffDiff v2.1.1 using a corresponding transposon annotation file (labshare.cshl.edu/shares/mhammelllab/www-data/TEToolkit/TE_GTF/) (Jin et al., 2015). All raw sequence data have been deposited in the SRA (www.ncbi.nlm.nih.gov/sra) under accession number PRJNA495606.

We thank W. Theurkauf, T. Tetsuya, D. McKearin, L. Gilboa and J. Brennecke for fly stocks and reagents. We also thank B. Irvine for critical reading of the manuscript.