Msl3 promotes germline stem cell differentiation in female Drosophila

Germ cells give rise to gametes, a fundamental requirement for sexual reproduction. The production of gametes is tightly controlled to ensure a constant supply throughout the reproductive life of an organism (Cinalli et al., 2008; Kimble, 2011; Lehmann, 2012; Spradling et al., 1997). Germline stem cells (GSCs) divide mitotically to both self-renew and generate differentiating daughters that can enter meiosis (Kimble, 2011; Lehmann, 2012; Spradling et al., 2011). Loss of differentiation and meiotic entry results in infertility (Hughes et al., 2018; Lesch and Page, 2012; Soh et al., 2015).

GSC differentiation is well characterized during Drosophila oogenesis (Ables, 2015; Collins et al., 2014). Drosophila ovaries are composed of individual egg-producing units called ovarioles. A structure called the germarium lies at the tip of each ovariole and houses GSCs, which are marked by round organelles called spectrosomes (Eliazer and Buszczak, 2011; Morris and Spradling, 2011; Morrison and Spradling, 2008; Spradling et al., 2011, 2001, 2008) (Fig. 1A). GSCs both self-renew and differentiate into cystoblasts (CBs), which divide and undergo incomplete cytokinesis to give rise to two-, four-, eight- and 16-cell cysts, which are marked by branched structures called fusomes (Chen and McKearin, 2003a,b; Xie, 2013).

Early germ cell differentiation is controlled by both intrinsic and extrinsic factors (Flora et al., 2017; Spradling et al., 2011). The somatic niche of the germarium provides Decapentaplegic (DPP) signaling that leads to phosphorylation of Mothers against DPP (pMad) in GSCs, and transcriptional repression of the differentiation factor bag of marbles (bam) (Chen and McKearin, 2003a,b; Kai and Spradling, 2003). After GSC division, the CB is displaced from the niche, allowing for Bam expression (Chen and McKearin, 2003a,b). Bam is sufficient to promote the transition from CB to a differentiated eight-cell cyst (McKearin and Ohlstein, 1995; McKearin and Spradling, 1990).

In the eight-cell cyst, expression of cytoplasmic isoforms of RNA-binding Fox protein 1 (Rbfox1) leads to translational downregulation of self-renewal factors to promote expression of Bruno (Bru) (Carreira-Rosario et al., 2016; Tastan et al., 2010). Bru, in turn, translationally represses mitotic factors, which promote cyst divisions, and regulates entry into a meiotic cell cycle (Parisi et al., 2001; Sugimura and Lilly, 2006; Wang and Lin, 2007). Multiple cells in the cysts initiate meiosis, but only the oocyte will commit to meiosis; the other 15 cells acquire a nurse cell fate in the 16-cell cyst stage (Carpenter and Sandler, 1974; Huynh and St Johnston, 2004; Mach and Lehmann, 1997; Navarro et al., 2001; Theurkauf et al., 1993). The oocyte and the 15 nurse cells are encapsulated by somatic cells to form a developing egg chamber and eventually an egg (Fig. 1A1). Although Rbfox1 expression in the germline is essential for entry into a meiotic cell cycle and oocyte specification, how it is induced is unclear (Carreira-Rosario et al., 2016).

Another hallmark of meiosis, apart from a specialized cell cycle, is homologous chromosome recombination. This process is regulated by the formation of the synaptonemal complex (SC) (Ables, 2015; Hughes et al., 2018). The SC starts to assemble on homologs in two cells with four ring canals, and to a lesser degree in the other 14 cells of the cyst, and it then becomes restricted to the future oocyte (Page and Hawley, 2001; Von Stetina and Orr-Weaver, 2011) (Fig. 1A). Although proper homologous recombination is not required for differentiation, it is required for fertility (Collins et al., 2014; Handel and Schimenti, 2010). How transcription of SC components is activated during meiosis is not well understood.

GSC differentiation during Drosophila oogenesis requires the histone methyltransferase SET domain containing 2 (Set2), which confers histone H3 lysine 36 trimethylation (H3K36me3) (Larschan et al., 2007; Mukai et al., 2015; Stabell et al., 2007). H3K36me3 typically marks transcriptionally active genes (Bannister and Kouzarides, 2011; Dong and Weng, 2013; Keogh et al., 2005). How H3K36me3 regulates GSC differentiation is not clear. Interestingly, in male Drosophila, H3K36me3 facilitates recognition of the X chromosome by the Male-Specific Lethal (Msl) complex, which leads to hyper-transcription of the male X and gene dose compensation with females, which have two X chromosomes (Bell et al., 2008; Conrad et al., 2012b; Larschan et al., 2007). Within the MSL complex, the chromodomain (CD) of Msl3 reads H3K36me3 marks and the histone acetyl transferase (HAT) Males absent on the first (MOF) deposits acetylation of histone H4 lysine 16 (H4K16ac) (Bone et al., 1994; Conrad et al., 2012a; Gu et al., 2000; Hilfiker et al., 1997; Larschan et al., 2007). Female flies do not assemble the MSL complex because key components, including Msl2, are not expressed at sufficient levels (Bashaw and Baker, 1997; Kelley et al., 1997; UCHIDA et al., 1981). However, whether individual MSL proteins in Drosophila regulate gene expression beyond their role in dose compensation is not known.

Here, we find that Set2, Msl3 and a HAT complex, Ada2a-containing (ATAC), promote GSC differentiation. We discovered that Msl3 is expressed in the early stages of oogenesis, where it promotes the HAT-mediated transcription of several members of the SC, as well as a germline-specific paralog of eukaryotic Ribosomal protein S19 (eRpS19/RpS19). Expression of RpS19b (S19b) helps increase overall levels of RpS19, which is then required for translation of Rbfox1 and thus promotes GSC differentiation into an oocyte.

To determine how Set2 promotes oogenesis in Drosophila, we stained control and Set2 depleted fly gonads with antibodies against Vasa, a germline marker, and 1B1, a marker of somatic cell membranes, spectrosomes and fusomes. Compared with controls, Set2-depleted gonads displayed a loss of spectrosome containing GSCs, an accumulation of fusome-containing cysts and a loss of proper egg chamber formation (Fig. 1B-D; Fig. S1A-B′). The Set2-depleted germaria accumulated eight- and 16-cell cysts, as well cysts that were multinucleated up to 64 cells (n=30). The egg chambers that do form contain undifferentiated and differentiating cells marked by spectrosomes and fusomes that fail to develop further (100% in Set2 RNAi compared with 0% in nosGAL4; P<2.2E-16, n=50), resulting in females that are infertile. Additionally, Set2-depleted germ cells had significantly reduced H3K36me3 levels compared with the control, consistent with previous reports (Mukai et al., 2015) (Fig. S1C-E).

The accumulation of cyst-like structures upon germline depletion of Set2 could be due to GSCs that divide but fail to undergo cytokinesis, resulting in GSC cysts or in differentiating cysts that cannot progress further in development (Carreira-Rosario et al., 2016; Mathieu and Huynh, 2017; Sanchez et al., 2016). To discern between these two types of cyst, we stained for pMad, a marker of GSCs. In addition, we crossed a bam transcriptional reporter, bam-GFP, into the Set2 RNAi background and independently assayed for Bam protein (Chen and McKearin, 2003b). We found that Set2 RNAi germaria accumulated differentiating cysts, which transcribed and then translated Bam, and were pMad negative (Fig. 1E-F′; Fig. S1F-I′). Thus, Set2 is required in the germline downstream of bam to promote the differentiation of Bam-expressing cysts into egg chambers.

Although germline depletion of Set2 leads to both loss of GSCs and accumulation of cysts, here we focus on the cyst accumulation phenotype. Loss of Set2 results in cysts that do not properly express the oocyte-specific protein Orb (Mukai et al., 2015), but loss of Orb does not phenocopy loss of Set2, suggesting that Orb downregulation is a consequence of the differentiation defect (Barr et al., 2019; Christerson and McKearin, 1994; Huynh and St Johnston, 2000). Similar to loss of Set2, loss of Rbfox1 results in the accumulation of Bam-expressing cysts that do not differentiate into proper egg chambers (Carreira-Rosario et al., 2016; Tastan et al., 2010). To test whether Set2 regulates Rbfox1 and Bru expression, we stained separately for Rbfox1 and Bru along with Vasa and 1B1 in control and germline Set2-depleted ovaries. Although control germaria express Rbfox1 robustly in eight-cell cysts, Set2-depleted germ cells exhibited a significantly lower level of Rbfox1, while somatic levels were unchanged (Fig. 1G-I). Furthermore, Bru levels were reduced and enrichment in the oocyte was ablated in Set2 RNAi germaria compared with controls (Fig. S1J-L). Thus, Set2 is required after Bam expression to promote proper differentiation via Rbfox1 expression.

As germline depletion of Set2 results in reduced levels of Rbfox1 and Bru, we hypothesized that Set2-depleted cysts do not properly enter meiosis or specify an oocyte. We stained control and Set2-depleted germaria with antibodies against a SC member, Crossover suppressor on 3 of Gowen [C(3)G], and Vasa (Anderson et al., 2005; Page and Hawley, 2001). The control had several C(3)G-positive germ cells in 16-cell cysts but only the most posterior germ cell in the egg chamber was marked with C(3)G. In Set2 germline-depleted germaria, the majority of cells displayed perturbed C(3)G expression; an irregular number of cells were C(3)G positive and C(3)G appeared fragmented (Fig. 1J-K1′). To determine whether the oocyte is properly specified, we stained for the oocyte determinant Egalitarian (Egl), as well as for Vasa and 1B1 (Carpenter, 1994; Huynh and St Johnston, 2000; Mach and Lehmann, 1997). Although control 16-cell cysts had a single Egl-positive cell, Set2 germline-depleted germaria showed diffuse staining of Egl without enrichment in a single cell (Fig. S1M-N′). Thus, the H3K36me3-depositing enzyme Set2 is required for proper differentiation and oocyte specification.

To identify readers of H3K36me3 that activate transcription downstream of Set2, we screened 18 known Chromodomain (CD)-containing proteins, which recognize lysine methylation marks, for loss-of-function phenotypes that phenocopied Set2 (Allis and Jenuwein, 2016; McCarthy et al., 2018; Navarro-Costa et al., 2016). Unexpectedly, of the 18 CD-containing proteins that we examined by RNAi knockdown, only Msl3 displayed a cyst accumulation phenotype (see Table S1). Msl3 is a reader of H3K36me3 mark that promotes dose compensation in the soma in male flies, where it acts within the MSL complex (see Lucchesi and Kuroda, 2015), but whether Msl3 has a role in the female Drosophila germline has not been examined.

As a first test, we investigated Msl3 expression in ovaries. We analyzed msl3 transcript levels at different stages of oogenesis, using RNA-seq libraries that we enriched for GSCs, CBs, cysts and adult whole ovaries, as previously described (McKearin and Ohlstein, 1995; Xie and Spradling, 1998; Zhang et al., 2014). We found that msl3 mRNA is expressed during oogenesis, consistent with what was reported on FlyBase (The Modencode Consortium et al., 2010) (Fig. S2A). We also examined ovaries from a fly line expressing GFP-tagged Msl3 under endogenous control (Strukov et al., 2011), by staining for GFP and 1B1. We found that Msl3-GFP was expressed in the germline in single cells marked by spectrosomes and early cysts marked by fusomes (Fig. 2A-A′; Fig. S2B). Thus, unlike Msl2 (Bashaw and Baker, 1997; Parisi et al., 2001), Msl3 is expressed at appreciable levels during early stages of oogenesis at both the RNA and protein level.

To verify our RNAi knockdown experiments indicating a role in oogenesis, we examined validated msl3 mutants (Bachiller and Sánchez, 1989; Sural et al., 2008; UCHIDA et al., 1981). As background mutations in msl3 mutant stocks have been described to cause synthetic lethality (Erickson, 2016; Gladstein et al., 2010), we analyzed three independent alleles from different genetic backgrounds. We created trans-allelic combinations of independent msl3 mutations to eliminate potential deleterious effects of homozygous background mutations; these combinations lead to cyst accumulation monitored by accumulation of fusome-positive cells, germline loss monitored by absence of Vasa-positive cells, and a reduction in formation of proper egg chambers (Fig. 2B-D; Fig. S2C-D′). Furthermore, as our screen indicated depletion of msl3 in the germline alone, using RNAi resulted in the accumulation of cysts, which phenocopied msl3 mutants (Fig. 2D, Fig. S2E-F′). The cysts that accumulate upon msl3 germline depletion expressed bam but were pMad negative and failed to properly express Rbfox1 or Bru, phenocopying Set2 germline depletion (Fig. 2E-I; Fig. S2G-M). In addition, these cysts failed to specify an oocyte, as monitored by Egl, and showed reduced expression of the synaptonemal protein C(3)G (Fig. 2J-K1′; Fig. S2N-O′). Last, expression of msl3 in the germline of msl3 mutant females was sufficient to rescue the differentiation defect (Fig. 2D,L,M). Taken together, we conclude that Msl3 is required in the female germline to promote proper differentiation.

To determine whether the H3K36me3 writer Set2 and the H3K36me3 reader Msl3 act together to promote oogenesis, we generated flies heterozygous for Set2 and msl3. Although the trans-heterozygous flies were viable, their germaria displayed severe germline loss compared with single heterozygous controls (Fig. S3A-C). Although loss of Msl3 did not affect H3K36me3 levels, loss of Set2 attenuated Msl3 expression (Fig. S3D-H). Together, these data suggest that Set2 and Msl3 impinge upon the same developmental pathway(s), with Msl3 acting downstream of Set2 to promote proper differentiation.

This function of Msl3 during oogenesis is independent of its role in the MSL complex, as we found that msl2, roX1 and roX2 are barely expressed (<1 TPM), consistent with previous reports (Bashaw and Baker, 1997; Meller et al., 1997; Parisi et al., 2004). Additionally, validated msl1, msl2 and mle mutants (Bachiller and Sánchez, 1989; Belote, 1983; UCHIDA et al., 1981) did not result in early oogenesis defects (Fig. S3I), nor did loss of germline MOF (Sun et al., 2015).

We asked whether the H3K36me3 reader Msl3 cooperates with another HAT-containing complex to regulate cyst differentiation. Using an RNAi screen, we found that members of the Ada2a-containing (ATAC) complex phenocopy loss of Set2 and msl3 in the germline (McCarthy et al., 2018; Spedale et al., 2012; Suganuma et al., 2008) (Fig. 3A-C; Fig. S4A-H). The ATAC complex contains 13 proteins, some shared with other complexes, including Gcn5 (Spedale et al., 2012). Depletion of six members, four of which are specific to the ATAC complex, resulted in the accumulation of cysts and germline loss (Fig. S4A-H), Of these ATAC complex members, we chose to focus on Negative Cofactor 2β (NC2β), as its defect was highly penetrant but maintained sufficient germline for transcriptomic analysis.

Loss of NC2β in the germline led to GSC loss monitored by loss of spectrosome containing cells and accumulation of cysts-like structures that were marked by fusomes (Fig. 3A-C; Fig. S4A-B′). To determine whether the cysts that accumulate due to loss of NC2β phenocopy cysts that accumulate due to loss of Set2 and msl3, we independently probed for GSC marker pMad and differentiation markers Bam, Rbfox1 and Bru. In addition, we also probed for meiotic marker C(3)G. We found that cysts that accumulate upon germline depletion of NC2β expressed Bam, did not contain pMad-positive cells or properly express Rbfox1 or Bru (Fig. 3D-H; Fig. S5A-G). In addition, loss of NC2β leads to meiotic defects and oocyte specification, as monitored by C(3)G localization and Egl, respectively (Fig. 3I-J1′; Fig. S5H-I′). These data suggest that the ATAC complex is required for proper differentiation, as well as oocyte specification that phenocopies loss of Set2 and msl3.

As components of the ATAC complex phenocopy loss of the H3K36me3 writer Set2 and the H3K36me3 reader msl3, we asked whether the ATAC complex acts downstream of H3K36me3 mark to promote proper meiosis. To test this, we stained for H3K36me3 in NC2β RNAi flies and found that H3K36me3 levels were unaltered, suggesting that ATAC complex acts downstream of Set2 (Fig. S5J-L). In addition, we made use of a mutant of the active HAT in the ATAC complex, Atac2, as there were no NC2β mutants available. We generated flies heterozygous for both Atac2 and msl3, and found that although trans-heterozygotes flies were viable, their germaria had severe defects compared with controls (Fig. S5M-O). Thus, the ATAC complex works downstream of Set2, and Atac2 genetically interacts with msl3. Taken together, our data suggest that Set2, Msl3 and the ATAC complex impinge upon the same developmental pathway(s) to regulate differentiation in the Drosophila female germline.

To determine how Set2, Msl3 and ATAC promote differentiation, we compared the transcriptomes of Set2, msl3 and NC2β germline-depleted ovaries to a developmental control that accumulates cysts. To enrich for cysts, we induced bam expression under the control of a heat-shock (hs) promoter in the background of germaria depleted for bam (bam RNAi;hs-bam) (Ohlstein and McKearin, 1997; Zhang et al., 2014). We found 662 significantly downregulated and 65 significantly upregulated RNAs in Set2-depleted germaria compared with bam RNAi;hs-bam ovaries [fold change >|4|; False discovery rate (FDR)<0.05] (Fig. 4A). There were 283 significantly downregulated RNAs and 302 significantly upregulated RNAs in msl3 RNAi compared with bam RNAi;hs-bam (Fig. 4A′). Last, there were 466 RNAs significantly downregulated and 277 upregulated in NC2β RNAi compared with the developmental control (Fig. 4A″). Set2 and ATAC are part of the generalized transcriptional machinery and are expected to regulate more genes than Msl3 alone. This is evident in the RNA-seq results, where we find more downregulated targets upon the loss of Set2 (662 targets) and NC2β (466 targets) than loss of msl3 (283 targets) (Fig. 4A-A″). Of those transcripts that were differentially expressed in Set2-, msl3- and NC2β-depleted germ cells compared with the bam RNAi;hs-bam control, there were 29 shared RNAs that were downregulated (Fig. 4B) and 11 shared RNAs that were upregulated (see Table S2). This small but significant overlap of targets suggests a specific subset of Msl3 targets are co-regulated by Set2 and ATAC.

Although Rbfox1 protein is not properly expressed upon loss of Set2, Msl3 and NC2β, Rbfox1 mRNA was not among the shared downregulated RNAs (Fig. 4C). We verified that Rbfox1 mRNA was present in the germline of msl3-depleted ovaries by in situ hybridization (Fig. S6A-B′). Thus, Set2, Msl3 and ATAC do not regulate Rbfox1 mRNA levels to promote differentiation. In contrast, we found that several SC member genes were among the shared downregulated genes, including orientation disruptor (ord), sisters unbound (sunn) and corona (cona) (Hughes et al., 2018), but C(3)G was not among those genes (Fig. 4D,E; Fig. S6C,D). To validate the loss of SC components, we crossed an Ord-GFP line that is under endogenous control (Balicky et al., 2002) into msl3 mutants and found that msl3 mutant ovaries had both lower levels and an altered location of Ord-GFP compared with controls (Fig. S6E-F′). Although this change can be mediated by both RNA levels or protein stability, we found ord nascent mRNAs levels are also downregulated (see below). The shared downregulated targets also included 11 candidate genes (CGs) of unknown function, and the ribosomal protein paralog RpS19b (S19b), but not RpS19a (S19a) (Fig. 4B,F; Fig. S6G-J′).

We hypothesized that Msl3 and Msl3-regulated mRNAs would be expressed at the same stages, from GSCs until the cyst stages. To test this hypothesis, we analyzed mRNA levels of the 29 targets in RNA-seq libraries enriched for either GSCs, CBs, cysts or unenriched wild-type ovaries. Indeed, transcript levels overlapped with Msl3 expression and then dropped off (Fig. 4G). Last, to determine whether SC genes and RpS19b were affected at the level of transcription, we measured nascent mRNA (pre-mRNA) levels, using qRT-PCR as a proxy for transcription. We found that, upon depletion of msl3, nascent mRNA levels of SC genes and RpS19b are significantly lower (Fig. 4H). Taken together, these data suggest that Set2, Msl3 and ATAC regulate transcription of SC components and RpS19b, but not Rbfox1, during differentiation. Although loss of SC components leads to loss of recombination, it does not cause a differentiation defect (Collins et al., 2014). Proper cyst differentiation is mediated by Rbfox1 (Carreira-Rosario et al., 2016; Tastan et al., 2010). Altogether, this suggests that a target of Set2, Msl, and ATAC regulates not Rbfox1 mRNA levels but another aspect of expression of the gene – potentially its translation to promote differentiation.

RpS19b is a ribosomal protein and is one of two RpS19 paralogs in Drosophila. Given that loss of Set2, Msl3 and NC2β decreased Rbfox1 protein levels without affecting Rbfox1 mRNA levels, we hypothesized that the reduced RpS19b expression we observed in these mutants resulted in the decreased translation of Rbfox1 mRNA. If RpS19b is required for Rbfox1 translation, then the RpS19b protein should be present when Rbfox1 protein expression occurs. We examined lines expressing RpS19b-GFP and RpS19a-HA from under endogenous control. RpS19b-GFP was germline enriched, whereas RpS19a-HA was expressed in both the germline and soma of the gonad (Fig. 5A-A1; Fig. S7A,B). In the germline, RpS19b-GFP was expressed at high levels in single cells and gradually decreased in cyst stages, which overlapped with the protein expression of Msl3 and Rbfox1 (Fig. 5B).

If RpS19b acts downstream of Msl3 to promote the translation of Rbfox1 mRNA, then loss of RpS19b should result in reduced Rbfox1 protein levels and a germline phenotype that phenocopies that of msl3. We used RNAi to specifically deplete RpS19b but not RpS19a in the germline (Fig. S7C-F′) and found that RpS19b-depleted germaria accumulated bam-positive cysts that lack Rbfox1 protein (Fig. 5C-H; Fig. S7G-H′). To test whether RpS19b is one of the main targets of msl3, we asked whether the addition of RpS19b could rescue the differentiation defect of msl3 mutants. Adding one copy of RpS19b-GFP in msl3 mutant flies rescued the early cyst defect, including Rbfox1 expression, and led to egg chamber formation (Fig. 5I-M). Overexpression of RpS19b could also rescue the differentiation defect upon germline depletion of msl3, leading to egg chamber formation (Fig. 5N,O). Thus, our data suggest that Msl3 promotes the expression of RpS19b and thus Rbfox1 translation and proper differentiation.

Our model predicts that the Msl3-mediated transcription of the SC members ord, sunn and cona is independent of Rbfox1 protein expression. To test this model, we examined the localization of the SC component C(3)G in msl3 mutants that express RpS19b (Anderson et al., 2005; Page and Hawley, 2001). C(3)G was not a co-regulated target and thus allowed us to visualize SC defects. We found that while msl3 mutants with restored RpS19b expression make egg chambers, C(3)G does not properly localize to the oocyte nucleus in egg chambers and the females are infertile (Fig. 5P-Q′). Thus, RpS19b is not involved in Msl3-mediated transcriptional regulation of SC members to promote recombination during meiosis.

To validate the role of RpS19b in oogenesis, we generated a CRISPR null mutant of the gene (RpS19b^CRISPR). This null is viable but unexpectedly did not display any oogenesis defects in contrast to RpS19b RNAi (Fig. S7I-K). Studies in organisms, including zebrafish, have reported transcriptional compensation in mutants, but not upon depletion by RNAi (El-Brolosy et al., 2019). To determine whether the RpS19b^CRISPR mutants display transcriptional changes, we performed RNA-seq of CB-enriched ovaries from control and RpS19b^CRISPR flies, both with a bam-depleted background to enrich for the undifferentiated stages in which RpS19b is primarily expressed. We found that loss of RpS19b resulted in 79 downregulated genes and 934 upregulated genes, with ninefold downregulation of RpS19b but no increase in RpS19a RNA levels (Fig. S7L). As RpS19a mRNA levels were not altered, we then asked whether RpS19b mutants have proper development because they have increased levels of RpS19a protein. Using a RpS19 antibody that detects both paralogs, we found that levels of RpS19 were not downregulated in RpS19b^CRISPR mutant gonads compared with the control (Fig. S7M-P). Furthermore, germline depletion of RpS19a in RpS19b^CRISPR mutants results in the complete loss of the germline, although there is no defect at all in homozygous RpS19b^CRISPR mutants and only the accumulation of cysts upon RpS19a depletion alone (Fig. S7Q-R). In contrast, RpS19b depletion in RpS19b^CRISPR mutants resulted in no defect (Fig. S7S-T). Thus, we conclude that full loss of RpS19b can be compensated for by increased levels of RpS19a in RpS19b^CRISPR but not in RpS19b RNAi, via as yet unknown mechanisms.

These data suggest that what is required for Rbfox1 translation is a critical level of RpS19, which can be provided by either the RpS19a or S19b isoforms, or by a combination of both. To test this, we depleted RpS19a from the germline and found that the germaria accumulate bam-positive cysts that have significantly reduced levels of Rbfox1 (Fig. S8A-L′). Ectopic expression of RpS19a-HA in msl3-depleted ovaries that have reduced levels of RpS19b restored Rbfox1 protein expression and egg chamber formation (Fig. S8M-P). Consistent with our previous data showing that Rps19b cannot rescue expression of SC components in the msl3 mutant, these females were also infertile and had perturbed C(3)G localization (Fig. S8Q-R). Furthermore, expression of human RpS19 in the germline of msl3-depleted germaria also was able to rescue the cyst accumulation phenotype, leading to the production of normal egg chambers (Fig. S8S-T1). Thus, our data taken together suggest that a proper dose of RpS19 is essential for the translation of Rbfox1 protein.

We next examined the requirement for RpS19 in Rbfox1 translation. First, to determine whether RpS19a and S19b were incorporated into actively translating ribosomes, we performed polysome profiling followed by western blot analysis using wild-type ovaries and those enriched for undifferentiated germ cell cysts (bam RNAi). Although RpS19a-HA is present in polysome fractions in wild type ovaries, RpS19b-GFP appears to be preferentially enriched in polysomes from ovaries enriched for undifferentiated germ cells, consistent with its expression pattern early in oogenesis (Fig. 6A-B′). To test whether RpS19 paralogs affect translation in cysts, we pulsed gonads with a puromycin analog, O-propargyl-puromycin (OPP), that is incorporated into translated peptides and can be detected using Click-chemistry (Sanchez et al., 2016). We found that cysts that accumulate upon the loss of RpS19a and RpS19b have decreased translation compared with cysts of control ovaries (Fig. 6C-F). Thus, we find that RpS19a and S19b are present in the translating polysomes and are important for translation in cysts.

To directly test whether RpS19b is required for Rbfox1 translation specifically, we then performed polysome-seq on germaria depleted of RpS19b compared with control germaria enriched for cysts using bam RNAi;hs-bam (Fig. 6G,H). Depletion of germline RpS19b did not significantly affect the translation of the germline-specific mRNA nanos, but there was a reduction in Rbfox1 cytoplasmic isoform mRNA translation compared with the control (Fig. 6I-J). Additionally, depletion of RpS19b using RNAi, did not reduce the levels or translation efficiency of RpS19a (Fig. 6K). Taken together, our data suggest that there is an increased expression of RpS19 during early development that is required for translation of Rbfox1 mRNA.

We have identified Set2, Msl3 and the ATAC complex as transcriptional regulators of differentiation and proper meiosis in Drosophila. We find that Set2, Msl3 and ATAC together regulate oogenesis downstream of the differentiation factor Bam, but upstream of the crucial differentiation factor Rbfox1. Although we find significant shared targets between Set2, Msl3 and NC2β that control differentiation, the total number of shared targets is small. One reason for this could also be the ability of Msl3 to read H4K20me1 marks (Kim et al., 2010), allowing it to affect a distinct set of targets than those marked by the H3K36me3 laid down by Set2. Similarly, Msl3 could recruit a different HAT from the one associated with NC2β, one that functions in the undifferentiated stages (McCarthy et al., 2018). Thus, we hypothesize that Set2, Msl3 and ATAC work in concert on a select subset of targets to regulate meiosis and oogenesis.

We find that recruitment by Msl3 of the basal transcriptional machinery, ATAC to the Set2-mediated H3K36me3 marks, would lead to enhanced transcription of a subset of genes that promote differentiation at the proper stage. We do not know what controls the expression of Msl3 itself during oogenesis. msl3 mRNA is present as part of the maternal contribution deposited into the developing oocyte (Eichhorn et al., 2016; Hua et al., 2014) This suggests that msl3 mRNA is also transcribed in the later stages of oogenesis and is likely post-transcriptionally regulated in the later stages. Although our data demonstrates that Msl3 expression is required for proper differentiation in female Drosophila, we do not think Msl3 expression is sufficient for differentiation, as overexpression of msl3 does not lead to precocious differentiation (Fig. 2L,M).

We find that Set2, Msl3 and ATAC regulate oogenesis in two ways: (1) they transcriptionally upregulate members of the synaptonemal complex that are crucial for recombination; and (2) they promote transcription of the germline-enriched RpS19 paralog RpS19b. The expression of RpS19b then controls the translation of Rbfox1, which is required for exit from the mitotic cell cycle and entry into the meiotic cell cycle (Fig. 7). Although several components of the synaptonemal complex, Cona, Ord and Sunn, are regulated at the transcriptional level by Set2, Msl3 and ATAC, components such as C(3)G and Crossover suppressor on 2 of Manheim [C(2)M] are not. This suggests that some synaptonemal complex members, such as C(2)M and C(3)G, might be regulated at the post-transcriptional level. Taken together, these data argue that the Set2-Msl3-ATAC complex coordinates the transcription of several crucial factors of the recombination machinery and the translation of a meiotic cell cycle regulator to promote proper oogenesis (Fig. 7).

Flies were grown at 25-31°C and dissected between 1 and 5 days post-eclosion. The following RNAi stocks were used in this study (if more than one line is listed, then both were quantitated and the first is shown in the main figure): Set2 RNAi [Bloomington Drosophila Stock Center (BDSC) #33706 (#1) and #42511 (#2)], msl3 RNAi (BDSC #35272), NC2β RNAi [BDSC #57421 (#1) and VDRC #v3161 (#2)], Ada2a RNAi (BDSC #50905), Atac1 RNAi [Vienna Drosophila Resource Center (VDRC) #v36092], Atac2 RNAi (VDRC #v16047), D12 RNAi (VDRC #v29954), wds RNAi (BDSC #60399), NC2α RNAi (BDSC #67277), bam RNAi (BDSC #58178), hs-bam/TM3 (BDSC #24637), RpS19b RNAi (VDRC #v22073 and #v102171) and RpS19a RNAi (BDSC #42774 and VDRC #v107188).

The following mutant and overexpression stocks were used in this study: Set2¹/FM7 (BDSC #77916), msl3¹/TM3 (BDSC #5872), msl3^KG/TM3 (BDSC #13165), mls3^MB/TM3 (BDSC #29244), msl1^γ216/CyO (BDSC #5870), msl1^kmB/CyO (BDSC #25157), msl2²²⁷/CyO (BDSC #5871), msl2^kmA/CyO (BDSC #25158), mle¹/SM1 (BDSC #4235), mle⁹/CyO (BDSC #5873), Hel89B⁰⁸⁷²⁴/TM3 (BDSC #11732), Hel89B Df/TM6 (BDSC #7982), Atac2^e03046/CyO (BDSC #18111), RpS19b^EY00801 (BDSC #15043), RpS19b^CRISPR (this study), UAS-hRpS19-HA (BDSC #66014) and UAS-msl3-GFP (this study).

The following tagged lines were used in this study: msl3-GFP (Kuroda Lab, Brigham and Women's Hospital, Boston, MA, USA), RpS19a-3xHA (this study), RpS19b-GFP (this study), and ord-GFP (Bickel Lab, Dartmouth College, Hanover, NH, USA).

The following tissue-specific drivers were used in this study: UAS-Dcr2;nosGAL4 (BDSC #25751), UAS-Dcr2;nosGAL4;bam-GFP (Lehmann Lab, Whitehead Institute, MIT, Cambridge, MA, USA) and If/CyO;nosGAL4 (Lehmann Lab). We used UAS-Dcr2;nosGAL4 for all RNAi experiments, as this combination as it gave us the most consistent phenotype with the most germline in all genotypes indicated.

Ovaries were dissected and stained as previously described (McCarthy et al., 2018). The following primary antibodies were used: mouse anti-1B1 (1:20; DSHB), rabbit anti-Vasa (1:1000; Rangan Lab, University at Albany, SUNY, USA), chicken anti-Vasa (1:1000; Upadhyay et al., 2016), rabbit anti-GFP (1:2000; Abcam, ab6556), guinea pig anti-Rbfox1 (1:1000; Tastan et al., 2010), mouse anti-C(3)G (1:1000; Hawley Lab, Stowers Institute, Kansas City, MO, USA), rabbit anti-H3K36me3 (1:500; Abcam, ab9050), rabbit anti-pMAD (1:150; Abcam, ab52903), mouse anti-BamC (1:200; DSHB, Supernatant), rabbit anti-Bru (1:500; Lehmann Lab), rabbit anti-Egl (1:1000; Lehmann Lab), rat anti-HA (1:500; Roche, 11 867 423 001) and rabbit anti-RpS19 (1:20; Proteintech, 15085-1-AP). Anti-RpS19 was pre-cleared at 1:20, the supernatant was then diluted at 1:2.5 for staining. The following secondary antibodies were used at a dilution of 1:500: Alexa 488 (Molecular Probes), and Cy3 and Cy5 (Jackson Labs).

The tissues were visualized and images were acquired using a Zeiss LSM-710 confocal microscope under 20×, 40× and 63× oil objectives.

We quantified loss of germline by counting ovarioles that did not have Vasa-positive cells. For loss of stem cells, we quantified loss of spectrosome-containing cells. For cyst differentiation defect, we counted the number of germaria with cysts containing more than 16 cells and/or accumulation of 16-cell cysts close to the niche.

To quantify antibody staining intensities for Rbfox1, H3K36me3, Bruno, GFP, HA and RpS19 or in situ probe fluorescence in germ cells, images for both control and experimental germaria were taken using the same confocal settings. Z stacks were obtained for all images. Similar planes in control and experimental germaria were chosen, the area of germ cells positive for the proteins or in situs of interest was outlined and analyzed using the ‘analyze’ tool in Fiji (ImageJ). The mean intensity and area of the specified region was obtained. An average of all the ratios (mean/area), for the proteins or in situs of interest, per image was calculated for both control and experimental. Germline intensities were normalized to somatic intensities or, if the protein or in situ of interest is germline enriched and not expressed in the soma, they were normalized to Vasa or background. The highest mean intensity between control and experimental(s) was used to normalize to a value of 1 arbitrary unit (AU) on the graph. A minimum of five germaria was used for quantitation.

Assays were conducted in cages with females under testing and wild-type control males. Cages were maintained at 25°C. All flies were 1 day post-eclosion upon setting up the experiment and analyses were performed on four consecutive days. The number of eggs laid were normalized to the total number of females.

Ovaries from flies of various genotypes were dissected in 1×PBS. For enriching GSCs, we over overexpressed TKV in the germline, for CBs, we used bam mutant and for cysts we used bam mutant also carrying a hsbam transgene. Post heat-shock CBs converted to cysts. RNA was isolated using TRIzol (Invitrogen, 15596026), treated with DNase (TURBO DNA-free Kit, Life Technologies, AM1907) and then run on a 1% agarose gel to check the integrity of the RNA. To generate mRNA-enriched libraries, total RNA was treated with poly(A)tail selection beads (Bioo Scientific, NOVA-512991) and then following the manufacturer's instructions of the NEXTflex Rapid Directional RNA-seq Kit (Bioo Scientific, NOVA-5138-08), except that RNA was fragmented for 13 min. Single-end mRNA sequencing (75 base pair) was performed on biological duplicates from each genotype on an Illumina NextSeq500 by the Center for Functional Genomics (CFG).

After quality assessment, the sequenced reads were aligned to the Drosophila melanogaster genome (UCSCdm6) using HISAT2 (version 2.1.0) with the RefSeq-annotated transcripts as a guide (Kim et al., 2015). Raw counts were generated using featureCounts (version 1.6.0.4) (Liao et al., 2014). Differential gene expression was assayed by edgeR (version 3.16.5), using a false discovery rate (FDR) of 0.05, and genes with fourfold or higher differential expression were considered significant. The raw and unprocessed data for RNA-seq generated during this study have been deposited in GEO under accession number GSE143728.

Adult ovaries (five ovary pairs per sample per experiment) were dissected and fixed as previously described. The ovaries were washed with PT [1× phosphate-buffered saline (PBS) and 0.1% Triton-X 100) three times for 5 min each. Ovaries were permeabilized by washing once with increasing concentrations of methanol for 5 min each (30% methanol in PT, 50% methanol in PT and 70% methanol in PT) then incubating in methanol for 10 min. Ovaries were then post-fixed by washing once with decreasing concentrations of methanol for 5 min each (70% methanol in PT, 50% methanol in PT and 30% methanol in PT). Ovaries were then washed with PT three times for 5 min and then pre-hybridized in wash buffer for 10 min (10% deionized formamide and 10% 20× SSC in RNase-free water). Ovaries were incubated overnight in hybridization solution (10% dextran sulfate, 1 mg/ml yeast tRNA, 2 mM RNaseOUT, 0.02 mg/ml BSA, 5×SSC, 10% deionized formamide and RNase-free water) at 30°C. The hybridization solution was removed, and ovaries washed with wash buffer twice for 30 min at 30°C. Wash buffer was removed, and ovaries were mounted using Vectashield with 4′,6′-diamidino-2-phenylindole (DAPI).

Templates were amplified with gene-specific primers (listed below) and then fluorescently labeled probes were generated by following the manufacturer's instructions (Thermo Fisher's FISH tag RNA kit, F32954): Rbfox1 F, 5′-CGTAGCGCCTTTTCCGGG-3′; Rbfox1 R, 5′-TAATACGACTCACTATAGGGCCACAGCCGCCACTTGAATA-3′; RpS19b F, 5′-TGCCTGGAGTCACAGTAAAGG-3′; RpS19b R, 5′-TAATACGACTCACTATAGGGTGTTGGCTATGCGATCCAAGT-3′; RpS19a F, 5′-ATGCCAGGCGTCACAGTGAA-3′; RpS19a R, 5′-TAATACGACTCACTATAGGGTTACTTGGAAATAACAATGGGCCC-3′.

Protein synthesis was detected using a short-term ovary incorporation assay, Click-iT Plus OPP (Invitrogen, C10456). Ovaries were dissected in Schneider's Drosophila media (Thermo Fisher, 21720024) and then incubated in 50 μM OPP reagent for 30 min. Tissue was washed in 1× PBS and then fixed for 15 min in 1× PBS plus 5% methanol-free formaldehyde. Tissue was then permeabilized with 1% Triton X-100 in 1× PBST (1× PBS with 0.2% Tween 20) for 30 min, and samples were then washed in 1× PBS and incubated in Click-iT reaction cocktail following the manufacturer's instructions. Samples were washed with Click-iT reaction rinse buffer and then immunostained following previously described procedures.

To generate the RpS19b mutants, guide RNAs were designed using https://flycrispr.org/target-finder/ and synthesized as 5-unphosphorylated oligonucleotides, annealed, phosphorylated and ligated into the BbsI sites of the pU6-BbsI-chiRNA vector using the primers listed below (Gratz et al., 2013). Homology arms were synthesized as a gene block (IDTDNA) and cloned into pHD-dsRed-attP (Gratz et al., 2015; Addgene plasmid #51019) using Gibson Assembly (gene blocks listed in the supplementary Materials and Methods). Guide RNAs and the donor vector were co-injected into nos-Cas9 embryos (Rainbow Transgenics): RpS19b gRNA1 F, 5′-CTTCGCATGCCTGGAGTCACAGTAA-3′; RpS19b gRNA1 R, 5′-AAACTTACTGTGACTCCAGGCATGC-3′; RpS19b gRNA2 F, 5′-CTTCGTAGTGATAATCATGGAAAC-3′; RpS19b gRNA2 R, 5′-AAACGTTTCCATGATTATCACTAC-3′.

RpS19a3x-HA- (referred to as RpS19a-HA throughout the text) and RpS19b-GFP-tagged lines were made using a combination of in vivo bacterial recombineering and GatewayTM Technology as previously described (Shalaby et al., 2017).

RNA was extracted from w¹¹¹⁸ ovaries and made into cDNA using the SuperScript II-Strand Kit (Thermo Fisher, 18064014). msl3 CDS was amplified, attB sites and tagged sequence were amplified into the PCR product using the primers listed below. PCR products were cloned into pDONR (Thermo Fisher, 11789-020) and swapped into pENTR (Thermo Fisher, 11791-020) using BP and LR reactions, respectively. The plasmid was sent for injection into w¹¹¹⁸ flies (Genetic Services). PCR primers are as follows: msl3 CDS F, 5′-ATGACGGAGCTAAGGGACGAGAC-3′; msl3 CDS R, 5′-CTAAGCAGCAATCCCATCCAGGG-3′; attB F, 5′-GGGGACAAGTTTGTACAAAAAAGCAGGCTTCATGACGGAGCTAAGGGACGAGAC-3′; attB R, 5′-GGGGACCACTTTGTACAAGAAAGCTGGGTCCTAAGCGTAATCTGGCACATCGTATGGGTAAGCAGCAATCCCATCCAGGG-3′.

Polysome profiling of ovaries from bam RNAi;hs-bam and UAS-Dcr2;nosGAL4 >RpS19b RNAi flies was adapted from Flora et al., (2018) and Fuchs et al. (2011). Two-hundred ovary pairs were dissected in Schneider's media and immediately flash frozen with liquid nitrogen. Ovaries were homogenized in lysis buffer, 20% of lysate was used as input for mRNA isolation and library preparation (as described above). Samples were loaded onto 10-50% CHX-supplemented sucrose gradients in 9/16×3.5 PA tubes (Beckman Coulter, 331372) and spun at 35,000 g in SW41 for 2.45-3 h at 4°C. Gradients were fractionated with a Density Gradient Fractionation System (Brandel, 621140007). RNA was extracted using acid phenol-chloroform and precipitated overnight. Pelleted RNA was resuspended in 20 μl water and libraries were prepared as described above.

Fifty to 200 C-enriched RpS19b-GFP ovaries and 30 adult RpS19b-GFP;RpS19a-HA ovaries were dissected and prepared as described above, except sucrose solutions were supplemented with either 100 μg/μl CHX or 2 mM puromycin with 1 mg heparin prior to making gradients. Following fractionation, protein was extracted by ethanol precipitation and run on a TGX pre-cast gradient gel (BioRad, 456-1094). Blots were blocked with 5% milk in 1× PBST and incubated in primary antibody in 5% BSA in 1× PBST. Following 1× PBST washing, blots were incubated in secondary antibody in 5% milk in 1× PBST. Blots were washed with 1× PBST and then imaged with chemi-luminescence kit (BioRad, 170-5060). The following primary antibodies were used: rabbit anti-GFP (1:4000; Abcam, ab6556), rat anti-HA (1:3000; Roche, 11 867 423 001), rabbit anti-RpS25 (1:1000; abcam, ab40820) and rabbit anti-RpS19 (1:1000; Proteintech, 15085-1-AP). The following secondary antibodies were used: anti-rat HRP (1:10,000; Jackson Labs, 112-035-003) and anti-rabbit HRP (1:10,000; Jackson Labs, 111-035-144).

Relative fluorescence signals were compared between control and experimental groups using parametric tests (Student t-test or one-way ANOVA). Horizontal lines on scatter dot plots represent mean with 95% confidence interval and stars on stacked bar graphs represent statistical significance of corresponding color data set. Reported P-values correspond to unpaired and two-tailed tests. Analysis of percentage defect were compared between control and experimental groups using Fisher's exact test. All analyses were performed using Prism 8 software (GraphPad) and are reported in figure legends.

Fly food was made using previously described procedures (Upadhyay et al., 2018).

We are grateful to Drs Marlow and Siekhaus for comments. We thank Drs Kuroda, Bach, Lehmann, Hawley, Bickel and Fuchs for reagents. We thank the Bloomington Drosophila Stock Center, the Vienna Drosophila Resource Center and the Transgenic RNAi Project (NIH/NIGMS R01-GM084947) for fly stocks.

The peer review history is available online at https://journals.biologists.com/dev/article-lookup/doi/10.1242/dev.199625