WD40 protein Wuho controls germline homeostasis via TRIM-NHL tumor suppressor Mei-p26 in Drosophila

Stem cell self-renewal and differentiation must be balanced for proper tissue homeostasis (Nakada et al., 2011; Simons and Clevers, 2011; Spradling et al., 2011). This balance is known to be coordinated by transcriptional and post-transcriptional mechanisms (Simons and Clevers, 2011; White-Cooper and Caporilli, 2013), which have not been fully described at a molecular level. An excellent model for studying genetic regulation of the cell fate transition from stem cell to differentiated progeny is the Drosophila ovary, as its germline stem cells (GSCs) and differentiated progeny are well characterized in cell biology (Fig. 1A,A′) (Wong et al., 2005). Although the physiology of Drosophila egg production is well described at a cellular level, the molecular regulatory mechanisms are still an area of active investigation.

Wuho (Wh; meaning ‘no progeny’ in Chinese) is an evolutionarily conserved protein comprising five WD40 domains (Cheng et al., 2016; Wu et al., 2006), which mediate protein-protein interactions (Xu and Min, 2011). Homologs of Wh have been shown to exert a wide variety of functions via interactions with m7G46 tRNA methyltransferase, Flap endonuclease 1 (FEN1) and Culin-Ring ubiquitin ligase 4 (Alexandrov et al., 2002; Cheng et al., 2016; Wang et al., 2017). Interestingly, wh mutant males are sterile because their spermatids are not properly elongated to make functional spermatozoa, and female flies are semi-sterile for unknown reasons (Wu et al., 2006), suggesting that Wh may potentially play an important role in the molecular regulatory circuitry of the GSC lineage.

Interestingly, the ovarian phenotypes we report in Wh mutants are strikingly similar to those in Mei-p26 mutants. Mei-p26 is a member of the tripartite motif and Ncl-1, HT2A and Lin-41 domain (TRIM-NHL) family of proteins, which is highly conserved among metazoans (Sardiello et al., 2008). TRIM-NHL proteins are known to control developmental transitions through mechanisms such as the promotion of stem cell differentiation by suppressing proliferation (Chen et al., 2014). The molecular action of TRIM-NHL proteins is typically ubiquitination and translation silencing via E3 ligase RING domains; meanwhile, the NHL domains mediate protein-protein interactions (Tocchini and Ciosk, 2015). In the Drosophila germline, Mei-p26 controls GSC maintenance and differentiation depending on its expression level (Li et al., 2012; Neumüller et al., 2008; Page et al., 2000). However, the regulators of Mei-p26 in the GSC lineage are not known. In this study, we show that Wh is a key regulator of Mei-p26, and that these proteins function together in multiple contexts to control GSC maintenance and differentiation for germline homeostasis. Our results document a potentially generalizable role for WD40 proteins as a bridge between TRIM-NHL proteins and other cellular components, a function that is necessary to balance self-renewal and differentiation in the GSC lineage.

Wh was suggested to be a modulator of female fly fertility, as females homozygous for the wh⁷ null allele produce low numbers of progeny (Wu et al., 2006). However, wh⁷ is a truncated mutation generated by imprecise P-element excision, which resulted in the deletion of the wh gene as well as the 5′ end of the neighboring top3β gene (Fig. 1B). To test whether wh deficiency alone is sufficient to produce the fertility defects seen in wh⁷ mutant females, we utilized CRISPR/Cas9 to generate a new wh null mutant allele, wh^CRISPR. In this mutant, a CRISPR cassette containing stop codons and RFPs driven by an eye-specific promoter was inserted 50 to 60 nucleotides after the wh transcriptional start site (Fig. 1B). wh^CRISPR and wh⁷ mutant females exhibited similarly dramatic reductions in egg laying when compared with controls or wh mutants bearing a wild-type wh transgene (Fig. 1C), suggesting a role for Wh in fecundity control.

To understand how Wh affects egg production, we first generated antibodies for western blot (see Materials and Methods) and for immunofluorescence (antibody generated by T.-S.H.), and examined Wh expression in the ovary. Wh was highly expressed and localized in the cytoplasm of nearly all germ cells and somatic cells (Fig. 1D,E), but was absent from the wh mutant, demonstrating the specificity of the anti-Wh antibody. We also examined the cell compositions in wh mutant ovarioles by labeling fusomes (Fig. 1F,G) (Lin et al., 1994) with 1B1 antibody against Hu-li tai shao, an Adducin-like protein (Zaccai and Lipshitz, 1996). In the wild-type ovariole (Fig. 1F), fusomes were only present in the germarium; GSCs and cystoblasts (CBs) contained round-shaped fusomes, whereas germ-cell cysts carried a branched fusome. In contrast, nearly 70% of wh⁷ mutant ovarioles (more than hundreds of ovarioles were examined) exhibited highly branched fusomes mislocalized in the germarium and egg chambers (Fig. 1G). Given that fusomes grow during germ cell division (McKearin, 1997; Storto and King, 1989), these data suggest that wh deletion results in overproliferation of germ cells and coincident fecundity defects.

To investigate whether Wh controls germ cell homeostasis, we examined GSC number in wh⁷ mutant germaria by counting the number of fusomes adjacent to niche cap cells with nuclear envelopes labeled by LamC (Hsu and Drummond-Barbosa, 2009). We found that wh⁷ mutant germaria carried fewer GSCs and the number further decreased with age (Fig. 2A,B,G), whereas niche cap cell numbers remained comparable with those in wild-type germaria (Fig. S1). The completion of cytokinesis between GSCs and CBs was marked by closure of the ring canal and the exclamation point morphology of the GSC fusome (Fig. 2C; see also Fig. 1A′) (Hsu et al., 2008; Kao et al., 2015). Strikingly, 67±6% (mean±s.d.) of wh⁷ mutant germaria (n=250) carried GSCs that were connected to several daughter cells via long branched fusomes with open ring canals between cells, forming stem-cysts (Fig. 2B,D,H). The number of wh⁷ mutant germaria with stem-cysts increased with age (Fig. 2G). As expected, the wild-type CB underwent four rounds of incomplete division to sequentially become a 2-, 4-, 8- and then 16-cell cyst (see insets in Fig. 2A), and cells in the 8-cell cyst were synchronous in terms of cell cycle progression (Fig. 2E,E′), according to the level of BrdU signal (S phase marker) in germ cells. In contrast, germ cells in the wh⁷ stem-cyst were asynchronous, as not all germ cells within an 8-cell cyst were positive for BrdU incorporation (Fig. 2F,F′). Further, the largest cyst size in wild-type germaria was always 16 cells, whereas the cyst size in wh⁷ mutant germaria varied. In some cases, stem-cysts in the mutant contained odd numbers of cells and/or the cysts had more than 16 cells (Fig. 2I), indicating a loss of the 2ⁿ rule for germ cell division.

To further examine the requirement for Wh in germ cell homeostasis, we disrupted wh expression specifically in the germline using an RNAi line driven by the nanos (nos)-GAL4 germ cell driver (Rørth, 1998). Wh expression was undetectable in germline cells of nos>wh^RNAi germaria (Fig. 2J,K), indicating high efficiency knockdown in the wh^RNAi line. As expected, GSCs were dramatically reduced and stem-cysts were present in 77.7% of nos>wh^RNAi germaria (n=53) (Fig. 2L-N). Similar results were also obtained from a clonal analysis (Fig. S2), wherein GSCs with the wh⁷ or wh^CRISPR homozygous mutation induced by the Flipase/Flipase recognition target system exhibited poor maintenance (Fig. S5A,B,E), formed stem-cysts (Fig. S2D) and progeny did not follow the 2ⁿ rule (Fig. S2B,C). These results indicate that Wh controls GSC maintenance and abscission, as well as synchronous division of germ cells in a cell-autonomous manner.

Because massive branched fusomes occupied the entire wh mutant germarium and were also present in early egg chambers (Fig. 1E and Fig. 2), we hypothesized that wh mutant germ cells may not properly exit mitosis and initiate meiosis. To test this idea, we examined expression patterns of several mitotic markers in wh mutant germaria: phospho-Histone H3 (PHH3) labeled M phase, BrdU labeled S phase and Cyclin B (CycB) labeled G2 and M phases (Hsu et al., 2008; Kao et al., 2015). In the wild type (Fig. 3A), 22% and 68% of germaria (n=50) carried PHH3- and BrdU-positive germ cells, respectively. These positively stained germ cells were mitotic and present in regions 1 (PHH3- and BrdU-positive germ cells) and 2a (BrdU-positive germ cells), but not in region 3, in which germ cells undergo meiosis (Lin and Spradling, 1993; Wettstein and Sotelo, 1971). In contrast, 50% and 91% of wh⁷ mutant germaria (n=64) carried PHH3- or BrdU-positive germ cells, respectively, with mitotic germ cells present in the posterior region. Note that wh mutant germaria were not able to be defined into three regions as they had abnormal cell arrangement and structure (Fig. 3B). Similar phenomena were also observed in germaria labeled with CycB (Fig. 3C,D) and γ-tubulin (Fig. 3E,F), which labels centrosomes that mark the onset of mitosis (Khodjakov and Rieder, 1999). These results indicate that wh⁷ mutant germ cells are mitotically active and persist in mitosis.

We further examined whether wh⁷ mutant germ cells underwent meiosis by labeling C3G, a component of the synaptonemal complex formed at the onset of meiosis (Page and Hawley, 2001), and γ-H2Av, a marker for double strand breaks (DSBs) that are induced during meiosis (Jang et al., 2003). As described in previous reports, meiosis was initiated in the early 16-cell cyst at region 2a, with small punctate C3G foci corresponding to the zygotene stage of meiosis (Page and Hawley, 2001). However, the zygotene stage is too brief to be observed in every germarium (Manheim and McKim, 2003). In agreement with these reports, C3G expression in the wild-type germarium was concentrated in two germ cells (pro-oocytes at pachytene of meiotic prophase I) of the lens-shaped 16-cell cyst, and only present in the oocyte of the egg chamber (Fig. 3G,G′). On the other hand, γ-H2AV foci were observable in region 2a/b but absent in the late region 2b of the wild-type germarium (Fig. 3G), suggesting that DSBs are repaired as meiosis proceeds (Hughes et al., 2018). In contrast, 40% of wh⁷ mutant germaria (n=83) were negative for both C3G and γH2AV (Fig. 3H). Perhaps the germ cells in these germaria did not enter meiosis or the 16-cell cyst stage was not reached. Nevertheless, ∼25% of wh⁷ mutant germaria carried germ cell cysts with punctate expression of C3G (Fig. 3I,I′), suggesting that some germ cells may be arrested in the zygotene stage of meiosis. About 30% of wh⁷ mutant germaria showed one or more large γ-H2Av foci (Fig. 3J), indicating DNA damage (Kuo and Yang, 2008; Lake et al., 2013). Consistent with these findings, we also saw that ∼30% of wh⁷ mutant germaria carried apoptotic germ cells (Fig. S3). Notably, these phenotypes – massive branched fusomes, stem-cysts, meiosis failure and germ cell death – were also observed in wh^CRISPR mutant germaria (Fig. S4). As a consequence, formation of oocytes [marked by Orb; Lantz et al., 1994] was disrupted (indicated by Orb not being specified to a single germ cell at a right gradient) in 24% or absent in 66% of wh⁷ mutants (n=60) compared with controls (Fig. 3K-P). It is also noteworthy that some wh mutant egg chambers showed a tumor-like phenotype, as they contained more than 16 germ cells (Fig. 3O,P), probably because of excessive germ cell mitosis. Thus, at a fundamental level, our results show that Wh is required for proper functioning of germ cell mitosis and meiosis programs.

Surprisingly, we found that the same phenotypes observed in wh⁷ mutant germaria, including stem-cysts, germ cell outgrowth with massive branched fusomes and fertility defect, are also present in mei-p26 mutants, i.e. mei-p26^mfs1/mei-p26^fs1 (Fig. 4A,B; Table 1) (Li et al., 2012; Neumüller et al., 2008). These similar phenotypes suggest that Wh and Mei-p26 may work in the same pathway to control germ cell homeostasis. To determine whether Wh acts either upstream or downstream of Mei-p26 to affect germ cell homeostasis, we first examined expression of Mei-p26 in wh mutant germaria and expression of Wh in mei-p26 mutant germaria. Mei-p26 expression was present in wh⁷ mutant germaria (Fig. 4C,D), and Wh expression levels and patterns were similar in wild-type and mei-p26 mutant germ cells (Fig. 4E,F). These results suggested that Wh might work with Mei-p26 in the same complex to maintain germ cell homeostasis, as both Wh and Mei-p26 bear protein-protein interaction domains. To test this hypothesis, we used ovaries carrying a wh-gfp transgene to perform an immunoprecipitation assay, pulling down GFP and examining whether Mei-p26 was present in Wh-GFP complex. Our results clearly showed that Wh interacts with Mei-p26 in the ovary (Fig. 4G).

As both Wh and Mei-p26 are highly conserved, we suspected that the Wh-Mei-p26 complex may be observable in other species. We thus expressed human orthologs of Wh tagged with Myc (WDR4-Myc) and Mei-p26 tagged with Flag (TRIM32-Flag) in human embryonic kidney 293T cells, a cell line often used for studying protein interactions. As expected, after pulling down Flag, WDR4-Myc was present in the TRIM-Flag complex (Fig. 4H), indicating an interaction between human WDR4 and TRIM32. To test whether human WDR4 can also function in Drosophila germ cells, we overexpressed WDR4 in wh mutant germ cells using the UAS-GAL4 system. Compared with 7-day-old wild-type ovaries bearing mature eggs with two dorsal appendages (as indicated by asterisks) (Fig. 4I, n=10), wh⁷ mutant ovaries were extremely small and rarely carried mature eggs (Fig. 4J, n=10). Overexpression of WDR4 in the germline of wh⁷ mutant ovaries largely rescued ovary size, but not the number of mature eggs (Fig. 4K, n=15). In addition, massive branched fusomes (Fig. 4L,M), GSC loss (Fig. 4N) and stem-cyst formation (Fig. 4N′) were significantly suppressed by the expression of WDR4 in wh mutant germ cells. Further, overexpression of WDR4 in germ cells also rescued Orb enrichment in the oocyte of wh⁷ mutant egg chambers (Fig. 4O,P). Most importantly, sterility of wh⁷ mutants showed a trend toward rescue, although the difference did not reach statistical significance (P=0.06) (Fig. 4Q and wild-type egg production in Fig. 1C). This mild rescue phenotype could be due to a requirement for Wh in ovarian somatic cells or only partial functional replacement of Wh by WDR4. Nevertheless, our results suggest that the interaction between WD40 protein, Wh, and TRIM-NHL protein, Mei-p26, is evolutionarily conserved, and the interaction may be crucial for germ cell function in other organisms.

Mei-p26 is known to control GSC self-renewal via BMP signaling (Li et al., 2012). Upon GSC receipt of ligands from niche cap cells, BMP stemness signaling is activated, causing the phosphorylation of Mad (pMad) (Song et al., 2004), which suppresses transcription of the differentiation factor bag of marbles (bam) (McKearin and Ohlstein, 1995) (Fig. 5A). Expression of Mad is negatively regulated by Brat, and translation of Brat is suppressed by a Mei-p26 interaction with Nos (Fig. 5A) (Li et al., 2012). To test whether Wh affects Mei-p26-facilitated BMP signaling, we first examined expression of pMad and Brat. In wh⁷ mutant germaria, GSCs expressed similar levels of pMad compared with controls (Fig. 5B,D); however, expression of pMad was remarkably low in GSCs of stem-cysts (Fig. 5C,D). Compared with wild-type germaria (Fig. 5E), wh⁷ mutant germaria exhibited a dramatically increased Brat signal in germ cells (Fig. 5F) and ovary extracts (Fig. S5). Interestingly, Wh interacted with Nos in the ovary (Fig. 5G), suggesting that Wh might work in the same complex with Mei-p26 and Nos to silence brat translation. We proceeded to examine bam transcription using a bamP: bam-gfp transgene, which encodes a Bam-GFP fusion protein driven by the bam promoter (Chen and McKearin, 2003). Consistent with a previous report (Chen and McKearin, 2003), Bam-GFP was absent from GSCs and expressed in differentiated dividing germ cell cysts (2-8 cell cysts) in the control germarium (Fig. 5H). On the other hand, all germ cells within the wh⁷ stem-cyst expressed Bam-GFP (Fig. 5I), suggesting that these cells may be only partially differentiated. This result was confirmed by in situ hybridization to directly detect bam RNA distribution (Fig. 5J,K), although fusomes could not be clearly labeled in RNA-hybridized germaria. Thus, the upregulation of Bam expression in wh⁷ mutant GSCs of stem-cysts indicates that BMP signaling was disrupted in these cells.

To test whether upregulated Brat and Bam account for GSC loss in wh mutants, we attempted to reduce expression of Brat in wh⁷ flies. However, we failed to obtain adult wh⁷ mutant flies bearing either a copy of the brat null or hypomorphic allele, suggesting that Wh and Brat may play a crucial role during development. We then introduced a copy of the bam null allele (bam¹) and a vasa-gfp transgene to mark germ cells in wh⁷ homozygous mutants (Kai et al., 2005). Compared with controls (Fig. 5J), flies with bam reduced by one copy did not show rescued abnormal fusome morphology in differentiated cysts and distribution in the germarium (Fig. 5L), nor did reduction of bam rescue oocyte formation (Fig. 5M). However, removal of one copy of bam from wh⁷ mutant females completely restored GSC number (Fig. 5N; Fig. 2G). In addition, the number of wh⁷ mutants with stem-cysts was also significantly reduced (Fig. 5N′). These results indicate that upregulation of Bam in wh mutant germ cells is required for formation of stem-cysts, which precedes GSC loss.

In differentiated cysts, Mei-p26 participates in the differentiation program via multiple regulatory actions (Fig. 6A). First, Mei-p26 forms a complex with Bgcn, Sxl and Bam to silence the translation of Nos (Li et al., 2013); nos deletion has been shown to cause GSC loss (Wang and Lin, 2004). Second, Mei-p26 represses ribosomal biogenesis (promoting cell proliferation; Donati et al., 2012) by suppressing expression of dMyc (also known as Myc) (McKearin, 1997), a master regulator of ribosomal biogenesis (van Riggelen et al., 2010). Third, Mei-p26 inhibits expression of a subset of microRNAs that are involved in differentiation (Neumüller et al., 2008), via an unknown mechanism. To understand the extent to which Wh works with Mei-p26 in differentiated germ cells, we examined whether Wh deficiency affects the known molecular functions of Mei-p26.

To test whether Wh promotes differentiation by the same mechanism as Mei-p26, we first examined microRNA levels in wh and mei-p26 mutant ovaries by microRNA-seq analysis. Indeed, expression of the same subset of microRNAs was increased in wh⁷ mutant ovaries (Fig. 6B) and mei-p26^mfs1/mei-p26^fs1 mutant ovaries (Fig. 6B); this subset also agreed with a previous report (Neumüller et al., 2008). We validated this result with qRT-PCR for three microRNAs, mir-1, mir-289, and mir-316 (Fig. 6C). These results suggest that Wh may interact with Mei-p26 to suppress expression of several microRNAs. We also examined Nos expression in wh mutant germaria. In the control germarium (Fig. 6D; Table S2), Bam (bamP: bam-gfp) and Nos were expressed in different cysts. However, Bam and Nos were co-expressed in wh⁷ mutant germ cells (Fig. 6E), suggesting that translational silencing of Nos by Bam is dependent on the presence of Wh. To test this hypothesis, we examined whether Wh is required for the Bam-Sxl-Bgcn-Mei-p26 complex to target the 3′ untranslated region (UTR) of nos mRNA using RNA-immunoprecipitation followed by RT-PCR. As previously reported (Li et al., 2013), we found that nos mRNA could be immunoprecipitated with Sxl antibody but not IgG in wild-type ovary lysate, but this immunoprecipitation did not occur in wh⁷ mutant ovary lysate (Fig. 6F). In addition, a pull-down assay using ovary extracts revealed that Wh weakly interacted with Bgcn (Fig. 6G); however, we failed to pull down Bam and Sxl (data not shown), possibly owing to a weak interaction between the proteins or because Wh only promotes Mei-p26 inclusion in the Bgcn-Bam-Sxl complex. This result suggests that the overlapping expression of Nos and Bam in wh⁷ mutant germ cells occurs because, in the absence of Wh, the Mei-p26-Bgcn-Bam-Sxl complex is not functional or not formed.

We also examined ribosomal biogenesis by labeling Fibrillarin to mark nucleoli (Fomproix et al., 1998). Ribosomal biogenesis takes place within nucleoli, and bigger nucleolar size represents higher ribosomal biogenesis activity (Grewal et al., 2005; Rudra and Warner, 2004). In the wild-type germarium (n=12) (Fig. 6H), the sizes of nucleoli (0.7±0.05 µm²) in GSCs and early germ cells were larger than those in differentiated cysts (0.4±0.02 µm², P<0.001), consistent with a previous report that showed ribosomal biogenesis is negatively correlated with differentiation status (Neumüller et al., 2008). In contrast, in wh⁷ mutant germaria and nos>wh^RNAi germaria (Fig. 6I,L), nucleoli in germ cells had larger or comparable sizes with those in wild-type GSCs or early germ cells, and large nucleoli were distributed throughout the germarium. RNA-seq analysis also showed that several ribosomal genes were upregulated (Fig. 6J). Interestingly, immunoprecipitation with ovary lysates showed that Wh interacted with dMyc (Fig. 6K), which is targeted by Mei-p26 for degradation in the wing disc (Ferreira et al., 2014). Removal of one dmyc copy from nos>wh^RNAi germaria significantly rescued the number of large nucleoli present in GSCs or early germ cells (Fig. 6L-N). In addition to the reduction in number of large nucleoli, massive branched fusomes and germ cell number (revealed by smaller germarial size) was also partially rescued; however, meiosis was still impaired (Fig. S6). These results suggested that Wh controls germ cell mitosis at least in part via ribosomal biogenesis. Taken together, our results show that Wh and Mei-p26 work in the same complex, and the function of Mei-p26 to maintain germline homeostasis relies on Wh.

Homeostatic regulation of stem cells (Hannezo et al., 2014; Jörg et al., 2019 preprint), the very foundation of tissue homeostasis, remains poorly understood at a molecular level. Using the Drosophila GSC lineage as an in vivo model to study stem cell biology, we found that Wh, a WD40 protein, controls GSC self-renewal and differentiation via Mei-p26, a TRIM-NHL protein (Fig. 7). An interaction between the proteins was identified in the ovary, and striking similarities between wh⁷ and mei-p26 mutants were observed at both phenotypic and molecular levels (Table 1). Based on our findings and published results regarding Mei-p26 function, we propose that, in wild-type GSCs, Wh, Mei-p26 and Nos form a complex to inhibit brat translation, allowing Mad to be stabilized and phosphorylated by BMP signaling. pMad then translocates to the nucleus, where it suppresses transcription of bam, a master regulator of differentiation. On the other hand, Wh also promotes differentiation of GSC progeny by multiple mechanisms. First, Wh, Mei-p26, Bgcn, Sxl and Bam form a complex that binds to the 3′ UTR of nos to silence its translation, possibly helping to turn off BMP signaling in the differentiating GSC progeny. Second, Wh interacts with Mei-p26, an E3 ubquitin ligase, to control dMyc protein levels and allow proper ribosomal biogenesis. Third, the interaction between Wh and Mei-p26 also limits expression of a subset of microRNA, which may contribute to differentiation. These functions of Wh appear to be especially important for the last step of cytokinesis (abscission), which is completed between the GSC and its daughter cell after early G2 phase, as revealed by closure of the ring canal.

In wh mutant GSCs, brat translation is not suppressed, decreasing the level of pMad and increasing the expression of the differentiation factor Bam. This sequence of events causes premature differentiation of GSCs and leads to GSC loss, consistent with the known roles of Brat and Bam in germ cell differentiation. Mutation of brat or bam increases GSC number, whereas overexpression of brat or bam in germ cells causes germ cell depletion by forcing GSC differentiation (Harris et al., 2011; McKearin and Ohlstein, 1995; Ohlstein and McKearin, 1997). Mutation of wh in germ cells also results in incomplete abscission between GSCs and daughter cells, leaving open ring canals that create stem-cysts. Although we did not determine the molecular mechanism by which Wh controls GSC abscission, removing a copy of bam significantly reduced stem-cyst number in wh mutant germaria, which suggests a role for Bam in Wh control of GSC abscission. In wh mutant germaria, stem-cysts simultaneously express Nos (a GSC maintenance factor) and Bam; thus, the growths display characteristics of both GSCs and daughter cells. In addition, ribosomal biogenesis is promoted via upregulated dMyc and drives germ cell overproliferation. Lastly, some differentiation-associated microRNAs are increased in the mutant ovaries, although their functions are not yet clear. We do not know whether a defect in meiosis is a consequence of overproliferative wh mutant germ cells, or whether Wh has a separate role in meiosis.

Our results show that Wh is required for Mei-p26 function in germ cell homeostasis. However, it is unclear whether Wh directly interacts with Mei-p26 and which domains in the two proteins mediate the interaction. It is possible that Wh serves as a bridge for Mei-p26 to interact with its known partners. We observed an interaction between human orthologs of Wh (WDR4) and Mei-p26 (TRIM32), suggesting that the interaction is evolutionarily conserved. Thus, the interaction between WD40 and TRIM-NHL proteins may be crucial for stem cell regulation in other organisms.

The last step in cell division, cytokinesis, is completed by abscission, which physically separates the two daughter cells. Cytokinesis starts by ingression of the cleavage furrow, constricting the plasma membrane onto the spindle midzone to form an electron-dense structure, the midbody, which comprises a thin membrane channel bridging two nascent daughter cells (Nähse et al., 2017). The stem-cyst forms owing to a failure of GSCs to separate from daughter cells. Two possible mechanisms may produce such an abscission failure. First, a stem-cell-specific defect may prevent GSC-CB abscission. Second, GSCs may exhibit characteristics of differentiating cells that cause them to adopt incomplete cell cytokinesis programs. In addition to controlling chromosome orientation and segmentation, Aurora B is known to intrinsically regulate the timing of cell abscission (Nähse et al., 2017), including in Drosophila female GSCs (Mathieu et al., 2013). During abscission, Aurora B in GSCs is targeted to the midbody and triggers membrane abscission via Endosomal sorting complex required for transport III (ESCRT-III) machinery (Eikenes et al., 2015; Matias et al., 2015; Nähse et al., 2017). Aurora B negatively controls ESCRT-III, i.e. when Aurora B is active, ESCRT-III activity is low and abscission is delayed, and vice versa. It has also been shown that ribosomal biogenesis coordinates with ESCRT-III in GSCs to promote GSC abscission (Sanchez et al., 2016). Increasing Aurora B activity or disrupting ESCRT-III generates stem-cysts with germ cells that undergo synchronous division, yielding 32 germ cells in most egg chambers (Eikenes et al., 2015; Mathieu et al., 2013; Matias et al., 2015). However, wh mutant germ cells within stem-cysts divide asynchronously with elevated ribosomal biogenesis, and wh mutant egg chambers carry various numbers of germ cells. In addition, in wh mutant stem-cysts, decreased pMad expression, upregulated Bam expression and branched fusomes are all hallmarks of differentiating cysts. These results suggest that wh mutant stem-cysts may adopt the abscission program of differentiating cysts. Interestingly, removing a copy of bam from wh mutants or from shrub (a subunit of ESCRT-III) mutants partially rescues stem-cysts (Sanchez et al., 2016), suggesting that GSC abscission is coupled with cell fate. Further studies will be required to understand the molecular targets of Wh that control GSC abscission, and how GSCs and differentiating cysts acquire different abscission programs. Nevertheless, our study has shown that Wh participates with Mei-p26 to regulate fate determination in stem cells and daughter cells. This novel interaction may be conserved in other species and introduces the idea that WD40 proteins may participate with TRIM-NHL proteins in cell fate decision.

Drosophila stocks were maintained at 22-25°C on standard medium. w¹¹¹⁸ was used as a wild-type control. Null wh⁷, bam¹ (a gift from A. Spradling, Carnegie Institution for Science, Washington, DC, USA), mei-p26^mfs1 [Bloomington Drosophila Stock Center (BDSC), BL5919], brat¹¹ (a gift from C.Y. Lee, University of Michigan, USA), dmyc⁴ (BDSC, BL64769), hypomorphic mei-p26^fs1 (a gift from P. Lasko, McGill University, Quebec, Canada) and brat^Df alleles (a gift from C.Y. Lee) have been described (Arama et al., 2000; Hsu et al., 2008; Page et al., 2000; Pierce et al., 2004). wh^CRISPR is a genetic null allele generated by the CRISPR/Cas9 system (WellGenetics, Taipei, Taiwan). p53R-gfp (nls) (a gift from J.M. Abrams, University of Texas Southwestern Medical Center, USA) is a p53 sensor containing a well-characterized conserved p53 binding site placed upstream of a nucleus-targeted GFP (Wylie et al., 2014). bamP: bam-gfp contains a bam-gfp coding sequence driven by the bam promoter (Chen and McKearin, 2003), and bgcnP: bgcn-gfp (a gift from M. Buszczak, University of Texas Southwestern Medical Center, USA) is a bgcn-gfp coding sequence driven by the bgcn promoter (Li et al., 2009). p[wh-gfp] is a genomic control bearing the wh gene with in-frame fusion of gfp to its C terminus (Wu et al., 2006). dmyc-gfp (BDSC, BL38633) was used to monitor dMyc expression, although we found that dMyc-GFP did not mimic the dMyc expression pattern, which was highly expressed in GSCs and early germ cells, but decreased in 16-cell cysts (Neumüller et al., 2008). Efficiency of the UAS-RNAi lines against wh (BDSC, BL61281) was tested (Fig. 2J,K). The nanos-GAL4-VP16 was used to express transgene in the germline (Rørth, 1998). Flies expressing RNAi were cultured throughout development at 29°C to allow for GAL4 expression. For the egg laying assay, newly eclosed females were cultured with w¹¹¹⁸ males for 1 day at 25°C, then transferred into plastic bottles capped with 35 mm molasses petri dishes supplied with a layer of wet yeast (replaced daily). Small holes were drilled in the bottom of the bottle to allow air flow. To measure egg production, five pairs of flies per bottle, which were placed upside down, were cultured for a period of 4-5 days and the number of eggs on the molasses petri dishes was counted every 24 h in triplicate. Other genetic elements are described in FlyBase (http://flybase.org).

Genetic mosaic clones were generated by FLP/FRT-mediated mitotic recombination (Xu and Rubin, 1993). Flies of genotypes hs-flp¹²²ubi-gfpFRT19A/FRT19A, hs-flp¹²²ubi-gfpFRT19A/wh⁷FRT19A and hs-flp¹²²ubi-gfpFRT19A/wh^CRISPRFRT19A were generated from standard crosses. The newly eclosed females with genotypes described above were fed on yeast diets for 1 day at 25°C, then subjected to heat shock for 1 h at 37°C twice daily at 10-12 h intervals for three days. After heat shock, flies were cultured at 25°C until dissection; food with dry yeast was changed daily. Homozygous mutant cells were identified by the absence of GFP.

A full length of human WDR4 was amplified from pRK5M-WDR4 (Wang et al., 2017) by PCR using a pair of primers carrying SpeI and BamHI sites (primer sequences are listed in Table S1). The amplified fragment was digested with SpeI and BamHI and subcloned into the pUASpattB vector to create pUASp-WDR4. Transgenic lines were generated as described previously (Spradling and Rubin, 1982).

Ovaries were dissected in Grace's insect medium (Lonza, 04-457F0) at room temperature (RT) within 15 min and fixed with 5% (w/v) paraformaldehyde (PFA; Alfa Aesar, 30525-89-4) for 13 min at RT, after which samples were washed and stained as previously described (Su et al., 2018). The following primary antibodies were used at the indicated dilutions in blocking buffer (Goal Bio, GBW-3400): mouse anti-1B1 [1:30, Developmental Studies Hybridoma Bank (DSHB), 7H9 1B1], mouse anti-LamC (1:40, DSHB, LC28.26), mouse anti-BrdU (1:20, BD Biosciences, clone B44), mouse anti-Fasciclin III (1:40, DSHB, 7G10), rabbit anti-Vasa (1:250, Santa Cruz Biotechnology, sc-30210), mouse anti-C(3)G (1:500, a gift from R.S. Hawley, Stowers Institute, USA), rabbit anti-Mei-p26 (1:500, gift from P. Lasko), rabbit anti-Cleaved caspase III (Asp175) (1/1000, Cell Signaling Technology, 9661), rabbit anti-gfp (1:1000, Torrey Pines Biolabs, TP401), rabbit anti-Histone H2AvD pS137 (1:1000, Rockland, 600-401-914), rabbit anti-Phospho-Histone H3 (Ser10) (1:250, Millipore, 06-570), rabbit anti-Brat (1:200, a gift from J. Noblich, IMBA, Austria), rabbit anti-Vasa (1:250, Santa Cruz Biotechnology, sc-30210), rabbit anti-Nos (1:1000, a gift from K.H. Nakamura, Kumamoto University, Japan), rabbit anti-Drosophila Wh (1:500, generated by T.-S. Hsieh at Duke University, USA), rabbit anti-phospho-Mad (1:200, a gift from E. Laufer, Columbia University, USA) and rabbit anti-dMyc (1:5000, a gift from D. Stein, The University of Texas at Austin, USA). Note, we did not observe expression patterns similar to those previously described, where dMyc expression is high in GSCs/early germ cells and decreased in 16-cell cysts (Neumüller et al., 2008). ApopTag^® Fluorescein In Situ Apoptosis Detection Kit (Merck, S7110) was used as previously described (Hsu and Drummond-Barbosa, 2009). BrdU incorporation was performed as previously described (Hsu et al., 2008). Ovaries were incubated with primary antibodies at 4°C overnight (O/N), washed in PBS with 0.1% Triton X-100 (PBST) 3 times, each for 20 min, and then stained with secondary antibodies, as follows: Alexa Flour 488 anti-rabbit (1:500, Invitrogen, A1008), Alexa Flour 568 anti-mouse (1:250, Invitrogen, A11077), Alexa Flour 633 anti-rabbit (1:250, Invitrogen, A21070) and Alexa Flour 633 anti-mouse (1:250, Invitrogen, A21094). Samples were incubated in 0.5 μg/ml DAPI, mounted in mounting solution [2% N-propyl gallate (Sigma-Aldrich, P-3130), 85% glycerol] and analyzed using a Zeiss LSM 700 confocal microscope. Z-stack images were acquired from the anterior part of the ovary on a Zeiss LSM 700 confocal microscope with a 40× or 60× oil immersion lens using identical acquisition conditions for each individual experiment. The image analysis was performed as experimental replicates. At least twenty flies were randomly examined per experiment and 5-7 germaria per ovary were picked to increase the frequency in the population. The total number of examined germaria is indicated for each individual experiment.

GSCs were characterized by placement of the fusome (spectrosome, labeled by 1B1) adjacent to the cap cell niche (oval-shaped cells labeled by LamC). GSCs interconnected with more than one germ cell by round/branched-shaped fusomes were identified as stem-cysts, in which GSC abscission was not complete. To quantify pMad expression, ImageJ was used to calculate the average fluorescence intensity in confocal z-stacks at the largest peripheral of GSCs or stem-cysts. To quantify the size of nucleoli labeled by Fibrillarin, ImageJ was used to calculate the area. All data were recorded in Excel, and two-tailed Student's t-test or chi-squared test were used to evaluate statistically significant differences; *P<0.05, **P<0.01, ***P<0.001.

To generate bam probes for in situ hybridization, the coding region of bam (980 bp) was amplified from the ovary cDNA (described below), using primer pairs (Table S1), and subcloned into the pGEM-T-Easy vector (Promega). Antisense RNA probes labeled with digoxigenin (DIG)-UTP (Roche) were synthesized from 1 µg of pGEM-T-bam plasmids digested with NcoI using the ampliCap™ SP6 high-yield message marker kit (Cell Script). Fluorescent RNA in situ hybridization was performed as previously reported (Tseng et al., 2018), with slight modifications. In brief, ovaries were dissected in Grace insect medium (Lonza) and fixed using 4% PFA in PBS (DEPC-treated) with freshly added 1% DMSO for 20 min at RT or at 4°C overnight. Samples were washed in PBS and dehydrated in an ethanol series (diluted in PBS: 25%, 50%, 75% and 100%) and were stored at −20°C for least one night. Samples were then rehydrated through an ethanol series (100%, 75%, 50% and 25%), rinsed with PBS, treated with 50 μg/ml Proteinase K (in PBS; Sigma-Aldrich, 003115836001) for 5 min at RT. After post-fixing with 4% PFA in PBST (1× PBS with 0.1% Tween 20) for 30 min at RT, samples were washed in 1 ml PBT (0.1% Tween 20, 0.1% DEPC) twice for 15 min, and three times for 5 min and prehybridized in hybridization buffer (HYB⁺) (50% formamide, 5× SSC, 50 μg/ml heparin, 0.1% Tween 20, 100 μg/ml tRNA, 10 μg/ml salmon sperm DNA) for 1 h at 60°C. Hybridization was performed in HYB⁺ containing denatured DIG-labeled RNA probes (100-150 ng) at 60°C O/N. Samples were washed with a series of HYB⁻ (50% formamide, 5× SSC with 0.1% Tween 20) solutions in 2× SSC (0.3 M NaCl, 30 mM sodium citrate; 75%, 50% and 25%) at 65°C and a series of 0.2× SSC solutions (75%, 50% and 25%) at 68°C, followed by rinsing with PBST at RT. Samples were treated with 3% H₂O₂/PBT for 1 h at RT to inactivate endogenous perioxidase and then blocked in 2× Roche Blocking solution for 1 h at RT. Ovaries were incubated with anti-Dig-POD (1:500, Roche, 11207733910) in blocking buffer at 4°C overnight, washed well, and incubated in 1:200 TSA/amplification buffer (TSA Plus Fluorescence Kits, PerkinElmer) for 30 min to develop signals. After washing, ovaries were blocked with blocking solution (GOAL Bio) and then subjected to the immunostaining procedure.

Recombinant Wh protein with a hexahistidine tag at the C terminus was produced and purified from Rosetta™(DE3) pLysS Competent Cells (Novagen, 70956-4CN) carrying the wh-expressing ET23b vector (Wu et al., 2006). In brief, Wh-His expression was induced in Rosetta cells by adding 1 mM IPTG for 4 h at 37°C. The cells were harvested and re-suspended in lysis buffer [100 mM NaCl, 10 mM Tris/HCl, 5 mM β-ME, 10% glycerol, protease inhibitor mixture (cOmplete Tablet, Mini EDTA-free, EASYpack, Roche)]. After sonication, the supernatant from the cell lysate was applied to equilibrated Ni Sepharose™ 6 Fast Flow beads (GE Healthcare). Wh-His protein-bound Ni beads were first washed with Tris buffer [300 mM NaCl, 10 mM Tris/HCl, 8 M urea (pH 5.9)] to remove low-affinity bound proteins. Next, Wh-His was eluted from the beads with a lower pH Tris buffer [300 mM NaCl, 10 mM Tris/HCl, 8 M urea (pH 4.5)]. The purified protein was used as an antigen to immunize rabbit (Abnova, Taiwan). As shown in Fig. 1E, Wh protein is absent in wh mutant nucleus and cytoplasmic fractions, demonstrating the specificity of the Wh antibody.

Ten or 40 pairs of ovaries from full grown or newly eclosed flies were homogenized in 100 μl EBI buffer [10 mM HEPES (pH 7.5), 1.5 mM MgCl₂, 10 mM KCl] plus 1% NP40 and protease inhibitor mixture (cOmplete™, Sigma-Aldrich). The homogenate was vortexed vigorously on ice for 5 min and centrifuged at 1700 g for 3 min at 4°C. The precipitant (nuclear pellet) was washed with 1 ml EBI plus 1% NP40 and protease inhibitors (cOmplete™, Sigma-Aldrich) by pipetting, and centrifuged at 1700 g for 3 min at 4°C again before storing on ice. The supernatant (cytoplasmic fraction, 110-120 µl) was collected and clarified by centrifugation at 18,000 g for 10 min at 4°C. Most (100 µl) of the clarified supernatant was added to 20 µl 6× Laemmli Sample Buffer (Bio-Rad), and the nuclear pellet was added to 120 µl of 1× Laemmli Sample Buffer. Samples were boiled for 6 min at 95°C. Both fractions were loaded and separated by 10% or 13% SDS-PAGE, blotted onto PVDF membranes and probed with primary and secondary antibodies, as follows: anti-histone H3 (1:2000; DSHB, ab1791; nuclear marker), anti-beta-tubulin (1:1000, DSHB, clone EP1332Y; cytoplasmic marker), rabbit anti-Wh antibody (1:2000, see above), horseradish peroxidase (HRP)-conjugated goat anti-rabbit IgG (1:5000, Jackson ImmunoResearch, 111-035-003) and HRP-conjugated goat anti-mouse IgG (1:5000, Millipore, AP124P). The blot was developed using a chemiluminescent reaction (Amersham ECL) following the manufacturer's instructions.

Dissected ovaries were frozen by immersion in liquid nitrogen and stored at −80°C until use. Total RNA was extracted using a standard Trizol (Ambion) extraction protocol (Su et al., 2018). In brief, 10-20 pairs of frozen ovaries were homogenized in 500 µl of Trizol, followed by a phase separation step with 500 µl 1-bromo-3-chloropropane (Sigma-Aldrich). RNA was precipitated with 500 µl isopropanol, washed with 75% ethanol, and resuspended in 30 µl RNase-free water. Genomic contamination was removed using TURBO™ DNase (Invitrogen, AM2238), according to the manufacturer's instructions for complete DNA digestion. A phenol/chloroform extraction was performed to inactivate TURBO™ DNase. MicroRNA levels were quantified using the TaqMan^® MicroRNA Assay kit (Applied Biosystems, 4440885), following the manufacturer's instructions. In brief, 10 ng RNA were reversed transcribed using master mix to generate specific cDNA for each targeted microRNA, with primer sets designed from Applied Biosystems. snoRNA227 was used as a control reaction. The cDNAs were then mixed with a small RNA assay buffer containing designed primers/probes specific for each cDNA sample (Applied Biosystems) and TaqMan^® Universal PCR Master Mix II (Applied Biosystems) for PCR reaction. The TaqMan™ microRNA assay miRBase ID and assay ID used in this study are listed in Table S2.

Frozen adult ovaries (400-600 pairs) were lysed in 300-350 µl of EB1 buffer [10 mM Hepes (pH 7.5), 10 mM KCl, 0.1% Nonidet P40, 1.5 mM MgCl₂, 0.2% Triton X-100] plus protease inhibitor mixture (cOmplete™, Sigma-Aldrich). After centrifugation (16,100 g, 10 min at 4°C), an equal volume of EB2 buffer [20 mM Hepes (pH 7.5), 420 mM NaCl, 1.5 mM MgCl₂, 0.2 EDTA, 25% glycerol] was added to the supernatant; 10% of the supernatant volume was stored at −20°C as the input control. Next, 15 µl of GFP-Trap^®_MA beads (ChromoTek) were equilibrated with 1 ml binding buffer [EB1]:[EB2]=1:1 three times and pull-down was performed using a MagJET Separation Rack (Thermo Fisher Scientific). The supernatants were incubated with 25-30 µl binding buffer containing equilibrated beads with gentle rotation at 4°C overnight. The beads were pulled down and washed three times, each for 1 min at 4°C, with binding buffer (freshly prepared) plus protease inhibitor mixture (cOmplete™, Sigma-Aldrich). Then 40 µl of 1× Laemmli Sample Buffer (Bio-Rad) was added to beads, and 20 µl of 2× Laemmli Sample Buffer was added to input. Both samples were boiled for 5 min, loaded and separated by SDS-PAGE, and analyzed using western blotting. Immunoblots were probed with rabbit anti-Wh (1:2000, see above), rabbit anti-Mei-p26 (1:200, a gift from P. Lasko), mouse anti-dMyc (1:20, a gift from P. Gallant, University of Würzburg, Germany) and rabbit anti-GFP (1:10,000, GenLab Biotech). Secondary antibodies included horseradish peroxidase (HRP)-conjugated goat anti-rabbit IgG (1:5000, Jackson ImmunoResearch, 111-035-003) and HRP-conjugated goat anti-mouse IgG (1:5000, Millipore, AP124P).

We transfected 1×10⁶ 293T cells with 15 µg empty vector (CMV-TRIM32-FLAG) or CMV-TRIM32-FLAG Plus CMV-WDR4-dmyc with the calcium phosphate method (the 293T cell line has recently been authenticated and has been routinely tested for free of mycoplasma contamination). Co-immunoprecipitation (co-IP) was carried out 24 h after plasmid transfection. Cells were harvested and washed with 5 ml ice-cold PBS. Then, 500 µl cold lysis buffer (0.75% NP40, 1 mM DTT in PBS) containing 1 mM phenylmethylsulphonyl fluoride (PMSF), 1 µg/ml aprotinin and 1 µg/ml leupeptin was added, and the cells were removed using a cell scraper. The mixtures were then collected and centrifuged at 13,000 rpm (15,871 g) for 15 min at 4°C. We supplemented 2 mg of total protein with co-IP lysis buffer to make a total 900 µl volume, and 100 μl M2 magnetic beads (Sigma-Aldrich) was added to the sample and mixed at 4°C for 2 h. The beads were collected by centrifugation at 5000 rpm (2348 g) for 1 min at 4°C. The protein-bead complexes were washed with 800 µl of co-IP lysis buffer five times. The protein-bead complexes were then denatured by heating at 95°C for 5-10 min in the presence of 1× loading buffer, which consisted of 6× loading buffer with co-IP lysis buffer. Equal amounts of samples were loaded into 10% Tris-glycine PAGE followed by SDS electrophoresis and immunoblotting.

IP was used in combination with RT-PCR as previously described (Li et al., 2012), with some modifications. In brief, 60-90 pairs of 7-day-old wild-type and wh⁷ ovaries were extracted in 300 µl lysis buffer [50 mM Tris-HCl (pH 8.0), 150 mM NaCl, 0.1% Nonidet P40, 5 mM EDTA, 5 mM DTT]. Protease inhibitor mixture (cOmplete™, Sigma-Aldrich) and RNase inhibitor (40 U/μl; Roche) were added to the lysate. Supernatant was obtained by centrifugation (16,100 g, 10 min at 4°C). Mouse anti-IgG Ab (1:250, Santa Cruz Biotechnology, sc-2025) or mouse anti-Sxl Ab (1:35, DSHB, M1114) was added and the samples were incubated at 4°C for 3 h. Then, the ovary lysate with antibody was incubated with pre-equilibrated Protein G Sepharose Fast Flow beads (GE Healthcare) at 4°C overnight. Beads were pelleted by centrifugation at 12,000 g for 1 min and washed briefly with 1 ml of fresh lysis buffer. Genomic DNA was removed using TURBO™ DNAase, according to the manufacturer's manual. RNA was extracted from beads by Trizol, and cDNA was made as described above. cDNA was synthesized from total RNA (100 ng) from whole ovaries and used as the template. Thirty cycles of PCR using Phusion^® High-Fidelity DNA Polymerase (New England Biolabs) were performed on 1 μl of cDNA to amplify nos 3′ UTR using the primers listed in Table S1; the resulting products were analyzed on 1% agarose gels.

For RNA sequencing analysis, thirty pairs of ovaries were collected and dissected from 7-day-old nos>gfp^RNAi, nos>wh^RNAi, w¹¹¹⁸ and wh⁷ mutant flies. nos>gfp^RNAi and nos>wh^RNAi flies were cultured at 29°C throughout development. The analysis was performed in duplicate. Total RNA was extracted and measured by absorbance at 260 nm using an ND-1000 spectrophotometer (Nanodrop Technology). RNA quality was evaluated using a Bioanalyzer 2100 (Agilent Technology) with an RNA 6000 labchip kit (Agilent Technology). All values, OD_260/280 and OD_260/230 were above 2.15. All RNA-seq procedures were performed in accordance with the manufacturer's instructions (Illumina). The library was constructed with Agilent's SureSelect Strand Specific RNA library Preparation Kit for 75SE (Paired-End) sequencing on a Solexa platform. Sequencing-by-synthesis technology with a TruSeq SBS kit was used for the sequencing. Raw sequences were obtained from the Illumina Pipeline software bcl2fastq v2.0 and expected to generate 12.5 million reads (M) per sample. Trimmomatic software was implemented to trim or filter the reads according to the quality score. The gene expression level was calculated as fragment per kb of transcript per million mapped reads (FPKM). CummeRbund was used to perform statistical analysis of gene expression profiles for differential expression analysis. The reference gene annotations were retrieved from Flybase. For microRNA sequencing analysis, thirty pairs of ovaries were collected and dissected from 7-day-old, w¹¹¹⁸, mei-p26 and wh⁷ mutant flies and cultured at 29°C throughout developmental stages. The analysis was performed in duplicate. RNA quantity and quality were evaluated as mentioned above. All values for OD_260/280 and OD_260/230 were above 2.1. Sample libraries were prepared using the QIAseq miRNA Library kit (Qiagen) according to the manufacturer's protocol. Adaptors were ligated sequentially to the 3′ and 5′ ends of miRNAs. Subsequently, universal cDNA synthesis with UMI assignment, cDNA cleanup, library amplification and library cleanup were performed. Libraries were sequenced on an Illumina instrument (75-cycle single-end read, 75SE) and expected to generate 10M per sample. Sequencing data was processed using the Illumina software BCL2FASTQ V2.20 for small RNA sequencing analysis. Initially, the generated sequences went through a filtering process to obtain qualified reads. Trimmomatic was implemented to clip the 3′ adaptor sequence, trim or remove the reads according to the quality score and discard trimmed reads in the range of (22-35) nucleotides. Qualified reads after filtering low-quality data were analyzed using miRDeep2 software for aligning reads to the mirBase 21 database. Notably, the ovaries contained a highly abundant 30 nt rRNA fragment (the 2S rRNA), which masked the microRNA and limited analysis to only 2% of total microRNAs, similar to a previous report (Fowler et al., 2018). Gene expression level was calculated as FPKM. Heatmaps were generated by the color-scale method using Excel.

We would like to thank A. Spradling, P. Lasko, J.M. Abram, M. Buszczak, H.W. Pi, C.Y. Lee, R.S. Hawley, E. Laufer, J. Knoblich and K.H. Nakamura, for providing flies and antibodies. We also thank the Bloomington Drosophila Stock Center (National Institute of Health, P40OD018537), Fly Stocks of National Institute of Genetics, the Vienna Drosophila RNAi Center and the Developmental Studies Hybridoma Bank for Drosophila stocks and antibodies. We also thank J.-R. Huynh for the valuable discussion and M. Calkins for English editing.