Main

D. melanogaster ovarian GSCs are an attractive model for investigating stem-cell–niche interactions at the molecular and cellular level3. In the germarium of the D. melanogaster ovary, two or three GSCs can be easily identified by their direct contact with cap cells and a spherical fusome (also known as a spectrosome), which is apically anchored6,7,8. Bam controls GSC daughter differentiation by forming a protein complex with Bgcn9,10,11,12. Niche-activated Bmp signalling is necessary and sufficient for repressing bam gene transcription in GSCs, thereby maintaining GSC self-renewal7,13,14. Immediate differentiating GSC daughters (also known as cystoblasts) also contain a spherical spectrosome and begin bam transcription6,15. These cells can further divide without cytokinesis to form 2-cell, 4-cell, 8-cell and 16-cell cysts, which harbour a branched fusome16. Although Bam is a master regulator of GSC differentiation10, it remains largely unclear whether and how Bam inactivates self-renewal factors in differentiating GSC progeny. In this study, we show that Bam converts the function of the COP9 complex from self-renewal to differentiation by sequestering Csn4.

In the yeast two-hybrid screen described in our previous study11, the carboxy-terminal 121 amino acid region of Csn4 was identified to interact with Bam (Extended Data Fig. 1a). In yeast (Saccharomyces cerevisiae) and D. melanogaster S2 cells, Bam also interacts with full-length Csn4 (Extended Data Fig. 1b–d). Using yeast two-hybrid interaction experiments, we found that the 151–350 amino acid central domain of Bam interacts with Csn4 (Extended Data Fig. 1b, c). Additionally, nos-GAL4-driven germline-specific expression of haemagglutinin (HA)-tagged Csn4 pulled down endogenous Bam in vivo, but germline-expressed Flag-tagged Csn5 failed to pull down Bam (Fig. 1a). Finally, we generated P[acman]-based bacterial artificial chromosome (BAC) transgenes17, in which a Flag tag was added to the coding region of Csn4 and Csn5 (Extended Data Fig. 2), and we showed that the Csn4 and Csn5 proteins exhibit similar ubiquitous expression patterns in germ cells and are co-expressed with Bam in mitotic cysts (Fig. 1b–d). These results demonstrate that Bam interacts with Csn4 to form a protein complex in vivo.

Figure 1: Csn4 physically interacts with Bam and antagonizes its function during GSC development.
figure 1

a, HA–Csn4, but not Flag–Csn5, pulled down endogenous Bam in vivo, as determined by co-immunoprecipitation. IB, immunoblot; IP, immunoprecipitation. b–d, Csn4–Flag (c) and Flag–Csn5 (d) transgenes showed ubiquitous expression in GSCs (solid circles) and in their progeny and cap cells (dashed ovals) compared with the control (b). Csn4 and Csn5 were co-expressed with Bam in mitotic cysts (arrows). Arrowheads indicate non-specific staining. e–h, Heterozygous Csn4n (f) and Csn4k (g) mutations suppressed the differentiation defect of the hypomorphic bam mutant (e) based on spectrosomes (arrowheads) and branched fusomes (arrows). h, Quantification of germaria with GSC tumour phenotypes in severe (e), moderate (f) and weak (g) phenotypes. Scale bars, 10 μm.

PowerPoint slide

Full size image

In our genetic screen, Csn4 was identified as a dominant suppressor of the germ cell differentiation defect in the hypomorphic bamZ/bamΔ86 mutant. Immunostaining for Hts labels spectrosomes and fusomes, allowing the identification of GSCs, cystoblasts and differentiated cysts16. In the bamZ/bamΔ86 mutant ovaries, the germaria and most of the egg chambers were filled with undifferentiated cystoblasts (Fig. 1e). By contrast, inactivation of one copy of the Csn4 gene using a deletion allele (Csn4n) or a P element insertion allele (Csn4k)18 dramatically reduced the number of undifferentiated cystoblasts and dramatically increased the number of differentiated cysts containing a branched fusome in the bamZ/bamΔ86 germaria and the number of normal egg chambers per ovariole (Fig. 1f–h and Extended Data Fig. 3a, b, b′). These results suggest that Csn4 and Bam have opposite functions in regulating germ cell differentiation.

To determine whether Csn4, Csn5 and Nedd8 are intrinsically required to maintain GSCs, we used FLP-mediated FRT recombination to inactivate the functions of Csn4, Csn5 and Nedd8 in GSCs by constructing control and mutant Csn4, Csn5 and Nedd8 GSC clones marked by the absence of Ubi-GFP (green fluorescent protein (GFP) controlled by the Ubiquitin promoter) or armadillo-lacZ as in our previous studies7,19. Csn5n and Nedd8AN015 represent a deletion allele and a hypomorphic allele, respectively18,20,21. In contrast to the marked control GSCs, which exhibited a slow natural turnover during the first 3-week period after clone induction (ACI) (Fig. 2a, b), the marked mutant Csn4k, Csn4n, Csn5n and Nedd8AN015 GSCs were lost much faster, and most of them were therefore lost from the niche at 3 weeks ACI (Fig. 2b, c). This finding is consistent with the previous finding that Nedd8 and Csn mutants exhibit similar phenotypes in D. melanogaster and plants22,23. Then, we used the combined Gal4–UAS and FLP-Out system to knock down the functions of the other Csn genes in adult GSCs by RNA interference and then examined the GSC numbers in germline-specific knock-down (GSKD) germaria 3 days (3 d), 7 d and 14 d after heat shock (AHS)24 (Extended Data Fig. 4a). In contrast to control Csn1aGSKD (a pseudo gene) and GFPGSKD germaria, which maintained two or three GSCs 3 d, 7 d and 14 d AHS, Csn1bGSKD, Csn2GSKD, Csn3GSKD, Csn6GSKD and Csn7GSKD germaria showed a significant GSC loss, and most of them contained only one GSC on average at 7 d and 14 d AHS (Fig. 2d, e and Extended Data Fig. 4b–g). Noticeably, Csn8 knock down yielded no GSC loss phenotype, which might have been due to a low knock-down efficiency (Fig. 2e and Extended Data Fig. 4g). Taken together, these results indicate that the whole COP9 complex is likely to be intrinsically required for GSC maintenance.

Figure 2: COP9 is an intrinsic controller of GSC self-renewal and proliferation.
figure 2

a–c, In contrast to the marked control GSCs that are maintained at 3 weeks (3 w) ACI (a), the marked Csn4n mutant GSC (c) detected 1 w ACI (top) has been lost by 3 w ACI (bottom). The solid circles highlight unmarked GSCs (GFP+), whereas the dashed circles indicate marked GSCs (GFP−) or mutant GSCs. The arrowheads indicate marked cysts (a). The changes in the percentage of germaria carrying a marked control, Csn4, Csn5 or Nedd8 mutant GSC at 1, 2 or 3 w ACI are shown (b). d, Csn3GSKD germaria contained two GSCs and one GSC at 3 days (3 d) and 14 d AHS, respectively. The arrowheads indicate differentiated germ cells. e, Csn1b, Csn2, Csn3, Csn6 and Csn7 (but not Csn8, GFP or Csn1a) knock-down germaria contained significantly fewer GSCs at 7 d and 14 d AHS than at 3 d. *, P < 0.05; ***, P < 0.001. Scale bars, 10 μm.

PowerPoint slide

Full size image

TUNEL (terminal deoxynucleotidyl transferase dUTP nick end labelling) results showed that marked mutant Csn4k, Csn4n and Csn5n GSCs and cysts exhibited no dramatic increase in apoptosis in comparison with marked control GSCs and cysts (Extended Data Fig. 5a–d). Interestingly, marked mutant Csn4n or Csn5n GSCs also produced fewer differentiated cysts than did marked control GSCs (Fig. 2a, c). Additionally, fewer Csn4 and Csn5 mutant GSCs than control GSCs were positive for 5-bromodeoxyuridine (BrdU) labelling and phosphorylated histone 3 (p-H3, a mitotic marker) (Extended Data Fig. 5e–h). Although COP9 regulates different signalling pathways in organisms ranging from plants to animals4,5, Csn4 and Csn5 are dispensable for Bmp signalling and E-cadherin-mediated cell adhesion in GSCs (Extended Data Fig. 6). These results suggest that Csn4 and Csn5 are intrinsically required to control GSC self-renewal and proliferation independently of Bmp signalling and E-cadherin.

To understand how Csn4 mutations suppress the bam differentiation defect, we examined the protein interactions among Bam, Bgcn and Csn4. In yeast cells, Bgcn binds to the same central Bam domain (amino acids 151–350) as Csn4 does but has a weaker interaction with Bam (Extended Data Fig. 7a–d, d′). The presence of increasing concentrations of Csn4 gradually weakened the Bam–Bgcn interaction, whereas a truncated Csn4 protein lacking the Bam-interacting domain was unable to compete with Bgcn for Bam binding (Fig. 3a and Extended Data Fig. 7e). These results support the Csn4–Bam–Bgcn competition model, which predicts that Csn4 and Bam antagonize each other’s function in the regulation of germ cell differentiation.

Figure 3: Csn4 inhibits cystoblast differentiation by competing with Bgcn for binding to Bam.
figure 3

a, Csn4 outcompetes Bgcn for binding to Bam in a concentration-dependent manner in S2 cells, as determined by co-immunoprecipitation. Three independent experiments are represented in the histogram on the right, and P values are indicated. The numbers above the blot on the left indicate the amounts of DNA constructs in micrograms. b–d, The nos-GAL4 bamΔ86/UAS-Csn4 germarium (c, bottom) had more cystoblasts (CBs) and two-cell pairs (arrowheads) than the nos-GAL4/+ (b, top), nos-GAL4;UAS-Csn4 (b, bottom) and nos-GAL4 bamΔ86/+ (c) germaria. GSCs are highlighted by ovals. Quantification and P values are shown in d. Roman numerals on the x axis indicate sample sizes. e, The nos-GAL4 hs-bam/+ germarium had lost all GSCs by 1 week AHS (top), but the nos>>HA-Csn4 hs-bam (UASp-HA–Csn4/+;nos-GAL4 hs-bam/+) germarium retained two GSCs (circles) 1 week AHS (bottom). Scale bars, 10 μm.

PowerPoint slide

Full size image

The numbers of cystoblasts and two-cell pairs (dividing cystoblasts and two-cell cysts) were then used to quantify germ cell differentiation defects because it is difficult to distinguish two-cell pairs that are dividing symmetrically to generate cystoblasts from normal two-cell cysts (Extended Data Fig. 7f). Interestingly, germ-cell-specific Csn4 overexpression in the bamΔ86 heterozygous germaria but not in wild-type germaria significantly increased the number of cystoblasts and two-cell pairs compared with that in bamΔ86 heterozygous germaria, but the germaria of both genotypes showed no significant differences in GSC numbers, indicating that excess Bam in differentiating GSC progeny suppresses the antagonizing function of Csn4 and thereby ensures their differentiation (Fig. 3b–d and Extended Data Fig. 7g). Ectopic Bam expression in GSCs via hs-bam induced the rapid differentiation of GSCs and their departure from the niche10. Indeed, a 1-h Bam induction induced 83–95% of the germaria to completely lose GSCs (Fig. 3e and Extended Data Fig. 7h). However, germline-specific Csn4 overexpression allowed 65% of the germaria to retain GSCs, indicating that Csn4 can also antagonize the differentiation-promoting function of ectopic Bam in GSCs (Fig. 3e, bottom, and Extended Data Fig. 7h). Taken together, these results suggest that differentiating GSC progeny express an excess of Bam, which neutralizes the self-renewal function of Csn4 and allows the formation of Bam–Bgcn protein complexes.

Although Csn4 mutant GSC progeny develop normally into germ cell cysts, some of the Csn5 mutant or knock-down GSC progeny failed to differentiate and accumulate as cystoblasts (Fig. 4a–c). In contrast to control germaria, which contained one or two cystoblasts, some of the Csn1bGSKD, Csn2GSKD, Csn3GSKD, Csn7GSKD and Nedd8GSKD germaria contained extra cystoblasts, exhibiting germ cell differentiation defects (Fig. 4c and Extended Data Fig. 8a–g). Consistent with this finding, the mutations in Csn3 (Csn3FS), Csn5 (csn5n) and Csn7 (csn7e and csn7MB) drastically and significantly enhanced the differentiation defect of the bam heterozygous mutant based on the number of cystoblasts and two-cell pairs (Fig. 4d and Extended Data Fig. 8h, i). Additionally, the Csn4 and bam transheterozygous ovaries had significantly fewer cystoblasts and two-cell pairs than did single heterozygous ovaries, further suggesting that Csn4 can antagonize Bam function in differentiating GSC progeny (Extended Data Fig. 8k–m). We ruled out the possibility that the differentiation defects are caused by an excess of GSCs (Extended Data Fig. 8j, m′). Finally, mutations in Csn5 and Nedd8 also drastically enhanced the differentiation defect of bamZ/bamΔ86 (Extended Data Fig. 3c, d). Together, these results demonstrate that Csn proteins, except for Csn4, promote cystoblast differentiation.

Figure 4: Csn proteins promote cystoblast differentiation in the absence of Csn4.
figure 4

a, b, Marked Csn4 mutant GSC progeny developed into 16-cell cysts (arrows, a), whereas marked Csn5 mutant (b, top) or Csn5 knock-down (b, bottom) GSC progeny remained as cystoblasts (arrowheads). Scale bar, 10 μm. c, Germline-specific knock down of Csn1b, Csn2, Csn3, Csn5 and Csn7 significantly increased the number of germaria carrying four or more cystoblasts (CBs) in comparison with GFP and Csn1a knock-down negative controls. d, A heterozygous Csn5n mutation significantly enhanced the differentiation defect of the bamΔ86/+ mutant. e, Bam is associated only with Csn4, but not Csn5, when co-expressed in S2 cells, as determined by co-immunoprecipitation. The asterisk indicates a non-specific band. f, The presence of Bam protein significantly decreased the ability of Csn4 to co-immunoprecipitate with Csn5, Csn6 and Csn7 in S2 cells. g, A working model illustrating how Bam converts the function of the COP9 complex from self-renewal to differentiation by sequestering Csn4 via protein competition.

PowerPoint slide

Full size image

One possible explanation for the opposite effects of Csn4 and other Csn proteins on cystoblast differentiation is that the sequestration of Csn4 by Bam allows other Csn proteins to have differentiation-promoting functions. The PCI (proteasome, COP9, initiation factor) domain in Csn4 is crucial for its assembly into COP9 through interaction with Csn5, Csn6 and Csn7 (refs 4, 25). Interestingly, Flag–Bam precipitated only Myc–Csn4 and not Myc–Csn5 and Myc–Csn6 in S2 cells, even in the presence of overexpressed Flag–Csn4, suggesting that Bam is associated only with Csn4 but not with the entire COP9 complex (Fig. 4e and Extended Data Fig. 9a). Additionally, the presence of Bam significantly interfered with the ability of Csn4 to interact with Csn5, Csn6 and Csn7 but did not affect the ability of Csn6 to interact with Csn5 and Csn7 (Fig. 4f and Extended Data Fig. 9b–d). These results suggest that the sequestration of Csn4 by excess Bam in differentiating GSC progeny inactivates the self-renewal function of COP9, as well as allowing other Csn proteins to carry out differentiation-promoting functions.

This study has provided important insight into how protein competition controls the balance between GSC self-renewal and differentiation in the D. melanogaster ovary (Fig. 4g). In GSCs, Csn4 works within the COP9 complex to maintain GSC self-renewal. In differentiating GSC progeny, upregulated Bam proteins sequester Csn4 from the COP9 complex via protein competition, allowing other Csn proteins to promote germ cell differentiation, as well as inactivating COP9 self-renewal function. In addition, excess Bam can also form protein complexes with Bgcn to promote cystoblast differentiation by repressing self-renewal factors. Interestingly, Csn proteins have been identified to be required for maintaining human embryonic stem cells, as shown in a genome-wide RNA interference screen26. Because many intrinsic self-renewal factors are expressed in both stem cells and in their differentiated progeny, protein competition could also be employed as a common mechanism for balancing stem cell self-renewal and differentiation in various stem cell systems.

Methods

D. melanogaster stocks

The following D. melanogaster stocks used in this study are described in FlyBase, unless specified: bamZ3-2884 (bamZ), bamΔ86, bam–GFP, Csn4k08018 (Csn4k), Csn4EY08080 (Csn4EY), Csn7MB01896 (Csn7MB), Csn7e02176 (Csn7e), Csn3FS, nos-GAL4, UASp-Csn4, FRT42D, FRT82B, armadillo-lacZ, Ubi-GFP, Csn4null (Csn4n), Csn5null (csn5n), hs-FLP, nos>STOP>GAL4 (provided by Y. Yamashita). The UASp-short hairpin RNA transgenic lines against GFP, Csn1a, Csn2, Csn3, Csn5, Csn6, Csn7 and Csn8 were inserted at the attP2 site on the third chromosome, as described previously24. All stocks were cultured at 25 °C on standard cornmeal/molasses/agar medium unless specified.

Genetic screen for modifiers of the bam hypomorphic mutant combination

The sensitized bam genetic background bamZ/bamΔ86 was used to test whether mutations in any of the ‘stemness’ genes identified in mammalian systems modify the differentiation defect of the bam mutant. In addition, mutations in the genes encoding various kinases and signalling molecules were also tested. In this screen, 11 enhancers were identified based on their ability to reduce the fecundity of bam mutant females, but Csn4 was the only suppressor based on improvement in the low fecundity of bam mutant females.

Genetic clonal analysis

The marked control, Csn4 and Csn5 mutant GSC clones were generated using the FLP-mediated FRT recombination technique, as described previously7. The following genotypes were used for clonal analysis: (1) hs-flp/+; FRT42D/FRT42D Ubi-GFP; (2) hs-flp/+; FRT42D Csn4n/FRT42D Ubi-GFP; (3) hs-flp/+; FRT42D Csn4k/FRT42D Ubi-GFP; (4) hs-flp/+;; FRT82B/FRT82B armadillo-lacZ; (5) hs-flp/+;; FRT82B Csn5n/FRT82B armadillo-lacZ; and (6) hs-flp/+; FRT42D Csn4n/FRT42D armadillo-lacZ; bam-GFP/+.

Germline-specific RNA-interference-mediated knock down (GSKD) of Csn genes

To knock down the functions of Csn genes in adult GSCs and their progeny, hs-flp; nos>STOP>GAL4 UAS-GFP/CyO virgins were crossed with males of UASp-RNAi lines against Csn genes and GFP. The female progeny were collected and heat-shocked in a 37 °C water bath (two 1-h treatments with an 8-h interval). Ovaries were isolated 3, 7 and 14 d AHS or 1, 7 and 13 d AHS for examining GSCs and cystoblasts after being labelled for Hts. For random sampling of flies, we anaesthetized the flies with carbon dioxide, mixed them and randomly picked enough females for ovary dissection. After the stained ovaries had been mounted on slides, we randomly selected and examined the germaria for GSCs and cystoblasts. All of the statistical analyses in this study were carried out by two-sided Student’s t-test.

Generation of Flag-tagged Csn4 and Csn5 transgenes using the P[acman] system

To generate Flag-tagged Csn4 and Csn5 transgenes, the P[acman] BAC clones CH322-141C03 (Csn4) and CH322-113L01 (Csn5) were chosen because they contain the nearby genes on both the 5′ and 3′ sides and thus should contain all the regulatory sequences (Extended Data Fig. 2). The Flag-tagged versions of Csn4 and Csn5, Csn4–Flag (C-terminal tag), Flag–Csn5 (N-terminal tag) and Csn5–Flag (C-terminal tag), were generated by following previously reported experimental procedures27. Confirmed plasmids were further used to make transgenic fly strains.

BrdU labelling of GSCs

At 1 week ACI, the freshly isolated D. melanogaster ovaries were incubated with 75 μg ml−1 BrdU in Grace’s Insect Medium for 1 h at room temperature. Then, the ovaries were fixed and processed for immunostaining with anti-BrdU and anti-LacZ or anti-GFP antibodies for the detection of clones according to our published procedures7.

Immunohistochemistry

The following antisera were used: monoclonal mouse anti-Hts antibody 1B1 (1:3; DSHB); rabbit polyclonal anti-β-galactosidase antibody (1:300; Cappel); monoclonal mouse anti-β-galactosidase antibody (1:100; Promega); rabbit polyclonal anti-p-Smad3 (1:200; Epitomics); mouse monoclonal anti-BrdU antibody (1:20; Sigma-Aldrich); rat monoclonal anti-E-cadherin antibody (1:4; DSHB); rabbit polyclonal anti-GFP antibody (1:200; Molecular Probes); and rat anti-Vasa antibody (1:3; DSHB). Secondary antibodies including goat anti-rabbit, anti-mouse or anti-rat IgG conjugated to Alexa Fluor 488 or 568 (Molecular Probes) were used at 1:200. All micrographs were taken with Leica confocal microscopes.

Plasmid construction

Invitrogen Gateway technology was used to construct the Myc- and HA-tagged Csn4, Myc-tagged Csn5, Myc-tagged Bgcn and Flag-tagged Bam plasmids for expression in S2 cells and for making transgenic flies. The coding regions or various truncations of bam, bgcn, Csn4 and Csn5, which were amplified from a D. melanogaster ovarian cDNA library, were cloned into the pENTR TOPO cloning vector and completely sequenced. These pENTR vectors were subsequently recombined with Flag-, Myc- or HA-tagged destination vectors (from T. Murphy) by using LR Clonase (Invitrogen).

Co-immunoprecipitation and western blotting

D. melanogaster S2 cells (Invitrogen) were cultured at 25 °C in Schneider’s Drosophila Medium according to the manufacturer’s manual. S2 cells were transfected with the indicated amounts of plasmids using Cellfectin (Invitrogen) according to the manufacturer’s manual. The S2 cells were harvested 36–48 h after transfection for immunoprecipitation and western blot analysis. For immunoprecipitation, 4 × 106 cells were lysed with 800 μl ice-cold lysis buffer (20 mM Tris.HCl, pH 7.5, 100 mM NaCl, 1% Nonidet P-40, 1 mM EDTA, 10% glycerol and a mixture of protease inhibitors). For immunoprecipitation experiments, anti-Flag monoclonal antibody M2 agarose (Sigma) was preincubated with 5% BSA at 4 °C for 1 h, and then 50 μl 50% agarose slurry was used in each experiment. For immunoprecipitation with monoclonal anti-Myc and anti-HA antibodies, cell lysates were incubated with 2 μg anti-Myc antibody <<?ENTCHAR lpar?>9E10, Sigma; 9B11, Cell Signaling Technology) or anti-HA antibody (26D11, Abmart; C29F4, Cell Signaling Technology) at 4 °C overnight, and then 20 μl precleared Protein A+G agarose (Sigma) was added in each experiment. Then, the immunoprecipitation was carried out exactly as previously described28. Monoclonal anti-Myc (9E10, Sigma) and monoclonal anti-Flag M2 peroxidase (Sigma) antibodies were used for western blots.

Yeast two-hybrid screen and interaction assays

The coding region and various truncations of bam amplified from a D. melanogaster ovarian cDNA library were cloned into the pGBKT7 vector as the baits and completely sequenced. The coding regions of bgcn and Csn4, which were amplified from a D. melanogaster ovarian cDNA library, were cloned into the pGADT7 vector and completely sequenced. The baits pGBKT7-bam/truncations and pGADT7-bgcn or pGADT7-Csn4 were cotransformed into the Y187 yeast strain to make the two-hybrid system. Positive interactions were screened using Leu-Trp-His triple selection markers.

For α-galactosidase unit measurement, interaction samples (Bam–Csn4 and Bam–Bgcn) were cultured in synthetic defined medium without His, Leu and Trp, whereas negative controls (BK-Csn4, BK-Bgcn and Bam-AD) were cultured in synthetic defined medium without Leu and Trp. Supernatant from the liquid culture was used to quantify yeast extracellular α-galactosidase activity resulting from expression of the MEL1 reporter gene in strain AH109. The catalytic activity of α-galactosidase was monitored colorimetrically by measuring the rate of hydrolysis of the chromogenic substrate PNP-α-Gal (N0877, Sigma) at 405 nm. This assay is a sensitive colorimetric method for quantifying the interaction in a two-hybrid system using strain AH109. α-galactosidase = (mU ml−1 × cell number) = OD405 × Vf × 1,000/(10.5t × Vi × OD600), where OD is optical density, Vf is the final volume of the assay, Vi is the volume of culture medium supernatant added and t is the elapsed incubation time.