Pelota controls self-renewal of germline stem cells by repressing a Bam-independent differentiation pathway

A stem cell is characterized by its ability to self-renew and generate differentiated cells throughout the lifetime of an organism. Understanding how stem cell self-renewal is controlled is an important issue in stem cell biology and will help realize potentials of stem cell-based therapies. Studies from diverse systems indicate that stem cell self-renewal is controlled by both extrinsic factors (niche signals) and intrinsic factors(Fuchs et al., 2004; Spradling et al., 2001). Niche signals have been shown to control GSC self-renewal by directly repressing expression of differentiation-promoting genes in the Drosophila ovary(Chen and McKearin, 2003a; Song et al., 2004). Therefore,the identification of pathways and genes that repress stem cell differentiation is crucial for understanding how stem cell self-renewal is controlled.

In the Drosophila ovary, GSCs reside in a structure called the germarium, which is at the anterior end of an ovariole(Lin, 2002; Xie and Spradling, 2001). At the anterior tip of the germarium, three types of somatic cells (terminal filament cells, cap cells and inner sheath cells) constitute a niche that supports two or three GSCs (Lin,2002; Xie and Spradling,2001; Xie and Spradling,2000) (Fig. 1A). One GSC divides to generate two daughter cells: the daughter cell maintaining contact with the cap cells through DE-cadherin-mediated cell adhesion renews itself as a stem cell, while the daughter cell moving away from the cap cells differentiates into a cystoblast (Song et al., 2002). The cystoblast divides four times with incomplete cytokinesis to form a 16-cell cyst, in which one cell becomes an oocyte and the rest become nurse cells (Spradling,1993). Bmp signals produced by the niche, Dpp and Gbb, have essential roles in controlling GSC self-renewal, as reduction of Bmp signaling activity results in the loss of GSCs by differentiation and overexpression of dpp in the germarium produces GSC-like tumors(Song et al., 2004; Xie and Spradling, 1998). Bmps from the cap cells function as short-range signals that directly repress the transcription of bam in GSCs to maintain their self-renewal, and also allow cystoblasts lying one cell diameter away to differentiate(Chen and McKearin, 2003a; Song et al., 2004). bam is necessary and sufficient for germ cell differentiating in the Drosophila ovary (Ohlstein and McKearin, 1997). In addition, two other genes, Yb and piwi, function in the somatic niche cells to control GSC(Cox et al., 2000; King et al., 2001). Yb encodes a novel protein and directly regulates expression of piwi and hh in TFs; hh signaling also modulates GSC self-renewal though it is not essential(King et al., 2001). piwi encodes a family of conserved RNA-binding proteins and is required in the niche cells for controlling GSC self-renewal and inside GSCs for their division (Cox et al.,1998; Cox et al.,2000). Two recent studies have shown that piwi also maintains GSC self-renewal by repressing bam expression through regulation of either the Bmp signaling pathway or a Bmp-independent signaling pathway (Chen and McKearin,2005; Szakmary et al.,2005). However, it remains unclear how piwi controls GSC division intrinsically.

Two translational repressors, Nanos (Nos) and Pumilio (Pum), have been shown to be required for the maintenance of ovarian GSCs by preventing differentiation (Forbes and Lehmann,1998; Wang and Lin,2004). Pum/Nos repress differentiation of PGCs and GSCs through a Bmp-independent pathway, as their expression is not regulated by Bmp signaling and their mutations cannot suppress hyperactive Bmp signaling-induced PGC proliferation (Gilboa and Lehmann,2004). It is likely that Nos and Pum are involved in repressing translation gene products that are important for germ cell differentiation and thereby for controlling GSC self-renewal. To identify further intrinsic factors that are required for Bmp-mediated GSC self-renewal, we performed a genetic screen to identify dominant suppressors of the dpp-induced GSC-like tumor phenotype (C.D. and T.X., unpublished). One of the suppressors is pelota (pelo), which has been studied for its role in the regulation of Drosophila male meiosis. Cellular and molecular analysis showed that pelo is required for the progression through meiosis in spermatogenesis and encodes an evolutionarily conserved protein that contains an eukaryotic release factor 1 α (eRF1α)-like domain at its C terminus (Eberhart and Wasserman,1995). The studies on the yeast pelo homolog dom34 suggest that Pelo is involved in translational regulation. Deletion of Dom34 causes growth retardation, defective sporulation and reduced polyribosomes (Davis and Engebrecht,1998). dom34 has a strong genetic interaction with RPS30A, which encodes ribosomal protein S30A; overexpression of RPS30A rescues the growth defects and reduced polyribosomes of dom34mutants (Davis and Engebrecht,1998). Moreover, Dom34 specifically interacts with Hbs1, a small GTPase that is also implicated in translational regulation(Carr-Schmid et al., 2002). It has been recently shown that pelo knockout mice exhibit early embryonic lethality with defects in cell division and proliferation(Adham et al., 2003). Taken together, pelo may be involved in the regulation of meiosis and mitosis possibly through regulating translation. In this study, we have revealed an unexpected new role of Pelo in the control of GSC self-renewal and division in the Drosophila ovary, possibly through regulating translation.

To make a UASp-pelo construct, the pelo cDNA insert from an EST clone (AT07625) was subcloned to pBlueScript with restriction enzymes EcoRI and PstI, and was then subcloned to a UASp vector with KpnI and XbaI. The Gateway cloning technology (Invitrogen)was used to generate UASp-tagged pelo constructs. The destination vectors used in this study were kindly provided by Terence Murphy (for more information, see http://www.ciwemb.edu/labs/murphy/Gateway%20vectors.html). Briefly, the pelo-coding region was amplified from the EST using the pfu DNA polymerase (Stratagene). The PCR product was cloned into an ENTR vector using a TOPO gateway cloning kit (Invitrogen). The ENTR clone was subsequently used in combination with different destination vectors to generate constructs of the pelo gene with different tags under the control of UASp promoter using the LR reaction kit (Invitrogen). Inverse PCR mutagenesis was used to make mutations on the putative nuclear localization sequence. All constructs were confirmed by sequencing before they were introduced into flies.

Unless otherwise stated, flies were reared at 25°C on standard molasses-based food. The fly stocks used for this study include: pelo¹, a strong P-element-induced allele(Eberhart and Wasserman,1995); bam⁸⁶, a null allele(McKearin and Ohlstein, 1995);Sec61α-GFP (ZCL0488) (Kelso et al.,2004); Dad-lacZ(Tsuneizumi et al., 1997); bam-GFP (Chen and McKearin,2003b); nosgal4VP16(Van Doren et al., 1998); and c587-gal4 (Song et al.,2004).

Marked clones were generated using the FLP-mediated FRT recombination technique according to the published procedures(Song et al., 2002; Xie and Spradling, 1998). These marked clones were analyzed and quantified 3 days, 10 days, 17 days and 24 days after clone induction (ACI). For examination of Dad-lacZ and bam-GFP expression in the marked pelo¹ mutant GSCs, the ovaries were analyzed 14 days ACI. The genotypes used for clonal analysis in this study are shown as follows:

1. hs-flp/+; FRT 40A pelo¹/ FRT40A arm-lacZ

2. hs-flp/+; FRT 40A pelo¹/ FRT40A ubi-GFP; Dad-lacZ/+

3. hs-flp/+; FRT 40A pelo¹/ FRT40A arm-lacZ; bam-GFP/+.

Antibody staining of ovaries was performed using our published protocol(Song et al., 2002). The following antisera and dilutions were used: rabbit anti-β-galactosidase(1:100; Molecular Probes); mouse anti-β-galactosidase (1:50; Molecular Probes); rabbit anti-GFP (1:100; Molecular Probes); monoclonal anti-Hts 1B1(1:4; DSHB); monoclonal anti-HtsRC (1:4; DSHB); monoclonal anti-Orb (1:30;DSHB); and rabbit anti-Vasa antibody (1:1000; a gift of Dr Paul Lasko). Secondary antibodies including goat anti-rabbit or anti-mouse IgG conjugated to Alexa 488 or Alexa 568 (Molecular Probes) were used at a dilution of 1:200.

Ovaries from 2-day-old females for ApopTag staining and BrdU labeling were dissected in Grace's medium. ApopTag in situ apoptosis detection kit(Serologicals, Clarkston, GA) was performed according to the manufacture's manual. For BrdU labeling, the ovaries were incubated in the medium at a final concentration at 75 μg/ml at room temperature for 1 hour. Fixation and BrdU detection were described previously (Zhu and Xie, 2003).

As we reported previously, dpp overexpression in inner sheath cells driven by the c587-gal4 driver completely blocks germ cell differentiation, resulting in the formation of GSC-like tumors and consequently female sterility (Song et al., 2004). Such female sterility is very sensitive to genetic changes of any dpp downstream components; for example, removal of one copy of any Bmp downstream gene such as punt, mad and Medcan sufficiently reverse the sterility phenotype, leading to fertile females(Doan and Xie, data not shown). To identify the genes that are potentially involved in Bmp signaling in GSCs, we performed a dominant suppressor screen using the existing deficiency kit, which covers 65% of the Drosophilagenome. In the screen, we have identified a small deficiency, Df(2L)s1402(30C-30F), that can rescue dpp overexpression-induced female sterility. Among 19 existing mutations in the genomic region, pelotais identified as dominantly suppressing dpp overexpression-induced sterility (Doan and Xie, data not shown). Although it has been identified for its role in Drosophila male meiosis, its mutant females are also semi-fertile, and ovaries are overtly small(Eberhart and Wasserman,1995), suggesting that it is also involved in the regulation of Drosophila oogenesis. However, it remains unclear whether the pelo mutation affects GSCs.

To determine if pelo is required for maintaining GSCs, we quantified GSCs in pelo¹ homozygous mutant germaria. pelo¹ is a P-element-induced strong or null allele, which is based on the evidence that it produces a truncated pelo transcript deleting most of the eRF1α domain(Eberhart and Wasserman,1995). The ovaries from 2-day-old or 7-day-old pelo¹ homozygous and heterozygous females were immunostained with anti-Vasa and anti-Hts antibodies. Vasa is expressed specifically in germ cells (Lasko and Ashburner, 1988), while Hts is preferentially rich in spectrosomes(GSCs and cystoblasts), fusomes (2-, 4-, 8- and 16-cell germline cysts) and membranes of somatic follicle cells in the germarium and egg chambers(Lin et al., 1994). GSCs can be reliably identified at the tip of the germarium by their anteriorly localized spectrosome and direct contact with cap cells(Lin, 2002; Xie and Spradling, 2001). In pelo¹ heterozygous control females, 2-day-old and 7-day-old germaria had an average of 2.35 and 2.40 GSCs, respectively, and more than 97% of the germaria contained two or more GSCs(Table 1; Fig. 1B), which closely resembles wild-type. The 2-day-old pelo¹ homozygous germaria contained an average of 1.4 GSCs, with some having two GSCs(Fig. 1C) and the other containing only one GSC (Fig. 1D), indicating that pelo is required for GSC establishment or GSC maintenance. Furthermore, the 7-day-old pelo¹ germaria had an average of 0.47 GSCs with 57.3%(51/89) of them containing no GSC (Table 1; Fig. 1E). Consistent with the previous study showing that pelo¹ is a strong or null allele (Eberhart and Wasserman, 1995), the pelo¹/df(2L)s1402 mutant ovaries behaved just like the homozygous pelo¹ mutant ovaries in terms of GSC numbers at different ages (Fig. 2E). This also suggested that the mutation in pelo is responsible for GSC loss phenotype in the pelo¹ homozygous mutant ovaries. Together, these results demonstrate that Pelo is required for maintaining GSCs in the Drosophila ovary.

Pelo could maintain GSCs by controlling either self-renewal or survival. Thus, the GSC loss in pelo mutant ovaries could be due to either differentiation or cell death. To differentiate these two possibilities, we examined cell death of GSCs in the pelo mutant ovaries using the ApoTag cell death labeling system, which has been successfully used in our previous studies (Zhu and Xie,2003). In this experiment, the spectrosomes and fusomes were also labeled for facilitating identification of GSCs, cystoblasts and germline cysts. Among 156 pelo¹ heterozygous control germaria, no dying GSCs (>250 GSCs examined) were observed, and there were eight dying cysts. This is consistent with the previous result that some germline cysts die naturally (Drummond-Barbosa and Spradling, 2001). Among 264 pelo¹ homozygous germaria, there were also no dying GSCs detected (>300 GSCs examined) but 49 dying cysts were observed (Fig. 1F). Clearly, the pelo mutation does not affect GSC survival but appears to increase dying cysts [from 5% of the 2-day-old wild-type germaria carrying dying cysts (n=40) to 18% of the 2-day-old pelo¹ mutant garmaria carrying dying cysts(n=89)]. These observations suggest that Pelo is required for GSC self-renewal and also for cyst survival.

We also noticed that the pelo mutant germaria that still harbored two GSCs were extremely small and contained very few germline cysts(Fig. 1C). As the pelomutation resulted in only 18% of the germaria carrying one or more dying cysts, the pelo mutant GSC division rate probably also decreased. Then, we used BrdU incorporation to label S-phase cells to determine whether pelo is required for controlling GSC division. In the 2-day-old females, 10.7% of pelo heterozygous control GSCs (n=525)were BrdU positive, whereas only 2.5% of pelo homozygous GSCs incorporated BrdU (n=202). This result shows that pelo is also required for controlling GSC division.

Pelo could control GSC self-renewal by acting either inside the GSC, in the niche or both. Our mRNA in situ hybridization and gene expression profiles of agametic ovaries show that pelo mRNAs were ubiquitously expressed at lower levels throughout the germarium, suggesting that pelo could function in GSCs or the somatic niche cells, or both (data not shown). We used FLP-FRT-mediated mitotic recombination to determine whether Pelo functions inside GSCs for controlling their self-renewal(Xu and Rubin, 1993). The FLP-mediated FRT recombination has been used to generate marked mutant GSCs and determine their loss rates for deducing the role of a particular gene in GSC maintenance (Xie and Spradling,1998). According to published experimental procedures(Xie and Spradling, 1998; Song et al., 2002), the ovaries of the females of appropriate genotypes were dissected at 3, 10, 17 and 24 days after clone induction (ACI) mediated by heat-shock treatments, and marked wild-type and pelo mutant GSCs were identified by the lack of arm-lacZ expression and the presence of an anteriorly anchored spectrosome. In the wild-type control, 55.0% of the marked GSCs (from 52.8% of the germaria carrying one or more marked control GSCs at 3 days ACI to 28.9%of the germaria carrying one or more marked control GSCs at 24 days ACI) were still maintained for 3 more weeks (Xie and Spradling, 1998) (Fig. 2A,B). By contrast, 97% of the marked pelo mutant GSCs(from 48.6% of the germaria carrying one or more marked mutant GSCs at 3 days ACI to 1.7% of the germaria carrying one or more marked mutant GSCs at 24 days ACI) were lost during the same 3- week period(Table 2; Fig. 2C,D). These results demonstrate that pelo is required in GSCs for controlling their self-renewal. It appears that these marked pelo mutant GSCs are lost slower than the GSCs in the homozygous pelo mutant ovaries. This could be explained by the possibility that Pelo is very stable or has functions in both GSCs and soma. If Pelo is very stable, it takes longer for residual wild-type Pelo that is made before clone induction to be degraded in the marked pelo mutant GSCs. Later, we will try to address whether pelo has a function in the soma to control GSC maintenance.

To confirm further that Pelo is indeed required intrinsically for controlling GSC self-renewal and that the pelo mutation is responsible for the GSC loss phenotype, we expressed a UASp-pelotransgene specifically in the germ cells using a germ cell-specific GAL4 driver, nos-gal4VP16, in pelo mutant ovaries. A UASp-pelo construct was made to be expressed in germline or soma using different GAL4 drivers (Rorth,1998); nosgal4-VP16 can drive a UAS transgene to be expressed in both GSCs and later germ cell cysts(Van Doren et al., 1998). Interestingly, introduction of one copy of UASp-pelo transgene alone into the pelo homozygous females was able to partially rescue the GSC loss phenotype, suggesting that there is a leaky expression of the transgene even without a GAL4 driver (Fig. 2E). The stem cell loss phenotype in pelo mutant ovaries was fully rescued by nosgal4VP16-driven pelo expression in the germline, including GSCs, restoring the normal GSC number(Fig. 2F). This result shows that the mutation in pelo is responsible for the GSC loss phenotype,and further confirms that pelo is required intrinsically for controlling GSC self-renewal.

Although pelo is required intrinsically for controlling GSC self-renewal, it does not rule out the possibility that a somatic function of pelo is also involved in regulating GSC self-renewal as it is also expressed in the somatic cells of the germarium. First, we examined whether the pelo mutation affects the survival of cap cells. A hh-lacZ enhancer trap line used in this study has been used to label terminal filament cells and cap cells in the Drosophila ovary(Forbes et al., 1996). In the pelo homozygous germaria carrying one or no GSCs, the cap cell number appeared to be normal, ranging from five to seven(Fig. 2G,H), indicating that Pelo function is at least not required for the formation or survival of cap cells. However, pelo could still be required in the surrounding somatic cells for controlling GSC function through regulating production of signals. To further test whether pelo has a somatic function in GSC regulation, we used the c587-gal4 driver, which is expressed in IGS cells and follicle progenitor cells, and the same UASp-pelo transgene to test whether somatic expression pelo can rescue the pelomutant GSC loss phenotype. However, c587-driven pelo expression in the somatic cells could only confer very limited rescue of the pelomutant GSC loss phenotype, in addition to the partial rescue conferred by the UAS-pelo transgene alone, suggesting that pelo has little role in IGS cells and follicle cell progenitors for GSC self-renewal(Fig. 2E). As we have not tested whether pelo expression in terminal filaments/cap cells can mitigate the GSC loss phenotype of the pelo mutant ovaries, our results could not completely rule out the possibility that Pelo has a function in somatic cells for controlling GSC self-renewal.

As pelo is expressed in all the somatic cells, including SSCs, we then determined whether pelo has a role in SSC maintenance using FLP-mediated FRT recombination to generate marked pelo mutant SSC clones. The marked SSCs were identified as the arm-lacZ-negative somatic cells residing at the 2a/2b boundary and generating arm-lacZ-negative (marked) follicle cells in the germarium and egg chambers, according to our previous studies(Song and Xie, 2002; Song and Xie, 2003). In the control, 68% of marked wild-type SSCs were maintained for 3 weeks, supporting the fact that SSCs have a slow natural turnover(Table 2). Similarly, 62% of the marked pelo SSCs were maintained for 3 weeks, indicating that pelo plays little or no role in SSC maintenance(Table 2). The marked pelo¹ mutant follicle cell clones exhibited a very minor phenotype: they appeared slightly thinner compared with wild-type follicle cells (data not shown). Although pelo is ubiquitously expressed throughout the germarium, the main function of pelo is primarily restricted to GSCs and their progeny in the ovary.

Drosophila Pelo has a putative nuclear localization signal sequence (PRKRK) at its N terminus(Eberhart and Wasserman, 1995; Nair et al., 2003); this sequence is perfectly conserved from Drosophila to human, raising an interesting possibility that Pelo is a nuclear protein. If Pelo indeed functions in the nucleus, we would expect that the disruption of the putative nuclear localization sequence leads to loss of its function. To directly test the idea, we generated a mutant version of Pelo with the replacement of PRKRK by RSRS as ablation of helix-breaking residue proline and reduction of basic residues can abolish the function of the nuclear localization signal(Conti et al., 1998). In S2 cells, the mutant Pelo protein tagged with 3xFlag and 6xMyc at its N terminus[F-M-Pelo(nls^*)] was localized in the cytoplasm in the same way as the wild-type version tagged with the same tags at its N terminus (F-M-Pelo)(data not shown). To further determine whether the NLS of Pelo is important for Pelo function in controlling GSC self-renewal in vivo, we generated transgenic flies carrying either UASp-F-M-Pelo or UASp-F-M-Pelo(nls^*). Two independent insertion lines for UAS-F-M-Pelo could fully rescue the pelo mutant GSC loss phenotype when they were driven to be expressed specifically in the germ cells by nos-gal4VP16, indicating that FLAG and MYC tags do not interfere with Pelo function (Fig. 2E). Similarly, two independent transgenic lines of the NLS mutated version of Pelo could also fully rescue the pelo GSC loss phenotype(Fig. 2E), further supporting that the putative NLS is not important for Pelo function in GSCs and that Pelo functions in the cytoplasm to control GSC self-renewal(Fig. 2E). Interestingly, one of the transgenic lines [UASp-FM(nls^*)#1] showed complete rescue for the pelo GSC loss phenotype with the nosgal4VP16 driver,whereas it exhibited little rescue for the GSC loss phenotype of the pelo mutant ovaries with the c587 driver or without any gal4 driver, further supporting our earlier conclusion that the mutation in the pelo gene is responsible for the GSC loss(Fig. 2E).

As the tagged Pelo is functional, we further determined the Pelo subcellular localization in GSCs using Flag or Myc antibodies. As in S2 cells,Pelo was mainly expressed in the cytoplasm in the germ cells(Fig. 3A). As Pelo protein has a domain showing homology to translation release factor 1a (RF1), we next examined whether Pelo is associated with the ER. Sec61α-GFP, a GFP fusion protein localized to ER membrane, was localized to ER membrane rich in the peri-nuclear area and spectrosome area of the GSC(Snapp et al., 2004)(Fig. 3B), while Pelo was evenly distributed throughout the cytoplasm of the GSC(Fig. 3C,D). There was only limited pattern overlap between Sec61α-GFP and F-M-Pelo, indicating that Pelo is primarily localized to the cytoplasm away from the ER membrane(Fig. 3D). Consistently, Pelo was also localized to the cytoplasm in S2 cells (data not shown). These findings show that Pelo functions in the cytoplasm of GSCs to control their self-renewal.

As discussed earlier, pelo was identified in a genetic screen looking for genes that can suppress dpp overexpression-induced GSC-like tumors, suggesting that pelo must somehow genetically interact with the dpp/Bmp pathway. To further reveal the relationship between pelo and Bmp signaling, we carefully examined the dose effect of pelo on dpp-induced GSC-like tumor formation. As reported previously (Xie and Spradling,1998; Song et al.,2004), all the ovarioles overexpressing dpp by the c587 gal4 driver (n=77) contained only single germ cells resembling GSCs (Fig. 4A). Among the dpp-overexpressing ovarioles also carrying one copy of the pelo¹ mutation, 36% of them (n=206) showed the same tumor phenotype, but the rest of the ovarioles contained differentiated germline cysts, developing egg chambers and even mature eggs(Fig. 4B), which could explain why pelo was identified in our suppressor screen. Among the dpp-overexpressing ovarioles also carrying two copies of the pelo¹ mutations, only 13.8% of them (n=261)contained only GSC-like single germ cells, 49.8% of them had a mixture of single germ cells and developing cysts(Fig. 4C). Interestingly, the rest (36.4%) were reminiscent of the pelo GSC loss phenotype only(Fig. 4D). These results suggest that pelo functions as one of the Bmp downstream components or in a pathway parallel to the Bmp signaling pathway to control GSC self-renewal.

To further understand how pelo modulates Bmp signaling activity,we examined the expression of a Bmp direct target gene, Dad, in the pelo mutant GSCs. Dad-lacZ is a lacZ enhancer trap line for Dad (Tsuneizumi et al.,1997). Its expression is the strongest in the GSCs, and is quickly downregulated in the differentiating cystoblasts(Kai and Spradling, 2003; Song et al., 2004). The pelo¹ mutant GSCs marked by loss of ubi-GFPexpression were generated by the FLP-mediated FRT recombination, and then were analyzed for Dad-lacZ expression 2 weeks after clone induction. Consistent with the idea that pelo is involved in modulating Bmp signaling, 69% (n=36) of the marked mutant pelo GSCs (GFP negative) showed the downregulation of Dad-lacZ expression in comparison with their neighboring wild-type GSCs (GFP-positive)(Fig. 4E,E′). We further determined whether pelo is also involved in Bmp-mediated bamrepression in GSCs as Bmp signaling directly represses bamtranscription in GSCs (Chen and McKearin,2003a; Song et al.,2004). A bam-GFP transgene (a GFP reporter driven by a bam promoter) is repressed in GSCs, while its expression is upregulated in the differentiating cystoblasts(Chen and McKearin, 2003b). The marked pelo mutant GSCs (lacZ negative) were generated by the FLP-mediated FRT recombination and were examined for bamexpression according to our previously published procedures(Song et al., 2004). Only about 5% (n=55) of the marked pelo mutant GSCs(lacZ negative) showed slight upregulation of bam-GFP in comparison with their neighboring unmarked wild-type GSCs (lacZpositive), while the rest of the marked pelo¹ mutant GSCs did not upregulate bam-GFP expression(Fig. 4F,F′). These findings indicate that Pelo is involved in modulating Bmp signaling in GSCs but plays little or no role in regulating Bmp-mediated bamrepression, and further suggest that it functions in one branch of the responses of the Bmp signaling pathway to regulate GSC self-renewal.

bam is both essential and sufficient for cystoblast differentiation (McKearin and Ohlstein,1995; Ohlstein and McKearin,1997). Our observation that pelo¹ mutant GSCs are lost because of differentiation but do not upregulate bamexpression suggests that pelo mutant GSCs differentiate using a bam-independent pathway. If so, we should expect bam mutant germ cells to be able to differentiate in the absence of pelofunction. To test this idea, we investigated the genetic relationship between bam and pelo. The pelo¹ homozygous GSCs that were heterozygous for bam^Δ86, a deletion allele of bam, were still lost rapidly as in the pelo¹ mutant GSCs (data not shown); the bam^Δ86 homozygous germ cells that were also heterozygous for pelo¹ still failed to differentiate, as did the bam^Δ86 mutant one(Fig. 5A). Interestingly, in pelo¹; bam^Δ86 double homozygous germaria, most of the germ cells were cysts with branched fusomes,and some of them still retained a round spectrosome(Fig. 5B). The round spectrosomes in the remaining single germ cells were unusually larger than those in bam mutant single germ cells, suggesting the single germ cells could be growth-arrested cystoblasts but can continue to grow their spectrosome. The morphology of the branched fusomes of the double mutant cysts appeared abnormal. To further determine whether the oocytes form in these pelo; bam mutant cysts, we examined the expression of Orb protein in the double mutant germaria. Orb normally starts to accumulate in newly formed wild-type oocytes (Fig. 5C); however, no obvious Orb expression was detected in the pelo; bam mutant germaria, indicating that there is no oocyte formation in the double mutant cysts(Fig. 5D). As pelomutant cysts can still form the oocyte, bam is probably required late for oocyte formation. These findings show that pelo mutant cystoblasts can differentiate without functional bam, and further suggest that pelo must repress a bam-independent differentiation pathway to maintain GSC self-renewal.

We also noticed that the pelo mutant females had some fertility after eclosion and quickly lost their fertility. Interestingly, in the 2-day-old pelo mutant ovaries, early egg chambers looked largely normal, and even a few mature eggs were present(Fig. 6A). By contrast, in the 7-day-old ovaries, egg chambers older than stage 9 were rarely observed, and even those early stage egg chambers exhibited condensed nurse cell DNA, which suggest that they are in the process of undergoing apoptosis(Fig. 6B,C). The egg chamber degeneration is not due to oocyte formation defect because in these egg chambers the oocyte was present (Fig. 6B). In addition, the size of some egg chambers exhibiting condensed nurse cell DNA was also smaller than normal(Fig. 6C). The Pelo function also appeared to be germ-cell specific in egg chambers, as marked mutant pelo germ cells failed to grow to the normal size and became apoptotic (Fig. 6D). These results indicate that pelo is also required in the germ cells for their survival or normal differentiation during later oogenesis.

In the Drosophila ovary, GSCs have been previously shown to be controlled by niche-derived Bmps and intrinsic factors such as Pumilio and Nanos. The Bmp signal controls GSC self-renewal by directly repressing Bam expression. In this study, we have revealed a new function of pelo in controlling GSC self-renewal in the Drosophila ovary. Our genetic clonal analysis and rescue experiment results show that Pelo is required intrinsically for controlling GSC self-renewal. It is involved in the regulation of some Bmp response in GSCs but is not required for repressing bam expression. Genetic analysis further indicates that it controls GSC self-renewal by repressing a bam-independent differentiation pathway (Fig. 7). Although translational control has been implicated in the regulation of GSC self-renewal, this is the first time it has been shown that a translational release factor-like protein controls GSC self-renewal. As Pelo is highly conserved from Drosophila to human, it is also possible that its mammalian homologs might also be involved in the control of stem cell self-renewal.

Drosophila Pelo belongs to a family of evolutionarily conserved proteins with their primary function in the regulation of cell division. In the yeast, dom34 (pelo) mutant cells grow slowly and have defects in the entry of meiosis, indicating that it is required for mitosis and meiosis. In mice, disruption of the pelo gene causes early embryonic lethality and defects in cell cycle progression(Adham et al., 2003). Although pelo is ubiquitously expressed throughout Drosophiladevelopment (Eberhart and Wasserman,1995), the pelo mutants survive to adulthood without obvious defects in the body. In Drosophila, pelo has been shown to be required to control meiotic cell cycle progression in male germ cells. In this study, we show that pelo is required intrinsically for controlling self-renewal and division of GSCs but not SSCs in the ovary, which is supported by rescue and stem cell clonal analysis experiments. Even though Pelo members are required for regulating cell cycle progression from yeast to mammals, it remains unclear how they accomplish this function.

The only clue to the potential cellular function of Pelo comes from its high homology to the translation release factor 1. Its likely function as a translational regulator is further complicated by the presence of a highly conserved NLS sequence (Eberhart and Wasserman, 1995). Using an epitope-tagged Pelo that can rescue pelo mutants, we demonstrate that Pelo is mainly localized to the cytoplasm of both S2 cells and germ cells. Furthermore, the pelo gene with a mutated putative NLS is still fully functional. The yeast Pelo, Dom34,is also localized to the cytoplasm (Davis and Engebrecht, 1998; Huh et al., 2003). These findings support the idea that Pelo proteins function in the cytoplasm as translational regulators in different organisms. If Pelo truly functions as a translational release factor, they must directly interact with ribosomes that are either associated with ER or in the cytoplasm. Consistent with this idea, some Pelo proteins are associated with ER membranes though the majority of Pelo proteins are not associated with ER membranes. In yeast, dom34 mutants have dramatically reduced polyribosomes and can be rescued by a high-copy of the ribosomal protein S30A gene, indicating that Dom34 is involved in translation(Davis and Engebrecht, 1998). In addition, expression of Drosophila pelo in dom34 mutants can rescue growth defects, indicating its conserved function during evolution. Therefore, Pelo is also likely a translational regulator in Drosophila, and is involved in regulating translation of a specific class of mRNAs that are important for germ cell function. In the future, it will be important to investigate whether Pelo is indeed involved in translational regulation and to identify its targets in germ cells.

Before this work, pelo has not been shown to be involved in regulating any signaling pathways in any organisms. The Bmp pathway is a major signaling pathway that is essential for controlling GSC self-renewal and division in the Drosophila ovary(Song et al., 2004; Xie and Spradling, 1998). The Bmp signaling activities can be reliably monitored by expression of Dad in GSCs (Kai and Spradling,2003; Song et al.,2004). We anticipate that it must somehow interact with the Bmp pathway in controlling GSC self-renewal as pelo was also identified as a dominant suppressor of Dpp overexpression-induced GSC-like tumors. In this study, we show that the GSCs mutant for pelo downregulate Dad. These findings indicate that Pelo participates in Bmp signaling to control expression of dpp target genes in GSCs such as Dad.

In the Drosophila ovary, one of the ways in which Bmp signaling controls GSC self-renewal is to directly repress bam expression in GSCs. bam is necessary and sufficient for cystoblast differentiation in the Drosophila ovary (McKearin and Spradling, 1990; Ohlstein and McKearin, 1997). In this study, we show that Pelo is essential for controlling GSC self-renewal but is not involved in repressing bam expression. pelo mutant GSCs have normal bamrepression but their progeny can still differentiate without bam,suggesting that pelo maintains GSCs by repressing a bam-independent pathway. During the preparation of this manuscript,two studies were published showing that pum controls GSC self-renewal by repressing a bam-independent pathway(Chen and McKearin, 2005; Szakmary et al., 2005). Pum is known to work together with Nos, which is also essential for Drosophila ovarian GSC self-renewal(Gilboa and Lehmann, 2004; Wang and Lin, 2004), to repress gene translation in the embryo(Barker et al., 1992; Forbes and Lehmann, 1998). As Pum/Nos does not participate in Bmp signaling(Gilboa and Lehmann, 2004) and Pelo is a translational release factor-like protein, we propose that Pelo works in a parallel genetic pathway with Pum in repressing the same or different Bam-independent differentiation pathways through regulating translation (Fig. 7). Although it is essential for repressing a Bam-independent pathway(s) in GSCs, Pelo is not so sufficient for doing so as Bmp signaling is for repressing bamas overexpression of pelo has no effect on the GSC maintenance and differentiation. In the future, it will be important to molecularly and genetically characterize the Bam-independent pathway repressed by Pelo and to further understand how Pelo represses it in relation to Pum.

We thank L. Cooley, P. Lasko, D. McKearin, T. Murphy, Bloomington Stock Center and Developmental Studies Hybridoma Bank for reagents. We also thank members of the Xie laboratory for stimulating discussions and critical comments on manuscripts, and J. Haynes for administrative assistance. This work is supported by NIH (1R01 GM64428-01) and Stowers Institute for Medical Research.