TSC1/2 tumour suppressor complex maintains Drosophila germline stem cells by preventing differentiation

Stem cells are characterized by the ability of self-renewal and generating daughters that give rise to one or more types of differentiated cells. These properties enable them to maintain homeostasis of many adult tissues throughout life. Stem cells in adult tissues are commonly regulated by extrinsic factors from the micro-environment or niche, and intrinsic factors that function cell-autonomously in stem cells (for reviews, see Li and Xie, 2005; Morrison and Spradling, 2008). Elucidating how the self-renewal and differentiation of stem cells are balanced by extrinsic and intrinsic factors is crucial in understanding tissue homeostasis control, the dysregulation of which is linked to many human diseases.

The Drosophila ovarian germline stem cells (GSCs) provide an attractive system in which to study the molecular mechanisms underlying stem cell self-renewal and differentiation (for a review, see Kirilly and Xie, 2007). A single stem cell-niche unit is found at the anterior tip of each germarium, in which terminal filament cells, cap cells and escort cells together form a functional GSC niche to host two or three GSCs (Fig. 1A) (Xie and Spradling, 2000). The niche physically anchors GSCs by forming DE-cadherin-mediated adherens junctions (Song et al., 2002). Additionally, the niche cells produce crucial signal molecules to govern GSC self-renewal and prevent differentiation. Previous studies have demonstrated a central role of BMP signalling in maintaining GSC self-renewal. Two ligands of BMP signalling, Decapentaplegic (Dpp) and Glass bottom boat (Gbb), which are expressed in the niche cells, directly act on GSCs through canonical pathway components to regulate the transcription of target genes (Song et al., 2004; Xie and Spradling, 1998). A major target is bag of marbles (bam) (Chen and McKearin, 2003a; Song et al., 2004), a differentiation-promoting gene (McKearin and Ohlstein, 1995; Ohlstein and McKearin, 1997). In GSCs, the activated/phosphorylated transcriptional factor Mad forms complexes with Medea (the Drosophila Smad4) and subsequently translocates to nucleus where they recognize a silencer element on the bam promoter and maintain transcriptional silencing of bam (Chen and McKearin, 2003a; Chen and McKearin, 2003b; Song et al., 2004). In cystoblasts, where bam repression is relieved, Bam functions together with Bgcn to inhibit the translation of Nanos (Nos), a translational repressor that prevents GSC differentiation; as a consequence, differentiation can proceed (Forbes and Lehmann, 1998; Li et al., 2009; Wang and Lin, 2004). Other factors are also required in the GSCs, such as Pelota (Pelo), a translational release factor-like protein that essentially helps in preventing GSC differentiation by modulating both BMP-Bam-dependent and -independent pathways (Xi et al., 2005). Furthermore, the microRNA pathway components, including Dcr-1, Ago1 and Loquacious are also cell-autonomously required for GSC division and maintenance (Hatfield et al., 2005; Jin and Xie, 2007; Park et al., 2007; Yang et al., 2007), whereas Mei-P26, a Trim-NHL containing protein, functions to antagonize the microRNA pathway and allow germline differentiation (Neumuller et al., 2008; Page et al., 2000). Therefore, in addition to transcriptional regulation by BMP signalling, post-transcriptional regulation is crucial for GSC self-renewal, although functional relationships among those regulators and pathways are not yet well understood.

Tuberous sclerosis complex is a human disease characterized by the presence of benign tumour cells termed hamartomas in multiple organs, and usually caused by mutations in either Tsc1 or Tsc2 genes (for a review, see Inoki et al., 2005). Studies led in Drosophila have demonstrated that TSC1 and TSC2 form a physical and functional complex to control cell growth (Gao and Pan, 2001; Ito and Rubin, 1999; Potter et al., 2001; Tapon et al., 2001). It functions downstream of PI3K and Akt, and upstream of TOR (target of rapamycin) and S6K. Biochemically, TSC1 and TSC2 function as a GAP (GTPase activating protein) to inhibit Rheb (Ras homolog enriched in brain), a small G protein that positively regulates TOR (for reviews, see Inoki et al., 2005; Kwiatkowski and Manning, 2005). TOR is a Ser/Thr kinase conserved from yeast to mammals that controls multiple cellular processes, such as cell growth, proliferation and metabolism. TOR functions via two complexes, TORC1, which is sensitive to rapamycin and performs most of TOR functions, and TORC2, which is not sensitive to rapamycin and regulates cytoskeleton organization. TORC1 has two well-studied targets, S6K and 4EBP, which can be phosphorylated by TORC1 and regulate protein synthesis. TSC1/2-TORC1 signaling has been implicated in mediating many environmental cues, such as growth factor signals, amino acids and ATP, and could function as a signal integration point to regulate cell growth and metabolism (for reviews, see Kwiatkowski and Manning, 2005; Wullschleger et al., 2006).

In this study, we have revealed a novel role of TSC1/2-TORC1 signaling in regulating GSC maintenance and differentiation in the Drosophila ovary. Our study suggests a model whereby, in addition to the general role in cell growth control, TSC1/2-TORC1 signaling could also orchestrate GSC self-renewal and differentiation through regulating both BMP-bam-dependent and - independent differentiation programs. These results underline the importance of translational regulation in controlling GSC behaviour.

Unless otherwise stated, flies were reared at 25°C on standard media with fresh yeast paste added to the food surface. The fly stocks used for this study include: Tsc1^Q87X, Tsc1^R453X, gig¹⁹² (Tapon et al., 2001), gig⁵⁶ (Ito and Rubin, 1999), Tsc1²⁹ (Gao and Pan, 2001), bam-GFP (Chen and McKearin, 2003b), bam^Δ86 (McKearin and Ohlstein, 1995), Tor^P1 (P[lacw]l(2)k17004), Tor^ΔP (Zhang et al., 2000) and S6k^I-1 (Montagne et al., 1999).

Wild-type and mutant GSC clones were generated by FLP/FRT-mediated mitotic recombination (Xu and Rubin, 1993), and the genotypes of flies studied are as follows:

• hsflp/ +; FRT82B arm-lacZ/ FRT82B;

• hsflp/ +; FRT82B arm-lacZ/ FRT82B Tsc1^Q87X;

• hsflp/ +; FRT82B arm-lacZ/ FRT82B Tsc1^R453X;

• hsflp/ +; FRT82B arm-lacZ/ FRT82B Tsc1²⁹;

• hsflp/ +; arm-lacZ FRT 80B/ FRT80B;

• hsflp/ +; arm-lacZ FRT 80B/ gig¹⁹² FRT80B;

• hsflp/ +; arm-lacZ FRT 80B/ S6k^I-1 FRT80B;

• hsflp/ +; arm-lacZ FRT 80B/ S6k^I-1 gig⁵⁶ FRT80B;

• hsflp/ +; bam-GFP/ +; FRT82B arm-lacZ/ FRT82B Tsc1^Q87X;

• hsflp/ +; bam-GFP/ +; FRT82B arm-lacZ/ FRT82B Tsc1^R453X;

• hsflp/ +; FRT82B arm-lacZ/ FRT82B bam^Δ86;

• hsflp/ +; FRT82B arm-lacZ/ FRT82B Tsc1²⁹ bam^Δ86;

• hsflp/ +; arm-lacZ FRT 40A/ Tor^P1T40A;

• hsflp/ +; arm-lacZ FRT 40A/ Tor^ΔP FRT40A;

• hsflp/ +; Tor^ΔP / +; FRT82B arm-lacZ/ FRT82B Tsc1^R453X.

Flies with FRT80B were treated with a stronger heat-shock regime because the recombination rate is relatively low for FRT80B. Flies were heat shocked in a 37°C water bath three times per day for 3 days for 80 minutes each time. Flies with other FRT sites were heat shocked twice a day for 3 days for 60 minutes each time, as previously described (Song and Xie, 2002; Xie and Spradling, 1998). For time-course clonal analysis, flies were randomly grouped and were analyzed at days 4, 6, 8 and 11 after clone induction (ACI), respectively. To examine pMAD and Bam expression level in Tsc1 mutant GSCs, flies were analyzed at day 6 ACI. To examine the effect of Tor/ + on the efficiency of rapamycin treatment in Tsc1 mutant GSCs, flies were analyzed at days 4, 8 and 20 ACI, respectively. For pMad detection in the imaginal discs, larvae at the first or second instars were heat shocked for 120 minutes and dissected in 2-5 days.

Different doses of rapamycin were used for different sets of experiments. For all experiments, rapamycin dissolved in ethanol was directly added to the food surface and the liquid was air-dried before used. In Tsc1 mutant rescue experiment, 100 μl of 500 μM rapamycin was added to each vial. For TOR inhibition in wild-type flies, 100 μl of 4 mM rapamycin was added to each vial. For experiments to test the effect of Tor/ + on rapamycin treatment, 100 μl of 100 μM rapamycin was added to each vial. Fresh food with rapamycin was replaced every other day until dissection.

Antibody staining of ovaries was performed as previously described (Zhao et al., 2008). Briefly, ovaries were fixed in 4% Paraformaldehyde for 15 minutes, and blocked in 5% normal goat serum in PBT [10 mM NaH₂PO₄/Na₂HPO₄, 175 mM NaCl (pH 7.4), plus 0.1% Triton X-100]. Primary and secondary antibodies were then added according to standard procedures. The following antisera and dilutions were used: rabbit anti-galactosidase (1:6000; Cappel); mouse anti-galactosidase (1:100; Molecular Probes); rabbit anti-GFP (1:300; Molecular Probes); mouse anti-α-spectrin (1:50; DSHB); rat anti-DE-cadherin (1:10; DSHB); mouse anti-orb (1:50; DSHB); mouse anti-HtsRC (1:50; DSHB); rabbit anti-Bam (1:100, a gift from D. Chen); and rabbit anti-pMad (1:3000; a gift from E. Laufer). Secondary antibodies, including goat anti-rabbit or anti-mouse IgG conjugated to Alexa 488 or Alexa 568 (Molecular Probes), were used at a dilution ratio of 1:300. For DAPI (4′, 6′-diamidino-2-phenylindole) staining, ovaries were incubated in PBT with 0.2 μg/ml DAPI (Sigma) for 6 minutes. All images were taken by Zeiss Imager Z1 equipped with ApoTome system, then processed in Adobe Photoshop and Illustrator.

Ovaries carry Tsc1 mutant GSCs were dissected at 6 days ACI, and in situ cell death detection kit (Roche) was used to detect apoptosis according to the manufacturer's manual and as previously described (Zhao et al., 2008).

As Tsc1 or Tsc2 (also named gigas, gig) homozygous mutants died at early larval stages (Ito and Rubin, 1999; Tapon et al., 2001), we used the Flipase (FLP)-mediated mitotic recombination technique to generate marked mutant GSCs and then studied their function in adult GSCs (Xu and Rubin, 1993). The marked wild-type or mutant GSCs were induced by heat-shock treatments, and were recognized by their position, the absence of lacZ expression (Fig. 1B-J, green), and the presence of spherical-shaped fusomes, a germline-specific organelle in cytoplasm (recognized by anti-α-spectrin, red) (Lin et al., 1994). Tsc1^Q87X, Tsc1^R453X, Tsc1²⁹ and gig¹⁹² alleles were used, all of which are strong or genetic null alleles (Ito and Rubin, 1999; Tapon et al., 2001). Flies of appropriate genotypes were subjected to time-course clonal analysis after clone induction (ACI) as previously described (Xie and Spradling, 1998).

At day 4 ACI, marked GSCs and follicle cell clones were readily observed, with 31%, 30%, 32%, 33% and 34% of germaria containing marked GSCs, respectively for wild-type, Tsc1^Q87X, Tsc1^R453X, Tsc1²⁹ and gig¹⁹² clones. The follicle cells in mutant clones of Tsc1 and gig showed increased cell sizes (Fig. 5E and data not shown), consistent with their roles in cell growth control. However, the size of mutant GSCs did not show obvious enlargement (Fig. 1E-J). After a week, there was no significant drop in the percentage of germaria with wild-type GSC clones (Fig. 1C,D; Table 1), which is consistent with previous observations that wild-type GSCs have a slow turn over rate (Xie and Spradling, 1998). However, Tsc1 and gig mutant GSCs showed rapid and progressive loss from their niches during this short period, with only about 2% to 3% of germaria containing marked GSCs for all mutant alleles at day 11 ACI (Fig. 1E-J; Table 1). Tsc1 and gig mutant germline cysts and egg chambers were readily observed and seemed to be able to differentiate normally, as 15 nurse cells and one oocyte were developed within each developing germline cyst, although their sizes were larger compared with wild-type germline cysts and egg chambers at the same developmental stages (Fig. 5A,C; see Fig. S1 in the supplementary material), suggesting that TSC1/2 also controls the growth of developing germline cysts. By calculating the number of germline cysts produced by Tsc1 mutant GSCs that were still maintained at day 8 ACI, we found that Tsc1 mutant GSCs have a similar cell division rate compared with wild-type GSCs (see Fig. S2 in the supplementary material).

The presence of differentiated germline clones in Tsc1/2 mutants implies that the loss of Tsc1/2 mutant GSC is caused by the commitment to differentiation without self-renewal, rather than to cell death. To verify this hypothesis, we performed TUNEL labelling to detect apoptosis in Tsc1 mutant GSCs. None of the Tsc1^Q87X (n=106) and Tsc1^R453X mutant GSCs (n=98) examined were positive for TUNEL (Fig. 1B); thus, the mutant GSCs are not apoptotic. Potentially, GSC loss could also be the consequence of out-competition by the neighboring wild-type GSCs, where cadherin-mediated cell adhesion plays a crucial role (Jin et al., 2008). Nevertheless, there was no obvious difference in DE-cadherin level at the stem cell/niche cell interface between Tsc1 mutant and wild-type GSCs (Fig. 2A). Besides, when a niche contains only Tsc1 mutant GSCs, they still differentiated into germline cysts (Fig. 2B), suggesting that Tsc1 mutant GSC loss is not caused by cell competition. Therefore, TSC1/2 functions cell-autonomously for GSC maintenance by preventing precocious differentiation.

As the BMP-Bam pathway is known to keep GSCs in undifferentiated state, we investigated whether BMP signaling is compromised in Tsc1/2 mutant GSCs. We used antibodies against phosphorylated Mad (pMad) to detect BMP pathway activity in Tsc1 mutant GSCs. Normally, high levels of nuclear pMad are detected in GSCs but not in the differentiating cystoblasts or early-stage cystocytes. Strikingly, pMad level was significantly decreased in Tsc1^Q87X [86% of mutant clones showed downregulation (32/37)] and Tsc1^R453X [73% (19/26)] mutant GSCs compared with the neighboring wild-type GSCs (Fig. 3A), indicating that TSC1/2 is required for BMP signaling activity in GSCs. We then tested whether TSC1/2 could also affect BMP signaling in other tissues. In imaginal wing discs, Dpp is expressed at the AP boundary and functions as a morphogen to pattern the AP axis. Tsc1 mutant clones in the wing discs showed similar levels of pMad expression compared with wild-type cells (see Fig. S3A in the supplementary material). Similarly, pMad expression was not altered in Tsc1 mutant clones in the eye discs (see Fig. S3B in the supplementary material). These observations suggest that TSC1 is required for BMP signaling specifically in GSCs.

To further test whether the reduced BMP signaling in Tsc1 mutant GSCs could cause bam upregulation and subsequently GSC differentiation, we examined bam transcription using a GFP reporter driven by the bam promoter (bam-GFP) (Chen and McKearin, 2003b). bam-GFP can reliably reflect bam transcription level: GFP is undetectable in GSCs, whereas it is detectable in cystoblasts, will reach a high level in early-stage cystocyts and then decrease in the late-stage cysts (Chen and McKearin, 2003b). Interestingly, in Tsc1^Q87X mutant GSCs, there was no obvious upregulation of bam-GFP (Fig. 3B) (38 out of 40 GSCs examined) compared with neighboring wild-type GSCs. For further confirmation, we also examined Bam protein expression in Tsc1 mutant GSCs. Normally, Bam protein is undetectable is the cytoplasm of GSCs, but accumulates as cystoblast differentiate, reaches highest level in eight-cell cysts, and then disappears dramatically in 16-cell cysts (McKearin and Ohlstein, 1995). Similarly, we did not observe Bam expression in all Tsc1 mutant GSCs examined (Tsc1^Q87X, n=50; Tsc1^R453X, n=50) (Fig. 3C). In addition, both bam-GFP and Bam protein expression pattern seemed to be normal in Tsc1 mutant cysts (Fig. 3B,C and data not shown), indicating that Tsc1 mutation did not affect bam expression in the differentiating germ cells either. Therefore, Tsc1 mutant GSCs display compromised BMP signaling activities, but bam transcription seems to be properly repressed.

To evaluate further the contribution of BMP-Bam pathway on Tsc1/2 mutant GSC loss, we tested whether bam elimination could rescue the precocious differentiation of Tsc1/2 mutant GSCs by generating Tsc1 bam double mutant (Tsc1²⁹ bam^Δ86) GSC clones.

As previously reported, bam elimination in wild-type GSCs (bam^Δ86, a null allele) completely blocked differentiation, leading to the development of GSC-like tumours (McKearin and Ohlstein, 1995). We observed that the mutant GSCs were well maintained in the niche, consistent with a previous observation that bam mutant GSCs are very competitive in occupying the niche (Fig. 4A,B,E) (Jin et al., 2008). Intriguingly, many Tsc1 bam double mutant germ cells showed enlarged and branched fusomes (Fig. 4D). After staining with anti-Orb, an oocyte marker that is initially detected in the cytoplasm of eight-cell or early 16-cell cystocytes (Lantz et al., 1994), and with HtsRC, a ring canal marker that starts to be expressed at around 16-cell stage (Robinson et al., 1994) (Fig. 4F), we found that the expression of both markers was not observed in Tsc1 bam double mutant germline clones (Fig. 4G), suggesting that the double mutant germ cells could only undergo limited differentiation into early cystocytes. Moreover, bam mutation could not rescue the loss of Tsc1 mutant GSCs, as double mutants still displayed the loss of GSC phenotype, although their loss was delayed compared with that of Tsc1 mutant GSCs (Fig. 4E). Taken together, we conclude that, although required for BMP signaling, TSC1/2 also represses GSC differentiation independently of bam; this might represent the primary mechanism by which TSC1/2 maintains GSCs.

It is well established that TORC1 functions downstream of TSC1/2 in controlling cell growth and metabolism. To test whether TSC1/2 also represses GSC differentiation through TORC1, we treated flies with rapamycin, a specific TORC1 inhibitor. The follicle cells from mutant flies treated with rapamycin were comparable in size with wild-type follicle cells (Fig. 5F). However, follicle cells in untreated Tsc1 mutant clones were enlarge, demonstrating that Tsc1 mutant cell growth was effectively suppressed by rapamycin treatment. Strikingly, the loss of GSC phenotype was also significantly rescued. After rapamycin treatment, more Tsc1 mutant GSCs were maintained from day 4 to day 11 ACI [97% (0.38/0.39) for Tsc1^Q87X and 65% (0.26/0.40) for Tsc1^R453X], compared with fewer than 1% (0.03/0.32 and 0.02/0.30 respectively) in controls (Table 1, Fig. 5A-D).

To further confirm the specificity of rapamycin in inhibiting TORC1 activity, we tested whether eliminating one copy of Tor could help to alleviate GSC loss in Tsc1 mutants. When using a low concentration of rapamycin, Tsc1^R453X mutant GSCs with only one copy of wild-type Tor were better maintained than those with two copies (Table 2). Thus, the loss of GSC phenotype caused by Tsc1/2 mutations is probably due to TORC1 hyperactivation. As S6K is a well-studied target of TORC1 that can be phosphorylated by TORC1 in order to regulate protein synthesis, we asked whether mutation in S6k could rescue the loss of GSCs in Tsc2 mutants. S6k mutant (S6k^I-1, a null allele) GSCs did not show any obvious defects in GSC maintenance (Table 1), although the mutant follicle cells were relatively smaller in size (not shown), which is consistent with its role in the control of cell growth (Montagne et al., 1999). Strikingly, gig S6k double mutant (S6k^I-1 gig⁵⁶, both null alleles) GSCs were properly maintained, with about 80% (0.33/0.41) of mutant GSCs maintained from day 4 to day 11 ACI, compared with fewer than 1% (0.02/0.34) of gig mutant GSCs maintained (Fig. 5H and Table 1). Similarly, S6k mutation also efficiently suppressed the overgrowth of both germline cysts and somatic follicle cells (Fig. 5G,H). Therefore, TORC1 functions downstream of TSC1/2 in both maintaining GSCs and controlling cell growth in the Drosophila ovary.

As previously stated, Tsc1 mutant GSCs showed reduced BMP signaling activity. We therefore asked whether TORC1 inhibition could restore the activity back to its normal level, in Tsc1 mutant GSCs. After rapamycin treatment, Tsc1 mutant GSCs were properly maintained and levels of pMad expression were also comparable with those in neighboring wild-type GSCs (Fig. 5I). This suggest that TSC1/2 also functions by inhibiting TORC1 in order to maintain BMP signaling activity in GSCs. Consistent with the notion that TSC1/2 controls both bam-dependent and -independent differentiation pathways via TORC1, rapamycin treatment prevented differentiation of Tsc1 bam double mutant GSCs, as branched fusomes in the mutant germ cells were no longer observed in the treated flies (Fig. 5J).

To test whether TOR is required to negatively regulate BMP signaling activity in GSCs, we treated wild-type flies (w¹¹¹⁸) with a high concentration of rapamycin to repress TORC1 activity and examined its effect on BMP signaling activity. Late oogenesis from treated females was blocked because few late-stage egg chambers or eggs were observed in the treated ovaries, suggesting that the rapamycin treatment inhibits ovary growth. In the germaria from treated flies, GSCs and their niches seemed to be morphologically normal and no obvious defect in GSC maintenance was found (Fig. 6A,B). Similar to GSCs in untreated flies, pMad was specifically enriched in the anterior tip of GSCs, but not in differentiating germ cells (Fig. 6A). Consistently, bam-GFP was fully repressed in GSCs, and derepressed in cystoblasts and in differentiating germline cysts (Fig. 6B).

Interestingly, there were more cystoblasts accumulated in the germarium of the treated flies than in untreated flies (Fig. 6B), indicating that rapamycin treatment could cause delay in cystoblast differentiation. We then examined Tor mutant flies to test directly the functional requirement for TOR in germline differentiation. Reducing TOR function by Tor^P1 mutation, which carries a P-element insertion at the 5′-UTR that causes reduced Tor expression (Zhang et al., 2000), did not significantly affect germline differentiation. The mutant germline cysts showed relatively normal morphology (Fig. 6C), although the development of mutant egg chambers was arrested in mid-oogenesis (Fig. 6D). The daughters generated by Tor^ΔP GSCs, a null allele of Tor, could also differentiate into cysts, but the mutant cysts soon arrested in growth and degenerated (Fig. 6E; data not shown). These observations indicate that TOR is not required for germline cyst differentiation, but may be essential for germline growth and survival. Consistent with the notion that TOR is not required for germline differentiation, S6k mutant GSCs were also able to produce daughters that could properly differentiate into germline cysts and egg chambers (Fig. 6F). Therefore, the accumulated cystoblasts in rapamycin-treated flies could be caused by non-specific effects of rapamycin. We conclude that TOR is not required for the normal BMP signaling pattern in the germline or for germline differentiation; however, it essential for germline growth and survival.

TSC1 and TSC2, the gene products of tuberous sclerosis complex human disease, are evolutionarily conserved cell growth regulators in metazoans. In this study, we have illustrated that, in the Drosophila ovary, in addition to controlling the growth of germline cysts and somatic follicle cells, TSC1/2 also regulates GSC maintenance. Clonal analysis demonstrates that TSC1/2 functions intrinsically in GSCs for their maintenance by preventing precocious differentiation. Mechanistically, TSC1/2 maintains BMP signaling activity and represses differentiation programs independently of bam in the GSCs, via inhibition of TORC1-S6K. Therefore, this study reveals a novel and potentially general mechanism of stem cell maintenance by TSC1/2, which may also help in better understanding the pathology of tuberous sclerosis complex and other human diseases associated with TOR dysregulation.

TSC1/2 is known to regulate cell growth via inhibition on TORC1. In this study, we have demonstrated that it also functions by inhibiting the activity of TORC1 to maintain GSCs. Treatment with rapamycin, a TORC1-specific inhibitor, can completely rescue GSC loss in Tsc1 mutants. In addition, eliminating S6K, which functions downstream of TORC1 in regulating protein translation, could also completely rescue GSC loss in Tsc2 mutants. Interestingly, the daughters of Tor mutant GSCs can differentiate into germline cyst properly, indicating that TOR is normally not required for differentiation, but its hyperactivation in Tsc1/2 mutants drives precocious GSC differentiation. The simplest explanation of the delayed cystoblast differentiation in rapamycin-treated females might be a non-specific effect of drug treatment. However, it is also possible that TORC1 inhibition by rapamycin might cause repression of some, but not all, aspects of TOR function, which leads to uncoordinated development and/or differentiation of cystoblasts in response to GSC division. Consistently, we also observed accumulated cystoblasts when overexpressing both Tsc1 and Tsc2 in the germline (data not shown). Together with the observation that TSC1/2-TORC1 signaling controls cell growth of germline cysts, our study suggests that TSC1/2-TORC1 may serve as a signaling integration point that orchestrates germline division, differentiation and development in order to control egg production in response to the local micro-environment and the system environment of the animals.

In the Drosophila ovary, BMP signaling from the niche directly suppresses bam expression in GSCs to prevent differentiation. This signaling is crucial for GSC maintenance. As revealed by pMad expression, BMP signaling activity is significantly downregulated in Tsc1 mutant GSCs. We also demonstrated that downregulation of pMad in Tsc1 mutant GSCs is mediated by TORC1 hyperactivation, as rapamycin treatment is able to restore the downregulated pMad level. However, TOR is not required for proper BMP signaling activity because pMad expression is not altered in rapamycin-treated germaria. Therefore, only TORC1 hyperactivation could inhibit BMP signaling in GSCs through unknown mechanisms, and this inhibitory effect occurs specifically in GSCs, as BMP signaling activity is not altered in Tsc1 mutant imaginal disc cells.

Logically, bam expression could be derepressed in Tsc1 mutant GSCs as a consequence of BMP pathway downregulation. Surprisingly, we could not detect significant upregulation of bam-GFP expression in mutant GSCs, although in other GSCs that were compromised by BMP signaling, such as tkv mutant and mad mutant GSCs, bam transcription is significantly upregulated (Song et al., 2004). Nevertheless, there might still be residual BMP signaling activities in Tsc1/2 mutant GSCs that are sufficient to suppress bam expression. Consistent with this notion is the observation that bam-GFP is not obviously upregulated in aged GSCs, even if BMP signaling activity has been significantly reduced (Zhao et al., 2008). Together with the observation that bam mutation could not rescue the differentiation of Tsc1 mutant germ cells, we suggest that the compromised BMP signaling activity may not be primarily responsible for Tsc1/2 mutant GSC loss. It is not clear why the effect of TSC1/2 on BMP signaling occurs specifically in GSCs. Possibly, Tsc1/2 mutant GSCs, once induced, have already primed for differentiation through a Bam-independent mechanism, which may trigger a positive feedback signal to inhibit BMP signaling activity, in order to facilitate differentiation.

Our study also reveals a BMP-Bam-independent mechanism that probably underlies the major role of TSC1/2-TORC1 signaling in GSC maintenance. The phenotype of Tsc1 bam double mutant germ cells differs from the bam alone mutant germ cells, as the double mutant GSCs can still become lost from the niche over time and undergo further differentiation into early cystocytes. Interestingly, the phenotype of Tsc1/2 mutant GSCs is similar to that of pelota (pelo) mutants. Pelo encodes a translational release factor-like protein and may regulate GSC maintenance at the translational level (Xi et al., 2005). In pelo mutant GSCs, there is also a downregulation of BMP signaling but no obvious upregulation of bam expression, and bam pelo double mutant germ cells are able to undergo similar limited differentiation into cystocytes, suggesting that TSC1/2 and Pelo might function in the same or parallel pathway to control GSC differentiation. We propose, as illustrated in Fig. 7, that similar to Pelo, TSC1/2 might function in a parallel pathway with the BMP-Bam pathway to control GSC differentiation, possibly by regulating the translation of differentiation-related mRNAs.

Pum and Nos, which are known to function together to repress translation of the target mRNAs in embryos, are also essential for GSC maintenance (Forbes and Lehmann, 1998; Wang and Lin, 2004). Recent genetic and biochemical studies suggest that Bam/Bgcn may directly inhibit the function of Pum/Nos to allow cystoblast differentiation (Chen and McKearin, 2005; Kim et al., 2009; Li et al., 2009; Szakmary et al., 2005). However, BMP signaling activation is able to prevent differentiation of nos mutant primordial germ cells, indicating that Pum/Nos could also function in parallel with the BMP-Bam pathway to control germ cell differentiation (Gilboa and Lehmann, 2004). In the future, it would be important to determine the functional relationships between the TSC1/2-TORC1 pathway, Pelo and Pum/Nos in regulating GSCs, and whether these factors, together with the microRNA pathway, target similar mRNAs to control GSC differentiation.

In this study, we have identified a novel role of TSC1/2 in controlling GSC maintenance and differentiation in the Drosophila ovary. Increasing evidence also suggests similar roles for TSC1/2-TOR signaling in regulating adult stem cell differentiation in mammals. For example, TSC1/2-mTOR signaling is also required for maintaining the quiescence of haematopoietic stem cells (HSCs), as Tsc1 deletion drives HSCs from quiescence to rapid cycling, which compromises HSC self-renewal (Chen et al., 2008). Thus, TSC1/2-TOR signaling could have an evolutionarily conserved role in regulating stem cell maintenance and differentiation from Drosophila to mammals.

We thank D. Chen, I. Hariharan, E. Laufer, T. Neufeld, G. Thomas, T. Xu, T. Xie, the Bloomington Stock Center and Developmental Studies Hybridoma Bank for fly stocks and reagents, members of the Xi laboratory for helpful comments and stimulating discussions, and S. Wang for help with manuscript preparation. This work was supported by the Chinese Ministry of Science and Technology.