Bmp signals from niche cells directly repress transcription of a differentiation-promoting gene, bag of marbles, in germline stem cells in the Drosophila ovary

Stem cells in adult tissues have the ability to self-renew and generate differentiated cells that maintain tissue homeostasis. Specific regulatory microenvironments, also known as niches, are thought to regulate many stem cell types by producing signals important for stem cell proliferation and differentiation (Watt and Hogan,2000; Spradling et al.,2001). Stem cells usually divide asymmetrically to generate parent stem cells and differentiated cells, although they can also undergo symmetric cell division to replenish lost stem cells or expand the stem cell pool(Xie and Spradling, 2000; Zhu and Xie, 2003). Even though many niche signals have been identified for different stem cell types in many systems, it is still largely unknown how niche signals control stem cell self-renewal and differentiation(Spradling et al., 2001). Therefore, it is essential to reveal the link between niche signals and intrinsic factors that are essential for stem cell self-renewal and differentiation in order to gain a better understanding of how stem cell behavior is controlled.

The Drosophila ovarian germline stem cells (GSCs) have become an attractive system to study stem cells and their relationship with niches(Xie and Spradling, 2001; Lin, 2002). Two or three GSCs are located at the tip of the ovariole, also known as the germarium, and are surrounded by terminal filament cells, cap cells and inner sheath cells that form a niche for GSCs. GSCs and their progeny in the germarium can be reliably identified and distinguished by a germ cell-specific structure, called the fusome, which is rich in membrane skeletal proteins, such as Hu li tai shao(Hts) and α-Spectrin (Lin et al.,1994; de Cuevas et al.,1997). In GSCs and their immediate differentiating daughters,cystoblasts, the fusome is spherical in shape, and is also known as the spectrosome. A cystoblast will undergo synchronous mitotic divisions with incomplete cytokinesis to generate two-, four-, eight- and sixteen-cell cysts,in which the fusome is branched to interconnect individual cystocytes(Lin et al., 1994). GSCs are invariably anchored to cap cells through adherens junctions(Song et al., 2002). Loss of adherens junctions between cap cells and GSCs causes GSCs to migrate away from cap cells and undergo differentiation(Song et al., 2002). Upon GSC division, the original GSC remains anchored to cap cells and retains stem cell identity, whereas the cystoblast moves away from cap cells and undergoes differentiation. As a GSC is lost, a neighboring GSC can generate two daughter cells that both contact cap cells and remain as GSCs, thus replenishing a vacant niche space (Xie and Spradling,2000). Functioning as a GSC niche, terminal filament/cap cells express piwi, dpp, fs(1)Yb (also known as Yb) and hedgehog (hh), which are essential for maintaining GSC asymmetric cell division (Xie and Spradling, 1998; King and Lin,1999; Cox et al.,2000; Xie and Spradling,2000; King et al.,2001). Intrinsic factors in GSCs, including pumilio, nanos,dpp receptors and downstream components, are also important for GSC maintenance (Lin and Spradling,1997; Forbes and Lehmann,1998; Xie and Spradling,1998). Two intrinsic factors, bag of marbles(bam) and benign gonial cell neoplasm (bgcn), are required in cystoblasts for their proper differentiation(McKearin and Spradling, 1990; McKearin and Ohlstein, 1995; Lavoie et al., 1999). However,in GSCs the interplay between genes involved in self-renewal versus differentiation remains unclear.

The functions of dpp signaling and bam in the maintenance of GSCs and the differentiation of cystoblasts seem to be directly opposing. Loss of bam function completely eliminates cystoblast differentiation, similar to that caused by dpp overexpression(McKearin and Spradling, 1990; Xie and Spradling, 1998). By contrast, forced overexpression of bam in GSCs causes their elimination, similar to that observed when dpp signaling is disrupted in GSCs (Ohlstein and McKearin,1997; Xie and Spradling,1998). These observations can be explained by a simple model wherein dpp, functioning as a short-range signal, directly promotes GSC self-renewal and suppresses bam expression in GSCs, while allowing cystoblasts to express bam and differentiate.

Several studies have supported this model(Xie and Spradling, 1998; Chen and McKearin, 2003a; Kai and Spradling, 2003). bam mRNA is absent in GSCs, but quickly accumulates in cystoblasts and mitotic cysts (McKearin and Spradling,1990). Overexpression of dpp completely suppresses the expression of BamC protein in germ cells, thus preventing cystoblasts from differentiating (Xie and Spradling,1998). A recent study by Kai and Spradling showed that dpp signaling activity is restricted to GSCs and cystoblasts(Kai and Spradling, 2003).

The asymmetric distribution of bam between GSCs and cystoblasts could be due to transcriptional regulation and/or mRNA stability. The recent elegant bam promoter analysis has revealed that its transcription is actively repressed through a silencer (Chen and McKearin, 2003a). However, whether and how dppsignaling directly represses bam transcription remains unknown. In this study, we provide genetic and molecular evidence to support the model proposing that Bmp signaling represses bam transcription through binding of its downstream transcriptional effectors, Mad and Medea (Med), to the defined bam silencer.

The following fly stocks used in this study were described either in FlyBase or as otherwise specified in the Results section:

• punt¹⁰⁴⁶⁰;

• punt¹³⁵;Med²⁶; Dad-lacZ;dpp^hr4;

• dpp^hr56; gbb⁴, gbb^D4;

• gbb^D20; bam-GFP (GFP gene driven by the bam promoter);

• vasa-GFP; c587-gal4;

• hs-gal4; UAS-dpp; and

• UAS-gbb; hsFLP; FRT_82Barmadillo-lacZ.

Most stocks were cultured at room temperature. To maximize their mutant phenotypes, dpp, gbb and punt mutant adult females were cultured at 29°C for 2-7 days. To achieve the uniform GSC-like phenotype, c587-gal4;UAS-dpp females were also cultured at 29°C for 7 days.

Clones of mutant GSCs were generated by Flp-mediated mitotic recombination,as described previously (Xu and Rubin,1993; Xie and Spradling,1998). To generate the stocks for making mutant GSC clones and examining bam-GFP expression, 2-day old hsFLP; bam-GFP/+;FRT_82Bpunt¹³⁵/FRT_82Barmadillo-lacZ and hsFLP; bam-GFP/+;FRT_82BMed²⁶/FRT_82Barmadillo-lacZfemales were heat-shocked at 37°C for 3 consecutive days with two one-hour heat-shock treatments separated by 8-12 hours. The ovaries were removed 3 days after the last heat-shock treatment and then processed for antibody staining.

To construct the stocks for overexpressing dpp or gbb,the females that carried hs-gal4, and either UAS-dpp or UAS-gbb, were heat-shocked at 37°C for different lengths of time for particular experiments, as indicated in the Results. The females that carried c587-gal4 and UAS-dpp or UAS-gbb were cultured at room temperature or at 29°C for 7 days. For examining the expression of bam-GFP in the ovary overexpressing dpp or gbb, the females that carried c587-gal4 or hs-gal4,and UAS-dpp or UAS-gbb, also carried a bam-GFPtransgene.

To measure stem cell loss in gbb mutant and control ovaries, the germaria with different numbers of GSCs, ranging from three to none, were counted from the ovaries of 2-day- and one-week-old bam-GFP gbb⁴/gbbD4, bam-GFP gbb⁴/gbb^D20 or bam-GFP (control) females. The 2-day-old control and gbbmutant females were cultured at room temperature after they eclosed at 18°C, whereas the one-week-old control and gbb mutant females were cultured at 29°C. Values are expressed as the average GSC number per germarium and the percentage of germaria with no GSCs.

To examine bam-GFP expression in dpp, gbb or punt mutant germaria, we generated females with the following genotypes at 18°C: bam-GFP gbb⁴/gbb^D4, bam-GFP gbb⁴/gbb^D20, bam-GFP dpp^hr56/dpp^hr4, bam-GFP;punt¹⁰⁴⁶⁰/punt¹³⁵ or bam-GFP (control) females. All the control and mutant females were cultured at 29°C for 4 days before their ovaries were isolated and immunostained, to compare bam-GFP expression under identical conditions.

The following antisera were used: polyclonal anti-Vasa antibody (1:2000)(Liang et al., 1994);monoclonal anti-Hts antibody (1:3); polyclonal anti-β-galactosidase antibody (1:100; Cappel); polyclonal anti-GFP antibody (1:200; Molecular Probes); and polyclonal anti-pMad antibody (1:200)(Tanimoto et al., 2000). The immunostaining protocol used in this study was described previously(Song et al., 2002). All micrographs were taken using a Leica SPII confocal microscope.

Total RNA from the ovaries of different genotypes or treatments was isolated using Trizol (Invitrogen), and biotin-labeled cRNA probes were produced using an RNA transcript labeling kit (Enzo BioArray). The Drosophila GeneChips were purchased from Affymetrix, and were hybridized, stained and detected according to the manufacturer's instructions.

After sorting GFP-positive cells by using Cytomation MoFlo, total RNA was prepared using Trizol (Invitrogen) from these isolated cells. The RNA samples were further amplified using the GeneChip Eukaryotic Small Sample Target Labeling Assay Version II (Affymetrix). After the RNA amplification, 100 ng of total RNA was reverse-transcribed (RT) using the SuperScriptIII First-Strand Synthesis System for RT-PCR, according to manufacturer's protocol(Invitrogen). The following primers were used in this study:

• dpp, 5′-AGCCGATGAAGAAGCTCTACG-3′ and 5′-ATGTCGTAGACAAGCACCTGGTA-3′;

• vasa, 5′-ATCGAGGAGGAAATCGAGATGGA-3′ and 5′-GGAAGCTATGCCACTGCTGAATA-3′;

• gbb, 5′-AGATGCAGACCCTGTACATAGAC-3′ and 5′-CTCGTCGTTCAGGTGGTACAGAA-3′; and

• rp49, 5′-GTATCGACAACAGAGTCGGTCGC-3′ and 5′-TTGGTGAGCGGACCGACAGCTGC-3′.

PCR was performed as follows: 94°C for 2 minutes; 35 cycles of 94°C for 30 seconds, 45°C for 30 seconds and 72°C for 45 seconds; and 72°C for 7 minutes. RT-PCR products were electrophoresed on a 2% agarose gel in the presence of ethidium bromide.

The GST-Mad construct was described previously(Kim et al., 1997). Med was PCR-amplified from its cDNA, with the introduction of XhoI sites at both ends, then subcloned into a pGEX-4T2 vector (Amersham Pharmacia Biotech). Its sequence was confirmed by sequencing. GST-Mad, GST-Med and GST proteins were purified by affinity chromatography using Glutathione Sepharose™ 4B according to the manufacturer's protocol (Amersham Pharmacia Biotech), and confirmed by western blots.

A Cy5 5′-modified oligonucleotide containing the bipartite bam silencer element (+17 to +54) was used as a probe. Binding reactions were performed according to the published protocol(Kim et al., 1997). Specificity of binding was determined by the addition of 100-fold molar excess of unlabeled competitor DNA corresponding to the bam silencer element, with site A and/or B. DNA-protein complexes were resolved on a 5%(w/v) non-denaturing polyacrylamide gel using 0.5×TBE running buffer at 150 V for 3 hours at 4°C. Gels were imaged on a Typhoon 8700 (Amersham Biosciences).

Previous studies have shown that a dpp signal produced by somatic cells is essential for maintaining GSCs but not for cystoblast development(Xie and Spradling, 1998; Xie and Spradling, 2000). bam transcription is active in young, differentiating germ cells but is repressed specifically in GSCs in the ovary(Chen and McKearin, 2003a). This raises the interesting possibility that dpp signaling and bam expression directly oppose each other. To investigate this possibility, we examined the correlation between dpp signaling activity and bam expression in the germarium. dpp signaling activity is usually monitored by Dad and phosphorylated Mad (pMad)expression (Tsuneizumi et al.,1997; Tanimoto et al.,2000). Dad is a dpp target gene, and a Dad-lacZ line recapitulates its expression(Tsuneizumi et al., 1997). A bam-GFP transgene (with the GFP gene driven by the bampromoter) has been generated to study bam transcription(Chen and McKearin, 2003a). Throughout this study, an anti-Hts antibody was used to label spectrosomes and fusomes, and a DNA dye, DAPI, was used to label nuclei. Cap cells can be reliably identified by bright DAPI staining, and by their unique position and nuclear morphology. GSCs are identified by the presence of a spectrosome (a spherical fusome) on their anterior side and by their direct contact with cap cells; cystoblasts also contain a spectrosome but fail to be associated with cap cells (Fig. 1A).

Similar to the observations recently made by Kai and Spradling(Kai and Spradling, 2003), Dad-lacZ was expressed in GSCs and some cystoblasts at high levels,but in the other cystoblasts and mitotic cysts at much lower levels(Fig. 1B). As reported by Chen and McKearin (Chen and McKearin,2003a), bam transcription was repressed in GSCs and some cystoblasts, but was active in the other cystoblasts and dividing cystocytes(Fig. 1C). In germaria carrying bam-GFP and DadlacZ, GSCs and the cystoblasts that had strong Dad expression did not show bam-GFP expression(Fig. 1D), whereas the cystoblasts and mitotic cysts that had weak or no Dad expression showed obvious bam-GFP expression(Fig. 1D). Similarly, GSCs and the cystoblasts that showed strong pMad expression did not express bam-GFP, whereas the cystoblasts and mitotic cysts that showed weak or no pMad expression expressed bam-GFP(Fig. 1E,F). These results further support the idea that the dpp signaling pathway is activated in GSCs at high levels, whereas bam transcription is actively repressed.

The dpp^hr56/dpp^hr4temperature-sensitive mutant was chosen to investigate the expression of bam-GFP in dpp mutant GSCs because it shows gradual loss of GSCs within two weeks at a restrictive temperature (29°C)(Xie and Spradling, 1998). After bam-GFP and bam-GFP dpp^hr56/dpp^hr4 females were cultured at 29°C for 2, 4 or 7 days, the ovaries were immunostained with anti-GFP and anti-Hts antibodies to visualize bam-GFP and fusomes, respectively. In the germaria from the bam-GFP females, the bam-GFP expression pattern was completely normal, and was absent in GSCs even one week after being cultured at 29°C(Fig. 2A). However, even two days after being cultured at 29°C, 28% of the bam-GFP dpp^hr56/dpp^hr4 germaria that contained GSCs started to express bam-GFP in one or more GSCs(n=283; Fig. 2B). After 4 days and 7 days, 66% (n=35) and 89% (n=19) of the mutant germaria that still had at least one GSC expressed bam-GFP in one or more GSCs, respectively (Fig. 2C,D). To further confirm the role of dpp signaling in repressing bam transcription, we also compared the levels of bam mRNA in wild-type and dpp mutant ovaries using a microarray approach. bam mRNA was dramatically upregulated in dpp^hr4/dpp^hr56mutant ovaries in comparison with wild type(Table 1; samples were normalized with an internal control, the Actin 42A gene). These results demonstrate that the dpp signal is required to repress bam transcription in GSCs.

Next, we investigated whether elevated bam transcription in dpp mutant GSCs can be correlated with reduction of pMad expression. After 4 days at 29°C, control germaria maintained the normal number of GSCs and showed the normal pMad expression pattern(Fig. 2E,F). By contrast, many dpp mutant germaria completely lost their GSCs, and in the remaining GSC-containing germaria in which bam-GFP was also upregulated in GSCs, pMad was severely reduced but not completely eradicated in the GSCs(Fig. 2G,H). In the germaria in which bam-GFP was not obviously upregulated, levels of pMad were relatively higher but less than normal (data not shown). These results indicate that dpp signaling contributes, at least in part, to pMad production in GSCs, and could be responsible for repressing bamtranscription.

Our previous study showed that overexpression of dpp throughout the germarium completely inhibits cystoblast differentiation and causes the accumulation of GSC-like cells that fail to express BamC(Xie and Spradling, 1998). Our experiments described above suggest that the dpp signal is likely to be restricted to the tip of the germarium, adjacent to cap cells. To test whether GSCs are competent to respond to dpp signaling outside their niches, dpp was specifically overexpressed in somatic cells other than cap cells, using the c587-gal4 line to drive a UAS-dpptransgene. The c587-gal4 line can drive expression of a UAS-GFP transgene in inner sheath cells and early follicle cells(Fig. 3A). When UAS-dpp expression was driven in inner sheath cells and follicle cells, germaria were filled with single germ cells with a spectrosome,suggesting that germ cells distant from their niche are still capable of responding to dpp (Fig. 3B). To further test whether dpp overexpression is sufficient to inhibit bam expression, we examined bam-GFPexpression in dpp-induced GSC-like tumors. In dpp-overexpressing ovaries, bam-GFP was not expressed in the single germ cells either close to (Fig. 3C) or away from (Fig. 3D) the germarial tip. These results indicate that dppsignaling is sufficient to inhibit bam transcription.

The c587-gal4 driver is expressed in somatic cells during early gonadal development, and overexpression of dpp also inhibits germ cell differentiation at early developmental stages(Zhu and Xie, 2003). To exclude the possibility that early dpp overexpression produces abnormal GSCs whose progeny cannot differentiate normally and thus fail to express bam, we examined bam-GFP expression at the adult stage when dpp was overexpressed using UAS-dpp driven by the hs-gal4 driver (the promoter of a heat-shock protein 70 gene fused with the gal4 gene). Without any heat-shock treatments, all the germaria had the normal GSC number and the normal bam-GFP expression pattern (Fig. 3E). After three consecutive days of 2-hour heat-shock treatments, the anterior half of the germaria were filled with single spectrosome-containing germ cells, and showed no obvious bam-GFP expression(Fig. 3F). These results further support the idea that dpp signaling is sufficient for directly or indirectly repressing bam transcription. Owing to the fact that GFP protein is stable, we could not determine how fast dppoverexpression can diminish bam mRNA using the bam-GFPtransgene. Thus, we measured the quantity of bam mRNA 2 hours after a pulse of heat-shock-induced dpp overexpression using the microarray approach. Interestingly, 2 hours after a pulse of dpp overexpression, bam mRNA was below detection(Table 1), indicating that dpp signaling rapidly represses bam transcription and/or causes rapid degradation of bam mRNA. This result further suggests that dpp signaling might directly repress bamtranscription.

In addition to Dpp, another Bmp-like molecule, Glass bottom boat (Gbb),exists in Drosophila and resembles human BMPs 5, 6, 7 and 8(Wharton et al., 1991; Doctor et al., 1992). It has been shown that synergistic signaling by dpp and gbbcontrols wing growth and patterning in Drosophila(Haerry et al., 1998; Khalsa et al., 1998). To investigate the possibility that gbb could also be involved in the regulation of GSCs, we first used RT-PCR to determine whether gbbmRNA was present in different cell types of the germarium. Inner sheath cells and early follicle cells were isolated from c587-gal4;UAS-GFP females using fluorescent-activated cell sorting (FACS). Agametic ovaries were isolated from newly eclosed females that developed from ovo^D1rS1 homozygous embryos lacking germ cells(Oliver et al., 1990). The agametic ovary is composed of terminal filament cells, cap cells and early follicle cells but lacks inner sheath cells(Margolis and Spradling,1995). Single germ cells, resembling GSCs, were isolated from c587-gal4; vasa-GFP/UAS-dpp females using FACS. vasa is a germ cell-specific gene (Hay et al.,1988; Lasko and Ashburner,1988), and vasa-GFP is specifically expressed in the germ cells (Nakmura et al., 2001). dpp is expressed in the somatic cells of the germarium but not in germ cells(Xie and Spradling, 2000). vasa mRNA was present in germ cells but not in inner sheath cells and agametic ovaries (Fig. 4A),whereas dpp mRNA was present in inner sheath cells and agametic ovaries but not in germ cells (Fig. 4A), indicating that the different cell types in germaria were properly isolated. gbb mRNA was detected in inner sheath cells and agametic ovaries but not in the GSC-like germ cells(Fig. 4A), indicating that gbb is expressed in the somatic cells. These results indicate that gbb could be another somatic signal for controlling GSCs.

We next determined whether mutations in gbb cause GSC loss in the ovary. Two allelic combinations of gbb, bam-GFP gbb⁴/gbb^D4 and bam-GFP gbb⁴/gbb^D20,and a wild-type strain carrying bam-GFP were allowed to develop to adulthood at 18°C and were then shifted to room temperature or 29°C. The germaria from the wild-type females 2 days after being cultured at room temperature, or 7 days after being cultured at 29°C, had a normal number of GSCs, two or three GSCs (Fig. 4B). However, 2 days after being shifted to room temperature, the germaria from the gbb⁴/gbb^D4 and gbb⁴/gbb^D20 females had an average of 1.0 and 1.5 GSCs, respectively(Table 2). One week after being shifted to 29°C, 88% of the gbb⁴/gbb^D4 mutant germaria and 60% of the gbb⁴/gbb^D20 mutant germaria completely lost their GSCs in comparison with 36% and 7% 2 days after being cultured at room temperature, although the rest usually had one GSC left(Fig. 4C-E). These results demonstrate that gbb is essential for maintaining GSCs in the Drosophila ovary.

As dpp and gbb can function synergistically in other developmental processes, we examined whether bam-GFP expression was upregulated in gbb mutant germaria. As described earlier, bam-GFP was not expressed in GSCs in the wild-type females after being cultured either at room temperature or at 29°C(Fig. 2A). Two days after being cultured at room temperature, the GSCs rarely expressed bam-GFP in the mutant gbb⁴/gbb^D4 germaria(one out of the total 49 germaria) and in the gbb⁴/gbb^D20 mutant germaria (two out of the total 55 germaria)(Fig. 4F). By contrast, one week after being cultured at 29°C, most of the GSCs expressed bam-GFP in the gbb⁴/gbb^D4 (five out the six germaria carrying one or more GSCs) and gbb⁴/gbb^D20 (13 out of the 16 germaria carrying one or more GSCs) mutant germaria(Fig. 4G,H). These results demonstrate that gbb is also essential for repressing bamtranscription in GSCs.

Having established that dpp is sufficient to repress bamexpression in GSCs, we then asked whether gbb overexpression was sufficient to repress bam transcription in germ cells. Similarly, bam-GFP expression was studied in germaria overexpressing gbb using the C587 driver and the UAS-gbb transgene, which has been used to effectively overexpress gbb in the wing disc(Khalsa et al., 1998). The germaria overexpressing gbb had the normal number of GSCs and cysts(Fig. 4I), indicating that GSC maintenance and division, and germ cell differentiation, appeared to be normal. Similarly, the bam-GFP expression pattern was also normal in the gbb-overexpressing germaria(Fig. 4I). These results suggest that gbb overexpression, unlike that of dpp, is not sufficient to inhibit bam transcription.

It appears that gbb uses the same downstream components as dpp does in regulating wing development(Haerry et al., 1998; Khalsa et al., 1998). dpp signaling results in the production of pMad(Newfeld et al., 1997; Tanimoto et al., 2000). To investigate whether gbb is also involved in the production of pMad in GSCs, we examined pMad accumulation in gbb mutant GSCs, and the relationship between pMad accumulation and bam transcription. As expected, pMad and bam-GFP expression patterns in GSCs and cystoblasts remained normal four days after the control females were cultured at 29°C (Fig. 5A,A′). Four days after being cultured at 29°C, the expression of pMad in the GSCs in both gbb⁴/gbb^D4and gbb⁴/gbb^D20females was generally reduced (Fig. 5B-F′). Some of the mutant gbb germaria that had moderately reduced pMad expression in GSCs showed no bam-GFPexpression in GSCs (Fig. 5B-D′), whereas the germaria that had severely reduced levels of pMad in GSCs showed significant bam-GFP upregulation in GSCs (Fig. 5E-F′). There appeared to be a good correlation between levels of pMad and bam-GFPexpression in gbb mutant GSCs. These results indicate that gbb signaling also results in the phosphorylation of Mad and that levels of pMad in GSCs seem to correlate with levels of bamrepression.

punt encodes a type II serine/threonine kinase receptor for dpp and also possibly for gbb(Letsou et al., 1995; Ruberte et al., 1995). A temperature-sensitive punt allelic combination, punt¹⁰⁴⁶⁰/punt¹³⁵,can develop to adulthood at 18°C and exhibits mutant phenotypes at 29°C (Theisen et al.,1996). Newly eclosed punt¹⁰⁴⁶⁰/punt¹³⁵females at 18°C had a normal number of GSCs and a normal bam-GFPexpression pattern in their germaria (Fig. 6A). Some punt¹⁰⁴⁶⁰/punt¹³⁵mutant GSCs started to express bam-GFP two days after being shifted to 29°C (Fig. 6B). One week after being cultured at 29°C, the GSCs in 75% of the mutant germaria (a total of 97 germaria were examined) that still carried one or more GSCs had already expressed bam-GFP (Fig. 6C), and 53% of the mutant germaria (a total of 123 germaria were examined) had only one or no GSC (Fig. 6C,D). After four days at the restrictive temperature, pMad in most punt mutant GSCs was severely reduced and bam-GFP was upregulated (Fig. 6E-H). These results further show that defective Bmp signaling results in the derepression of bam transcription in GSCs and that levels of pMad are correlated with the repression status of bam transcription in GSCs.

So far, bam expression has been examined only in dpp, gbband punt mutant germaria in which Bmp signaling is defective in both somatic cells and germ cells. To determine whether direct Bmp signaling in GSCs is necessary for repressing bam transcription, we used the FLP-mediated FRT recombination technique to generate punt and Med mutant GSCs marked by loss of expression of the armadillo (arm)-lacZ transgene, and then examined bam-GFP expression in the marked mutant GSCs(Xu and Rubin, 1993; Xie and Spradling, 1998). Med encodes a common Smad 4 for Tgfβ-like signaling pathways,and Med²⁶ is a strong Med mutant(Das et al., 1998; Wisotzkey et al., 1998). 54%of the marked three-day old punt¹³⁵ GSCs expressed bam-GFP (a total of 37 marked GSC clones were examined),and 65% of the marked three-day old Med²⁶ mutant GSCs showed obvious bam-GFP upregulation (a total of 48 marked GSC clones were examined) (Fig. 6I-L). These results demonstrate that direct Bmp signaling is necessary for repressing bam transcription.

We have so far shown that Bmp signaling mediated by Dpp and Gbb is essential for repressing bam transcription in GSCs. This bamtranscriptional repression could be directly or indirectly controlled by Bmp signaling. As shown recently by Chen and McKearin, a silencer located at the 5′ UTR of the bam gene is both necessary and sufficient for repressing bam transcription in GSCs(Chen and McKearin, 2003a). In Drosophila, the brinker (brk) gene is actively repressed by dpp signaling through a transcriptional silencer(Campbell and Tomlinson, 1999; Jazwinska et al., 1999; Minami et al., 1999; Marty et al., 2000; Muller et al., 2003). Interestingly, bam and brk silencers show remarkably similar sequences: 13 out of 19 base pairs are identical in A and B sites(Fig. 7A). The brksilencer has been shown to be directly occupied by a complex containing Mad,Med and Schnurri (Shn), and its repression requires shn and functional dpp signaling (Muller et al., 2003). shn is known to be required in GSCs for their maintenance, and loss of shn function results in GSC loss(Xie and Spradling, 2000). All the evidence suggests that the bam silencer could be directly occupied by a complex containing Mad, Med and possibly Shn.

We performed electrophoretic mobility shift assays to test whether Med and Mad can bind directly to the bam silencer in vitro using a Cy5-labeled bam silencer element(Fig. 7B), and purified bacterially expressed GST-Mad and GST-Med(Fig. 7C). GST-Mad (a fusion between GST and the N-terminal DNA-binding domain and linker region of Mad)was shown to bind to the dpp responsive elements in vitro(Kim et al., 1997), whereas GST-Med is a fusion of GST with the full-length Med. Interestingly, both Mad and Med could bind to the silencer but with different affinities. It appeared that Med bound to the silencer with a higher affinity than Mad(Fig. 7D). The binding specificity of Mad and Med to the silencer was demonstrated by a competition experiment with an unlabelled DNA fragment containing A and B sites(Fig. 7B,D). The unlabeled DNA fragment containing either an A or a B site could almost completely compete for binding of Mad to the labeled silencer. However, the unlabeled DNA fragment with the A site, but to much less extent, with the B site could compete for binding of Med to the labeled silencer. These data suggest that Mad occupies both the A and B sites, whereas Med preferentially binds to the A site. pMad accumulates in the GSC nucleus at high levels(Kai and Spradling, 2003)(this study). As described earlier, Med is also required in GSCs for repressing bam transcription. These in vitro binding results suggest that a protein complex containing Mad and Med, stimulated by Bmp signaling,directly binds to the bam silencer to repress its transcription in GSCs.

Stem cells are located in a niche, which provides extracellular cues that control stem cell self-renewal, division and differentiation. Dpp is a signaling molecule that originates from the niche and is necessary for maintaining GSCs in the Drosophila ovary(Xie and Spradling, 1998; Xie and Spradling, 2000). This study has identified another Bmp-like molecule, Gbb, as an essential niche signal for maintaining GSCs. It is likely that Dpp and Gbb function cooperatively as short-range signals in the GSC niche, as their signaling activities, monitored by pMad and Dad expression, are restricted to GSCs and some cystoblasts. Previous studies have demonstrated that bam is known to be necessary and sufficient for cystoblast differentiation(McKearin and Spradling, 1990; Ohlstein and McKearin, 1997). This study shows that upregulation of bam in GSCs is associated with stem cell loss in dpp and gbb mutants. Here we propose a model to possibly explain how Bmp signaling directly represses bamtranscription in Drosophila ovarian GSCs(Fig. 7E). The Bmp signals from the GSC niche activate their signaling cascade in GSCs, which leads to Mad phosphorylation, and then the translocation of the Mad and Med complex into the nucleus, which probably directly binds to the bam promoter and represses bam transcription in GSCs. This study demonstrates that Bmp signals maintain GSCs, at least in part, by repressing bamtranscription in GSCs in the Drosophila ovary.

In this study, a new function of gbb in the regulation of GSCs in the Drosophila ovary is revealed. Loss of gbb function leads to GSC differentiation and stem cell loss, similar to dpp mutants. gbb is expressed in somatic cells but not in germ cells, suggesting that gbb is another niche signal that controls GSC maintenance. Like dpp, gbb contributes to the production of pMad in GSCs and also functions to repress bam expression in GSCs. As in the wing imaginal disc (Haerry et al., 1998; Khalsa et al., 1998; Ray and Wharton, 2001), gbb also probably functions to augment the dpp signal in the regulation of GSCs through common receptors in the Drosophila ovary. In both dpp and gbb mutants, pMad accumulation in GSCs is severely reduced but not completely diminished. As the dpp or gbb mutants used in this study do not carry complete loss-of-function mutations, it remains possible that complete elimination of either dpp or gbb function is sufficient for eradicating pMad accumulation in GSCs. Alternatively, both dpp and gbbsignaling are required independently for full pMad accumulation in GSCs, and thus disrupting either one of them only partially diminishes pMad accumulation in GSCs. The lethality of null dpp and gbb mutants, and the difficulty in completely removing their function in the adult ovary, prevent us from further testing these possibilities directly.

Interestingly, dpp overexpression results in complete suppression of cystoblast differentiation and complete repression of bamtranscription in the germ cells, whereas gbb overexpression does not have obvious effects on cystoblast differentiation or bamtranscription. Even though the UAS-gbb transgene and the c587 driver for gbb overexpression have been demonstrated previously to function properly (Khalsa et al., 1998; Kai and Spradling,2003; Zhu and Xie,2003), it is possible that active Gbb proteins are not produced in inner sheath cells and somatic follicle cells because of a lack of proper factors that are required for Gbb translation and processing in those cells,which could explain why the assumed gbb overexpression does not have any effect on cystoblast differentiation. However, as active Dpp proteins can be successfully achieved using the same expression method, and Dpp and Gbb are closely related Bmps, it is unlikely that active Gbb proteins are not produced in inner sheath cells and follicle cells. Alternatively, dpp and gbb signals could have distinct signaling properties, and dpp may play a greater role in regulating GSCs and cystoblasts. Recent studies have indicated that Dpp and Gbb have context-dependent relationships in wing development (Ray and Wharton, 2001). In the wing disc, duplications of dpp are able to rescue many but not all of the phenotypes associated with gbbmutants, suggesting that dpp and gbb have not only partly redundant functions but also distinct signaling properties. In the wing and ovary, gbb and dpp function through two Bmp type I receptors, sax and tkv(Khalsa et al., 1998; Xie and Spradling, 1998) (this study). The puzzling difference between gbb and dpp could be explained by context-dependent modifications of Bmp proteins, which render them different signaling properties in different cell types. It will be of great interest to better understand what causes Bmps to have distinct signaling properties in the future.

All the defined niches share a commonality, structural asymmetry, which ensures stem cells and their differentiated daughters receive different levels of niche signals (Watt and Hogan,2000; Spradling et al.,2001). In order for a niche signal to function differently in a stem cell and its immediately differentiating daughter cell that is just one cell away, it has to be short-ranged and localized. This study, and a recent study by Kai and Spradling (Kai and Spradling, 2003), show that Bmp signaling mediated by Dpp and Gbb results in preferential expression of pMad and Dad in GSCs. In this study, we show that Bmp signaling appears to elicit different levels of responses in GSCs and cystoblasts, suggesting that the cap cells are likely to be a source for active short-ranged Bmp signals. These observations support the idea that Bmp signals are only active around cap cells. Consistently, when GSCs lose contact with the cap cells following the removal of adherens junctions they move away from the niche and then are lost(Song et al., 2002). As gbb and dpp mRNAs are broadly expressed in the other somatic cells of the germarium besides cap cells, localized active Bmp proteins around cap cells could be generated by localized translation and/or activation of Bmp proteins. As they can function as long-range signals(Podos and Ferguson, 1999), it remains unclear how Dpp and Gbb act as short-range signals in the GSC niche.

Previous studies, and this study, have shown that Bmp signaling and bam expression are directly opposing in Drosophila ovarian GSCs (McKearin and Spradling,1990; Ohstein and McKearin, 1997; Xie and Spradling, 1998). bam is actively repressed in GSCs through a defined transcriptional silencer (Chen and McKearin,2003a). These observations lead us to propose a model in which Bmp signals from the niche maintain adjacent germ cells as GSCs by actively suppressing bam transcription and thus preventing differentiation into cystoblasts (Fig. 7E).

In this study, we show that the levels of pMad are correlated with the amount of bam transcriptional repression in GSCs and cystoblasts. In the wild-type germarium, pMad is highly expressed in GSCs and some cystoblasts where bam is repressed. In other cystoblasts and differentiated germline cysts, pMad is reduced to very low levels, and thus bamtranscriptional repression is relieved. In the GSCs mutant for dpp,gbb or punt, pMad levels are severely reduced, and bambegins to be expressed. The repression of bam transcription as a result of dpp overexpression seems to be a rapid process as bam mRNA was reduced to below detectable levels two hours after dpp was overexpressed. This suggests that repression of bamtranscription by Bmp signaling could be direct. Furthermore, Med and Mad can bind to the defined bam silencer in vitro, which also supports the idea that Bmp signaling acts directly to repress bam transcription. Similar results have been obtained in a recent study(Chen and McKearin, 2003b). Similarly, Dpp signaling has been shown to repress brk expression in the wing imaginal disc and in the embryo(Campbell and Tomlinson, 1999; Jazwinska et al., 1999; Minami et al., 1999). The repression of brk expression by Dpp signaling is mediated by the direct binding of Mad and Med to a silencer element in the brkpromoter (Muller et al.,2003). As the brk silencer is very similar to the bam silencer, our results suggest that bam repression in GSCs is also mediated directly by Dpp and Gbb in a similar manner.

It remains unclear how the binding of Med and Mad to the bamsilencer results in bam transcriptional repression in GSCs. For the brk silencer, Dpp signaling and Shn are both required to repress brk expression in the Drosophila wing disc and embryo(Marty et al., 2000; Torres-Vazquez et al., 2001; Muller et al., 2003). Mad and Med belong to the Smad protein family, which are known to function as transcriptional activators by recruiting co-activators with histone acetyltransferase activity (reviewed by Massague and Wotton, 2000). In the wing disc, Shn is proposed to function as a switch factor that converts the activating property of Mad and Med proteins into a transcriptional repressor property (Muller et al.,2003). Possibly, the Mad-Med complex could also recruit Shn to the bam repressor element. Consistent with the possible role of Shn in repressing bam expression in GSCs is the observation that GSCs that lose shn function differentiate, and thus are lost(Xie and Spradling, 2000). Also, it remains possible that Mad and Med could recruit a repressor other than Shn when binding to the bam repressor element. In the future, it will be very important to determine whether Shn itself is a co-repressor for Mad/Med proteins or whether it directly recruits a co-repressor to repress bam transcription in GSCs.

We thank D. Drummond-Barbosa, S. Cohen, S. Kobayashi, P. Lasko, D. McKearin, A. Spradling, A. Laughon, K. Wharton, M. Hoffmann, P. ten Dijke, and the Drosophila stock center for reagents. We would also like to thank G. Call,J. Coffman, S. Hawley, S. Page, A. Spradling for critical comments on manuscripts, the Xie laboratory members for stimulating discussions, Jo Haynes for help with the manuscript preparation, C. Sonnenbrot, J. Haug, H. Newkirk and K. Zueckert-Gaudenz for technical help. This work is supported by the Stowers Institute for Medical Research and NIH (1R01 GM64428-01).