Notch signaling controls germline stem cell niche formation in the Drosophila ovary

Adult tissues undergo continuous cell turnover throughout an organism's lifetime. Stem cells, a group of undifferentiated cells residing in adult tissues, are responsible for generating differentiated cells for maintaining tissue homeostasis due to their unique self-renewal ability. The stem cells are controlled by their specialized local regulatory microenvironments, known as niches, that are formed by their neighboring stromal cells(Li and Xie, 2005; Spradling et al., 2001; Watt and Hogan, 2000). The signals from niche cells work with intrinsic factors to control stem cell self-renewal, proliferation and differentiation(Molofsky et al., 2004; Xi et al., 2005). Although the identification of stem cells remains challenging, due to their rarity and lack of unique molecular markers in mammalian systems, several niches are roughly defined based on their proximity to stem cells(Calvi et al., 2003; Nishimura et al., 2002; Tumbar et al., 2004; Zhang et al., 2003). However,little is known about how niche formation is genetically controlled, even though niche structure is defined. In this study, we show that Notch(N) signaling directly controls formation of the germline stem cell(GSC) niche in the Drosophila ovary.

The Drosophila ovary is one of the best-studied stem cell systems because of its easily identified stem cells and powerful genetic tools(Xie et al., 2005). There reside three types of stem cells: GSCs, somatic stem cells (SSCs) and newly identified escort stem cells (ESCs), which are responsible for generating differentiated germ cells, follicle cells and escort cells, respectively(Decotto and Spradling, 2005; Lin and Spradling, 1993; Margolis and Spradling, 1995; Wieschaus and Szabad, 1979). GSCs have been shown to be situated in the niche, which is composed of cap cells, and possibly terminal filament (TF) cells and ESCs(Cox et al., 1998; Decotto and Spradling, 2005; Kretzschmar et al., 1999; Xie and Spradling, 1998; Xie and Spradling, 2000)(Fig. 1A). Recent findings show that the number of cap cells is closely correlated with the GSC number in the normal ovary (Xie and Spradling,2000) and that GSCs must be anchored to cap cells in order to be maintained as stem cells through DE-cadherin-mediated cell adhesion(Song et al., 2002). This supports the idea that cap cells are an important component of the GSC niche. dpp, gbb, Yb [fs(1)Yb - FlyBase], piwi and hh, known to be important for GSC maintenance, are expressed not only in cap cells but also in TFs and/or inner germarial sheath(IGS) cells (Cox et al., 1998; Cox et al., 2000; Kiger and Fuller, 2001; King and Lin, 1999; King et al., 2001; Song et al., 2004; Song et al., 2002; Xie and Spradling, 1998). These findings point to a crucial function of cap cells in the GSC niche, but it remains unclear how cap cell formation is genetically controlled.

N signaling plays an important role in regulating proliferation and differentiation of many different cell types(Artavanis-Tsakonas et al.,1999; Lai, 2004). In the Drosophila ovary, it was first shown to be required for maintaining follicle cells in their precursor stage and for specification of polar cells that mark the ends of the egg chamber(Grammont and Irvine, 2001; Larkin et al., 1996; Xu et al., 1992). During late oogenesis, N signaling is required for the switch from the mitotic cycle to the endocycle and differentiation of follicle cells by negatively regulating the cut gene(Shcherbata et al., 2004; Sun and Deng, 2005), and it is also required for patterning the anterior egg shell(Dobens et al., 2005). In this study, we have shown, for the first time to our knowledge, that Nsignaling is necessary and sufficient for controlling formation of the GSC niche.

The following Drosophila stocks were used in this study: c587-gal4 (Zhu and Xie,2003); UAS-Dl^30B, UAS-Dl³⁶ and hs-gal4 (Bloomington Drosophila Stock Center); two UAS-N^int lines, UAS-N^B2A2 and UAS-N^33c3 (kindly provided by Dr Gary Struhl, Columbia University, New York City, NY); N^264-39, N^54l9,N^ts1 and hh-lacZ (Bloomington DrosophilaStock Center); UAS-Dl^DN(Parks et al., 2000); UAS-mam^H and UAS-mam^N(Helms et al., 1999); Dl-lacZ and E(spl)-CD2 (kindly provided by Dr Leonard Dobens, University of Missouri, Kansas City, MO).

IGS cells overexpressing dpp or N^ICD were generated using a technique that combines the FLP-FRT and UAS-GAL4 systems(Ito et al., 1997). hs-flp; AyGal4 (act≫y≫gal4) UAS-GFP/CyOvirgin females were crossed with either UAS-dpp/TM3,UAS-N^B2A2/CyO or UAS-N^33C3/TM3 males,respectively. Clones were induced by two 1-hour heat shock treatments of 2-day-old females at 37°C separated by an interval of 5 hours. The heat-shock-treated females were cultured at room temperature for 1 week with daily supplied fresh food, and their ovaries were dissected out and processed for immunostaining with monoclonal anti-Hts (1B1) and rabbit anti-GFP antibodies as described previously (Xie and Spradling, 1998).

The 2-day-old hh-lacZ/+ control, c587-gal4/+;UAS-Dl^30B/+;hh-lacZ/+, c587-gal4/+; UAS-Dl³⁶/hh-lacZ,c587-gal4/+;UAS-N^B2A2/+;hh-lacZ/+, c587-gal4/+;UAS-N^33c3/hh-lacZ females were fed on wet yeast paste mixed with 20 mg/ml BrdU solution for 3 consecutive days with fresh BrdU yeast paste each day. The ovaries from these flies were processed for immunostaining with anti-BrdU, anti-Hts and anti-β-galactosidase (β-gal) antibodies according to our published procedures(Song et al., 2002). The rest of the flies were then transferred to fresh food with yeast flakes containing no BrdU every day for 3 consecutive weeks, and their ovaries were processed for immunostaining with the same antibodies.

Immunostaining of the Drosophila ovaries was performed according to previously published procedures (Song et al., 2002). Primary antibodies used in this study are as follows: rabbit anti-β-galactosidase antibody (1:200, Cappel), rabbit anti-GFP antibody (1:200, Molecular Probes), mouse monoclonal anti-CD2 antibody (1:100, Serotec), mouse monoclonal anti-Hts antibody, 1B1 (1:4,DSHB), a mouse monoclonal anti-Dl antibody, c594.9B (1:3, DSHB), two mouse monoclonal anti-N antibodies, F461.3B and c458.2H (1:3, DSHB), rabbit anti-Vasa antibody (1:1000, a gift from Dr Paul Lasko, McGill University,Montreal, Canada), rat anti-DE-cadherin (1:4, DSHB) and sheep anti-BrdU antibody (1:100, Capralogies). The secondary antibodies used in this study are the Alexa 568-, Alexa 468- and Alexa 596-conjugated goat or donkey anti-mouse,rabbit, rat or sheep antibodies (1:200, Molecular Probes). All the images were taken using a Leica TCS SP2 confocal microscope.

As N signaling plays an important role in regulating specification of many different cell types(Artavanis-Tsakonas et al.,1999; Lai, 2004),we sought to test whether N signaling regulates the formation of cap cells in the Drosophila ovary by manipulating the signaling pathway in somatic cells of developing female gonads using the GAL4-UAS system(Brand and Perrimon, 1993). Ectopic expression of a truncated N intracellular domain(N^int) or its ligand Delta (Dl) can activate N signaling in ectopic locations(Struhl et al., 1993), and the c587-gal4 driver can drive a UAS-GFP specifically in most, but not all, of the somatic cells of developing female gonads(Zhu and Xie, 2003)(Fig. 1B,B′). A hedgehog (hh)-lacZ line (the bacterial lacZ gene inserted in the hh gene) is highly expressed in five to seven cap cells and eight to ten TF cells(Forbes et al., 1996a)(Fig. 1C). GSCs are identified by their direct association with cap cells and the presence of an anteriorly anchored spherical spectrosome (Fig. 1C). Their immediate daughters, cystoblasts, also contain a spherical spectrosome but are positioned away from the cap cells, while other further differentiated progeny, germ cell cysts, can be identified by the presence of a branched fusome (Fig. 1C). The spectrosome and the branched fusome are identified by their expression of Hu-li tai-shao (Hts)(de Cuevas et al., 1997). By contrast to the five to seven cap cells in a normal germarium(Fig. 1C), we observed more lacZ-positive cap cells at the tip of either Dl- or N^int-overexpressing germaria using two independent transgenic UAS-Dl or UAS-N^int lines(Fig. 1D,D′). Note that overexpression of Dl and N^int gave similar phenotypes, although overexpression of the latter generated a stronger phenotype than that of the former. In the germaria with increased cap cells,spectrosome-containing single germ cells, which were later shown to behave like GSCs, also increased at the germarial tip(Fig. 1D,D′). This result shows that N signaling is sufficient to induce cap cell formation and supports the idea that cap cells are a key niche component for controlling GSC self-renewal.

In addition to increased cap cells at the germarial tip, we frequently observed one or more patches of strongly lacZ-positive somatic cells away from the germarial tip when Dl or N^int was overexpressed in the developing gonads(Fig. 1E-F′). These lacZ-positive somatic cells appeared to be functional cap cells at ectopic locations, as spectrosome-containing single germ cells (later shown to be GSCs) were closely associated with them(Fig. 1E-F′). The ectopic GSCs associated with the ectopic cap cells also anchored their spectrosome on the side that contacts cap cells, as observed in a normal GSC context. Some of these ectopic cap cells were surrounded by IGS cells(Fig. 1E,E′) or somatic follicle cells (Fig. 1F,F′), and it appeared that both types of ectopic cap cells could sufficiently maintain GSCs. In rare cases of N^intoverexpression, lacZ-positive cap cells completely occupied the anterior half of the germaria instead of IGS cells, and consequently, GSCs were everywhere in the anterior half of the germarium(Fig. 1G,G′). Germ cells moving away from the cap cells could still differentiate, as indicated by the presence of the branched fusomes. These observations further indicate that signals from the GSC niche directly repress differentiation of GSCs close to cap cells, allowing germ cells moving away from the cap cells to differentiate, because they are beyond the influence of short-range signals from the cap cells. The ectopic lacZ-positive cells and their associated GSCs could persist for at least 5 weeks (the longest time we had tested), suggesting that these expanded or ectopic cap cells are stable and sustain GSCs like normal cap cells. Together, these results demonstrate that N signaling is sufficient to induce cap cell formation. Furthermore,our observation that ectopic cap cells without TF cells or IGS cells are able to sustain GSC self-renewal indicates that cap cells are a key component to establish the niche for sustaining GSC self-renewal.

To further verify that the ectopic hh-lacZ-positive cap cells exhibit known properties of normal cap cells, we examined the expression of other markers for cap cells. A wingless (wg) enhancer trap line, wg-lacZ, is known to be expressed in some but not all cap cells(Forbes et al., 1996b) (see Fig. S1A in the supplementary material). In agreement with our hypothesis,some of these ectopic cap cells expressed wg-lacZ whether they were surrounded by IGS cells or somatic follicle cells (see Fig. S1B,C in the supplementary material). Nuclear lamin C is expressed highly in nuclear membranes of TF cells and cap cells (Xie and Spradling, 2000) (see Fig. S1D in the supplementary material),and the cap cells in normal or ectopic locations highly express this marker(see Fig. S1E in the supplementary material). Cap cells also express and accumulate DE-cadherin at their junction with GSCs in keeping GSCs in the niche (Song et al., 2002). Indeed, DE-cadherin proteins significantly accumulated between ectopic cap cells and their associated GSCs (see Fig. S1F in the supplementary material),which might also function to anchor GSCs. Therefore, our molecular evidence strongly indicates that these ectopic cap cells behave like normal cap cells.

To investigate whether the spectrosome-containing single germ cells associated with expanded cap cells resemble normal GSCs, we examined the expression of bam-GFP and Dad-lacZ, which were used to monitor bam and Dad transcription(Chen and McKearin, 2003; Kai and Spradling, 2003; Song et al., 2004). As in the wild-type niche (Fig. 2A), the spectrosome-containing single germ cells associated with those ectopic cap cells did not express bam-GFP, which resembles the property of normal GSCs (Fig. 2B,C). Interestingly, germ cells lying one cell away from ectopic GSCs were often germline cysts, as they contained branched fusomes, indicating that the progeny of the ectopic GSCs probably undergo normal differentiation(Fig. 2B,C). However, too many cap cells (more than seven) at the normal location(Fig. 2D) or at an ectopic site(Fig. 2E) often caused the accumulation of spectrosome-containing single germ cells located two or more cell diameters away, and bam expression was also repressed in those single cells, indicating that these extra single cells also resemble GSCs. Normally, bam is only repressed in GSCs due to the short-range BMP signal, which may be caused by a limited amount of DPP protein produced by five to seven cap cells in a wild-type germarium. As we have previously demonstrated that dpp overexpression in the germarium causes the accumulation of single cells that are negative for bam expression(Song et al., 2004; Xie and Spradling, 1998),these observations further suggest that cap cells are the major source of BMP,and more cap cells can produce more BMP, which could help it diffuse farther than the normal distance.

A short-range BMP signal from cap cells specifically activates its signaling cascade in GSCs to activate expression of Dad(Casanueva and Ferguson, 2004; Kai and Spradling, 2003; Song et al., 2004)(Fig. 2F). As expected, all the GSCs that directly contact the expanded cap cells at the germarial tip highly expressed Dad-lacZ, indicating that the expanded cap cells have the capacity to produce a BMP signal in a similar manner to normal cap cells(Fig. 2G). Interestingly, many spectrosome-containing single germ cells not directly associated with the expanded cap cells (more than ten cap cells) also expressed Dad-lacZ,further supporting the idea that the increased number of cap cells leads to production of more diffusible BMP (Fig. 2G). When the ectopic cap cells are surrounded by IGS cells or near follicle cells, their associated GSCs also highly expressed Dad-lacZ as in normal GSCs (Fig. 2H), confirming the idea that ectopic cap cells can also emit the BMP signal like normal cap cells. These results suggest that cap cells are the source of active BMP and that the spectrosome-containing single germ cells associated with expanded or ectopic cap cells resemble GSCs.

A GSC generates two daughters that remain connected to each other via a contractile ring, through which an elongated spectrosome passes(de Cuevas and Spradling,1998). In our earlier experiments, we observed that many cases of ectopic GSCs carried an elongated spectrosome, indicating that those ectopic GSCs are probably capable of dividing and generating differentiated germ cells. To further confirm that these extra spectrosome-containing germ cells behave like GSCs, we labeled them with BrdU, a nucleotide analog, for 3 days and chased for 3 weeks. As a control, 96.4% of the germaria (n=57) of the hh-lacZ/+ heterozygotes contained one or more BrdU-labeled GSCs after 3 days of BrdU feeding, while 84.5% of the total GSCs (n=136)were labeled by BrdU. Consequently, most of the control germaria contained two or three labeled GSCs (Fig. 3A). After the females were fed for 3 more weeks on normal food without BrdU, the GSCs in the control germaria (n=35) completely lost their BrdU label (Fig. 3B),indicating that the BrdU label is completely diluted out as the labeled GSCs continuously divide for 3 weeks.

Similarly, 98.8% of the germaria (n=85) (developed from the female gonads overexpressing Dl or N^int) carried one or more BrdU-labeled GSCs close to the cap cells at the normal location or at ectopic cap cells (Fig. 3C,D). Among them, 94.1% of the expanded niches (more than seven cap cells carrying four or more GSCs; n=51) at the normal location (close to TFs)carried one or more BrdU-labeled GSCs (Fig. 3C,C′), while 96.0% of the ectopic niches (n=25)carried one or more BrdU-labeled GSCs (Fig. 3D), indicating that extra GSCs in the expanded niches as well as in the ectopic niches are mitotically active like normal GSCs. After the 3 week chase, 85.9% of the germaria (n=120) did not carry any BrdU-labeled GSCs at the normal location or at ectopic sites(Fig. 3E), whereas 14.1% of them carried one or more BrdU-labeled GSCs(Fig. 3F), also indicating that the expanded or ectopic GSCs remain active for over 3 weeks. It also appeared that those GSCs did not divide as frequently as normal GSCs. Perhaps this is due to their excessive number at each niche. These results strongly suggest that the GSCs at the expanded niche or at the ectopic niche can continuously divide and generate differentiated germ cells like normal GSCs.

Cap cells normally form during the larval-pupal transition and in the early pupal stage (Zhu and Xie,2003). To further determine when ectopic cap cells form, we used an hs-gal4 transgene (the gal4 gene under the control of the heat shock protein 70 promoter) to drive the expression of Dl during gonadal development by heat shock treatments. When Dl expression was induced during the second instar larval stage or after the mid-pupal stage,about 98% of the germaria carried five to seven cap cells in the normal location, just like in the wild type (n=603), whereas about 2% of the germaria had ectopic cap cells positioned away from TFs, as detected in the heat-shocked hsgal4; hh-lacZ control germaria (n=606). Also,overexpression of Dl in adult ovaries did not increase cap cell numbers and induce ectopic cap cells (n=212). Interestingly, when Dl was ectopically expressed during the third instar and early pupal stages, the number of cap cells at the tip of some germaria was increased(Fig. 4A,A′), and about 10% of these germaria carried ectopic cap cells (n=234)(Fig. 4B,B′), indicating that elevated N signaling can induce cap cell formation in the normal location as well as in the ectopic sites. In some of the germaria(n=27), cap cells were formed away from TFs and still maintained GSCs, but no GSCs were associated with TFs(Fig. 4B,B′), indicating that TFs alone are not sufficient to sustain GSC self-renewal. These results suggest that the IGS precursors in the developing gonad are competent to form cap cells in response to ectopic N signaling from the late third-instar larval stage to the early pupal stage. Note that these ectopic cap cells no longer expressed Dl or had the activated Nsignaling in the adult ovary but were stable and able to maintain GSCs,suggesting that ectopic activated N signaling by c587-gal4is not directly involved in controlling GSC self-renewal.

To rule out the possibility that the remaining activated N in the ectopic niches may directly control expression of dpp and thereby maintain ectopic GSCs, we used the previously established `flip-out' cassette to express dpp and the activated N in IGS cells of the adult Drosophila ovary, which were labeled by GFP expression(Ito et al., 1997). Interestingly, when dpp was induced to be expressed in several of the IGS cells of adult Drosophila ovaries, many spectrosome-containing single germ cells accumulated, and those GFP-positive dpp-expressing IGS cells still maintained their normal morphology, suggesting that dpp-expressing IGS cells may sufficiently create ectopic GSC niches without being transformed into cap cells (see Fig. S2A in the supplementary material). By contrast, overexpression of the activated N in the GFP-positive IGS cells did not affect the differentiation status of the underlying germ cells, indicating that activation of N signaling in adult IGS cells alone does not sufficiently create ectopic GSC niches, unlike dpp overexpression (see Fig. S2B,C in the supplementary material). In light of the evidence we have presented so far, we conclude that ectopic expression of the activated N during niche formation leads to formation of ectopic niches and GSCs, but the continuous expression of the activated N in adult IGS cells might not directly contribute to the maintenance of ectopic GSCs.

To gain further insight into how N signaling is involved in controlling cap cell formation, we examined N and Dlexpression in the developing gonads from the late third instar larval stage to the early pupal stage. A Dl-lacZ line (the lacZ gene inserted in the Dl locus to recapitulate its expression) and an anti-Dl antibody were used to determine Dl expression in the developing female gonads (Grossniklaus et al., 1989). Dl was primarily expressed in newly formed TFs and other somatic cells anterior to the primordial germ cell zone (PGCZ)at high levels but not in primordial germ cells (PGCs) during the late third-instar larval stage or the larval-pupal transitional stage(Fig. 4C-D′). Interestingly, N protein was expressed at high levels in TFs and cap cells as well as in the somatic cells that are mingled with PGCs at lower levels, but not in PGCs (Fig. 4E,E′). N signaling is known to regulate expression of E(spl) complex genes(Bailey and Posakony, 1995); an E(spl)mβ-CD2 transgene has been used to monitor N signaling activity in several tissue types(Cooper and Bray, 1999; de Celis and Bray, 1997). Consistent with the idea that TFs and cap cells are capable of activating N signaling due to presence of both Dl and N, E(spl)mβ-CD2 was expressed almost exclusively in TFs and cap cells but not in the somatic cells that are mixed with PGCs(Fig. 4F,F′). These expression results suggest that developing TFs and cap cells are active in N signaling, which may be dependent on Dl in TFs.

Our earlier observation that the somatic cells that mingle with PGCs express N raises the interesting possibility that activated N signaling transforms those somatic cells normally destined to form IGS cells and/or follicle cells into cap cells. To further investigate this possibility, we used the E(spl)mβ-CD2 as a marker to study whether ectopic expression of Dl or N^int could induce expression of E(spl)mβ-CD2 in the somatic cells that are mingled with PGCs. When N^int or Dl was expressed by c587-gal4 in most of the somatic cells in the gonad, clusters of the somatic cells that were also closely associated with PGCs started to express E(spl)mβ-CD2(Fig. 4G,G′). This suggests that ectopic N signaling in somatic cells, which is normally devoid of N signaling, induces them to assume cap cell fate.

To directly investigate whether N signaling is required for cap cell formation, we attempted to examine the number of cap cells in Nand Dl temperature-sensitive mutants. Unfortunately, all the existing temperature-sensitive Dl and N mutant third-instar larvae or early pupae did not survive at a restrictive temperature (29°C) to adulthood. As N is known to be haploinsufficient in several developmental processes (Artavanis-Tsakonas et al., 1999), we examined the cap cell number in 2-day-old heterozygotes carrying strong or null N mutants, N^54l9 and N^264-39, using hh-lacZ to label TFs and cap cells. Because there is no or little cap cell and GSC turnover in 2-day old Drosophila females, the number of cap cells and GSCs should truly reflect the number of cap cells that form during the pupal stage. By contrast, with 2-day-old wild-type germaria with 5.5±1.3 cap cells and 2.5±0.6 GSCs (n=121), the 2-day-old N^54l9 heterozygous germaria carried 4.1±1.5 cap cells and 1.8±0.6 GSCs (n=131), while the N^264-39 heterozygous germaria had 4.2±1.4 cap cells and 1.8±0.6 GSCs (n=97), indicating that decreased Nsignaling can significantly reduce cap cell number (P<0.001) and consequently GSC number (P<0.001). Moreover, about 36% of the N heterozygous germaria had three or fewer cap cells, by contrast to the five to seven cap cells of the wild-type germaria(Fig. 4H,H′). To gain further evidence supporting the idea that N signaling is required for cap cell formation, we used the c587-gal4 driver to overexpress dominant-negative mutants for N pathway components, such as Dl and mastermind (mam). The 2-day-old germaria developing from female gonads that overexpressed a dominant-negative Dl had an average of 3.3±1.6 cap cells and 1.6±0.6 GSCs(total 62 germaria examined), while the germaria developing from female gonads that overexpressed two dominant-negative forms of mam,mam^H and mam^N, carried an average of 3.2±1.2 cap cells and 1.6±0.6 GSCs (total 162 germaria examined)as well as 2.8±1.2 cap cells and 1.4±0.6 GSCs (total 128 germaria examined), respectively. As the dominant-negative forms of Dl and mam have been shown to specifically block Nsignaling in imaginal discs (Helms et al.,1999; Parks et al.,2000), our results further support the idea that Nsignaling is required for controlling cap cell formation.

Due to lack of the ability to remove Dl function specifically from TFs to test whether TF-expressing Dl controls cap formation, we investigated if cap cells could still form without germ cells. ovo^D1rS1 mutant females contain no germ cells(Rodesch et al., 1995). Among the 2-day-old agametic ovo^D1rS1 mutant germaria, many still contained one to five cap cells, although some completely lacked cap cells, which could have degenerated during the pupal period(Fig. 4I-J′). In the osk mutant agametic germaria, cap cells appear to form and persist(Margolis and Spradling,1995). Together, these findings argue that TF-expressed Dl is responsible for cap cell formation in the agametic germarium, and possibly in the normal germarium as well.

To investigate whether N signaling is also involved in regulation of niche and GSC functions in the adult ovary, we examined the expression of Dl using the Dl-lacZ line mentioned earlier. By contrast with the fact that Dl expression is restricted to the somatic cells anterior to the PGC zone in the third-instar larval stage, including newly formed TFs, Dl was also observed to be expressed at low levels in germ cells of adult germaria, including GSCs, in addition to its expression in TFs (see Fig. S3A in the supplementary material). E(spl)m7, a gene in the E(spl)complex, is a response gene of N signaling(Bray, 2006). E(spl)m7-lacZ was used to monitor Nsignaling activity in the germarium. Interestingly, E(spl)m7-lacZ was mainly detected in cap cells,possibly in GSCs at very low levels, suggesting that N signaling remains active in cap cells of the adult germarium (see Fig. S3B in the supplementary material). To further investigate whether or not Nsignaling is required for the maintenance or function of the GSC niche, we used a well-studied temperature-sensitive N allele(N^ts1) to determine the role of N signaling in the adult ovary. One previous study using this N allele showed that N signaling is required for proper differentiation of follicle cells and thus proper formation of egg chambers(Xu et al., 1992). As a control, the germaria from the wild-type females cultured at 29°C for 1 or 2 weeks had 2.4±0.5 (n=33) and 2.2±0.7 (n=39)GSCs, respectively, and germ cysts differentiated normally and egg chambers formed normally as well (see Fig. S3C,D in the supplementary material). By contrast, the germaria from the N^ts1 mutant females cultured at 29°C for 1 or 2 weeks were generally small, and contained 1.1±1.0 (n=27) and 0.5±0.8 (n=36) GSCs,respectively. By contrast with 6.0±0.9 (1 week) and 5.8±1.0 (2 weeks) cap cells for the wild-type ovaries, the N^ts1mutant ovaries from the females cultured at 29°C for 1 or 2 weeks carried 4.6±1.3 and 2.7±1.4 cap cells, respectively. As complete loss of GSCs does not quickly eliminate cap cells(Xie and Spradling, 2000), our results suggest that GSC loss in the N mutant ovaries is probably due to loss of cap cells. Indeed, the germaria from the females cultured at 29°C for 1 week had more cap cells than the ones cultured for 2 weeks (see Fig. S3E-H in the supplementary material), and in some of the germaria, cap cells completely disappeared (see Fig. S3H in the supplementary material). In light of the evidence that N signaling activity in the adult ovary is restricted to cap cells, these results further suggest that Nsignaling is required for maintaining cap cells and thus GSCs.

Although it has recently been shown that several adult stem cell types are controlled by their neighboring stromal cells(Li and Xie, 2005), it remains unclear what constitutes a functional niche and how its formation is genetically controlled. In this study, we have revealed one novel molecular mechanism underlying GSC niche formation in the Drosophila ovary(Fig. 5). The newly formed TFs express Dl, which is probably responsible for activating Nsignaling in their neighboring somatic cells and for inducing niche formation. Ectopic N signaling is sufficient to induce niche formation in a developmental stage-dependent manner, and reduced N signaling results in reduction of cap cell and GSC numbers, demonstrating that Nsignaling is important for controlling GSC niche formation. As ectopic niche cells in different locations are stable and able to sustain GSC self-renewal,this leads us to conclude that the niche does not have to function in a fixed position. Finally, N signaling is also required for the survival of niche cells in adults (see Fig. S3 in the supplementary material). By analogy, N signaling may control niche formation in other systems, including mammals.

At the onset of the larval-pupal transition, all of the 16 to 20 TF stacks have formed and initiate ovariole formation(Godt and Laski, 1995; King, 1970), while another group of somatic cells, cap cells, start to occupy a position between the TFs and the germ cells (Zhu and Xie,2003). The PGCs in direct contact with cap cells are further anchored through E-cadherin and are further expanded through symmetric division and develop into permanent GSCs in the adult ovary(Song et al., 2002; Zhu and Xie, 2003). Actin-filament regulator, Cofilin/ADF, and ecdysone signaling, are required for TF formation (Chen et al.,2001; Hodin and Riddiford,1998). However, no studies have been carried out to investigate the formation of cap cells, which are a key component of the GSC niche. In this study, we have investigated the role of N signaling in controlling cap cell formation.

N signaling controls cell fate determination in a variety of organisms (Artavanis-Tsakonas et al.,1999; Kadesch,2004; Lai, 2004). In this study, we show that in late third-instar larval female gonads, Dl is expressed in newly formed TFs, while the N receptor is expressed in all the somatic cells, including TFs and cap cells. Consequently, N signaling is active in newly formed TFs and cap cells and its activation is sufficient to induce cap cell formation, suggesting that TF-expressed Dlactivates N signaling to induce cap cell formation. To further support the idea that N signaling specifies cap cell fate, reduction of N signaling results in a reduced number of cap cells. Induction of cap cells by N signaling can only take place during the late third instar and early pupal stages, suggesting that active N signaling only promotes cap cell formation along with other factors provided at particular stages. Cap cells can still form without germ cells. This suggests that Dl is unlikely to be required in germ cells for cap cell formation. Therefore, we conclude that N signaling, activated probably by Dl from newly formed cap cells, specifies cap cell fate through direct induction. In this study, we also show that N signaling is required for maintaining the GSC niche in the adult ovary, as loss of N function results in rapid loss of cap cells and GSCs. Taken together, the results of this study demonstrate that N signaling is important for controlling niche formation as well as niche maintenance.

Although niches have been defined for GSCs in the Drosophila ovary and testis, as well as in several tissue types of the mammalian systems, it remains unclear whether they still function properly for controlling stem cell self-renewal after their location and size are changed. In this study, we have provided two pieces of experimental evidence supporting the idea that expanded niches are functional for controlling GSC self-renewal. First, increased cap cells in the normal location express all known cap cell markers, such as hh-lacZ, wg-lacZ, Lamin C and E-cadherin, and behave like normal cap cells. Second, these expanded cap cells can support self-renewal of extra GSCs, which behave similarly to normal ones based on Dad-lacZ and bam-GFP expression, and their ability to self-renew and generate differentiated germ cells. Even when cap cells cover the anterior half of the germarium, the GSCs associated with the cap cells also appear to be capable of self-renewing and generating differentiated germ cells. Our findings show that GSC niche size can be expanded by adding more niche cells.

This study has also demonstrated that the GSC niche could function in ectopic locations. Ectopic cap cells, which are surrounded by IGS cells or follicle cells, also express known cap cell markers and sufficiently support functional GSCs, supporting the idea that TFs and IGS cells are not essential components of the GSC niche. This is consistent with early published studies showing that the numbers of cap cells and GSCs are closely correlated and that TFs and cap cells express the genes important for GSC self-renewal such as piwi, Yb, hh and dpp/gbb(Chen and McKearin, 2005; Cox et al., 2000; King et al., 2001; Song et al., 2004; Szakmary et al., 2005; Xie and Spradling, 1998). In light of the recent evidence that ESCs in direct contact with cap cells and GSCs are required for maintaining GSCs(Decotto and Spradling, 2005),it remains formally possible that some unidentified ESC cells associated with expanded or ectopic cap cells contribute to niche function. In any case, this study demonstrates that the size and location of the GSC niche can be genetically manipulated while it maintains its functions. Our ability to manipulate niche location and size will further increase our capacity to investigate how niche formation is controlled and how the niche controls stem cell function in general.

One of the major unsolved issues for the GSC niche in the Drosophila ovary is how BMPs function as a short-range signal to control GSC self-renewal and allow the GSC daughter adjacent to the GSC to differentiate at the same time. Several previous studies from us and others have shown that BMP signaling activity is primarily restricted to the GSC based on Mad phosphorylation and Dad expression, two indicators of BMP signaling (Casanueva and Ferguson,2004; Kai and Spradling,2003; Song et al.,2004). Our early work has also shown that dpp is primarily expressed in TF and cap cells, while gbb is expressed in TF and cap cells as well as in IGS cells(Song et al., 2004). In this study we show that BMP signaling activity can spread two or more cell diameters based on expression of Dad-lacZ and bam-GFP when more cap cells exist. Furthermore, when more cap cells accumulate in ectopic sites, the GSCs associated with the cap cells as well as the germ cells lying two or three cells away are capable of activating BMP signaling and repressing bam expression. One of the explanations for these observations is that cap cells are the source of BMP and more cap cells would produce more BMP to diffuse further away to repress differentiation of germ cells lying two or more cell diameters away. Another explanation is that the ratio of cap cells to ESCs or escort cells increases so that BMP inhibitors, such as Sog,produced by ESCs or escort cells, are diluted or deterred by more cap cells,and consequently more active BMP is available for reaching and activating cells lying more than two cell diameters away. In Xenopusgastrula-stage embryos, an effective BMP-4 activity gradient is established,not by diffusion of BMP-4 protein but by the long-range effects of two BMP-4 inhibitors, Noggin and Chordin (Jones and Smith, 1998). Finally, it is also possible that a combination of both mechanisms contributes to restriction of BMP signaling activity to one cell diameter in the GSC niche. Our observations from this study have suggested that a limited amount of active BMP produced by cap cells is probably responsible for its short-range effect on GSC self-renewal in the GSC niche.

We would like to thank S. Artavanis-Tsakonas, L. Dobens, S. Hou, P. Lasko,D. McKearin, G. Struhl, B. Yedvobnick, and the Drosophila Stock Center and Developmental Studies Hybridoma Bank for reagents, the Xie laboratory members, R. Krumlauf and R. Lehmann for critical comments, B. Steinkuhler for administrative assistance and the fly food facility for fly food. This work is funded by NIH (2 R01 GM064428-06) and Stowers Institute for Medical Research.