Stonewalling Drosophila stem cell differentiation by epigenetic controls

Stem cells are partially defined by their ability to divide asymmetrically,producing a differentiating daughter cell and a self-renewing stem cell. This paradigm is used many times throughout development to produce and maintain mitotically active tissues, reserving stem cells as a source of cell progenitors throughout the life of an animal. Mechanisms regulating stem cell biology are relevant to studies of developmental biology as well as to regenerative medicine and to the pathogenic transformation of tissues. Loss of stem cells can disrupt normal tissue homeostasis and may cause premature aging, infertility or defects in tissue regeneration. By contrast,misregulation of stem cell self-renewal can produce a population of undifferentiated cells that may be susceptible to carcinogenic transformation.

The germline stem cells (GSCs) of the Drosophila ovary have emerged as a valuable model for studying stem cell behavior and its regulation. In contrast to stem cells in many other tissues, the ovarian GSCs are easily identified by location and by molecular markers. Two or three ovarian GSCs reside at the anterior of each ovariole, closely apposed to the somatic cap cells of the germarium. Mutational analysis in Drosophilahas revealed a suite of genes that is necessary for GSC maintenance(Gilboa and Lehmann, 2004a; Spradling et al., 2001). These studies have demonstrated that a microenvironment, or niche, composed of multiple cell types influences GSC self-renewal. Factors intrinsic to the stem cell, as well as signals from the somatic cells of the niche, are important for GSC maintenance.

The emerging view of GSC maintenance in Drosophila indicates that GSCs are maintained largely by preventing their differentiation(Kai et al., 2005; Li and Xie, 2005; Wang and Lin, 2004). Stromal cells produce Dpp locally to prevent GSC differentiation by initiating a Smad signaling cascade in GSCs that silences transcription of the key differentiation factor bag of marbles (bam)(Chen and McKearin, 2003a; Song et al., 2004; Xie and Spradling, 1998). Adherens junctions anchor GSCs to cap cells and are essential for GSC maintenance and for Smad signaling (Song et al., 2002). Other factors required in the cap cells, such as piwi (Cox et al.,2000) and fs(1)Yb(King and Lin, 1999), are also essential GSC maintenance factors, but their mode of action remains unknown.

Gene products that are required intrinsically in stem cells for their maintenance also function to prevent germline differentiation. For example,the translational inhibitors Nanos (Nos) and Pumilio (Pum) are required in stem cells (Forbes and Lehmann,1998; Lin and Spradling,1997; Wang and Lin,2004). The Nos-Pum complex is thought to repress the translation of key differentiation factors to prevent GSC differentiation, but the targets of Nos and Pum in GSCs are unknown. A recent report has also suggested that chromatin state influences GSC maintenance, because Iswi, a key chromatin-remodeling factor, is required in GSCs for their maintenance(Xi and Xie, 2005).

Here, we present evidence that the DNA-associated protein Stonewall (Stwl)is required in GSCs, and that it promotes their maintenance by repressing germ cell differentiation. Stwl is a nuclear factor that has protein domains that suggest an interaction with histone-modifying enzymes. We demonstrate that stwl mutants act as dominant suppressors of variegation, indicating a requirement for stwl in heterochromatin assembly or maintenance. We order the requirement for stwl with respect to other GSC maintenance factors and find that stwl is not essential for bamsilencing, but rather that it acts in opposition to bam activity to prevent germ cell differentiation. We propose that Stwl represses the expression of many genes, including those targeted by Nos-Pum translational inhibition. We identify a group of genes that are specifically upregulated in the absence of stwl and, among those genes, find candidates for Nos-Pum translational inhibition.

Genetic mutants identified below are also described at http://flybase.bio.indiana.edu. stwl mutants used included stwl^Z1, stwl⁹⁵ (K. Clark, PhD thesis, University of Texas Southwestern Medical Center, 1996) and stwl^EY05697(Bellen et al., 2004). Other mutations included nos^RC(Gavis and Lehmann, 1992),Df(nos) Df(3R)DL-FX1 (Sonoda and Wharton, 1999) (gift from R. Wharton), pum^ET9(Barker et al., 1992), bam^Δ86 (McKearin and Ohlstein, 1995), Su(var)3-9¹(Reuter et al., 1986) and EY00986 upstream of CG3919 (stwl neighbor) from Bellen(Bellen et al., 2004).

Transgenic flies used included P{bamP-GFP}(Chen and McKearin, 2003b),P{hsp70-stwl} (Clark and McKearin, 1996), P{FRT 79D} and P{hsFLP} from Bloomington Stock Center, P{nosP-Gal4:VP16} (a gift from R. Lehmann)(Van Doren et al., 1998), DX1 P{w+} (a gift from S. Henikoff)(Dorer and Henikoff, 1994),P{Nos-myc} (a gift from R. Wharton)(Verrotti and Wharton, 2000). Histone-GFP fusion flies were a gift from J. Duffy (Indiana University,Bloomington, IN).

Ectoptic Stwl was produced by combining the stwl^EPallele, which allows misexpression of stwl⁺ when a source of Gal4 transactivator is introduced, with the germline-specific Nos-Gal4:VP16. Flies were transferred to wet yeast and shifted to 29°C for a minimum of 18 hours prior to ovary dissection. For each genotype, a minimum of 30 ovarioles were scored. {HS-stwl} produced a similar, but more variable, stem cell expansion phenotype.

For clonal analysis, flies of the genotype {hsFLP}; {histoneGFP} {FRT79D}were crossed to stwl⁹⁵ {FRT 79D} flies. To make clones of wild-type alleles, such as for Lola expression experiments, mitotic recombination was induced in P{hsFLP}/+; {histoneGFP} {FRT 79D}/{FRT 79D}females. Adult females were subjected to heat-shock (37°C) for 1 hour,three times per day for 2 days. All fly stocks were raised on standard cornmeal molasses agar at 25°C, unless noted otherwise. Stem cell maintenance was assayed by retention of mutant stem cells over time. At 5, 7,9, 11 and 13 days post-clone-induction, ovaries were examined and scored for the number of GSC clones/ovariole. For each time-point and genotype, a minimum of 30 ovarioles were scored.

stwl bam double-mutant analysis used stwl⁹⁵bam⁸⁶ and stwl^Z1bam⁸⁶ or stwl^EYbam⁸⁶ chromosomes. All three stwl alleles acted as strong loss-of-function alleles and behaved similarly in combination with bam⁸⁶.

For eye pigment extraction, ten heads from male flies aged for 3 days were homogenized in 700 μl of methanol with 0.1% HCl and were then incubated overnight at 4°C. Following incubation, the extracts were cleared by centrifugation and measured for optical density at 488 nm. Each genotype was assayed in triplicate.

Ovaries were prepared for reaction with antibodies as described previously(Christerson and McKearin,1994). The antibodies and dilutions used were: rabbit anti-GFP(Invitrogen) (1:10,000 dilution); mouse anti-α-Hts (1B1)(Zaccai and Lipshitz, 1996)(1:750); mouse anti-BamC (A7) (McKearin and Ohlstein, 1995) (1:10); rat anti-Pum 1637(Macdonald, 1992) (1:500);rabbit anti-Vasa (Lasko and Ashburner,1990) (1:1000); rabbit anti-Bru(Sugimura and Lilly, 2006)(1:3000), rabbit anti-Lola (Giniger et al., 1994) (1:100); rat anti-Stwl F6 (K. Clark, PhD thesis,University of Texas Southwestern Medical Center, 1996) (1:100); and mouse anti-Myc (9E10, DSHB) (1:1000). Alexa Fluor secondary antibodies (Molecular Probes) were used at a 1:500 dilution.

Ovaries isolated from 3-day-old well-fed flies were incubated with anti-Hts and anti-GFP antibodies. Images were collected on a Zeiss LSM 510 Meta and projected to count the number of spherical spectrosomes/fusomes, the number of bam-GFP-negative cells and to identify differentiated cysts. Round spectrosome-containing cells that were negative for bam-GFP were scored as GSCs. GSC number was determined by scoring a minimum of 20 germaria per genotype.

Microarray analyses for stwl bam versus bam ovaries were performed at the UT Southwestern Microarray Core Facility using Affymetrix Drosophila Genome 2.0 chips, representing ∼18,500 transcripts. Ovaries from each genotype were hand-dissected and placed immediately into a Trizol RNA isolation reagent (Invitrogen). Total RNA was used to probe microarray chips. The expression of over 18,500 transcripts represented on the Affymetrix Drosophila Genome 2.0 microarray was examined in ovarian tissue samples obtained from stwl bam and bam mutant females. Triplicate hybridizations were performed for each sample. Overall,∼9000 transcripts were expressed in the tissues of each genotype.

Data analysis was performed to identify differentially expressed genes between stwl bam versus bam mutant genotypes. Using GeneSpring analysis software (Silicon Genetics), data were initially screened to identify signals counted as present and showing at least a twofold-change difference between the two genotypes. The data was analyzed using a parametric t-test and multiple correction method (Benjamini and Hochberg False Discovery Rate; P<0.05). We classified the differentially expressed transcripts into functional groups based upon GO(www.geneontology.org),and functional designation in NCBI and FlyBase notation.

A GSC maintenance defect typically results in a progressive decline in both germ cell number and in egg chamber production. In ovaries from stwlmutant females, we noticed that egg chamber number declined as the females aged. Taken together with our previous studies demonstrating that stwl acts early in germ cell development and is expressed in stem cells and cystoblasts (K. Clark, PhD thesis, University of Texas Southwestern Medical Center, 1996), we considered the possibility that stwl may play a role in stem cell maintenance. Ovaries from wild-type females retained germ cells for more than 3 weeks (Fig. 1A), but stwl mutant females showed a dramatic and rapid loss of germ cells, with approximately 50% of ovarioles lacking germ cells by 10 days post-eclosion (Fig. 1B,C). Akiyama had also reported germ cell decline while studying another collection of stwl mutant alleles and proposed that these stwl germaria lost GSCs (Akiyama,2002).

The microenvironment or niche that regulates GSC maintenance includes somatic cells of the cap and terminal filament, which produce survival and differentiation factors that are crucial for GSC maintenance. Stwl is expressed in both germ cells and in somatic cells (K. Clark, PhD thesis,University of Texas Southwestern Medical Center, 1996); thus, stwlmay be required in somatic cells to produce the microenvironment necessary for GSC maintenance and/or in the stem cell to respond to that microenvironment. To differentiate between these possibilities, we used mitotic recombination to eliminate stwl specifically in germ cells. Under conditions that produced GFP-negative GSCs (hereafter referred to as `clones') in approximately 35-40% of ovarioles (Fig. 1D,G), wild-type GSC clones were maintained for over 2 weeks,whereas most stwl^- GSC clones were rapidly depleted(Fig. 1E-G). TUNEL labeling in both wild-type and stwl^- ovaries failed to reveal any germ cell apoptosis in the more than 100 germaria that were examined (data not shown).

To distinguish further between stem cell loss by germ cell death or by differentiation, we followed cyst progression in mosaic ovaries. stwl^- clones were induced by heat-shock and animals were aged for a week or more to ensure that persisting clonal germ cells were derived from a mutant stem cell. Careful analysis of ovarioles containing stwl^- germline clones revealed germaria containing clonally related mutant germ cell clusters without a corresponding stem cell,indicating that the stem cell that gave rise to these germ cell clusters had been lost (Fig. 1D-F). Together with the lack of apoptotic germ cells, these studies suggested that stem cell loss in stwl mutants was not a consequence of cell death, but rather that stwl^- stem cells were lost by differentiation.

In addition to its requirement for GSC maintenance, stwl was also essential as a cell-autonomous factor for proper oocyte maturation. Developing egg chambers derived from stwl^- GSCs displayed several defects previously described in ovaries from stwl^- females(Clark and McKearin, 1996),including a failure of oocyte differentiation, as evidenced by abnormal oocyte karyosome formation and by a failure to concentrate Orb within a single cell(data not shown). Likewise, pycnotic nuclei and widespread TUNEL labeling in cystocytes of stage-4 egg chambers revealed that stwl^-germline clones eventually became apoptotic, as was described for homozygous stwl ovaries (Clark and McKearin,1996). By contrast, egg chambers containing stwl^- follicle cells but wild-type germ cells developed normally (data not shown), suggesting that, like GSC loss, stwl acts in germ cells rather than in soma to ensure oocyte development.

A corollary to the observation that stwl inactivation caused premature differentiation is that excess stwl expression might be sufficient to delay or block cystoblast differentiation. Although ovaries expressing ectopic Stwl in germ cells were morphologically similar to wild type, they contained extra GSC-like cells in the germarium, as measured by the presence of round fusomes (spectrosomes) and by the quiescence of a bam transcriptional reporter in germ cells at the anterior end of germarium (compare Fig. 2A with 2B). Ectopic Stwl resulted in two to three extra stem-cell-like cells per germarium, on average, compared with Gal4 control ovaries(Fig. 2C).

The extra GSC-like-cells phenotype induced by ectopic Stwl expression provided a valuable assay to identify genes that were required for Stwl action in the GSC differentiation pathway. We introduced single-copy mutations of several GSC maintenance genes into the Stwl overexpression background and scored GSC number. Introduction of single-copy mutations in the dpppathway (e.g. mad, punt, tkv) did not significantly alter the stwl overexpression phenotype (data not shown). By contrast,mutations in nos and, to a lesser extent, pum, specifically suppressed the stwl overexpression phenotype, reducing the average GSC number to three per germaria (Fig. 2C). As a control, we found that the number of GSCs in Nos-Gal4/+; nos^RC/+ and Nos-Gal4/+; Df(nos)/+ ovaries was the same as in Nos-Gal4/+; TM3/+ ovaries. These findings suggested that the action of Stwl on GSC fate depended more heavily on the contribution of Nos-Pum to the GSC maintenance pathway than on niche signaling mediated by Dpp.

Our data suggested that stwl acted autonomously in the stem cell to promote GSC maintenance or to repress cystoblast differentiation. Because the Stwl protein contains modified SANT domains, which have been implicated in chromatin regulation (Boyer et al.,2002), we considered the possibility that stwl may influence gene expression via the modification of chromatin structure or function. We noticed that Stwl was expressed in eye discs (data not shown),and took advantage of a well-described assay for chromatin-mediated gene repression to test for stwl-dependent chromatin modification(Dorer and Henikoff, 1994). Position-effect variegation (PEV) occurs when euchromatic genes are placed in,or adjacent to, heterochromatin, at which point they become subject to heterochromatic gene silencing. This silencing is often mosaic, or variegating, and can be relieved by reducing the concentration of heterochromatin-binding proteins or other associated transcriptional repressors (Karpen, 1994).

The DX1 chromosome contains a `heterochromatinized'white⁺ (w⁺) gene array that normally produces weak w⁺ expression(Dorer and Henikoff, 1994),resulting in mostly white eyes with a few spots of red eye pigment(Fig. 3A). Mutant alleles of known chromatin-regulatory genes, such as Su(var)3-9¹,which encodes a histone methyltransferase required for heterochromatin structure (Rea et al., 2000),suppressed the variegated phenotype by producing more-uniform expression of w⁺ (Fig. 3D). Surprisingly, inactivating stwl alleles acted as strong dominant Su(var) mutations(Fig. 3B,C). We quantified the degree of suppression by extracting A₄₈₈-absorbing eye pigment from fly heads of each genotype and found that a reduction in stwl gene dose resulted in a two- to three-fold increase in eye pigment, a reflection of w⁺ expression level(Fig. 2E). stwlmutants showed a similar dominant Su(var) affect when tested with the variegated w^m4 and bw^D alleles (data not shown).

One attractive candidate for a target silenced by stwl was the bam gene, because previous studies had shown that bamsilencing was required for GSC maintenance(Ohlstein and McKearin, 1997). If bam transcription were de-repressed in stwl mutants, GSCs would be lost because bam expression would cause GSC-to-cystoblast differentiation (Ohlstein and McKearin,1997). We therefore scored bam expression in stwl^- ovaries using both a Bam-specific antibody and a bam transcriptional reporter. We found, however, that bamtranscriptional silencing was properly maintained in stwl^-ovaries (Fig. 4A), indicating that Stwl is not required for bam silencing.

Mutations that cause GSC loss without disrupting bam silencing,such as those in pum and pelo, have been described previously (Chen and McKearin,2005; Szakmary et al.,2005; Xi et al.,2005). Tests of genetic epistasis have shown that double mutants of bam and the GSC-loss gene (pum bam or pelo; bam)can form differentiating germ cell cysts, indicating that bamfunction is dispensable when either pum or pelo is also absent (Chen and McKearin,2005; Szakmary et al.,2005; Xi et al.,2005). We similarly constructed stwl⁹⁵bam⁸⁶ double-mutant animals to determine whether stwl and bam also constituted an antagonistic gene pair. In approximately 80% of stwl bam germaria, we found germ cells interconnected by branched fusomes, indicating that cystoblasts had formed and initiated cyst differentiation (Fig. 4D). On average, 30% of germ cells in these ovaries were in clusters penetrated by branched fusomes. Taken together, these findings indicated that stwl antagonizes bam action by a mechanism that is independent of bam transcriptional silencing.

To identify Stwl targets whose transcriptional silencing is required for GSC maintenance, we carried out microarray analysis of undifferentiated germ cells that lack bam and compared them to stwl bam mutants. Both of these genetic backgrounds provided a nearly homogeneous population of cell types, because bam mutant cells failed to differentiate into cystoblasts and stwl bam germ cells arrested as partially formed cysts. The homogeneity and early arrest of the mutant germ cells improved the quality of microarray data by eliminating late-stage egg chambers that produce complex and abundant populations of mRNA, which can distort microarray analysis.

We identified 501 genes that were differentially expressed twofold or more(P<0.05) in stwl bam mutants relative to bammutant ovaries. Of these candidate transcripts, 235 were upregulated in stwl bam versus bam mutants. Differentially expressed transcripts designated as having functions in mRNA processing, transcription or other roles are listed in Table 1, and those with the largest differential change in expression are highlighted by bold. Transcripts that appeared as downregulated are listed in Table S1 in the supplementary material.

Because we expected Stwl to act as a transcriptional inhibitor, we concentrated our initial studies on those transcripts shown in Table 1. Among the genes upregulated in the absence of stwl, we expected to find those whose expression was directly influenced by stwl, as well as genes that are upregulated early in cyst development, because stwl bam mutant ovaries execute early cyst formation. Both classes were of interest, as few early germ cell differentiation factors have been identified to date.

Bruno (also known as Arrest - FlyBase), a translational repressor required for cyst formation (Parisi et al.,2001), appeared in greater abundance in stwl bamprofiles. Bruno accumulation did not change significantly in stwl^+/- versus stwl^-/- germline clones(Fig. 5A,B). Bruno accumulation did increase, however, throughout stwl bam germaria(Fig. 5E,F) but not in bam mutant germaria (Fig. 5C,D), correlating with cyst differentiation. Thus, Bruno was not a direct Stwl target gene, but bruno mRNA was present in greater abundance in the stwl bam transcript profile due to cyst differentiation.

Transcripts from lola, a gene implicated in chromatin organization(Zhang et al., 2003), also appeared in greater abundance in stwl bam profiles. Likewise, Lola protein was detected in 48% of the cells in stwl bam ovaries, in contrast to in 12% of bam cells and in 18% of early germ cells (GSCs to eight-cell cysts) in stwl^+/- ovaries. Lola is a dynamically expressed protein (Giniger et al., 1994), and its apparent increase in transcript profiles and in ovaries could be explained if a greater percentage of stwl bamcells was trapped at Lola-expressing stages. We therefore examined Lola expression more closely in stwl mutant and stwl mosaic backgrounds, in which cystocytes do not arrest differentiation until much later in oogenesis. Unlike Bruno, Lola protein was expressed in much greater abundance in germ cells of stwl homozygous mutant flies compared with stwl^+/- (data not shown). Analysis of stwl mosaic germlines provided the clearest demonstration of the dependence of Lola expression on stwl activity. As shown in Fig. 5G,H, almost all stwl^-/- cells in mosaic germaria expressed Lola, whereas most stwl^+/- germ cells were Lola-negative. Lola was detected in 84% of germ cells lacking stwl (54/64 stwl^-/- germ cell clones) and in only 26% of germ cells expressing stwl⁺ (15/52 stwl^+/+ germ cell clones). We concluded, therefore, that induction of lola mRNA reflected in the transcript profiling experiments was due to upregulation of lola transcription in the absence of Stwl, suggesting that Stwl directly repressed lola transcription.

As previously noted, genetic interactions had suggested a link between Stwl and Nos-Pum action. We examined the expression of Nos and Pum in stwlmutant ovaries, but found no change in the levels or localization of Nos or Pum in the absence of stwl (see Fig. S1 in the supplementary material). Because these proteins act in different cellular compartments(nucleus versus cytoplasm), we suspected that an overlapping set of targets might account for the genetic interactions. To identify Stwl-repressed mRNAs that might also be targets of Nos-Pum translational repression, we searched the 235 genes upregulated in the absence of stwl for those that contained consensus Pum-binding sites (termed Nanos response element or NRE:AUUGUA) within their 3′-UTRs (Murata and Wharton, 1995; Sonoda and Wharton, 1999). We identified 25 transcripts with one or more NRE consensus sites within their 3′-UTR, including six transcripts with two NREs and five transcripts with three NREs(Table 2). Although the number of transcripts containing one NRE in their 3′-UTR sequences was not different from random occurrence, the frequency of 3′-UTRs with multiple NREs was significantly higher in the induced group of transcripts than in a group whose levels did not change in the absence of stwl (4.7% versus 2%). Intriguingly, several of the multiple lola transcripts carried two predicted NREs within their 3′-UTR, suggesting that lolamay be a candidate for both Nos-Pum translational regulation and stwltranscriptional regulation.

Previous studies have shown that mechanisms that block differentiation are crucial for maintaining ovarian GSCs (Kai et al., 2005; Wang and Lin,2004; Wong et al.,2005). Dpp signaling within the GSC niche silences Bam expression(Chen and McKearin, 2003a; Li and Xie, 2005; Song et al., 2004), and the translational repressors Nos and Pum prevent translation of putative cystoblast-promoting mRNAs (Gilboa and Lehmann, 2004b; Wang and Lin,2004). Our finding that GSCs prominently express a marker of transcriptional quiescence is another indication that the stem cell state depends heavily on mechanisms that restrict gene expression. Similar conclusions have been made for several stem cell populations in mammals. For example, the transcriptional repressor Plzf (Zbtb16) is required for spermatogonial stem cell maintenance in the mouse(Buaas et al., 2004; Costoya et al., 2004). Plzf associates with the Polycomb family member bmi-1 (bmi1) and has been implicated in chromatin-mediated repression(Barna et al., 2002). Other adult stem cells are also likely to be regulated by chromatin state, as bmi-1 is required for the self-renewal of hematopoetic and neural stem cells in the mouse (Lessard and Sauvageau, 2003; Molofsky et al., 2003; Park et al.,2003).

Stwl carries copies of two modified SANT domains, the MADF and BESS motifs,which are found principally in flies and have been studied in several Drosophila proteins (Bhaskar and Courey, 2002; Cutler et al.,1998). SANT-domain proteins in humans and yeast are most frequently associated with chromatin-remodeling and histone-modifying activities (Boyer et al.,2002). In many instances, the SANT domains are essential for histone binding. Peterson (Boyer et al.,2004) has advanced the hypothesis that SANT domains bind histone tails in histone-modifying complexes, such as histone acetyl transferases(HAT) and histone de-acetylation complexes (HDACs). On the other hand, SANT domains in Myb, Adf1 and Dip3 have been shown to act as sequence-specific DNA-binding motifs (Bhaskar and Courey,2002; Cutler et al.,1998; Gabrielsen et al.,1991). Although we cannot exclude the possibility that Stwl acts as a sequence-specific DNA-binding protein, we favor the idea that it acts as a histone-interacting protein, because the Su(var) phenotype has been remarkably predictive of proteins involved in histone modification and chromatin remodeling (Ebert et al.,2004). For instance, Su(var)3-7 is a Drosophilaprotein with a modified SANT domain. This domain interacts with a major heterochromatin protein, HP-1 [Su(var)205](Delattre et al., 2000), and contributes to heterochromatic gene silencing(Delattre et al., 2004).

The Su(var) activity and stem-cell-loss phenotype of stwlled us to propose that Stwl is a component of a chromatin-modifying complex and that chromatin state may influence GSC fate. Indeed, the chromatin remodeler iswi also plays an essential role in maintaining GSCs(Xi and Xie, 2005), indicating a significant function for chromatin remodeling in GSC chromosomes. Our finding that stwl acts as a potent Su(var) and is required for GSC maintenance suggests that domains of chromatin silencing are established and maintained in these stem cells to regulate differentiation.

Remarkably, loss- and gain-of-function experiments documented that Stwl was both necessary and sufficient for GSC fate(Akiyama, 2002; this paper). Our data indicates that Stwl represses the expression of differentiation-promoting genes, probably by chromatin-mediated transcriptional silencing. Strikingly, imposing Stwl-mediated silencing in germ cells caused more of them to persist as GSC-like cells. These ectopic GSCs retained properties of stem cells, including bam silencing. Eventually, these supernumerary GSCs differentiated, probably when they exited the crucial niche region.

Only a few genes that can expand the number of GSCs have been identified and, in each case, these genes occupy crucial positions in the genetic circuits maintaining GSCs. Constitutive activation of the Dpp signaling pathway can block germ cell differentiation and maintain all germ cells in a GSC-like state (Casanueva and Ferguson,2004; Xie and Spradling,1998). Prolonged somatic expression of piwi can expand the number of GSC-like cells, but only within the anterior-most region of the GSC niche (Cox et al., 2000; Szakmary et al., 2005). Ourselves and others have proposed that the region defined by the limited expansion of GSCs represents a domain of graded Dpp signaling(Casanueva and Ferguson, 2004; Chen and McKearin, 2005; Szakmary et al., 2005). Overexpression of factors that enhance Dpp signaling, such as piwi(Chen and McKearin, 2005; Szakmary et al., 2005), expand the niche by `flattening' the gradient. We propose that Stwl represses GSC differentiation cell-autonomously, delaying germ cell differentiation and expanding GSC number within the limits of the niche. Our finding that an intrinsic transcriptional repressor can expand GSC number might have important implications in systems that require stem cell number to expand (for example,during growth), or in cases of cancer metastasis when cancer stem cell numbers might amplify.

In addition to Drosophila GSCs, epigenetic-silencing mechanisms have been implicated as maintenance factors in a few other stem cell systems(Buszczak and Spradling,2006). For example, the Polycomb proteins are bound to a suite of promoters in embryonic stem (ES) cells and the corresponding genes become activated when ES cells begin to differentiate(Boyer et al., 2006; Lee et al., 2006). A separate study identified another PcG protein, Ezh2, as necessary for maintenance of mouse hematopoietic stem cells (Kamminga et al., 2006).

The current paradigm for the mechanisms that regulate GSCs predicts that stem cells are maintained by blocking their differentiation. GSCs are subject to repressing activities by niche signaling(Chen and McKearin, 2003a; Song et al., 2004) and translational control (Gilboa and Lehmann,2004b; Wang and Lin,2004). Here, we propose that Stwl preserves GSCs by imposing epigenetic transcriptional quiescence that acts downstream of bam-promoter silencing and depends on full nos⁺and pum⁺ function for the preservation of GSC fate. We note that previous studies on transcriptional silencing in pole cells in Drosophila and germ cells in C. elegans have implicated Nos and its orthologous proteins in regulating epigenetic modifications(Schaner et al., 2003). In flies, nos⁺ was required to delay the accumulation in pole cell genomes of lysine 4 methylation of histone H3 (H3meK4) that would signal broad transcriptional activation (Schaner et al., 2003).

The fact that Stwl action depended on the genetic dosage of nos⁺ can be explained if Stwl silences the transcription of genes whose transcripts are also translationally inhibited by Nos-Pum. Alternatively, Stwl-dependent and Nos-dependent epigenetic mechanisms might collaborate to provide full-genome transcriptional silencing. Because transcripts bearing multiple NREs were over-represented in the class of Stwl-repressed genes, we favor the initial hypothesis, and present a model for Stwl function in Fig. 6. We propose that Stwl represses the transcription of a cohort of genes, including key differentiation factors also targeted by Nos-Pum translational repression. It is also possible that nos and stwl are both involved in regulating chromatin state in the stem cell. In pole cells, Nos activity has been shown to influence chromatin state, probably by repressing the translation of chromatin modifiers(Schaner et al., 2003). The possibility of dual mechanisms for silencing the expression of this group of genes might suggest that they can promote key steps in GSC differentiation and may require overlapping mechanisms to block their expression. Perhaps it will be possible to produce a chart of the earliest steps in differentiation from a stem cell to a differentiated daughter by noting genes that are common targets of multiple repressors.

We would like to thank Scott Cameron, Rick Jones and Bob Hammer for useful comments on the manuscript. The authors thank anonymous reviewers who offered very useful critiques to improve the manuscript. Rick Jones also provided valuable reagents and advice about gene regulation by chromatin modifications. M.W. extends her thanks to the UT Southwestern Summer Undergraduate Research Program and J.K.P. thanks the UT Southwestern Medical Scientist Training Program for advice and support. We thank members of the McKearin laboratory,past and present, as well as members of UTSW model organism laboratories for helpful discussions. We thank Mary Kuhn and Tiana Endicott for outstanding technical assistance. The UTSW Microarray Core Facility provided invaluable expertise and technical assistance. This work was supported by grants to D.M.M. from NSF (IOB-0346223) and NIH (GM45820).