A discrete transcriptional silencer in the bam gene determines asymmetric division of the Drosophila germline stem cell

Stem cells in adult tissues share two defining characteristics – cell division is self-renewing and asymmetric. In the Drosophila oogenic germline, the self-renewing aspect of the division produces a new stem cell that remains associated with a complement of somatic cells (terminal filament cells, cap cells and inner sheath cells; collectively, somatic tip cells) that maintain a supportive microenvironment(Spradling et al., 2001). Parallel investigations of the fly's spermatogenic lineage have revealed that a morphologically similar but molecularly distinct microenvironment supports spermatogenic stem cells (Tulina and Matunis, 2001; Kiger et al.,2001).

Studies from several laboratories have identified genes required in ovarian somatic tip cells that are necessary for germline stem cell (GSC) function,including piwi, dpp and fs(1)Yb(Cox et al., 2000;King and Lin, 1999;Xie and Spradling, 1998). Genes required within the GSCs include nanos (nos),pumilio (pum) and the Dpp-dependent receptors and transcription factors (Forbes and Lehmann,1998; Lin and Spradling,1997; Xie and Spradling,1998). Asymmetry of the GSC division produces a specialized cystoblast that will produce a clone of cells from which the oocyte proper will form. Mutations in such genes as pum or nos destroy asymmetry because both daughters differentiate as cystoblasts, causing the loss of GSCs (Forbes and Lehmann,1998; Lin and Spradling,1997). The opposite is true when cystoblast differentiation factors, such as bam or bgcn, are inactivated. In these cases, asymmetry is disrupted because both GSC daughters retain GSC characteristics (Gateff, 1982;Lavoie et al., 1999;McKearin and Ohlstein, 1995;McKearin and Spradling,1990).

The cystoblast is a differentiated cell because, unlike the GSC, it divides precisely four times to produce 16 daughter cystocytes, which eventually produce 15 nurse cells and one oocyte. Also, unlike the GSC, each cystoblast or cystocyte cell division executes only partial cytokinesis and thus the cystocytes remain interconnected as a cyst(de Cuevas et al., 1997).

Fig. 1 presents a schematic view of germarial Region 1 (the cyst-forming region) on which the pattern ofbam mRNA and protein accumulation is superimposed. Low levels of Bam protein are expressed in GSCs, where it accumulates in fusomes butbam mRNA is undetectable by in situ hybridization. These observations indicate that bam is transcribed at a very low level in GSCs and that this small amount of transcript produces the protein seen associated with GSC fusomes. In the cystoblast and young cystocytes,bam transcripts become readily detectable, reflecting an upregulation in transcription rate or increased stability of the mRNA. This more abundant pool of mRNA produces higher levels of Bam protein that begin to accumulate in the germ cells' cytoplasm and continues to decorate the growing fusome(McKearin and Ohlstein, 1995). As cysts mature, the abundance of bam transcripts declines such that mRNA is again undetectable in eight-cell cysts and Bam protein levels fall precipitously once the 16-celled cyst forms such that Bam is again restricted to the fusome (McKearin and Ohlstein,1995).

Although loss of bam or bgcn function produced hyperplasia of undifferentiated germ cells, ectopic Bam expression caused GSC elimination (Ohlstein and McKearin,1997) similar to that seen when GSC maintenance factors were inactivated (Xie and Spradling,1998). These data are most readily explained by assigningbam the role of a necessary and sufficient cystoblast differentiation factor (Ohlstein and McKearin,1997). A simple model to account for the experimental observations is that the microenvironment, or niche(Spradling et al., 2001),created by somatic tip cells maintains the most anterior germ cells as GSCs by suppressing bam⁺ expression, thereby preventing them from differentiating as cystoblasts.

The nature of intrinsic and extrinsic regulatory factors can provide clues about which aspects of gene expression are targeted by the signaling that sustains GSCs. The abrupt appearance of bam transcripts in cystoblasts and their persistence in cystocytes for only a few cell divisions is consistent with control of either bam transcription or mRNA stability or both (McKearin and Spradling,1990). The demonstration by Xie and Spradling(Xie and Spradling, 1998) that GSC maintenance depends on the Dpp signaling cassette suggests that Smad-dependent transcriptional control could regulate the choice between GSC or cystoblast fate (Xie and Spradling,1998; Xie and Spradling,2000). Alternatively, the asymmetric distribution of bammRNA could arise if bam transcripts were more stable in cystoblasts than in GSCs.

Previously, we had found that a reporter gene driven by a partialbam promoter could be transcribed in bgcn^–cells that could not become cystoblasts, implying that bamtranscriptional activation might precede full cystoblast differentiation(Lavoie et al., 1999). On this basis, we set out to determine the role of transcriptional control using promoter mutagenesis studies in transgenic flies. We did so with the expectation that, if transcriptional regulation was important for the GSC-to-cystoblast switch, identification of the essential promoter elements could provide significant insights into the signaling pathways responsible. Here we confirm that bam⁺ transcriptional control is necessary for proper GSC and cystoblast fate and we define the regions of thebam promoter that are required and sufficient for the GSC-low/cystoblast-high pattern of expression. We find that one transcriptional control element in the 5′-UTR acts as a silencer in the GSC and bam transcriptional silencing is required for GSC fate.

All fly stocks used in this paper were maintained under standard culture conditions. w¹¹¹⁸ flies were used as the host for all P-element-mediated transformation. Microinjections were performed as described by Rubin and Spradling (Rubin and Spradling, 1982), except that embryos were injected without chorion removal. bam^Δ86 is a nullbam allele (McKearin and Ohlstein, 1995) and was used for bam rescue tests.w; P[w⁺; UASp-GFP] was a kind gift from A. Ephrussi's laboratory.

Constructs testing promoter activity used either green fluorescent protein(GFP) alone or Bam:GFP recombinant protein. Although Bam:GFP protein could provide full rescuing activity, the distribution of GFP fluorescence from fusion protein did not entirely recapitulate native Bam protein localization because fusomes were GFP-negative even when the Bam:GFP transgene was the only source of functional Bam protein. The basis for this difference from native Bam distribution (McKearin and Ohlstein,1995) is currently the object of study.

Unless stated otherwise, all constructs were made using the CaSpeR4 transformation vector (Pirrotta,1988). The P[bam promoter-Gal4:VP16] transgene used a derivative of P[hsp70-CaSpeR] that contained the promoter and 3′UTR from the hsp70 gene (gift of D. Smith). The various constructs described in this paper were generated using primers listed inTable 1.

PCR reactions to recover promoters that were truncated from the 5′-side were anchored on the 3′-end by an antisense primer at position +133 (Table 1). These PCR fragments were joined to Bam:GFP or GFP reporter gene with appropriate restriction enzyme digestions and subsequently linked to the bam3′UTR in pCaSpeR4.

Fragments truncated from the 3′-end were generated by a set of antisense primers (Table 1) and were anchored at the 5′ side with oligonucleotides corresponding to positions –898, –799 or –698(Table 1). These PCR fragments were cloned upstream of a GFP reporter and linked to bam 3′UTR in pCaSpeR4.

Fragments for sufficiency tests were recovered by PCR or assembled by annealing synthetic oligonucleotides. Promoters containing small deletions were synthesized directly by PCR using primers that lacked the targeted residues. A two-step strategy was used to construct promoters with internal deletions; separate PCR products were ligated such that the targeted nucleotides were deleted and the ligation products were used as templates for subsequent PCR. We used nos and P-element minimal promoter sequences that had been identified previously (Van Doren et al., 1998; Rorth,1998) for constructing heterologous promoters.

A two-step PCR was performed to amplify the fragment containing thebam promoter and an inverted silencer element. Fly genomic DNA was used as template for first round PCR. After the first PCR reaction for 10 cycles, resulting PCR products were used as template for second round PCR. The final PCR products were subcloned into CaSPeR vector in front of a GFP reporter.

Two-step PCR method:

1. Sense primer: bam-898/KpnI.

2. Antisense primer 1 for first round PCR: bam-as/SE/s(5′-cgcagacagcgtggcgtcagctaacgtatcgatatcatg-3′).

3. Antisense primer 2 for second round PCR: bam/SE/s(5′-cgcaccggtattcgcagacagcgtggcgtc-3′).

Germline clones were produced by inducing the expression of autosomal P(hsp70-FLP) transgenes at 37°C for 1 hour. Heat shock was repeated at least two more times on the first day and the same flies were occasionally heat-shocked again by the same protocol on the next day to achieve higher production of clones. Cells made homozygous for brkmutant alleles were recognized as GFP-negative against a background of GFP-positive cells.

GFP fluorescent images from reporter transgenes and germline clones were collected on a Zeiss LSM410 confocal microscope. Samples were evaluated as single optical sections in the plane that included the terminal filament to make certain that GSCs were included in the image. The optical files were processed using Adobe Photoshop 6.0 and Deneba Canvas 7.0.

Total RNA was isolated from 0-3 h collections of w¹¹¹⁸embryos and digested with RNase-free DNase. cDNA was prepared by reverse transcription using an antisense bam primer and tagged with a poly(A)tail using terminal deoxynucleotide transferase according to manufacturer's instructions (Promega). The ends of tailed cDNA were amplified by PCR with oligo-dT and bam-specific primers. The resulting PCR products were cloned into the pGEM T-Easy vector (ProMega) and sequenced by a UT-Southwestern DNA Sequencing facility.

RNA in situ hybridization had previously shown that asymmetric bamexpression between the GSC and cystoblast could be seen at the level of mRNA expression (McKearin and Spradling,1990). This finding was consistent with transcriptional control ofbam expression. More recently, we found that a GFP reporter transgene, driven by a bam promoter fragment, was efficiently expressed in bam mutant cells but not in GSCs (not shown). This finding provided additional evidence that activation of bamtranscription preceded cystoblast differentiation because bam cells cannot progress to cystoblasts (McKearin and Ohlstein, 1995). Both of these observations supported the premise that bam transcriptional control was an early and crucial step in the differentiation of stem cell daughters into cystoblasts. This conclusion led us to undertake a detailed analysis of bam promoter elements.

To begin our studies, it was necessary to define the precise transcriptional start site. Initial RNAse protection assays had defined the start site at a sequence `TTCTGGG'(McKearin and Spradling,1990), but more precise RACE assays placed the transcriptional start site at `gccAtcattctt'. This site will be designated as `+1'in this paper for the purposes of orienting bam promoter elements.

Previously we found that a genomic fragment of 1008 bp immediately upstream from the +1-site plus the 5′-UTR could fully rescue bamphenotype (McKearin and Spradling,1990), and subsequently had refined essential promoter sequences to –898 bp plus 5′-UTR (B. Ohlstein and D. M., unpublished). Both promoters recapitulated the native bam pattern by driving expression in cystoblasts and young germ cells, but remained quiescent in GSCs. We concluded that all sequences necessary for bam⁺transcriptional control were contained within the sequences spanning from–898 bp 5′ to the bam transcriptional start site through 133 bp of the bam 5′-UTR.

We constructed P[UASp-Bam:GFP] transgenes to test the effects of expressingbam⁺ from a variety of different promoters. By itself, the construct produced no detectable GFP expression. When expressed from anos promoter in P[UASp-Bam:GFP];P[nosP-Gal4:VP16-nos3′UTR] flies, adult ovaries from young females contained only a few maturing eggs but no other germline cells(not shown). This was the same phenotype obtained whenbam⁺ was expressed from a heat shock promoter(Ohlstein and McKearin, 1997)or when any of several GSC maintenance genes were inactivated by mutation(Forbes and Lehmann, 1998;Lin and Spradling, 1997;Xie and Spradling, 1998;Xie and Spradling, 2000). Thus the nos promoter, which directed bam⁺ expression in all PGCs and adult germ cells including GSCs, caused GSC ablation. We also noted that the testes of [UASp-Bam:GFP];P[nosP-nos5′-Gal4:VP16-nos3′UTR] males were shriveled but did not determine the basis of the defect. In contrast to heterologous transcriptional control of bam⁺ expression,oogenesis was normal and GFP was expressed in cystoblasts and young cystocytes but not GSCs when a native bam promoter controlled transgene expression in P[bamP-bam5′-Gal4:VP16-hsp70-3′UTR];P[UASp-Bam:GFP] flies. This combination of transgenes also fully rescuedbam mutant flies. Because levels of GFP reporters suggested that the relative `strength' of the nos and bam promoters was similar in cystoblasts (Van Doren et al.,1998), the different results obtained with bam andnos promoters predicted that the timing of bam transcription was critical for GSC fate. In addition, as GFP expression in P[UASp-Bam:GFP];P[nos-Gal4:VP16] animals recapitulated the nos expression pattern and not the bam pattern, we concluded that the Bam ORF and 3′-UTR were not sufficient for the wild-type bamexpression.

To determine whether the 3′-UTR contained necessary regulatory signals, we examined the GFP expression in P[bamP-bam5′-Gal4:VP16-hsp70-3′UTR];P[UASp-GFP] flies and found that GFP expression pattern mimicked wild-typebam mRNA accumulation (not shown). Thus the bam 3′-UTR was not required for the bam expression pattern.

Consistent with the conclusion stated above, comparison of the expression of Bam:GFP and GFP reporters using bam promoters showed that the Bam ORF did not contain sequences that were necessary or sufficient for the wild-type bam expression pattern. Inclusion of the Bam ORF in reporter constructs, however, did affect perdurance of reporter protein in maturing cysts such that GFP fluorescence was much shorter-lived with Bam:GFP than GFP alone (for example, see Fig. 2). This probably reflects instability of Bam protein in maturing cystocytes, as has been postulated previously based on the distribution of native Bam protein (McKearin and Ohlstein,1995).

A series of deletions from the 5′-end of the bam promoter defined cis-acting regulatory elements required for bamexpression (Fig. 2). We assayed four aspects of expression of constructs that fused promoter fragments to a Bam:GFP chimeric protein: (1) germ cell specificity of GFP fluorescence,relative GFP levels in (2) GSCs and (3) cystoblasts and (4) rescuing activity of the tested transgene. The full bam promoter (–898) showed no GFP expression in GSCs and strong expression in cystoblasts and young cystocytes (Fig. 2A). For example, Fig. 2B.a shows an example GFP-negative GSC and a GFP-positive cystoblast within the same germarium. The GSC can be recognized because it is the anteriormost cell in the germarium and carries a spherical fusome (arrowhead) at the most anterior position, whereas the cystoblast is immediately posterior to the GSC and has a fusome (arrow) that is located more centrally within the cell.

Deletions of sequences between –898 to –198 showed no effect on the timing, germ cell specificity or relative intensity of reporter activity(Fig. 2A,B.b). In addition,bam/bam flies expressing Bam:GFP protein from these truncated promoters showed complete rescue because females developed wild-type ovaries and retained an active GSC population(Fig. 2A). Expression from a promoter fragment that contained sequences starting at –96 was very low or, in approximately half the ovaries assayed, was not detectable at all. Consistent with its apparent weak promoter activity, this construct was also unable to rescue bam/bam flies(Fig. 2A). Nevertheless, when Bam:GFP protein was produced at detectable levels, its accumulation was restricted to cystoblasts and the cystocytes of developing cysts(Fig. 2B.c). Thus the sequence between –198 and –96 provided element(s) that enhanced transcriptional output but did not effect the timing or cell specificity of transcription. Additional deletions refined the limits of quantitative element(s) to between –168 and –140(Fig. 2B.e,f).

When deletions from the 5′-end extended to the –47 position,all GFP expression was eliminated (Fig. 2A,B.d). Thus sequences between –96 and –47 contained elements that could direct germ cell-specific transcription and were essential for bam promoter activity. The –96 to –47 region could include germ cell enhancer sites or elements required for basal transcription or both, but the 5′-deletion series could not distinguish between these possibilities. Strikingly, the data from the 5′-deletion constructs could also not explain bam transcriptional asymmetry in GSCs and cystoblasts.

A series of deletions from the 3′-end of the bam promoter +5′-UTR fragment complemented the data obtained from the constructs described above. Promoters contained sequences as far upstream as –799 and were truncated at various positions within or immediately outside of the 5′-UTR to test for elements that might be included within the UTR region(Fig. 3A). Promoters that started 799 bp upstream of the transcriptional start site and included at least 55 bp of the 5′-UTR reproduced the GSC-negative,cystoblast-positive pattern of bam expression(Fig. 3A,B.a andFig. 3B.b). Note that, because these constructs included only the GFP reporter protein and were therefore not targets of Bam-dependent degradation, cystocytes in fully formed and developing cysts were GFP-positive.

Additional deletions in this series revealed a new feature of bamcis-regulatory sequences. When the promoter fragment extended only to +4,reporter activity expanded to include GSCs as well as germ cells in cysts and beyond (Fig. 3A,B.c). Although other deletions from the 3′-side that reached as far as –20 continued to express GFP in all germ cells including GSCs(Fig. 3A,B.d), GFP expression was eliminated when the 3′ deletion series reached the –47 position (Fig. 3A,B.e). Thus sequences between –46 and –20 were essential for any transcriptional activity, whereas sequences between +5 and +55 were necessary to silence bam expression in GSCs. Combining these observations with the results of deletions from the 5′-end of the bam promoter fragment, we concluded that sequences between –96 and –20 provided positively acting elements sufficient for young cystocyte-specificbam transcription, and that critical components of a silencer element that blocked bam expression specifically in GSCs lay between +5 and+55.

Previous studies had shown that bam misexpression from ahsp70 (Ohlstein and McKearin,1997) or nos promoter(Fig. 4A,B) that extended the domain of bam expression could cause specific GSC loss. The most attractive model to explain these effects was that inappropriate temporal control of bam transcription, specifically misexpression in GSCs, was responsible for GSC ablation. Transgenes that expressed Bam:GFP from promoter fragments that lacked the silencing element but retained other native promoter elements would provide a powerful test of this hypothesis. In fact, young females that expressed a functional Bam:GFP protein controlled by a promoter fragment from –799 to +4 showed the expected `few eggs – no germline' phenotype associated with GSC elimination(Fig. 4C,D). The effect was fully penetrant but showed variable expressivity when the transgene was present in one copy; two copies increased expressivity to 100%. These observations indicated that bam transcriptional quiescence was necessary for GSC fate and that regulation of the silencer element was the key component of the switch between GSC and cystoblast fate.

In order to delineate the silencer element more precisely, we tested the ability of various fragments to suppress activity of a promoter containing the germ cell enhancer and basal elements. We chose the (–799 to –20)promoter fragment as the germ cell `constitutive promoter' because it directed GFP expression in GSCs and remaining germ cells but not ovarian somatic cells(Fig. 3B.d). When an oligonucleotide containing (+5 to +55) was fused to this basal promoter,cystoblasts and cystocytes were GFP-positive but GSCs were negative (not shown), indicating that this fragment carried full silencer activity.

One of the 3′-deletion constructs (–799 to +34) had caused reproducible intermediate silencer activity(Fig. 5B). We suspected that this truncation might leave some of the silencer element intact and noted that the truncation breakpoint fell within an imperfect palindromic sequence(+27-CGCaGaCaGCG to +37; Site A). We probed the structure of the silencer element by constructing (–799 to +55) promoters that precisely lacked the palindromic sequence. The GFP reporter from this construct was active in GSCs as well cystoblasts and cystocytes(Fig. 5C), indicating that residues +27 to +37 were essential for full silencer activity. The palindromic element was not sufficient for silencer activity, however, because promoters constructed by fusing an oligonucleotide encoding only Site A to a (–799 to –20) basal promoter permitted GFP expression in both cystoblasts and GSCs (Fig. 5D). These assays demonstrated that the palindromic sequences were an essential part of the silencer element but were insufficient to block bam transcription in GSCs.

We discovered that the (+27 to +37) oligonucleotide used in this experiment carried the sequence `CGCAGACTGCG' instead of the native bam sequence`CGCAGACAGCG'. Although it seemed unlikely, we could not eliminate the possibility that silencer activity of this element might have been destroyed by the mutation that generated an even more symmetrical palindrome. We subsequently constructed a D. yakuba bam promoter-GFP reporter that allowed us to test the efficiency of this silencer sequence because the D. yakuba Site A sequence corresponds to the perfect palindrome. We found that the yakuba promoter showed the same GFP expression pattern as wild-type melanogaster bam and concluded that `CGCAGACAGCG' and`CGCAGACTGCG' palindromes were both efficient silencers in the context of thebam promoter sequences.

Although promoters that included only to +37 permitted GSC expression,promoters that extended to +55 did not, indicating that a fully active silencer must include sequences between +37 and +55. We noted that the eight base pairs directly adjacent to the palindromic element closely matched the consensus binding site (Site B) for a known sequence-specific transcriptional repressor, Brinker (Campbell and Tomlinson,1999; Jazwinska et al.,1999; Minami et al.,1999; Rushlow et al.,2001; Saller and Bienz,2001; Kirkpatrick et al.,2001; Zhang et al.,2001). We therefore tested whether Site B was part of the silencer element sequences. A (–799 to +55) construct that contained a deletion of base pairs +37 to +39 expressed GFP in GSCs(Fig. 5E), suggesting that Site B was important for full silencer activity. Note that this deletion altered Site B but reconstituted a complete Site A sequence. To determine whether Sites A and B were sufficient for the silencer element, we joined an oligonucleotide encoding (+25 to +45) (Materials and Methods) to the(–799 to –20) constitutive germ cell promoter. GFP expression from this construct was strong in cystoblast and cystocytes but was undetectable above background levels in GSCs (Fig. 5F), demonstrating that full silencer activity requires only sequences contained in Sites A and B.

Despite the fact that Site B matched a Brinker consensus binding element,the results of subsequent experiments made Brinker a poor candidate for an in vivo regulator of bam transcription. Ovarian β-galactosidase expression in brk[XA]/+ females that carry a P[lacZ]enhancer-trap transposon in the brk gene(Campbell and Tomlinson, 1999)was restricted to cap, terminal filament and inner sheath cells, but, notably,was not expressed in any germline cells (not shown). Furthermore, germline clones of brk[M68] loss-of-function allele produced wild-type egg chambers and GSCs were stable for at least two weeks after clone induction(Fig. 5G). Thus, germline cells neither express nor require Brinker for GSC maintenance or cyst development.

The fact that the silencer was in the 5′-UTR raised the possibility that it acted at a post-transcriptional level. We tested this possibility by making bam transcripts that included the silencer element but were transcribed from the nos promoter that is active in all germline cells including GSCs (Van Doren et al.,1998). Transgenic flies carrying the construct[nosP-bam5′-GFP-bam3′](Fig. 6A) expressed GFP in GSCs, cystoblasts and other cystocytes at approximately equivalent levels(Fig. 6B). We examined the structure of the [nosP-bam5′-GFP-bam3′]mRNA to verify that these transcripts included the silencer element motif using RT-PCR and primers that would bracket potential start sites(Fig. 6A). RT-PCR of RNA from[nosP-bam5′-GFP-bam3′] flies showed that the transcript produced from the transgene started very close to or precisely at the native bam gene transcription start site(Fig. 6C). Therefore, thebam silencer element did not act post-transcriptionally because the reporter mRNA carrying the element was stably expressed in GSCs.

We confirmed the transcriptional role of the silencer element by constructing a [bamP-SE^rev-GFP] transgene that carried an inverted silencer (Materials and Methods). This transgene expressed GFP in the same pattern as a wild-type construct (not shown), demonstrating that the silencer element was active in either orientation. Because the inverted silencer would encode a sequence that was the complement of wild-type in the[bamP-SE^rev-GFP] mRNA, we concluded that the silencer acts as a DNA, not RNA, element.

Although transcripts from the[nosP-bam5′-GFP-bam3′] transgene showed that the silencer did not act as an RNA cis-element, GFP expression in GSCs also revealed that the silencer element did not block transcription from the nos promoter. This observation suggested that effective transcriptional silencing by the bam element was sensitive to context. As an additional test of silencer properties, a bam fragment containing a fully active silencer (–16 to +133) was placed upstream of the basal bam promoter (–799 to –20). GFP was expressed in GSCs, cystoblasts and cystocytes in flies carrying these constructs,indicating the silencer was inactive in this position(Fig. 7A).

Previous experiments had shown that the fragment from (–96 to–20) was sufficient for promoter activity in germline cells and that fragments (–96 to –47) and (–46 to –20) were necessary(Fig. 3). We reasoned that these regions contained enhancer element(s) that directed germ cell-specific transcription and a core promoter element, respectively.

We sought to identify essential enhancer sequences experimentally by joining bam fragments to a previously identified core promoter from the Drosophila P-element (Rorth,1998). The transcriptionally inactive (–799 to –47)fragment (Fig. 4E) was joined to the P-element core promoter driving the GFP reporter gene(Fig. 7B). Flies carrying this construct expressed GFP in GSCs, cystoblasts and cystocytes just as had been observed for the (–799 to –20) bam promoter fragment(Fig. 4D). Thus the (–47 to –20) region could be replaced by a heterologous minimal promoter element and at least one germ cell-specific enhancer lies in the region between –96 and –47.

An additional deletion series refined the position of the enhancer by revealing that promoters lacking nucleotides from –86 to –61 deletion were inactive (compare Fig. 7C and D). Deletion of a GC-rich block of nucleotides within the required region (–68 to –61) also extinguished germ cell expression of the GFP reporter (Fig. 7E). Thus,`gcgacggc' block of nucleotides was essential for the bam enhancer element. Although the `gcgacggc' element matches the consensus binding site for the Mad protein (Kim et al.,1997), we note that Xie and Spradling(Xie and Spradling, 1998) have shown previously that mad^–/– cysts develop normally. It is therefore unlikely that this essential bam promoter element binds Mad in germ cells.

Although bam is expressed throughout cyst formation(McKearin and Ohlstein, 1995),it is first required for asymmetry of the oogenic stem cell division in which it acts as the primary cystoblast differentiation factor(McKearin and Ohlstein, 1995;McKearin and Spradling, 1990). Prior studies suggested that GSC and cystoblast differentiation was sensitive to the levels of Bam and that bam expression was tightly regulated(McKearin and Ohlstein, 1995;McKearin and Spradling, 1990). It was not clear, however, whether regulation was pre- or post-transcriptional or a combination of mechanisms. This is an important issue because GSC-intrinsic factors acting at the level of transcription (Smads) and post-transcriptionally (Pum/Nos) have been identified as critical regulators of GSC fate (Xie and Spradling,1998; Forbes and Lehmann,1998; Lin and Spradling,1997). Accumulation of bam mRNA to detectable levels correlates precisely with the choice between the stem cell and cystoblast cell fate, suggesting that mechanisms regulating mRNA abundance could be important factors in the GSC-to-cystoblast switch.

Our collection of transgenes allowed us to address the role of post-transcriptional control in regulating GSC to cystoblast differentiation. Transgenic reporters driven by bam promoters but in which GFP replaced the bam ORF and hsp70 3′-UTR replaced thebam 3′-UTR were expressed in cystoblasts and cystocytes but not in GSCs, demonstrating that sequences in the bam ORF and 3′-UTR were dispensable for proper regulation. Animals that produced abam5′-GFP-bam3′ reporter from a nospromoter, however, expressed GFP in GSCs as well as other germ cells. RT-PCR showed that the transcript produced from the P[nosP-bam5′-GFP-bam3′] transgene included all of the bam 5′-UTR sequences and would therefore contain all elements that could direct post-transcriptional regulation. Because GSCs were GFP-negative when the same reporter was expressed from fully active fragments of the bam promoter, we concluded that cell-specific transcriptional activation was the first and most significant level of control for regulating differential bam expression in stem cells and cystoblasts.

Alternative explanations for differential bam expression are that positively acting elements first become active in cystoblasts, that negatively acting elements function specifically in GSCs to repress bamtranscription or a combination of these mechanisms. The fact that bampromoters that lacked the silencer element could direct reporter expression in germ cells, including GSCs, demonstrated that all germ cells contained the factors necessary to activate the bam promoter but that only GSCs contained factors to repress bam transcription. Because Bam is necessary and sufficient for cystoblast differentiation(Ohlstein and McKearin, 1997),the choice between GSC or cystoblast fate can be reduced to understanding thecis- and trans-acting factors that activate and repressbam transcription.

The native pattern of bam mRNA expression alerted us to at least two features of bam transcription; it was active only in germ cells and it was undetectable in GSCs. The sum of control elements should explain activity in this pattern and we therefore examined putative promoter deletions for their capacity to (1) limit bam expression to germ cells and (2)limit bam expression to cystoblasts and very early cystocytes.

The 5′-deletion series showed that the minimal bam promoter was small because a fragment containing as few as 168 bp upstream of the transcriptional start site was sufficient to reproduce all aspects of nativebam expression. Additional deletions demonstrated that sequences between –168 and –140 accounted for quantitative effects, but additional functional assays are necessary to identify the specific elements that regulate transcriptional output.

Even fragments that could not produce wild-type levels of transcription,however, retained cell specificity of the remaining transcriptional activity. At a minimum, these elements (Fig. 8) include (1) an enhancer element between –86 and –61 that directs bam transcription in GSCs, cystoblast and germline cyst cells and (2) a silencer between +27 and +44 that blocks enhancer activity in GSCs, (3) sequences between –47 and –20 that are essential for transcriptional activation and can be replaced by a P-element core promoter.

The 3′ series of deletions showed that promoters that retained at least to –20 could reproduce the wild-type bam expression profile but promoters that retained only to –46 could not. Because fusion of the (–799 to –46) fragment to the P-element minimal promoter could recapitulate bam-like expression, we concluded that–46 to –20 contained the `basal elements' that would interact with the core RNA polymerase subunits(Kadonaga, 1998;Reinberg et al., 1998). We noted that the bam promoter did not comply with the rules for a TATA-dependent promoter because none of the sequences within the essential regions matched the TATA element consensus(Zenzie-Gregory et al., 1992). Residues –24 to –17 (TTAACAA)(Fig. 8) could represent a functional variation on the TATA motif(Singer et al., 1990), but the fact that promoters that included only to –20 were active means that the full motif is not required for normal transcriptional activity. `TATA'-less promoters that lack the `TATAA' box and bind core RNA polymerase proteins with Initiator (Inr) cis-acting sites have also been characterized(Smale, 1997). The efficiency and accuracy of Inr elements is increased in the presence of two other sequence motifs, the BRE and DPE (Kutach and Kadonaga, 2000). The sequences surrounding the bamtranscriptional start site, however, do not match the consensus for Inr, BRE or DPE elements. Thus, other elements between –46 and –20 might function as `TATA-less' RNA polymerase recruitment sites in the bampromoter or `TTAACAA' at –24 to –17 functions as a `TATAA' boxin vivo and can be adequately replaced in transgenic constructs.

The same transgenes that identified the core element(s) of the bampromoter revealed that a germ cell-specific enhancer(s) was located upstream of –47. Deletions from the 5′-side showed that the upstream border of this enhancer region was located at –96. Additional deletions showed that essential enhancer sequences lay between –86 and –61. The sequence `gcgacggc' (Fig. 8)matches the binding consensus site for the Dpp-activated transcription factor Mad (Kim et al., 1997) and specific deletion of that eight-base pair element inactivated the bampromoter (Fig. 7). Genetic studies, however, have shown that Mad is not required for cystoblast and cyst differentiation (Xie and Spradling,1998) and is therefore not essential for bamtranscription. We suspect, therefore, that another activating transcription factor with a Mad-like recognition sequence binds to this site inbam. The significance of other sequences within this enhancer region will be determined only by additional transgenic studies with site-directed mutant constructs.

Deletions that caused the spreading of reporter expression into GSCs and constructs that tested the sufficiency of specific bam promoter fragments revealed that an 18 bp silencer element (SE;Fig. 8) regulated bamexpression. Although the SE was located in the 5′UTR, experiments that distinguished between transcriptional and post-transcriptional effects demonstrated that the bam SE repressed transcription. Chimeric promoters that joined heterologous sequences to bam fragments revealed several other features of the silencer element. The bam SE could not block GSC transcription from a nos enhancer/promoter fragment nor when the SE was placed upstream of the bamenhancer/basal region. Both of these observations can be explained if the SE's range was limited because either construct would place the silencer much further from the relevant enhancer elements. Such short-range repressive effects have been described for the dCtBP family of transcriptional repressors(Chinnadurai, 2002;Mannervik et al., 1999). Other explanations are equally possible, however, and resolution of the silencer's mechanism of action must await biochemical studies with silencer-binding proteins.

The structure of silencer, as revealed by complementary sets of necessity and sufficiency tests, suggests that at least two proteins bind to the site. The palindromic part of the silencer element (Site A) was essential for activity but it does not correspond to any known transcriptional regulatory protein-binding site. The more 3′ part of the silencer (Site B)corresponds to a consensus site for Brinker binding(Sivasankaran et al., 2000;Zhang et al., 2001). Brinker has been identified previously as a transcriptional repressor of Dpp target genes and was therefore a reasonable candidate for a silencer binding protein. Brinker is an antagonist of the Dpp-signaling because it prevents the transcription of a subset of Mad activated genes(Kirkpatrick et al., 2001;Rushlow et al., 2001;Saller and Bienz, 2001), but Brinker is itself a Dpp target (Minami et al., 1999) because Mad and Schnurri heterodimers can blockbrk transcription in dpp-responsive cells(Marty et al., 2000). The paradigm established from these studies for regulation of brkexpression predicted that the gene would be transcriptionally repressed in cells with activated Dpp signaling, such as GSCs. Indeed, previous studies demonstrated that both Mad and Shn are active in GSCs(Xie and Spradling, 1998;Xie and Spradling, 2000) and are therefore in place to repress brk expression. In addition,enhancer trap expression and germline clonal analyses did not support a role for Brk in germ cells, making it unlikely that Brk could play a significant role in keeping bam transcriptionally quiescent in GSCs. Perhaps another repressor protein that recognizes Brk-like elements binds tobam silencer Site B and forms a complex with the palindrome-binding protein to block bam transcription in GSCs.

The discovery of the silencer element permitted refinement of a model explaining the GSC-to-cystoblast switch. Our dissection of the bampromoter shows that bam can be expressed in GSCs because promoters lacking the silencer element are transcribed in all germ cells. For the same reason, the silencer does not work by blocking an overlapping enhancer site because silencer sequences can be deleted without disrupting enhancer-dependent transcription (Fig. 3C). Our current working model(Fig. 9) is that the silencer is occupied in GSCs and bam is transcriptionally quiescent because the germ cell-specific enhancer(s) cannot be engaged. When the presumptive cystoblast is born, silencer element-dependent repression can be relieved,allowing enhancer(s) in bam promoter to activate transcription and promote cystoblast differentiation. The clear implication of this model is that bam silencer occupancy is the basis of the asymmetric GSC division and that identification of silencer binding proteins can provide a mechanism for the asymmetry.

Previous studies showing that somatic cell signaling maintained GSCs by suppressing cystoblast formation suggested that bam might be the target of these signals (King and Lin,1999; Xie and Spradling,1998). We propose that occupancy of the bam silencer element in the germ cells adjacent to cap cells (i.e. GSCs) is signal dependent and is based on asymmetric activation or distribution of silencer element-binding proteins (SEBPs). For example, with the division of the neuroblast mother cell as a paradigm (Doe,2001; Jan and Jan,2001), perhaps asymmetric partitioning of one or multiple SEBPs accounts for cystoblast-specific bam transcriptional activation. Alternatively, differences in extrinsic signaling could produce post-transcriptional modifications of SEBPs to account for differential activity in GSCs and cystoblasts. Distinguishing between these and other possibilities to explain the molecular details of the GSC to cystoblast switch requires the identification of SEBPs and understanding the signaling circuits that control the occupancy of the bam silencer element.

The authors thank Gerard Campbell, Anne Ephrussi and Dean Smith for generously providing reagents and fly stocks. We also gratefully acknowledge the resources provided by the Bloomington Drosophila Stock Center and Berkeley Drosophila Genome Project. Mary Kuhn was responsible for much of the immunohistochemistry and tissue dissection represented in the Figures. Cynthia Lavoie and Ben Ohlstein made important observations about thebam promoter, and provided valuable comments, along with anonymous reviewers, during the course of this work and preparation of the manuscript. Finally, we thank Ting Xie and Chip Ferguson for kindly providing relevant data prior to publication.