Downregulation of the tissue-specific transcription factor Fork head by Broad-Complex mediates a stage-specific hormone response

The steroid hormone 20-hydroxyecdysone (20E) coordinates the major developmental transitions during the life cycle of Drosophila melanogaster. Larval molts, puparium formation, pupation and metamorphosis are all controlled by pulses of this hormone (Riddiford, 1993). The primary effects of 20E are mediated by a heterodimeric nuclear receptor, formed by EcR and the Drosophila homologue of RXR, Ultraspiracle (Usp) (Koelle et al., 1991; Thomas et al., 1993; Yao et al., 1992; Yao et al., 1993). Like other nuclear receptors, EcR/Usp acts as a ligand-dependant transcription factor that modulates the activity of hormone-responsive target genes (Mangelsdorf et al., 1995; Riddiford et al., 2000; Thummel, 1995). In Drosophila, the hormone-receptor complex activates regulatory hierarchies that control development. These hierarchies have been investigated in some detail for two of the consecutive 20E pulses that initiate metamorphosis. The first of these pulses marks the end of larval development and signals the onset of metamorphosis, while the second pulse, in the late prepupa, is required for progression into the pupal stage, during which adult development is completed (Richards, 1997; Russell and Ashburner, 1996; Thummel, 1996). Each hormone pulse leads to the direct induction of overlapping but non-identical sets of a few ‘early’ genes. These genes are thought to control much larger sets of more than 100 ‘late’ genes (Ashburner et al., 1974). In support of this concept, three early genes, the Broad-Complex (BR-C), E74 and E75, each encode multiple forms of related transcription factors (Burtis et al., 1990; DiBello et al., 1991; Segraves and Hogness, 1990).

The BR-C is a complex early gene that encodes four classes of proteins, each characterized by a unique pair of C₂H₂ zinc-finger motifs (Z1, Z2, Z3 and Z4) (Bayer et al., 1996; DiBello et al., 1991). All BR-C proteins share a common N-terminal ‘core’ region that contains the evolutionarily conserved BTB/POZ domain, a protein interaction domain that is also found in other zinc-finger DNA-binding proteins (Bardwell and Treisman, 1994; Zollman et al., 1994). Genetically, the BR-C is defined by three lethal complementation groups, reduced bristles on the palpus (rbp), broad (br) and 2Bc, as well as a group of non-complementing npr1 (nonpupariating 1) alleles that behave like deletions of the BR-C (Belyaeva et al., 1980; Kiss et al., 1988). It has been established that rbp⁺ function is provided by Z1, 2Bc⁺ function by Z3, and br⁺ function by Z2 isoforms (Bayer et al., 1997; Crossgrove et al., 1996; Emery et al., 1994). Z2 can also provide partial 2Bc⁺ function and Z4 provides partial rbp⁺ function (Bayer et al., 1997). Null mutations of the BR-C (npr1) lead to larval lethality before puparium formation. Partial loss-of-function mutations that affect single BR-C functions have pleiotropic effects during metamorphosis, interfering with both histolysis of larval tissues and differentiation of adult structures, and lead to defects in chorion formation during oogenesis (Belyaeva et al., 1980; Brennan et al., 1998; Deng and Bownes, 1997; Kiss et al., 1988; Restifo and White, 1991; Restifo and White, 1992).

The BR-C is not only required for the control of late gene expression at puparium formation (Crossgrove et al., 1996), but also earlier, in the middle of the third larval instar, for activation of the intermolt or salivary gland secretion protein (Sgs) genes (Crowley et al., 1984; Karim et al., 1993; von Kalm et al., 1994). These genes are thought to be induced in response to a small, transient increase in 20E titer that precedes the late-larval 20E pulse (Hansson and Lambertsson, 1983). As the Sgs genes have a very specialized function, directing synthesis of a glue that adheres the pupa to a solid substrate, they are exclusively expressed in the salivary glands of mid- to late-third instar larvae. They have therefore been widely used as a model system to understand how a systemic hormone signal is refined into a strictly tissue- and stage-specific response (Lehmann, 1996). An exceptional feature of the Sgs genes is that they are not only activated by 20E but also repressed by the hormone at the end of larval development (Ashburner, 1972; Crowley and Meyerowitz, 1984).

Activation of Sgs genes requires the wild-type function of the BR-C, as well as intact recognition sites for the transcription factor Fork head (Fkh)(Bayer et al., 1997; Karim et al., 1993; Lehmann and Korge, 1996; Mach et al., 1996; von Kalm et al., 1994). Fkh is a member of the family of HNF-3/Fkh proteins (also referred to as winged helix proteins) that include critical developmental regulators not only in insects but also in mammals and nematodes (Gajiwala and Burley, 2000; Kaufmann and Knöchel, 1996). The fkh gene has homeotic functions and is required for development of the ectodermal portions of the gut and the salivary glands (Jürgens and Weigel, 1988; Weigel, 1989). In fkh mutant embryos, two salivary gland placodes form, but are removed early in development by apoptotic cell death (Myat and Andrew, 2000). The spatial restriction of Fkh expression in third-instar larvae suggests that it plays a pivotal role in determining the tissue specificity of the Sgs hormone response (Lehmann and Korge, 1996; Mach et al., 1996). The temporal control of the response is mediated by the rbp function of BR-C (Bayer et al., 1997; Karim et al., 1993; von Kalm et al., 1994) and direct binding of EcR/Usp to the Sgs regulatory regions (Lehmann and Korge, 1995; Lehmann et al., 1997). Factors that are likely to contribute to the tissue specificity of the response are SEBP3A and SEBP3B, two binding activities that contain the transcription factors dAP-4 and Daughterless (King-Jones et al., 1999).

Repression of the Sgs genes seems to be a direct response to the late-larval 20E pulse, as suggested by the rapid and cycloheximide-insensitive effect of 20E on Sgs transcript levels in cultured salivary glands (Crowley and Meyerowitz, 1984; Hansson and Lambertsson, 1989). However, the observation that intermolt genes are not properly repressed in 2Bc mutants of the BR-C indicates that the hormone-controlled regulatory hierarchy at puparium formation is also involved in repression of these genes (Guay and Guild, 1991; Karim et al., 1993). The importance of the hormone for repression was recently confirmed by the finding that Usp and EcR are required for proper downregulation of Sgs4 (Hall and Thummel, 1998; Lam and Thummel, 2000).

Based on the finding that bacterially produced BR-C proteins are able to bind to sites in the Sgs4 regulatory region in vitro, it has been proposed that Sgs genes are directly regulated by the BR-C (von Kalm et al., 1994). However, these sites can also be bound by Fkh protein extracted from salivary gland nuclei (Lehmann and Korge, 1996). As the Fkh consensus binding sequence is contained within the BR-C consensus binding sequences (Fkh consensus underlined; Z1 consensus, ATTTGTTTATTA; Z3 consensus, TTTAGTTTA)(Lehmann and Korge, 1996; von Kalm et al., 1994), it is uncertain if mutations can be found that will only affect binding of one of the factors and thus allow discrimination between their functions. The strong Fkh binding that was observed on salivary gland polytene chromosomes at the Sgs4 regulatory region and loci of other Sgs genes indicates that Fkh regulates these genes directly (Lehmann and Korge, 1996; Mach et al., 1996). However, while antibody staining showed that BR-C protein can be found at many loci of the polytene chromosomes (Emery et al., 1994), the question still has to be addressed of whether these loci include the Sgs genes and, in particular, the putative BR-C binding sites of Sgs4.

We show that downregulation of fkh at puparium formation is necessary for proper repression of Sgs4. The finding that fkh fails to be downregulated in 2Bc mutants of the BR-C suggests an unexpected indirect mechanism of action of the BR-C in Sgs gene repression. Moreover, the absence of BR-C protein at Sgs loci and transposed regulatory elements of Sgs4 in salivary gland polytene chromosomes suggests that not only repression but also activation of the Sgs genes by BR-C is accomplished by an indirect mechanism. Such a model of indirect BR-C action is further supported by the detection of BR-C protein at the fkh locus on polytene chromosomes from mid- and late-third instar larvae.

The wild-type strain Oregon R (Stanford) was used for polytene chromosome staining, analysis of fkh expression, and the heat shock control experiment shown in Fig. 3.

The 2Bc¹, 2Bc² and 2Bc⁴ mutants die as late prepupae or early pupae and are maintained by balancing the y marked 2Bc X chromosome over a Binsn X chromosome that carries a y⁺ allele. RNA for the analysis of fkh and Sgs4 expression was extracted from salivary glands of hemizygous mutant males that can be distinguished from their wild-type siblings by the yellow phenotype of their mouth hooks and denticle belts. Control RNA was extracted from salivary glands of y⁺ males hemizygous for the Binsn chromosome.

The transformant line T4/7-6, here referred to simply as T4, that was used for polytene chromosome stains was provided by A. Krumm. Transformant P[hs-Fkh¹¹¹] that allows expression of wild-type Fkh under heat shock control was a kind gift from H. Jäckle and M. Hoch.

Wandering larvae of the transformant line P [hs-Fkh¹¹¹] and, as a control, Oregon R were collected and incubated in a 37°C water bath for 60 or 30 minutes. After a recovery period of 3-4 hours at 24°C, salivary glands were dissected for RNA extraction from mid-third instar larvae (cells filled with glue granules), late-third instar larvae (glue granules secreted into gland lumen), as well as freshly formed prepupae. As an additional control, RNA was extracted from non-heat shocked animals of the same stages.

The detection of fkh mRNA by RT-PCR was carried out according to the protocol of Huet et al. (Huet et al., 1993) with a few modifications. RNA was extracted using the Trizol reagent method, according to the protocol of the manufacturer (Life Technologies/Gibco BRL). For reverse transcription and amplification, we used MMLV-RTase from Life Technologies/Gibco BRL and Biotherm Taq polymerase from Rapidozym (1 unit each/reaction), and we used 100 μM of each dNTP and 25 pmol of each primer per reaction. Taq polymerase alone (in the absence of reverse transcriptase) did not yield any reaction products, and DNase I treatment of the RNA before RT-PCR had no effect on the results. For amplification of fkh transcripts the oligonucleotide 5′-CCG CGC GTC CTT AAG CCA GC-3′ was used as an RT/PCR primer and the oligonucleotide 5′-GCA CGT GCC CAC TCA TCA CC-3′ as a PCR primer. RT-PCR detection of rp49 was performed as described by Huet et al. (Huet et al., 1993).

Total RNA was extracted from salivary glands using the Trizol reagent (Life Technologies/GIBCO BRL). RNA was then fractionated by formaldehyde gel electrophoresis and transferred to Hybond-N+ nylon membranes (Amersham Pharmacia Biotech) using standard procedures. After UV crosslinking, the filters were hybridized with ³²P-labeled DNA probes synthesized by random priming. For detection of Sgs4 mRNA, a 1.4 kb XhoI-HindIII fragment was used that contains the Sgs4 transcribed region (Lehmann and Korge, 1995). Rp49 mRNA was detected using a 650-bp EcoRI-HindIII restriction fragment from pHR0.6 (O’Connell and Rosbash, 1984).

For generation of anti-BR-C antibodies that would allow detection of all isoforms, we cloned a 2 kb EcoRI-HindIII fragment that encompasses the complete coding region for the ‘core’ region of the Z1 isoform from the pET-FM-derived expression plasmid pETCoQZ1 (von Kalm et al., 1994) into pET21 (Novagen). Recombinant protein for immunization of rabbits (Eurogentec) was essentially produced as described previously (King-Jones et al., 1999), with the exception that a native protein fraction was not used, as the bulk of the protein accumulated in inclusion bodies. The anti-Fkh antibody was a kind gift of P. Carrera and H. Jäckle, and has been described previously (Lehmann and Korge, 1996).

Western analysis and in vitro translation was performed according to standard procedures essentially as described in King-Jones et al. (King-Jones et al., 1999). In the Western analysis, an anti-guinea pig-immunoglobulin antibody conjugated to alkaline phosphatase (Dianova) was used as secondary antibody. Conditions and oligonucleotides used in the mobility shift DNA-binding assay are described by Lehmann and Korge (Lehmann and Korge, 1996). Non-radiolabeled competitor oligonucleotides were used at a 20-fold molar excess.

Immunofluorescent staining of polytene chromosomes was performed as described (Lehmann and Korge, 1996). Anti-BR-C antiserum SA4256 was used as the primary antibody at a 1:200 dilution, and an affinity-purified Cy3-conjugated goat anti-rabbit immunoglobulin antibody (IgG H+L; Dianova) was used as a secondary antibody at a 1:400 dilution. Staining of polytene chromosomes with the SA4256 pre-immune serum did not produce any fluorescent signals above background.

The high degree of overlap between the in vitro BR-C- and Fkh-binding sites within the Sgs4 regulatory region seems to preclude the simultaneous binding of these proteins to a single site. In addition, the fact that Fkh can be detected on salivary gland polytene chromosomes at the Sgs4 locus and at transposed Sgs4 regulatory elements suggests that at least a subset of these binding sites is occupied by Fkh in vivo (Lehmann and Korge, 1996). To test whether BR-C protein is also bound to Sgs4 in situ, we raised polyclonal antisera directed against the conserved core region that is common to all BR-C isoforms. On western blots, these antisera specifically recognize the protein used for immunization, as well as BR-C protein extracted from wild-type larvae but not from npr1³ larvae, with antibody SA4256 giving the strongest reaction (data not shown). Fig. 1A shows that this antibody does not stain the salivary gland polytene chromosomes of npr1³ larvae. By contrast, it detects BR-C protein at many loci in chromosomes from Oregon R wild-type larvae. Surprisingly, although strong binding was observed in its vicinity, no BR-C binding was detected at the Sgs4 locus 3C11 in Sgs-expressing animals (puff stage 1; Fig. 1B). At later stages (after regression of the puff at 3C11; puff stage 6), however, we sometimes observed a very weak immunofluorescence at this locus (data not shown). Such a weak fluorescence could be due to the higher degree of chromatin condensation at 3C11 after regression of the puff, leading to a more condensed antigen distribution. This, in turn, could allow the detection of low amounts of BR-C protein that were previously below the detection threshold. We therefore asked whether an amplification of the number of putative BR-C-binding sites would allow us to obtain a clear fluorescent signal. Previous studies of Sgs4 regulation have used a transgenic line, T4, that carries a transposon with four copies of the Sgs4 regulatory region on the fourth chromosome (King-Jones et al., 1999; Lehmann and Korge, 1996). The increased number of binding sites within the transposon led to strongly enhanced immunofluorescent signals for Fkh, dAP-4 and Daughterless proteins at the integration site compared with the Sgs4 locus at 3C11. We were, thus, able to confirm that these factors are recruited to this locus through the Sgs4 regulatory region. We therefore also looked at BR-C binding in this transgenic line. In a control stock that does not carry the transposon, a weak fluorescent signal was observed at 102D3-5, the integration site of the transposon (Fig. 1C). This weak signal was not enhanced or otherwise changed in the presence of the transposon, either in Sgs4-transcribing animals (puff stage 1) (Fig. 1C) or in animals after repression of Sgs4 transcription (puff stage 6; data not shown). Considering the strong enhancement of fluorescence that is usually observed at the P-element integration site (King-Jones et al., 1999; Lehmann and Korge, 1996) and the large number of putative BR-C binding sites in the Sgs4 regulatory region, these results strongly suggest that BR-C proteins do not directly bind to Sgs4 in vivo. This surprising finding prompted us to also inspect the loci of other Sgs genes for BR-C binding. None of these loci showed significant immunofluorescence (Fig. 1B). Even the 68C locus, which contains three Sgs genes (Sgs3, Sgs7 and Sgs8), did not accumulate detectable amounts of BR-C protein, suggesting that the BR-C acts on Sgs genes by an indirect mechanism. This conclusion is supported by mobility shift DNA-binding assays which show that the putative Sgs4 target sites, which are protected in footprinting experiments using large amounts of BR-C protein purified from overexpressing strains of bacteria (von Kalm et al., 1994), are unable to bind in vitro translated BR-C proteins or BR-C proteins extracted from salivary gland nuclei (Lehmann and Korge, 1996)(M. L., unpublished). The conditions used in these assays were sufficient to clearly detect binding of other proteins like Fkh, EcR/Usp, Pipsqueak or GAGA factor to their target elements (Lehmann and Korge, 1996; Lehmann et al., 1997; Lehmann et al., 1998).

If BR-C proteins do not bind to the Sgs genes directly, how do they exert their profound effects on Sgs expression? One obvious possibility is that they affect the expression of other factors that are known to be involved in Sgs gene regulation. One of these factors is SEBP2/Fkh (Lehmann and Korge, 1996; Mach et al., 1996). During our initial studies on this factor we found that the SEBP2 DNA-binding activity is absent in salivary glands of early prepupae (Fig. 2A). We therefore reasoned that lack of this binding activity might be due to decreased fkh expression after the larval-to-prepupal transition. To test this possibility, we isolated RNA from salivary glands of mid-third and late-third instar larvae as well as early prepupae. In this study, late-third instar larvae are defined as larvae after secretion of glue proteins into the gland lumen. The glands of these animals are easy to distinguish from those of mid-third instar larvae by their bloated and transparent appearance, providing a convenient means to select animals that have already experienced the late-larval 20E pulse (Boyd and Ashburner, 1977). The amounts of fkh transcripts in the RNA samples were determined using the RT-PCR assay devised by Huet et al. (Huet et al., 1993). Fig. 2B shows that the level of fkh mRNA indeed drops dramatically following the late-larval 20E pulse. This decrease is clearly evident in salivary glands of late-third instar larvae and lasts in early prepupae. Consistent with the drop in fkh mRNA, a western analysis showed that the amount of Fkh protein is severely reduced in salivary glands of white prepupae (Fig. 2C). We conclude that the lack of SEBP2/Fkh-binding activity in early prepupae is due to a rapid turnover of Fkh protein after the transcriptional downregulation of fkh in an apparent response to the late-larval 20E pulse.

Because Fkh-binding sites are required for proper activation of Sgs4 in mid-third instar larvae, the downregulation of fkh at the end of this larval instar suggests that this event may contribute to the 20E-triggered repression of Sgs4. We therefore asked if maintenance of a high level of Fkh expression in late-third instar larvae and early prepupae is sufficient to support continued Sgs4 expression. To address this question, we used a transgenic line, P[hs-Fkh¹¹¹], that expresses a wild-type fkh gene under control of the hsp70 heat shock promoter. Heat treatment of these animals led to elevated fkh mRNA levels that could still be detected by 4 hours after heat shock application (data not shown). When expression of Fkh from the transgene was induced in late-third instar larvae, the downregulation of Sgs4 that is normally observed at pupariation was clearly impaired (Fig. 3A). High levels of Sgs4 mRNA could still be detected in salivary glands of prepupae after heat shock induction of the transgene (Fig. 3A, lanes 15-17), while Sgs4 transcripts were almost completely absent from salivary glands of non-heat shocked animals (Fig. 3A, lanes 10-12), or wild-type control animals that had been heat shocked like P[hs-Fkh¹¹¹] (Fig. 3A, lanes 6 and 7). These data strongly suggest that the downregulation of fkh at the larval-to-prepupal transition indeed contributes to the repression of Sgs4, and raises the question of how this downregulation is achieved.

A continuation of intermolt gene expression into the prepupal stage, similar to that observed in the presence of ectopically expressed Fkh, has been reported for 2Bc mutants of the BR-C (Karim et al., 1993). This raises the possibility that the 2Bc product might be required for the downregulation of fkh and thus contributes only indirectly to the repression of Sgs4 and other Sgs genes. To confirm that proper Sgs4 repression indeed depends on the 2Bc function of the BR-C, we performed a northern analysis of Sgs4 expression in third instar larvae and early prepupae of three different 2Bc mutants (Fig. 3B). Early prepupae hemizygous for the mutant alleles 2Bc¹, 2Bc², and 2Bc⁴, all showed elevated levels of Sgs4 mRNA in their salivary glands when compared with the corresponding control animals. The effect on Sgs4 repression was not as dramatic as described for the intermolt gene Gene VII (Karim et al., 1993) or as dramatic in the presence of ectopically expressed Fkh (Fig. 3A). However, impaired repression of Sgs4 in the 2Bc mutants was reproducibly observed in three independent experiments. We therefore conclude that Sgs4, like other intermolt genes, is not properly repressed in animals that lack the 2Bc gene product(s).

If 2Bc product(s) indeed contribute indirectly to Sgs4 repression by downregulating fkh, fkh should not be properly repressed at the larval-to-prepupal transition in 2Bc mutants. We therefore isolated total RNA from salivary glands of mid- and late-third instar larvae as well as early prepupae of the 2Bc¹, 2Bc² and 2Bc⁴ mutants, and determined their content of fkh mRNA by RT-PCR. Although fkh transcripts are barely detectable by RT-PCR in late-larval and prepupal salivary glands of wild-type animals (Fig. 2B), these glands still contain fkh mRNA in 2Bc mutants (Fig. 4). There is only a slight decrease of fkh mRNA in the 2Bc¹ and 2Bc⁴ mutants, and fkh transcript levels remain virtually constant in the 2Bc² mutant. We therefore conclude that not only Sgs4 but also fkh is not properly repressed in 2Bc mutants of the BR-C. It is interesting to note that, in contrast to the 2Bc¹ mutant, in the 2Bc² mutant, a reinduction of intermolt genes is observed in an apparent response to the prepupal 20E pulse (Karim et al., 1993). Although the 2Bc¹ and 2Bc² alleles have been referred to as genetically amorphic alleles (Bayer et al., 1997), this finding suggests that 2Bc¹ might indeed be a hypomorphic allele and 2Bc² either a strong hypomorphic or a null allele with respect to Sgs transcription. The more severe effect of the 2Bc² mutation on fkh expression is consistent with this interpretation. Moreover, it strengthens our conclusion that the 2Bc wild-type function represses the Sgs genes indirectly by downregulating fkh expression.

Our finding that the downregulation of fkh at pupariation depends on the 2Bc function of the BR-C, as well as our failure to detect BR-C protein at Sgs loci in situ, suggests that the 2Bc product(s) might exert their effects on fkh expression (and thus on Sgs expression) by direct binding to the fkh gene. A prediction of this model is that BR-C protein is present at the fkh locus at 98D2-3 on the salivary gland polytene chromosomes. Indeed, we reproducibly observed a very strong fluorescent signal with our BR-C antibody at this locus on chromosomes from mid- and late third instar larvae (Fig. 5). While this finding is consistent with our model, further studies at a higher resolution will be required to unambiguously demonstrate that BR-C protein is associated with the fkh gene.

This study was conceived to help bridge the gap between our increasing knowledge of 20E-controlled regulatory hierarchies and the mechanisms that eventually lead to a specific biological response. One central player within the 20E-controlled hierarchies is the early gene BR-C. Genetic studies have implicated a key role for the BR-C in Sgs gene regulation and glue production (Crowley et al., 1984; Karim et al., 1993; von Kalm et al., 1994), however, the mechanism by which this control is exerted remains unclear.

DNaseI footprinting experiments with bacterially produced recombinant protein showed that BR-C proteins can bind to the control regions of the BR-C-dependent genes Ddc, L71-6 and Sgs4 (Crossgrove et al., 1996; Hodgetts et al., 1995; von Kalm et al., 1994). Consensus DNA-binding sequences have been defined in these studies for each of the four isoform-specific zinc-finger domains. However, the high variability of these A/T-rich sequences suggests that additional determinants are important for specific DNA recognition. One such determinant could be the BTB/POZ domain that is present in all BR-C isoforms. Nuclear proteins with this domain, like the GAGA or Mod(mdg4) proteins, are thought to act by remodeling chromatin structure (Dorn et al., 1993; Farkas et al., 1994; Gerasimova et al., 1995; Tsukiyama et al., 1994). The BR-C dependence of a DNase I-hypersensitive site in the hsp23 gene suggests that BR-C proteins might act in a similar manner (Dubrovsky et al., 1994). The BTB/POZ domain of the GAGA factor mediates strong cooperative DNA binding to multiple sites but, like other BTB/POZ domains, inhibits binding to single sites (Bardwell and Treisman, 1994; Katsani et al., 1999). BTB/POZ domains are thus likely to play a critical role in targeting proteins to specific chromosomal loci. Our data support this concept, as they show that DNA binding sites that are bound by BR-C proteins in vitro are not sufficient to bind BR-C in a chromosomal context. Such sites may, therefore, turn out to be irrelevant for the control of the gene they are connected with. Mutation of in vitro BR-C binding sites in the promoter of the death gene reaper (rpr), for instance, has no effect on the expression of a reporter gene. It was therefore concluded that the BR-C might only indirectly regulate rpr transcription in the larval salivary glands, which foreshadows the BR-C-dependent destruction of this tissue during metamorphosis (Jiang et al., 2000).

Mutations that eliminate BR-C binding to Sgs4 in vitro do lead to reduced reporter gene expression in vivo (von Kalm et al., 1994). However, these mutations also affect Fkh binding sites that are required for Sgs4 expression (Lehmann and Korge, 1996), and this effect may therefore be unrelated to the ability of these sites to be bound by BR-C. We favor this interpretation, not only because BR-C proteins cannot be detected at Sgs4 in situ, but also because BR-C proteins extracted from salivary gland nuclei or produced by in vitro translation are not able to recognize these sites in our mobility shift DNA-binding assay. It is important to note that we were not only unable to detect BR-C protein at the Sgs4 locus 3C11, but also at loci of other Sgs genes like 68C, which harbors Sgs3, Sgs7 and Sgs8. Interestingly, although expression of these three genes depends on BR-C function to a much larger extent than that of Sgs4 (Crowley et al., 1984), Fkh binding sites, but no BR-C binding sites, could be detected in the transcriptional control region of Sgs3 (K. Crossgrove, PhD thesis, University of Pennsylvania, Philadelphia, 1995) (Mach et al., 1996). Fbp1 is another example of a gene that is clearly under BR-C control, although it does not seem to be bound directly by BR-C proteins. Fbp1 is activated in response to 20E in the larval fat body at about the time when the Sgs genes are activated in the salivary glands. Expression of an Fbp1 transgene depends on different BR-C isoforms, but attempts to demonstrate direct binding of these isoforms to elements within the transgene were not successful (Mugat et al., 2000).

In this study, we present evidence not only for an indirect control of Sgs genes by the BR-C, but also for a possible mechanism that explains how this indirect control is achieved. Based on our results, we propose a new model for the regulatory interactions between BR-C, Fkh and the 20E receptor EcR/Usp in Sgs gene regulation (Fig. 6). We find that the 2Bc function of the BR-C is required at puparium formation for downregulation of the Fkh transcription factor, which is involved in tissue-specific activation of the Sgs genes. We further show that ectopic expression of fkh, which overrides repression of the endogenous gene, mimics the effects of 2Bc mutations on Sgs4 expression. The high level of Sgs4 mRNA that is maintained in the presence of ectopic Fkh clearly shows that the downregulation of fkh is necessary for Sgs repression. Together with downregulation of the Sgs activator SEBP3 (Fig. 2A; K. K.-J., unpublished), it may even be sufficient for repression, leading to the surprising conclusion that direct binding of a repressor may not be required for this process. Our study not only provides strong evidence for an indirect mode of action of the BR-C in Sgs gene repression but also suggests that activation might be indirect. BR-C proteins cannot be detected at Sgs loci when the genes are actively transcribed, while they are present at 98D, the cytogenetic region that includes fkh and other genes. The presence of BR-C protein at 98D in mid- and late-third instar larvae is consistent with our model of direct repression of fkh by the BR-C. However, as fkh seems to be normally expressed in mid-third instar larvae of BR-C null mutants (Lehmann and Korge, 1996), BR-C activation of Sgs genes is likely to be mediated by a factor other than Fkh. Alternatively, BR-C protein might transiently bind to the Sgs genes at an earlier time when the polytene chromosomes are not yet accessible for immunolocalization studies (Fig. 6).

A central player in our model is the Fkh protein, which plays an important role in development of the salivary glands as well as, later on, in tissue-specific gene control in this organ (Andrew et al., 2000). As fkh mutants show homeotic transformations, it has been suggested that Fkh is required for the establishment of tissue identity (Jürgens and Weigel, 1988). Downregulation of fkh by the BR-C 2Bc function at puparium formation may therefore have a more global effect on the salivary glands than just repression of glue genes, altering the determined state of the glands. Downregulation of fkh could be the first step towards destruction of the larval salivary glands by programmed cell death that occurs about 14 hours later, at pupation (Jiang et al., 1997; Jiang et al., 2000). Interestingly, fkh mutations result in removal of the embryonic salivary gland placode by apoptosis, suggesting that Fkh might function as a survival factor that prevents the salivary glands from being eliminated by programmed cell death (Myat and Andrew, 2000). Unfortunately, little is known about the mechanisms through which Fkh and other winged helix proteins exert their functions. Studies on the mouse serum albumin enhancer have implicated HNF3/Fkh proteins in chromatin organization (Cirillo and Zaret, 1999; McPherson et al., 1993). However, efforts to identify interacting proteins, including an attempt in our laboratory to demonstrate a direct interaction between Fkh and BR-C proteins, have failed so far (N. R., unpublished).

The study presented here integrates fkh in the 20E-controlled regulatory hierarchies that are active at the onset of metamorphosis and assigns BR-C to a new position within these hierarchies. Future studies will show if similar regulatory connections between the BR-C and tissue-specific factors exist in other responses to 20E signaling during development.

We thank C. Bayer for providing the BR-C mutant stocks, L. von Kalm for the BR-C expression plasmids, and H. Jäckle and P. Carrera for providing the P[hs-Fkh¹¹¹] transformant and the anti-Fkh antibody. We also thank C. S. Thummel and T. Kozlova for critical comments on the manuscript. We gratefully acknowledge the help of M. Brünner in performing the chromosome stains, and the assistance of G. Korge in localizing the immunofluorescent signals. Moreover, we thank G. Korge for continuous support that made the conclusion of this study possible. This work was supported by a grant from the Deutsche Forschungsgemeinschaft to M. L. (Le 870/2-2).