Brinker is a nuclear protein that antagonizes Dpp signalling in Drosophila. Its expression is negatively regulated by Dpp. Here, we show that Brinker represses Ultrabithorax (Ubx) in the embryonic midgut, a HOX gene that activates, and responds to, the localized expression of Dpp during endoderm induction. We find that the functional target for Brinker repression coincides with the Dpp response sequence in the Ubx midgut enhancer, namely a tandem of binding sites for the Dpp effector Mad. We show that Brinker efficiently competes with Mad in vitro, preventing the latter from binding to these sites. Brinker also competes with activated Mad in vivo, blocking the stimulation of the Ubx enhancer in response to simultaneous Dpp signalling. These results indicate how Brinker acts as a dominant repressor of Dpp target genes, and explain why Brinker is a potent antagonist of Dpp.
Signalling by Dpp and other TGF‐β family proteins affects phosphorylation of cytoplasmic Smad proteins. As a result, activated Smads translocate to the nucleus and stimulate the transcription of TGF‐β target genes (Massagué, 1998). Smad‐induced transcriptional switches determine multiple cell fates during normal animal development (Raftery and Sutherland, 1999). In mice and man, Smads and TGF‐β receptors also function as tumour suppressors (Massagué et al., 2000).
Recently, a nuclear protein called Brinker (Brk) has been discovered that antagonizes the Dpp response in Drosophila (Campbell and Tomlinson, 1999; Jazwinska et al., 1999a,b; Minami et al., 1999). Brinker negatively regulates Dpp target genes in multiple developmental contexts and is itself repressed by Dpp signalling. This repression is mediated by Schnurri, a nuclear zinc finger protein whose expression is induced by Dpp near Dpp signalling sources (Marty et al., 2000). Hence, the transcriptional activation of Dpp target genes reflects both a positive and a double‐negative event, namely the activation of the Mad agonist and the removal of the Brinker antagonist by the Schnurri repressor.
Among the direct Dpp target genes are the HOX genes Ultrabithorax (Ubx) and labial (Tremml and Bienz, 1992; Hursh et al., 1993; Thüringer and Bienz, 1993), which partake in Drosophila endoderm induction. This process is initiated in the visceral mesoderm by Ubx, which activates, and also responds to, the localized expression of Dpp and Wingless (Wg), while labial is the ultimate target of these extracellular signals in the subjacent endoderm (Bienz, 1996). The midgut enhancers of Ubx and labial contain binding sites for Mad (Kim et al., 1997) that mediate the response to Dpp stimulation (Szüts et al., 1998).
Here, we show that Brinker represses Ubx in the embryonic midgut. The functional target for Brinker repression within the Ubx midgut enhancer coincides with tandem Mad binding sites. We find that Brinker efficiently competes with Mad in vitro for binding to these sites and that it prevents activated Mad from stimulating Dpp target enhancers in vivo. These findings indicate a mechanism by which Brinker could dominate over activated Mad to block the activity of Dpp target genes.
The control of Ubx by Dpp and Wg signalling has been studied by functional dissection of a minimal midgut enhancer called Ubx B (Thüringer et al., 1993). This enhancer directs expression of a linked β‐galactosidase (lacZ) gene in parasegments (ps) 6–9, and also in ps3, of the midgut mesoderm as a result of stimulation by Dpp and Wg, which are expressed in or near these regions (Figure 1A). This stimulation requires distinct Dpp and Wg response sequences (DRS and WRS) within Ubx B (Eresh et al., 1997; Riese et al., 1997; Szüts et al., 1998) (Figure 1B). In addition, Ubx B is repressible by high Wg levels near the Wg signalling source (Yu et al., 1998), and is also repressed in the absence of Wg signalling in cells remote from the source (Riese et al., 1997; Waltzer and Bienz, 1998). The former repression is mediated by the WRS‐R, a sequence coinciding with the Mad binding sites within the DRS, the latter by the WRS, a binding site for dTCF (Drosophila T cell factor) (Figure 1B).
Ubx is repressed by Brinker in the embryonic midgut
Since Ubx is a Dpp target gene in the embryonic midgut, we asked whether this HOX gene might be under brk control. Thus, we stained brk mutant embryos with an antibody against Ubx and found weak ectopic Ubx staining in the posterior midgut mesoderm of these mutants (Figure 2A and B). Normally, the HOX protein Abdominal‐A represses Ubx in the posterior midgut (Bienz and Tremml, 1988) (Figure 1A), but evidently this is not sufficient to keep Ubx repressed in the absence of brk. However, we did not detect any Ubx derepression in the anterior midgut of brk mutants, probably because of the silencing of Ubx in this region by Polycomb (Immerglück et al., 1990). But we did find derepression in the anterior and posterior midgut of brk mutant embryos when examining lacZ expression conferred by an extensive Ubx midgut enhancer called RP9 (Figure 2C and D), the expression of which closely resembles Ubx expression in the midgut (Thüringer and Bienz, 1993). These stainings show that brk represses Ubx in the embryonic midgut.
To determine the target for brk‐mediated repression within Ubx, we asked whether the minimal Ubx B midgut enhancer is also under brk control. Indeed, we found that Ubx B is strongly derepressed at both ends of the midgut in brk mutant embryos, approximately in ps2 and 12 (Figure 2E and F). We note that each of these ps overlaps a domain of brk expression (Jazwinska et al., 1999b) (Figure 1A), indicating that Brinker represses Ubx and its reporters cell‐autonomously at both midgut ends.
Next, we tested a series of mutant versions of Ubx B that carry nested point mutations (Figure 1B). Most of these were still derepressed in brk mutants, e.g. BM1 (Figure 2G and H), which has a mutated MadB site. However, three mutant enhancers were no longer derepressed: B4, which has a mutated dTCF binding site (Figure 2I and J); B4R8, which carries a mutation in a conserved sequence motif (Figure 2K and L); and BM2, in which both Mad binding sites are mutated (Figure 2M and N). Formally, each of these mutations could define a target for Brinker repression. Alternatively, they define sequences that are essential for enhancer activation, in particular for ectopic activation at the midgut ends. This is a clear possibility since B4, BM2 and B4R8 are each considerably less active than Ubx B and other mutant enhancers such as BM1.
Brinker binds to the Mad binding sequence within Ubx B
We expressed full‐length and various fragments of Brinker (Figure 3A) as glutathione S‐transferase (GST) fusion proteins in bacteria, in order to test whether these fusion proteins can bind to the signal‐responsive sequence from Ubx B (residues shown in Figure 1B, plus 32 additional 3′ residues) in gel shift assays. This revealed that full‐length Brinker, or its N‐terminus alone, can bind to this sequence, whereas the C‐terminus cannot (Figure 3B, lanes 3–5). This is consistent with the suggestion that the N‐terminus contains a putative DNA binding domain similar to the homeodomain (Jazwinska et al., 1999a) (Figure 3A). Indeed, a minimal fragment spanning this domain (BRK44‐99) binds to the probe as well as full‐length Brinker (Figure 3B, lanes 2 and 5).
Next, we tested Brinker binding to mutant DNA probes (Figure 1B). Of these, BM2 and BM0 were the only mutants that no longer showed any binding to Brinker (Figure 3C). Likewise, Brinker binding to DNA can be competed with an excess of unlabelled wild‐type probe, but not with mutant BM2 probe (Figure 3D). This shows that Brinker binds to Ubx B in a sequence‐specific manner, and that the residues mutated in BM2 and BM0 are critical for Brinker binding.
While our manuscript was under review, Sivasankaran et al. (2000) published a consensus site for Brinker binding, GGCG C/T C/T, to which we find three perfect matches in Ubx B (Figure 1B). These are adjacent to one another, and each of them is mutated in BM2. Our results with BM1 indicate that the first of these matches (Brk bs1) is sufficient for Brinker function in vivo and in vitro. However, Brk bs3 alone is unlikely to be sufficient for function, given that Brinker cannot bind to the mutant probe BM0 (Figure 3C). Finally, our results indicate that Brk bs2 (perhaps together with bs3) can substitute for Brk bs1 and provide full function: BC2 is repressible by brk in vivo (also see below), and Brinker binds to BC2, BC and BM01 mutant probes (Figure 3C), all of which lack Brk bs1.
Interestingly, the three Brinker binding sites completely overlap the two Mad binding sites within the DRS (Figure 1B). Indeed, our previous work indicated that the Dpp response critically depends on MadA (Szüts et al., 1998); MadA fully overlaps Brk bs1, which we find to be sufficient for Brinker function in vitro and in vivo. We thus asked whether Brinker might be able to compete with Mad for DNA binding. We performed competitive DNA binding experiments using bacterially expressed DNA binding domains of Brinker and Mad. This revealed that the former is capable of competing successfully with the latter for DNA binding at a molar ratio of 1:150, and Brinker almost completely blocks Mad binding at a ratio of 1:15 (Figure 3F and G). Note that full‐length Mad binds to DNA ∼100× less efficiently than its isolated DNA binding domain (Kim et al., 1997), indicating that Brinker would be able to compete even more successfully with full‐length Mad. Thus, Brinker can block Mad binding to DNA in vitro in the presence of a considerable molar excess of Mad.
Ectopic Brinker targets the Mad binding sequence in vivo
To confirm that the above Brinker binding sites within Ubx are functional targets in vivo, we expressed Brinker throughout the midgut mesoderm with the GAL4 system (Brand and Perrimon, 1993). This revealed that expression of Ubx in the middle midgut is nearly eliminated in Brinker‐overexpressing embryos (Figure 4A and B). Instead, many of these embryos show an endodermal bulge in the middle midgut (Figure 4B, arrowhead) that is also observed in Ubx mutants (Tremml and Bienz, 1989). Furthermore, the first and second midgut constrictions are rudimentary at best, and often missing altogether (Figure 4E and F). Again, loss of the second constriction is indicative of mutations of Ubx and dpp (Bienz and Tremml, 1988; Panganiban et al., 1990), while loss of the first may reflect mutation of the dpp‐related gene gbb (Wharton et al., 1999). Finally, ectopic Brinker also drastically reduces dpp and wg expression in the middle midgut (Grieder et al., 1995; Marty et al., 2000; Figure 5E–H and data not shown), which is expected since their expression depends on Ubx. This indicates that Brinker, by virtue of repressing Ubx, is capable of blocking the whole process of endoderm induction that depends on this HOX gene.
LacZ staining from Ubx B was virtually eliminated in Brinker‐overexpressing embryos, except for two patches of strong staining in the head and some staining in the dorsal epidermis (Figure 4C and D). Of the mutant enhancers, we found BM2 to be the least affected by ectopic Brinker (Figure 4E and F): in this case, we observed individual cells in ps8 and 9 of the visceral mesoderm that retained strong lacZ staining. In contrast, other mutant enhancers such as B4, B4R8 and BM1 behave like Ubx B itself, at best showing traces of lacZ staining in Brinker‐overexpressing embryos (not shown). Thus, BM2 is unique in being refractory to Brinker repression, indicating that the Brinker binding sites that are mutated in this enhancer are functional in vivo.
We confirmed this by testing a minimal construct derived from Ubx B, called L‐CRE, which spans the WRS, and MadA/Brk bs1 and bs2 (Figure 1B), and which is signal‐responsive in both the visceral mesoderm and the endoderm (Riese et al., 1997). If we overexpress Brinker in the endoderm only (to avoid repressing wg and dpp in the visceral mesoderm), L‐CRE is completely repressed in this germ layer, while it remains active in the visceral mesoderm (Figure 4G and H). From this and our DNA binding studies we conclude that MadA/Brk bs1 and bs2 are direct targets for Brinker repression in the embryonic midgut.
Brinker successfully competes with activated Mad in the embryo
Our DNA binding experiments indicated that Brinker can prevent Mad from binding to functional target sites. Furthermore, we found that ectopic endodermal Brinker repressed L‐CRE in the endoderm, implying that this repression occurs in the presence of unimpaired signalling by Dpp and Wg.
To test directly whether Brinker repression can overcome Dpp‐ and Mad‐mediated target gene activation, we monitored Ubx B after simultaneous overexpression of Brinker and Dpp in the mesoderm. If Dpp is overexpressed alone, this leads to substantial stimulation of Ubx B expression in most regions of the midgut (Thüringer et al., 1993) (Figure 5A and B). However, Ubx B was repressed by ectopic Brinker whether or not Dpp was present (Figure 5C and D). Essentially the same was observed if a Dpp‐responsive dpp enhancer called BMPX (Sun et al., 1995) was tested in this way: BMPX was strongly stimulated by ectopic Dpp, but only in the absence of co‐expressed Brinker (Figure 5E–H). These results demonstrate that Brinker can repress Dpp‐responsive targets even in the presence of Dpp stimulation. This confirms our results from the DNA binding studies in vitro, and implies that Brinker is a potent repressor that can block activated Mad in vivo.
We have shown that Brinker binds to a specific sequence within the Ubx midgut enhancer, and that this sequence is a functional Brinker target in vivo. Ectopic Brinker represses Ubx in the middle midgut, and also blocks the expression of other Dpp target genes in this region, including dpp itself and wg. Our results indicate that Brinker is a direct repressor of Ubx, and thus a potent antagonist of the Dpp‐dependent process of endoderm induction. We note that Brinker is expressed in ‘signal‐free’ zones bordering the anterior and posterior limits of the midgut (Figure 1A). Its presence in these zones may have a barrier function, helping to block the spread of the Dpp response beyond the midgut limits.
Interestingly, the critical Brinker target site within Ubx B overlaps MadA, a functional Mad binding site that is required for the stimulation of this enhancer by Dpp signalling. Furthermore, Brinker competes effectively with Mad in binding to this site in vitro, and blocks activated Mad from stimulating Ubx B in vivo. This indicates that the mechanism by which Brinker repression dominates over stimulatory Dpp inputs is based on direct competition for binding to Dpp target enhancers. Given that most, if not all, Dpp signalling is mediated by Mad (Raftery and Sutherland, 1999), it seems likely that this competition‐based mechanism of Brinker repression is widespread and extends to genes that are Dpp targets in other developmental contexts.
Notably, MadA is also the target sequence for repression of Ubx B in response to high Wg levels in the middle midgut (Yu et al., 1998). MadA is thus a pivotal enhancer sequence that gauges and integrates positive inputs from Dpp and negative inputs from Brinker and Wg. Wg‐mediated repression in the middle midgut is mediated by the zinc finger protein Teashirt (Waltzer et al., 2001) and can be overriden by simultaneous Dpp stimulation (Yu et al., 1998). In contrast, Brinker‐mediated repression dominates over simultaneous Dpp stimulation (Figure 5). It thus appears that Brinker is a more potent repressor than Teashirt, and is designed to function as a signal‐antagonist even in the presence of high levels of Dpp signalling.
Jazwinska et al. (1999a) noticed that Brinker contains the sequence PMDLS (Figure 3A), which resembles the P‐DLS motif through which a number of transcription factors recruit the co‐repressor dCtBP (Schaeper et al., 1995; Nibu et al., 1998a,b; Poortinga et al., 1998). Indeed, using in vitro pull‐down assays, we found that dCtBP binds to full‐length Brinker as well as to an N‐terminal Brinker fragment that contains the PMDLS motif (not shown). This suggests that Brinker may recruit dCtBP to repress Dpp target genes in the embryo. Interestingly, dCtBP assists various transcription factors, such as Knirps, Snail and Krüppel, that act at short‐range to repress their target genes (Nibu et al., 1998a,b). These short‐range repressors bind to autonomous enhancers to quench nearby bound transcriptional activators, which prompted Levine and colleagues to suggest that dCtBP may be specifically designed to quench (Nibu et al., 1998b). Therefore, this quenching ability of dCtBP could enable Brinker to not only compete efficiently with activated Mad in the binding of DNA, but also out‐compete the activity of nearby transcription factors such as activated dTCF.
Fly strains and embryo staining.
The loss‐of‐function allele brkF124 (Lammel et al., 2000) was used for analysis. The following transformants were used: RP9 (Thüringer and Bienz, 1993); Ubx B (also called Bhz; Thüringer et al., 1993); B4 (Riese et al., 1997); BM1, BM2 and BC2 (Szüts et al., 1998); B4R8 (Yu et al., 1998); B24.GAL4 (Brand and Perrimon, 1993); 48Y.GAL4 (Martin‐Bermudo et al., 1997); UAS.Brk (Lammel et al., 2000); UAS.Dpp (Staehling‐Hampton et al., 1994). Mutant embryos and embryos overexpressing various proteins were identified by blue balancers or by their gut phenotypes.
GST fusion proteins, gel shift and pull‐down assays.
Full‐length Brinker and Brinker fragments were subcloned into pGEX‐2TK (Pharmacia) using standard PCR cloning procedures and expressed in Escherichia coli BL21 as described (Waltzer and Bienz, 1998). The plasmid encoding the DNA‐binding fragment of Mad (MH1+L or Mad1‐241) was also used (Waltzer and Bienz, 1998; see also Kim et al., 1997). All fusion proteins were purified by affinity chromatography and eluted from the GST–agarose beads in glutathione elution buffer (0.01 M reduced glutathione in 50 mM Tris–HCl pH 8.0). Note that the preparations of Mad (MH1+L) were variable; only two of six preparations gave satisfactory binding results (Figure 3E–G).
For band shifts, a probe spanning the signal‐responsive sequence from Ubx B (see above and Figure 1B) was labelled with [α32P]dATP and [α32P]dCTP by PCR. Binding reactions were typically performed with 10 μg (or as indicated in the figure legend) of purified GST fusion proteins incubated with 0.2 ng of probe in 20 μl of buffer containing 5 mM HEPES pH7.5, 10% glycerol, 70 mM KCl, 5 mM MgCl2, 2 mM EGTA, 1 mM EDTA, 2 μg poly(dI–dC), 1 mM dithiothreitol (DTT), 0.2 mM phenylmethylsulfonyl fluoride (PMSF) and protease inhibitors for 20 min at room temperature. Electrophoresis was carried out with 5% polyacrylamide gels/0.5× TBE containing 2.5% glycerol.
Pull‐down assays were performed as described (Waltzer and Bienz, 1998).
We thank Ann Kelley for help with the band‐shifts, Uwe Lammel for fly strains and plasmids, Lucas Waltzer, Jorge Bolivar and Fiona Townsley for advice and discussion. E.S. is supported by a long‐term fellowship from the Swiss National Science Foundation.
- Copyright © 2001 European Molecular Biology Organization