Decapentaplegic (Dpp) is an extracellular signal of the transforming growth factor‐β family with multiple functions during Drosophila development. For example, it plays a key role in the embryo during endoderm induction. During this process, Dpp stimulates transcription of the homeotic genes Ultrabithorax in the visceral mesoderm and labial in the subjacent endoderm. Here, we show that a cAMP response element (CRE) from an Ultrabithorax enhancer mediates Dpp‐responsive transcription in the embryonic midgut, and that endoderm expression from a labial enhancer depends on multiple CREs. Furthermore, the Drosophila CRE‐binding protein dCREB‐B binds to the Ultrabithorax CRE, and ubiquitous expression of a dominant‐negative form of dCREB‐B suppresses CRE‐mediated reporter gene expression and reduces labial expression in the endoderm. Therefore, a CREB protein may act as a nuclear target, or as a partner of a nuclear target, for Dpp signalling in the embryonic midgut.
During animal development, cells often instruct each other by secreting signals. Transforming growth factor‐β (TGF‐β)‐like growth factors such as activins and Drosophila Decapentaplegic (Dpp) are among the best studied extracellular signals that control development (Padgett et al., 1987; reviewed by Jessel and Melton, 1992; Smith, 1994; Massagué, 1996). These signals act in many developmental contexts, e.g. they organize the embryonic dorsoventral pattern (Irish and Gelbart, 1987; Ferguson and Anderson, 1992) and patterning of adult appendages in flies (Zecca et al., 1995; Lecuit et al., 1996; Nellen et al., 1996), and they function during mesoderm and endoderm induction in frogs and flies (Green and Smith, 1990; Smith et al., 1990; Hemmati‐Brivanlou and Melton, 1992; Bienz, 1994; Staehling‐Hampton et al., 1994; Frasch, 1995). In some of these events, the TGF‐β‐like signals have morphogenetic properties: they act at long range, and distinct and sharp cellular responses are elicited by multiple signalling thresholds (Green and Smith, 1990; Ferguson and Anderson, 1992; Green et al., 1992; Gurdon et al., 1994; Lecuit et al., 1996; Nellen et al., 1996; reviewed by Lawrence and Struhl, 1996). Ultimate decoding of these thresholds is likely to be achieved by transcription factors controlling expression of the signal target genes. None of these signal response factors has been identified as yet.
In the Drosophila embryo, dpp plays a key role during endoderm induction (reviewed by Bienz, 1994; Figure 1). Dpp is secreted from a localized source in the visceral mesoderm (VM) to stimulate transcription of the homeotic gene labial (lab) in the subjacent endoderm (Immerglück et al., 1990; Panganiban et al., 1990; Reuter et al., 1990; Neufeld et al., 1996). As a result, different cell types of the larval gut are specified (Hoppler and Bienz, 1995), for example the copper cells whose development depends on lab (Hoppler and Bienz, 1994). However, Dpp also signals within the VM where it stimulates expression of three different genes (Figure 1): its own (Hursh et al., 1993; Staehling‐Hampton and Hoffmann, 1994; Yu et al., 1996), expression of Wingless (Wg), an extracellular signal expressed in adjacent VM cells (Immerglück et al., 1990; Panganiban et al., 1990; Reuter et al., 1990), and of Ultrabithorax (Ubx), the homeotic gene expressed in the same VM cells as dpp (Panganiban et al., 1990; Hursh et al., 1993; Thüringer and Bienz, 1993). In turn, dpp expression is stimulated directly by Ubx (Sun et al., 1995), and is also stimulated by wg (Yu et al., 1996) which feeds back positively on Ubx expression as well (Thüringer and Bienz, 1993). Thus, dpp is part of an indirect autoregulatory loop by which Ubx, at the top of the inductive cascade, maintains its own expression (Thüringer and Bienz, 1993). Similar indirect autoregulatory feedback loops of cell fate‐determining genes have been observed in vertebrate development, e.g. in the chick limb bud (Niswander et al., 1994) and in the Xenopus embryo (Tada et al., submitted). They may be designed to stabilize developmental decisions in groups of cells (Bienz, 1994).
Previously, we have characterized short enhancer fragments from lab and Ubx which confer the response to dpp signalling in the endoderm and in the VM, respectively (Tremml and Bienz, 1992; Thüringer et al., 1993). Here, we identify the DNA target sequence in these enhancers which is necessary and to some extent sufficient for this response. In both cases, this sequence closely resembles the binding site for CREB (cAMP response element‐binding protein, see below), and we present evidence to suggest that a Drosophila CREB protein may be a target transcription factor, or a dimerization partner of such a factor, for dpp signalling in the embryonic midgut.
The Dpp response sequence in the Ubx midgut enhancer is a CRE
We previously have characterized a short Ubx enhancer, called B, which confers Wg‐ and Dpp‐dependent β‐galactosidase (lacZ) reporter gene expression in the VM. Staining mediated by Ubx B is in two stripes of cells in the VM, a wide prominent one in parasegments (ps) 6–9 and a narrow weak one in ps3 (Figures 1 and 2a; see also Thüringer et al., 1993). Our previous dissection of Ubx B led us to conclude that the target sequences for Dpp and Wg signalling within this enhancer are separable (Thüringer et al., 1993). To identify these signal target sequences, we carried out a footprint analysis of this Ubx enhancer, using crude nuclear protein extracts. We thus found eight distinct sequences to be protected by these extracts (to be described elsewhere in more detail; see also Figure 4a). We noticed that footprint 5 (FP5) partly overlaps a near palindromic sequence TGGCGTCA which closely resembles a typical cAMP response element (CRE) (TGACGTCA; Montminy et al., 1986) (Figures 1 and 4a). To test the function of this sequence, and of the adjacent sequence covered by FP5, we introduced a 3 bp substitution into the former (mutant construct BC) and a 4 bp substitution into the latter (mutant construct B5). We then examined the lacZ expression patterns mediated by these mutant enhancers in stably transformed embryos and compared them with that mediated by Ubx B.
In the case of B5, the two stripes of lacZ expression are widened significantly and stain more strongly than those conferred by the wild‐type B enhancer (Figures 1 and 2d). Conversely, in BC transformants, the wide stripe is narrowed to ps8/9 and stains only weakly, and the narrow stripe in ps3 is hardly detectable (Figures 1 and 2g). Thus, the sequence motif TGGCGTCA functions in the embryo to mediate transcriptional stimulation, whereas the adjacent FP5 sequence mediates transcriptional repression. We shall refer to the putative proteins which act positively or negatively through this region of the B enhancer as the CRE activator or the FP5 repressor, respectively.
dpp and wg synergize to stimulate Ubx expression in the VM (Thüringer et al., 1993). We note that the loss of expression due to the BC mutation coincides with the two main sources of dpp expression (in ps7 and 3; cf. St Johnston and Gelbart, 1987; Bienz, 1994; Figure 1). Moreover, the residual BC expression in ps8/9 coincides with the main source of wg expression in the middle midgut (in ps8; van den Heuvel et al., 1989; Figure 1). This suggests that BC still responds to Wg, but no longer to Dpp signalling. We tested this by monitoring the response of B, B5 and BC to ectopic expression of Dpp or Wg. In the case of B, ectopic Dpp or Wg each produces a slight widening of the lacZ stripes and an increase of their staining intensity; however, lacZ expression is still undetectable in certain midgut regions (e.g. in ps10/11; Figure 2b and c; Thüringer et al., 1993). In the case of B5, lacZ staining is strongly increased under both conditions, and staining induced by either signal extends throughout the midgut VM (Figure 2e and f). In contrast, in the case of BC, there is some additional lacZ staining in response to ectopic Wg (Figure 2h), but there is no significant change of the normal BC pattern in response to ubiquitous Dpp (Figure 2i). As expected, B5‐mediated lacZ staining, like B‐mediated staining (Riese et al., 1997), is substantially reduced in dpp mutants, whereas there is little change in the lacZ staining levels due to BC in these mutants (not shown). Most significantly, the BC mutation is the only one of 12 point mutations introduced into B (Riese et al., 1997; J.Riese and S.Eresh, unpublished data) which causes complete loss of responsiveness to Dpp. We therefore conclude that the sequence TGGCGTCA acts as a Dpp response sequence (DRS) in the VM. Conversely, since B5 responds readily to Dpp and Wg, it is unlikely that the FP5 repressor is negatively regulated by either signal. Instead, it appears to be a constitutive repressor which antagonizes the stimulating effects of the two signals.
We asked whether the DRS might be sufficient to respond to Dpp in the midgut. We oligomerized four copies of the CRE flanked by residues from FP5 (5CRE) or from FP4 (4CRE) (Figure 1), and we placed these adjacent to a canonical TATA box. 5CRE (Figure 2j) and 4CRE transformants (not shown) both show conspicuous lacZ stripes in the midgut, in each case a wide and strongly staining one in ps6/7, and a weak narrow stripe in ps3. Each stripe is near a source of Dpp, which implies that 5CRE and 4CRE might respond directly to Dpp signalling. Indeed, while 5CRE expression is not changed in response to ectopic Wg (Figure 2k), this construct responds very clearly to Dpp in that lacZ staining is stronger and expanded through most of the midgut as a result of ectopic Dpp (Figure 2l). 4CRE also responds to ectopic Dpp, although less extensively than 5CRE. To ascertain that the midgut staining in these constructs is due to the CRE, we made two mutant versions of 4CRE: we introduced base substitutions into each CRE copy within 4CRE (4CRE‐BC; Figure 1), or into the 5′ flanking sequences of the CREs (4CRE‐FL; Figure 1). As expected, in 4CRE‐BC transformants, we no longer observed any lacZ expression in the midgut, while the 4CRE‐FL transformants showed a midgut expression pattern indistinguishable from that of 4CRE (not shown). Finally, 4CRE‐mediated lacZ staining in the midgut is completely abolished in dpp mutant embryos (not shown). These results suggest that the DRS may be sufficient to mediate Dpp‐responsive expression in the embryonic midgut.
Multiple functional CREs in the lab midgut enhancer
Curiously, lacZ staining mediated by 5CRE and 4CRE is mostly endodermal (Figure 2j), whereby the main stripe in each case roughly coincides with the region in which lab expression is induced by Dpp (Figure 1). Indeed, the shortest enhancer fragment from lab which confers robust dpp‐dependent lacZ expression in the endoderm (HZ550; Tremml and Bienz, 1992) contains four sequences resembling the CRE consensus sequence TGACGTCA (cf. Materials and methods). Three of these are contained within a minimal 255 bp fragment (HZ255) which mediates a low level of dpp‐dependent lacZ staining in the endoderm (Tremml and Bienz, 1992). We therefore asked whether these CREs are required for the endodermal response of the lab enhancers to Dpp.
We introduced minimal base substitutions into each of the four CREs in HZ550 (mutant construct 550C), or into the three CREs in HZ255 (255C), and we compared the lacZ staining patterns of these with those produced by the corresponding wild‐type fragments. We found that, while HZ550 mediates strong lacZ staining in the region of the endoderm in which lab is expressed (Figure 3a; Tremml and Bienz, 1992), 550C produces at most residual lacZ staining in some of the cells in this endodermal region (Figure 3b). In two of the five 550C transformant lines, we saw even less endodermal staining (not shown). Also, the thin lacZ stripes in the lateral epidermis (within or overlapping posterior compartments; arrows in Figure 3a) are no longer visible in any of the 550C transformants (Figure 3b). Similarly, while the wild‐type HZ255 construct mediates low but reproducible lacZ staining in the endodermal cells in which lab induction is maximal (Figure 3c; Tremml and Bienz, 1992), only one of the three 255C transformant lines showed any lacZ staining in the endoderm. This staining was very low and sporadic (Figure 3d; the ectodermal staining due to HZ255, different from that seen with HZ550, does not disappear in the 255C transformants). There was no detectable endodermal lacZ staining in the other two 255C lines (not shown). Thus, the CREs within the lab 550 enhancer are critical for this enhancer's activity.
Evidently, the Ubx CRE can mediate the response to dpp signalling in both cell layers of the embryonic midgut, in the VM and in the endoderm. This implies that other transcription factors act through the Ubx B enhancer to confer its tissue‐specific response to Dpp in the VM. In our oligo constructs 5CRE and 4CRE, the Ubx CRE is detached from its normal enhancer context and thus avoids the constraints imposed by these factors. Supporting this notion, we find that an extended version of 4CRE (L‐CRE, including a binding site for lymphocyte enhancer‐binding factor 1, or LEF‐1) produces Dpp‐responsive lacZ expression not only in the endoderm, like 5CRE and 4CRE, but also in the VM (Riese et al., 1997). This and additional evidence led us to conclude that the CRE needs to cooperate with the LEF‐1‐binding site to respond to the Dpp signal in the VM. Why the CRE should be apparently sufficient to respond to Dpp in the endoderm, we do not presently understand.
Binding of dCREB to the DRS
In order to find out which transcriptional activator might act through the DRS to confer the Dpp response, we asked whether any of the putative CRE‐binding proteins known in Drosophila would bind to the Ubx CRE. Candidates for CRE‐binding proteins include CREB (Hoeffler et al., 1988) and CREB relatives, e.g. CREM (Foulkes et al., 1991) or ATF protein (Hai et al., 1989). CREB‐like proteins belong to the large family of basic region/leucine zipper (bZIP) transcription factors which bind to DNA as dimers (reviewed by Lalli and Sassone‐Corsi, 1994). Mammalian CREB‐related proteins can also heterodimerize with AP1 proteins (e.g. Hai and Curran, 1991; Masquilier and Sassone‐Corsi, 1992; van Dam et al., 1993).
Two genes encoding CREB‐like proteins are known in Drosophila, dCREB‐A and dCREB‐B/dCREB‐2 (Abel et al., 1992; Smolik et al., 1992; Usui et al., 1993; Yin et al., 1995); dCREB‐2 is closely related to, and may be an ancestral form of, mammalian CREB and CREM (Yin et al., 1995). Like CREM, dCREB‐2 encodes multiple differentially spliced isoforms (Yin et al., 1995) of which dCREB‐B is one (called dCREB‐2c by Yin et al., 1995; note that all known isoforms of dCREB‐2 have the same bZIP domain). dCREB‐B is expressed uniformly and at moderately high levels throughout the embryonic VM and endoderm (not shown; see Materials and methods, and also Usui et al., 1993), but there does not seem to be any dCREB‐A expression in the midgut (Smolik et al., 1992; Andrew et al., 1994; our unpublished observations). There are also two Drosophila AP1 proteins, D‐Jun and D‐Fos (Perkins et al., 1990), both of which appear to be expressed throughout the two cell layers of the midgut (Perkins et al., 1990; Tremml, 1991; unpublished observations), but it is not known whether these AP1 proteins can heterodimerize with Drosophila CREBs. Interestingly, D‐fos expression is elevated to high levels in the endoderm in the lab expression domain (Perkins et al., 1990), reflecting induction by dpp independent of, and in parallel to, lab (Tremml, 1991; J.Riese, G.Tremml and M.Bienz, submitted). Based on their expression patterns in the embryonic midgut, we shall consider dCREB‐2, D‐Jun and D‐Fos as candidate proteins which may act through the CRE to mediate the Dpp response.
We first tested whether any of these proteins could bind to the Ubx CRE, using bandshift assays. Indeed, recombinant dCREB‐B binds to the wild‐type Ubx CRE sequence (which is identical to CRE2 in the lab HZ550 enhancer; see Materials and methods), but not to the mutant sequence BC (Figure 4, lanes 2–6 and 9–13). As expected, the same is true for dCREB‐2a (not shown). However, neither recombinant D‐Jun nor D‐Fos by themselves bind to the CRE (Figure 4, lanes 7 and 8). We also do not see any evidence for binding of either of these in combination with dCREB‐B (Figure 4, lanes 10 and 11). However, these binding data do not rule out a low level of binding of a putative heterodimer between dCREB‐B and D‐Jun or D‐Fos: the signal from a putative dCREB‐B–D‐Jun heterodimer might have been obscured by the signal due to the similarly sized dCREB‐B homodimer, and a signal from a putative dCREB‐B–D‐Fos heterodimer might have been below detection levels because of the low binding activity of our D‐Fos extracts (see Materials and methods). As a control, we tested the binding of these proteins to a consensus AP1‐binding site. D‐Jun clearly binds this site (Figure 4, lane 16), while D‐Fos appears to bind to it only in combination with D‐Jun (see the additional smeary bands above the main band in Figure 4, lane 20, which we observe reproducibly if recombinant D‐Fos is included in the binding reaction; but see also Perkins et al., 1990). dCREB‐B also binds to the AP1‐binding site (Figure 4, lane 15). These binding data imply that dCREB‐2 isoforms are good candidates, whereas D‐Jun and D‐Fos are poor candidates, for transcriptional activators acting through the Ubx CRE.
Dominant‐negative effects of a truncated CREB protein in the midgut
In order to test whether dCREB‐2 or AP1 proteins can act through the DRS in vivo, we generated truncated versions of dCREB‐2, D‐Jun and D‐Fos, consisting in each case of the bZIP fragment (called Cbz, Jbz and Fbz; see Materials and methods). bZIP domains such as these are known to act dominant‐negatively as they are able to dimerize and bind DNA without being able to stimulate transcription (Lloyd et al., 1991; Bohmann et al., 1994). We expressed these bZIP fragments ubiquitously in the embryo, using the yeast GAL4 system (Brand and Perrimon, 1993), to see whether any of them would affect reporter gene expression, or lab expression itself (we did not expect to see any effect on Ubx expression as lack of dpp signalling only mildly reduces Ubx expression in the VM; Immerglück et al., 1990; Panganiban et al., 1990; Reuter et al., 1990).
We found that Cbz, if expressed with a strong hs.GAL4 driver line, virtually eliminated 5CRE expression in the endoderm (Figure 5b, compare with a). This effect was not seen if Cbz expression was limited to the VM, using the mesodermal driver line 24B.GAL4 (not shown), arguing that the effect of Cbz on endodermal 5CRE expression is autonomous and direct. Neither Jbz nor Fbz showed any reduction of 5CRE‐mediated lacZ staining in the endoderm (though we did see a slight widening of endodermal 5CRE expression in the case of Jbz; this, however, appears to be caused indirectly as a similar widening is caused non‐autonomously by Jbz expression in the VM). This lack of an effect of Jbz and Fbz on CRE‐mediated expression is not due to inactivity or instability of these bZIP protein fragments since both bZIP constructs strongly interfere with proper eye development when expressed in the eye imaginal disc (D.B. and D.B.J., unpublished results). More significantly, Fbz interferes with copper cell development when expressed in the embryonic endoderm (J.Riese, G.Tremml and M.Bienz, submitted; see also below).
We also stained embryos expressing each of these bZIP constructs with lab antibody. We found that, in the case of ubiquitous Cbz, lab staining in the endoderm was significantly reduced, and even absent in some endodermal cells in the ps6/7 region (Figure 5d, compare with c). This reduction of staining was not seen after mesodermal expression of Cbz, or after ubiquitous expression of Jbz. Ubiquitous expression of Fbz caused a reduction of lab antibody staining similar to ubiquitous Cbz expression (not shown). Consistent with this, endodermal expression of Fbz leads to copper cell defects in the larval gut (J.Riese, G.Tremml and M.Bienz, submitted; recall that copper cells require continuous lab function in order to develop; Hoppler and Bienz, 1994). Note, however, that the suppressive effect of Fbz on lab expression and on copper cell development most probably is not mediated by the lab CREs since we cannot detect any effect on 5CRE‐mediated lacZ staining under the very same conditions of expressing ubiquitous Fbz (see above; note that reporter gene expression is typically a more sensitive assay than expression of an endogenous gene; e.g. Tremml and Bienz, 1992; Riese et al., 1997; Yu et al., 1996). This result is fully consistent with our failure to detect binding of D‐Fos to the Ubx CRE. We therefore presume that the suppressive effect of Fbz on lab expression is mediated through AP1‐binding sites that are located outside the lab 550 enhancer (there are no AP1‐binding sites in the lab 550 enhancer fragment; Tremml, 1991; Tremml and Bienz, 1992).
Taken together, our results strongly indicate that Drosophila CREB proteins are capable of activating transcription through the Ubx and lab CREs in the midgut. Furthermore, although D‐Fos may have a function in stimulating lab expression in the endoderm, we found no evidence that either of the two AP1 proteins, D‐Fos or D‐Jun, can act through the Ubx CRE, the Dpp response sequence in the midgut.
Our work identifies a CRE within the Ubx midgut enhancer as a target sequence for Dpp signalling in the embryonic midgut. Two lines of evidence implicate a Drosophila CREB protein in the response of midgut cells to Dpp: firstly, Drosophila CREB isoforms bind to the Ubx CRE and, secondly, expression of the DNA‐binding bZIP domain of dCREB‐2 in stably transformed embryos acts dominant‐negatively to suppress expression from a DRS‐containing reporter gene and to reduce lab expression. Taking into account their uniform expression in the embryonic midgut, dCREB‐2 isoforms, rather than dCREB‐A, are good candidates for transcription factors acting through the DRS. Finally, we have shown that the DRS mediates transcriptional activation, and we have not found any evidence for a repressor acting through the Ubx CRE. As dCREB‐2a is the only dCREB‐2 isoform known to be a transcriptional activator (Usui et al., 1993; Yin et al., 1995), dCREB‐2a is currently the best candidate for a transcription factor involved in the response to Dpp in the embryonic midgut. Interestingly, dCREB‐2a is the only CREB isoform known to be signal responsive (Yin et al., 1995). However, we would like to point out that there may be additional dCREB‐2 isoforms and additional CREB‐like genes, unidentified as yet, that could be involved in this process.
Recently, we have identified a LEF‐1‐binding site within the FP4 region of the Ubx midgut enhancer as the target sequence for Wg signalling (WRS) in the embryonic midgut (Riese et al., 1997). We have shown that, in contrast to the DRS, the WRS is not sufficient to confer transcriptional stimulation on its own, but that it requires linkage to the DRS. These and additional results led us to propose that a Drosophila LEF protein mediates integration of Wg and Dpp signalling. Interestingly, mouse LEF‐1 by itself is not a transcriptional activator, but functions in concert with other enhancer‐binding proteins one of which is a CREB (Carlsson et al., 1993; Giese and Grosschedl, 1993). This is an additional, and independent, indication that the protein acting through the DRS may be a Drosophila CREB protein.
We did not find any evidence that Drosophila AP1 proteins could act through the DRS: we failed to detect binding of D‐Jun and D‐Fos to the Ubx CRE in vitro, and we also failed to see dominant‐negative effects of their bZIP domains on CRE reporter gene expression in the midgut. Interestingly, we did see a suppressive effect of D‐Fos bZIP on lab expression, indicating a role for D‐Fos in the transcriptional regulation of lab (J.Riese, G.Tremml and M.Bienz, submitted). However, our evidence does not support the idea that D‐Fos takes part in the direct transcriptional response to Dpp signalling; rather, it suggests that D‐Fos may act in parallel to Dpp signalling to stimulate lab transcription.
Our results raise the possibility that Dpp signalling may modify the activity of a CREB protein, or that of a CREB dimerization partner. Mammalian CREB is known to be phosphorylated, and thus activated, in response to cAMP (Gonzales and Montminy, 1989; Lee et al., 1990). Protein kinase A (PKA) phosphorylates CREB at a critical serine residue (conserved in dCREB‐2; Usui et al., 1993; Yin et al., 1995) which facilitates binding of CREB to the CREB‐binding protein CBP, a step that is thought to contribute to target gene activation (Chrivia et al., 1993). In the Drosophila midgut, we think it unlikely that PKA plays a significant role, as overexpression of a constitutively activated PKA catalytic subunit (Jiang and Struhl, 1995; Li et al., 1995) affects neither midgut morphology nor expression of Ubx, lab or their reporter genes (unpublished observations). However, CREB and CREM can also be phosphorylated by other kinases in vitro and in vivo (de Groot et al., 1993), including a Ras‐dependent CREB kinase (Ginty et al., 1994), implying that CREB‐like proteins are targeted by signals other than cAMP. Indeed, it has been reported that phosphorylation of CREB transfected into mammalian cells is increased after TGF‐β stimulation of these cells (Kramer et al., 1991). However, it remains to be seen whether Dpp signalling directly causes modification of a Drosophila CREB protein.
Recently, a gene called schnurri (shn) has been described which is required downstream of the Dpp signal in multiple developmental contexts including the embryonic midgut (Arora et al., 1995; Grieder et al., 1995). This led to the proposal that the shn product, a zinc finger protein, may be a target transcription factor of Dpp signalling (Arora et al., 1995; Grieder et al., 1995). However, preliminary results from in vitro DNA binding assays with individual shn zinc fingers suggest that these fingers bind neither to the Ubx CRE nor to the FP5 sequence with high affinity (M.Affolter, K.Arora and R.Warrior, personal communication). However, lacZ expression mediated by 5CRE is abolished in shn mutant embryos even if Dpp is resupplied with a heat‐shock promoter (M.Affolter, personal communication). This raises the possibility that the requirement for shn in the response to Dpp signalling may be an indirect one.
Finally, what is the role of FP5, the sequence overlapping the CRE? Evidently, this sequence antagonizes the activating effects of Dpp and Wg signalling on the Ubx enhancer, and our results argue that the FP5 repressor is constitutively active and not controlled by either signal. The close physical linkage of FP5 and the CRE suggests that there may be competition for transcriptional activation of Ubx between the CRE‐binding activator and the FP5 repressor at the level of DNA binding. As a consequence, the signal response of Ubx would be spatially limited. It is very common that cis‐regulatory elements controlling the spatial expression of developmental regulators contain arrays of closely linked or overlapping binding sites for transcriptional activators and repressors (e.g. Small et al., 1991). Such arrays constitute transcriptional switches that are eminently sensitive to small changes of repressor and/or activator availability (reviewed by Ptashne, 1986). A switch designed like CRE/FP5 is likely to confer a sharp response to signalling thresholds, and similar switches might account for the strikingly sharp responses to TGF‐β‐type signalling as observed for Xenopus embryonic cells (Green et al., 1992). Thus, such switches would appear to be ideal targets for extracellular signals and morphogens.
Materials and methods
The following fly transformants were used: Bhz (Thüringer et al., 1993); HZ550 and HZ255 (Tremml and Bienz, 1992); hs‐wg (Noordermeer et al., 1992); UAS.dpp and 24B.GAL4 (Brand and Perrimon, 1993; Staehling‐Hampton et al., 1994); and a strongly expressing hs.GAL4 line (Brand et al., 1994). The dpps4 allele (Immerglück et al., 1990) was used to test dpp dependence of reporter gene expression. Mutant embryos were identified by their midgut morphology; note that the dpps4 mutation selectively affects midgut expression of our reporter genes, but not their expression elsewhere, e.g. in the ectoderm (cf. Immerglück et al., 1990).
B5 and BC substitutions (Figure 1) were generated by standard procedures, using mutator oligomers, and mutant constructs were generated analogously as the wild‐type construct Bhz (Thüringer et al., 1993). For CRE5, four copies of the 5CRE oligomer sequence (Figure 1; one copy in the ‘non‐coding’ followed by three copies in the ‘coding’ orientation) separated by TCGA linkers were cloned into the SalI site of Bluescript and subcloned as an XbaI–XhoI fragment into the transformation vector cut with XbaI and KpnI (XhoI and KpnI blunt‐ended). The same was done for 4CRE, except that the linkers between individual oligomer copies were TTTC (between oligomer 1/2 and 3/4) and TCGACGGTATCGTCGAGGTCGA (between oligomer 2/3); the final orientation in the transformation vector was ‘non‐coding’. Two distinct mutant versions of 4CRE (4CRE‐BC and 4CRE‐FL) were generated by using oligomers with base substitutions as shown in Figure 1.
The lab HZ550 enhancer fragment contains four CRE‐like sequences (Tremml, 1991) which match the CRE consensus sequence in 7/8 (CRE1, CRE2) or 6/8 positions (CRE3, CRE4): TCACGTCA (CRE1), TGGCGTCA (CRE2; same sequence as the Ubx CRE; Figure 1), TGTGGTCA (CRE3), GAACGTCA (CRE4). The following base substitutions (indicated by lower case letters) were introduced into these: TagtactA (CRE1), TGctcgag (CRE2), atgcGcaA (CRE3) and GAggGcCc (CRE4). Mutant constructs with these substitutions (255C, CRE2–4 mutated; 550C, each CRE mutated) were generated analogously to the corresponding wild‐type constructs HZ550 and HZ255 (Tremml and Bienz, 1992; note that the HZ255 bp construct contains a NarI–ClaI fragment which constitutes the 3′ portion of the 550 bp ClaI fragment contained in HZ550, instead of a BstXI fragment from the central portion of HZ550 as indicated in Figure 1 of Tremml and Bienz, 1992; sequence available on request).
The Cbz, Jbz and Fbz constructs were generated using standard PCR‐based methods. These constructs encompassed amino acids 183–289 of D‐Jun, 252–337 of D‐Fos (Perkins et al., 1990) and 223–285 of dCREB‐B (Usui et al., 1993). A consensus translation initiation sequence (Cavener, 1987) was engineered at the 5′ end of these open reading frames. These constructs subsequently were cloned into the pUAST germline transformation vector (Brand and Perrimon, 1993). Full‐length pUAST constructs were also made for dCREB‐B (Usui et al., 1993), dCREB‐2a (Yin et al., 1995), D‐Fos and D‐Jun (Perkins et al., 1990; see also Bohmann et al., 1994); details of constructs available on request.
P‐element transformation and analysis of transformants
For each construct, 3–5 individual transformant lines were isolated and made homozygous for the transposon; cn; ry42 was used as a host strain for lacZ constructs, y w1118 for the bZIP constructs. Analysis of lacZ expression was done as described (using formaldehyde fixation and a monoclonal mouse antibody against lacZ; Busturia and Bienz, 1993; Thüringer et al., 1993). The rat polyclonal antiserum against lab protein was generated by Tremml (1991). We used the same hs‐wg strain and heat‐shock procedure as described (Thüringer et al., 1993), but we used the GAL4 system (UAS.dpp and 24B.GAL4, see above) to express dpp throughout the mesoderm since this produced a stronger and more reproducible Dpp response than the hs‐dpp strain previously used. To produce clear and strong effects with good penetrance in the case of bZIP constructs, both the hs.GAL4 driver and the bZIP‐encoding UAS transposon had to be homozygous (although the same effects were also observed with poor penetrance in the presence of just one copy each). The following heat‐shock conditions were used: 4–8‐h‐old embryos were subjected to four consecutive heat shocks at 37°C (20 min each; plates immersed in a waterbath) separated by 2 h at 25°C. UAS constructs expressing the full‐length dCREB‐B, dCREB‐2a, D‐Jun and D‐Fos proteins produced phenotypic effects in wings (to be described elsewhere); however, none of these produced any effects on reporter gene expression or on midgut morphology when ubiquitously expressed in the embryo as described.
Recombinant dCREB‐B (Usui et al., 1993) was purified by standard procedures by virtue of its His‐Tag, and injected into rats to produce a polyclonal antiserum. This serum recognizes recombinant dCREB‐B and dCREB‐2a, but not dCREB‐A on Western blots. Embryos stained with this antiserum show moderately high levels of stained nuclear antigen in most if not all embryonic cells, including VM and endodermal cells which show nuclear staining levels uniformly throughout the midgut, confirming earlier studies of dCREB‐B transcript expression by Usui et al. (1993). Uniform expression of D‐Jun throughout the VM and endoderm was also observed using a polyclonal rabbit antiserum against D‐Jun (Bohmann et al., 1994).
Crude protein extracts and DNA binding assays
For the preparation of crude protein extracts from embryonic nuclei (0–24‐h‐old embryos) and subsequent DNase I footprinting, the protocols of Biggin and Tjian (1988) were followed. As competitor DNA, 1 μg of poly(dIdC) was added into binding reactions of 50 μl; increasing amounts of protein extract were added into these reaction as follows: 1.7, 3.4, 5.1, 6.8, 8.5 and 10.2 μg (see Figure 4a). To identify the footprint regions (Figures 1 and 4a), the protection patterns from various experiments with different extracts were averaged.
Crude protein extracts containing recombinant dCREB‐B (Usui et al., 1993), dCREB‐2a (J.Yin, unpublished), D‐Jun (Peverali et al., 1996) and D‐Fos (F.Peverali and D.Bohmann, unpublished) were prepared essentially as described by Studier et al. (1990), with the following modifications. Harvested bacterial cells were resuspended in 100 μl of lysing buffer [1 mM phenylmethylsulfonyl fluoride (PMSF), 1 mM β‐mercaptoethanol, 0.2 mg/ml lysozyme in phosphate‐buffered saline pH 8.5], incubated for 20 min on ice followed by sonication on ice in 1% Triton X‐100 (final concentration). After centrifugation, the supernatant was used as a crude protein extract. These extracts were incubated for 10 min on ice in binding buffer (20 mM HEPES pH 7.9, 20% glycerol, 100 mM KCl, 0.1% NP‐40, 20 mM MgCl2, 0.5 mM dithiothreitol, 3 mg/ml bovine serum albumin) in a final volume of 20 μl. After addition of radiolabelled oligomer probe (15 000 c.p.m.), the mix was incubated for a further 20 min on ice. The resulting complexes were separated on 6% native polyacrylamide gels run in 0.5× Tris borate buffer. Oligomer probes were end‐labelled with [γ‐32P]ATP and T4 polynucleotide kinase and reannealed according to standard procedures. The following oligomer sequences were used: wild‐type CRE, GGGCTGGACTGGCGTCAGCGCCGG; BC mutant CRE, GGGCTGGACTGGgccCAGCGCCGG (base substitutions in lower case letters); AP1, GAGCCGCAAGTGACTCAGCGCGGGGCGTGTGCAGG.
We thank Mark Biggin for protocols and advice in preparing embryonic protein extracts, Sarah Smolik, Jerry Yin and Fiorenzo Peverali for plasmids and advice on bacterial expression, Gabi Tremml for the lab antibody, Michael Hoffmann and Andrea Brand for fly strains, and Matthew Freeman, Jürg Müller and Xiang Yu for comments on the manuscript. J.R. is an MRC student supported by the Boehringer Ingelheim Fonds, D.B.J. is a recipient of a TMR student fellowship.
↵† S.Eresh and J.Riese contributed equally to this work
- Copyright © 1997 European Molecular Biology Organization