Two different activities of Suppressor of Hairless during wing development in Drosophila

The Notch signalling pathway plays a central role during the assignation of cell fates in Drosophila. For example, during the selection of neural or muscle precursors from clusters of equivalent cells, the Notch pathway is required to ensure that only some of the cells of the clusters adopt such fates (reviewed in Artavanis-Tsakonas et al., 1999). In other instances, e.g. during wing development, Notch signalling provides an inductive signal essential for the growth and patterning of the wing (Speicher et al., 1994; Couso et al., 1995; Diaz-Benjumea and Cohen, 1995; deCelis et al., 1996a; Doherty et al., 1996; Jönsson and Knust, 1996; Klein and Martinez Arias, 1998a).

The current view of signalling through Notch involves the interaction of three core elements, a DSL signal, a LNG receptor and a CSL transcription factor. In Drosophila, two DSL proteins are known, Delta (DL) and Serrate (Ser). Both ligands are single transmembrane proteins with several EGF-like repeats and the DSL domain in their extracellular domain (see Greenwald, 1998; and Artavanis-Tsakonas et al., 1999, for further details). They can activate the only Drosophila LNG receptor Notch, which is also a single transmenbrane protein with several EGF repeats in its extracellular domain. In contrast to its ligands, Notch has a large intracellular domain,which includes a region of six CDC 10-like repeats that are essential for signal transduction. There is only one CSL factor in Drosophila, Suppressor of Hairless (Su(H)).

The activation of Notch by its ligands results in the cleavage and release of its intracellular domain (Nintra), which, together with the nuclear protein Su(H), generates a bifunctional transcription factor required for target gene regulation in the nucleus (Blaumueller et al., 1997; Pan and Rubin, 1997; Fortini and Artavanis-Tsakonas, 1994; Jarriault et al., 1995; Lecourtouis and Schweisguth, 1998; Struhl and Adachi, 1998, Schroeter et al., 1998). In this complex, Nintra probably acts as the transactivation domain and Su(H) as the DNA-binding domain of the binary factor (Jarriault et al., 1995; Hsieh et al., 1996; Wettstein et al., 1997; Lecourtois and Schweisguth, 1998; Kidd et al., 1998; Struhl and Adachi, 1998). Recent results suggest that the cleavage of Nintra requires the function of the transmembrane protein Presenilin (Psn) (De Stroopper et al., 1999; Struhl and Greenwald, 1999; Ye et al., 1999).

Despite the many similarities that exist between Notch signalling in different species, there appear to be some differences. For example, in the absence of Notch signalling, the vertebrate CSL factor CBF1 seems to act as a suppressor of gene expression (Hsieh et al., 1996; Kao et al., 1998), but suppression of gene expression only applies to Su(H) during formation of the midline of the embryonic CNS (Morel and Schweisguth, 2000). Further, in Drosophila, the product of the Hairless (H) gene can antagonize the function of Su(H) (Brou et al., 1994; Bang et al., 1995).

Here we have characterized the role of Su(H) in Notch signalling during two processes: the selection of the adult sensory precursors and the development of the wing. Our results show that during neural precursor selection, but not wing development, the activity of Su(H) can be mediated by the genes of E(spl)C alone. They also show that during wing development, Su(H) can activate expression of a subset of Notch target genes in the absence of the intracellular domain of Notch (Nintra), indicating that Su(H) can activate gene expression alone or in combination with factors other than Nintra. In other instances, Su(H) is capable of repressing gene expression; these targets, however, are activated by a fusion protein between Su(H) and the transcriptional activator VP16, suggesting that in these cases Su(H) acts in concert with Nintra, as a transcription activation domain. The results also suggest that Hairless is a suppressor of Su(H) activity during wing development. Based on our results, we suggest a model in which Su(H) function is activated by Notch through the release of Su(H) from an inhibitory complex containing Hairless.

The ap^UG035 allele is described in Cohen et al. (1992). The Su(H) alleles AR9 and SF8 are described in Schweisguth and Posakony (1992). Psn^B3 and Psn^I2 are null Presinillin mutants and were provided by Mark Fortini (Yeh et al., 1999; Lukinova et al., 1999). FRT82BH^E31 provided by F. Schweisguth is a null mutant described in Schweisguth and Lecourtois (1998). The insertion line A101 labels all sensory precursor in the imaginal discs and is described in Huang et al. (1991).

Expression of vg at the DV boundary was detected using the vestigial Boundary Enhancer is described in Williams et al. (1994) and referred to here as vgBE (for Boundary Enhancer). The vg-quadrant enhancer is described in Kim et al. (1996) and is referred to here as vgQE. For the expression of wg in the developing discs, we used a lacZ insertion in the wg gene on a CyO chromosome (Kassis et al, 1992). The Dl lacZ line is described in Klein and Martinez Arias (1998b), E(spl)m8 lacZ in Lecourtois and Schweisguth (1995). E(spl)mβCD2 is a gift of S. Bray and described in Dominguez and deCelis (1998).

Ectopic expression of the different genes was achieved through the GAL4/UAS system of Brand and Perrimon (1993). The UAS construct used were the following: UAS-GFP (Yeh et al., 1995), UAS-Su(H)VP16 (Kidd et al., 1998), UAS-m8 (Nakao and Campos-Ortega, 1996), and UAS-m7 (Ligoxygakis et al., 1999). The expression of the different UAS constructs was driven in the imaginal discs with various GAL4 inserts. patched Gal4 (ptc-Gal4) expresses UASX in a stripe along the AP boundary of the discs (Speicher et al., 1994). In the third instar, decapentaplegic Gal4 (dpp-Gal4) (Wilder and Perrimon, 1995) is expressed in a similar pattern to ptc-Gal4, although the expression is weaker over the ventral side (Klein and Martinez Arias, 1998a). The klu-Gal4 is expressed in all territories of bristle development and is described in Klein and Campos-Ortega (1997).

Stocks carrying various GAL4 and UAS combinations in wild type and mutants were generated. All stocks were balanced over the SM6a-TM6b compound balancer, which allowed the identification of larvae of the correct genotype because of the dominant larval marker Tb. Details of the stocks as well as the stocks themselves are available upon request. In the case of ap mutations, stocks were established with a CyO balancer carrying a P-lacZ insertion in wg and the mutant discs were checked for the absence of the wg expression pattern. In the case of some experiments involving ap mutants, mutant discs were identified by the aberrant morphology of the wing disc and the absence of the CyO wg-lacZ balancer.

Clones were induced using a UAS-Flp construct (kindly provided by N. Perrimon; Duffy et al., 1998) activated with ptc-Gal4. The Df(1)N81K FRT101 chromosome is a gift from K. Brennan and carries a null allele of Notch (Brennan et al., 1997). The y w (Ubq.GFP) (FRT101) chromosome is a gift from N. Perrimon. Both chromosomes are described in Brennan et al. (1999).

The following donated antibodies: S. Cohen (anti-Wg), Sean Carroll (anti-Vg), E. Knust (anti-Ser), F. Schweisguth (anti-Su(H)), S. Bray (anti-E(spl) – mAb 323), S. Artavanis-Tsakonas (anti-Notch – C17.9C6). The Cut antibody developed by G. Rubin was obtained from the Developmental Studies Hybridoma Bank developed under the auspices of the NICHD and maintained by the University of Iowa, Department of Biological Sciences, Iowa City, IA 52242. Staining were performed according to standard protocols. The secondary antibodies were purchased by Jackson Immuno Research Laboratories. Staining were performed according to standard protocols.

The degree of rescue of the Su(H) mutant phenotype by the different lines allowed us to select a strong expresser (UAS-Su(H)II) and a weak one UAS-Su(H)III). Anti-Su(H) stainings revealed that only the second chromosomal construct can be detected slightly above the normal level of expression in third instar wing discs, if expressed with ptc-Gal4. UAS-Su(H)III produces only very low levels of Su(H) expression. However, as described here, these levels seems to be sufficient to provide the full function)

UAS-Su(H) construct were made as follows: the 2.8 kb HindIII-EcoRI Su(H) cDNA fragment was inserted into Bluescript KS+ (KS12, Schweisguth and Posakony, 1992) and subcloned in pUAST (Brand and Perrimon, 1993) as a XhoI-XbaI fragment. Transformant flies bearing this construct were obtained using standard procedures.

To study the role of Su(H) in Notch signalling in Drosophila, we have established transgenic lines carrying the Su(H) gene under control of the yeast Gal4 targetted expression system (Brand and Perrimon, 1993; see Materials and Methods). Since Su(H) is expressed ubiquitously in all adult tissues (Gho et al., 1996), the expression of Su(H) with this system simply leads to an overexpression of the protein and consequently we shall use this term throughout this work. Lines carrying UAS-Su(H) were tested for functionality in two ways. First we tried to rescue the defects of Su(H) null mutants; secondly, we monitored the effects of overexpression of Su(H) on the development of the peripheral nervous system in the wild type (Fig. 1). In the absence of Su(H), the wing primordium is established but does not grow and lacks the definition of a margin (Schweisguth and Posakony, 1992; Couso et al., 1995; Klein and Martinez Arias, 1998a). In the notum, lateral inhibition fails and, as a consequence, there is an excess of neural precursors in each of the proneural clusters (Schweisguth and Posakony, 1992). Expression of a weak UAS-Su(H) line (see Materials and methods) in Su(H) mutants rescues both phenotypes (Fig. 1A-H). In one these experiments, we expressed UAS-Su(H) under the control of dpp-Gal4 and observed a rescue of the size of the wing pouch and of wg expression within the domain of dpp-Gal4 expression (Fig. 1A-D). Although we expressed Su(H) throughout the anterior compartment, the expression of wg was restricted to a narrow line of cells in the middle of the pouch, which probably corresponds to the dorsoventral (DV) boundary (Fig. 1C). This pattern is likely to reflect the interactions between the activity of Notch, which is restricted to the DV boundary, and Su(H) (Klein et al., 1998).

In another set of experiments, we used klu-Gal4 to express UAS-Su(H) in the proneural clusters of Su(H) mutants (see Fig. 1H; Materials and methods), and observed a strong suppression of the Su(H) mutant phenotypes both in neurogenesis and wing development (Fig. 1E-H). The rescue of the wing pouch is interesting, since klu-Gal4 is expressed only as a narrow stripe along the DV boundary (Klein and Campos-Ortega, 1997), and indicates that, for the development of the wing, the activity of Su(H) is only required at the DV boundary.

The observation that UAS-Su(H) can provide lateral inhibition function in Su(H) mutants led us to test if overexpression of Su(H) in the wild type can increase lateral inhibition. Overexpression of UAS-Su(H) with klu-Gal4 results in the loss of nearly all neural precursors and derived bristles in the notum and the wing margin of wild-type flies (data not shown). This loss of bristles is not due to a loss of proneural gene expression since Achaete is initially expressed in the notum of these discs (data not shown), and therefore must reflect an excess of lateral inhibition.

Altogether these results indicate that our Su(H) construct is able to mediate all aspects of Su(H) activity.

A variety of genetic results have suggested that the function of Su(H) during lateral inhibition is mediated by genes of the Enhancer of split complex (E(spl)C) (Jennings et al., 1994; Nakao and Campos-Ortega, 1996; Lecourtois and Schweisguth, 1995). To test whether or not this is the case, we have made use of the fact that the expression of klu-Gal4 is independent of Su(H) (see above) and thus drives expression of two members of the E(spl)C, m8 and m7, in Su(H) mutant wing discs (Fig. 2). Expression of either construct alone leads to a strong suppression of the neurogenic phenotype caused by loss of Su(H) function, but is unable to correct the defects of wing development and patterning of this mutant (data not shown). This result demonstrates that, as suspected from loss-of-function experiments, the expression of a single gene of the E(spl)C is sufficient to mediate the full activity of Su(H) during selection of neural precursors in the adult PNS (deCelis et al., 1996b; Heitzler et al., 1996). They also show that the bHLH proteins of the E(spl)C do not mediate the activity of Su(H) during the development of the wing blade (see also deCelis et al., 1996a).

Notch signalling relies on interactions between the intracellular domain of Notch and Su(H) (reviewed in Artavanis-Tsakonas et al., 1999) and it has been suggested that the stoichiometry of this interaction is an important element of this activity (Kidd et al., 1998). To test this in vivo, we have compared the abilities of Nintra and Su(H) on their own to elicit gene expression in a wild-type context. In these experiments, we have focused on the developing wing pouch and have monitored the expression of several targets of Notch signalling. Some of them, for example, cut, wg, Dl, Ser, the vg-boundary enhancer (vgBE) and E(spl)m8, are activated by Notch/Su(H) activity and one, the vg-quadrant enhancer (vgQE), is suppressed (Couso et al., 1995; Diaz-Benjumea and Cohen, 1995; Doherty et al., 1996; Kim et al., 1996; deCelis et al., 1996b; Klein and Martinez Arias, 1999). The results are summarized in Figs 2 and 3.

Surprisingly, overexpression of Su(H) on its own mimics only some of the effects of Nintra and can elicit ectopic expression of the members of the E(spl)C detected by mAb 323 (Jennings et al., 1994; Fig. 2C,D), Dl, Ser and vgBE (and Vg, data not shown, see Fig. 5J), also suppress the activity of the vgQE (Figs 2L-O, 3A-E).

We notice that, in these experiments, the overexpression of Su(H) induces ectopic Ser expression only in the dorsal half (Fig. 3B), an effect also produced by Nintra (Klein and Martinez Arias, 1998b). This effect suggests that the differences between dorsal and ventral cells are due not to the differential distribution of Fringe (Panin et al., 1997), but to intrinsic differences, which might have been inherited from the embryo (Klein and Martinez Arias, 1998b). We further observe that the overexpression of Su(H) elicits the typical overgrowth of the imaginal disc also observed with UAS-Nintra or UAS-Dl.

In contrast to UAS-Nintra or UAS-Dl, which activate Notch signalling constitutively, UAS-Su(H) it is not able to ectopically activate the expression of cut, wg, E(spl)m8 and E(spl)mβ in the wing pouch (Fig. 2A,B,E,F,G-I,J,K). This could be interpreted as a result of different thresholds of the Nintra/Su(H) activity required for the activation of the expression of different target genes. However, in the case of wg and E(spl)m8, overexpression of Su(H) in the wild type does not induce their expression, even when it is expressed at high temperature (29°C, Fig. 2H), at which the Gal4 system is highly active, or with other strongly activating Gal4 lines such as dpp-Gal4 or sd-Gal4 (Fig. 2H; data not shown). Furthermore, in the case of E(spl)m8, E(spl)mβ and cut (cut-lacZ and Cut protein), UAS Su(H) not only fails to induce gene expression, but actively represses it (even at high expression levels, see Fig. 2B,F,H,I). These results argue against the possibility that the levels of Su(H) overexpression are not sufficient to activate the expression of certain target genes.

One simple explanation for the suppression of expression of some genes by Su(H) is that Su(H) might ‘squelch’ other limiting proteins from promoters and therefore leads to their inactivation (Gill and Ptashne, 1988). Since it is not known if, for example, cut is a direct target of Notch signalling, it is possible that Su(H) squelches a protein from its promoter, which leads to loss of cut expression. However, squelching has so far only been observed for promoters that do not have a binding site for the overexpressed transcription factor (Gill and Ptashne, 1988). In the case of the vgBE and the E(spl)m8 promoter, it has been shown that Su(H) is required for their activation by direct binding to target sites in their control regions (Lecourtois and Schweisguth, 1995; Kim et al., 1996). Despite this requirement, vgBE expression is activated, whereas that of E(spl)m8 is suppressed. Therefore, these data suggest that the differences observed in the effects of UAS-Su(H) and UAS-Nintra are qualitative and that Su(H) cannot activate certain promoters in the absence of Notch signalling. They further suggest that the genes whose expression is activated through Su(H) overexpression are activated by Su(H) alone. The results therefore raise the possibility that Su(H) can activate expression of some genes without Notch activation.

The observation that some target genes of Notch signalling, e.g. wg, are not activated by overexpression of Su(H) alone, but are activated by ectopic expression of Dl or Nintra prompted us to test whether Notch activity might induce expression of some of these genes in the absence of Su(H). Evidence for such a Su(H)-independent Notch pathway has recently accumulated (see e.g. Artavanis-Tsakonas et al., 1999)

It has been reported that expression of a Nintra construct under the control of the heat-shock promoter is not sufficient to rescue the defects occurring in Su(H) mutant discs (Bailey and Posakony, 1995). However, in these experiments, Notch activity is induced by a relatively short heat shock and therefore might decay before eliciting some morphological changes in the Su(H) mutants. Further, in this experiment, only the expression of E(spl) genes was monitored and not wing formation. For this reason, we repeated this experiment providing sustained expression of UAS-Dl and UAS-Nintra in Su(H) mutant discs with dpp-Gal4, whose expression is independent of Su(H) (see above and Fig. 1C). In these experiments, we never observed any effect of Notch activation in the absence of Su(H) (Fig. 4A-C). This is not likely to be due to low levels of expression since, in ap mutants, the same combinations are sufficient to induce a wing pouch with a margin (Klein and Martinez Arias, 1998b).

One caveat of this experiment is that Su(H) is required for the activation of the pouch-determining gene vg early in development (Kim et al., 1996; see above) and genes like wg are activated by Notch only in the pouch, under the influence of vg (Klein and Martinez Arias, 1998a, 1999). Therefore, it is possible that the absence of vg in Su(H) mutants obscures any possible effect of Notch activity in these mutants. To account for this, we coexpressed UAS-Ser and UAS-vg in Su(H) mutant discs. In these experiments, we also found no significant activation of wg expression or any significant change in the morphology of the discs (Fig. 4D,E). This failure is not due to an inefficiency of Ser to activate Notch, since its expression in ap mutant discs in the same way leads to a good rescue of the wing pouch and margin (Klein and Martinez Arias, 1998b). Therefore, we conclude that Su(H) is absolutely required for the activation of all Notch target genes during wing development. However, for the expression of some target genes, like wg, cut and E(spl)m8 and mβ, it is not sufficient.

The ability of Nintra to activate genes like wg, cut, E(spl)m8 and mβ suggests that Nintra is required for the activation of these genes in addition to Su(H). It is likely that in these cases Nintra is acting as a transactivation domain for Su(H), as has been proposed before (Jarriault et al., 1995; Hsieh et al., 1996; Wettstein et al., 1997; Kidd et al., 1998). One prediction of this model is that Su(H) could activate expression of these targets, if a transactivation domain is provided to it. Therefore we expressed a Su(H) construct bearing the viral VP16 transactivation domain (Kidd et al., 1998) in the wild type and monitored the expression of various Notch targets. We found that this Su(H)-VP16 construct is able to ectopically activate the expression of wg, cut and the vgBE in a similar manner as Nintra (Fig. 4F-H). Furthermore, the expression of the vgQE is suppressed by this construct as when Su(H) is overexpressed (Fig. 4I), suggesting that this effect is mediated through a factor activated by Su(H). The effects of Su(H)-VP16 expression on cut expression are especially interesting, because on its own Su(H) represses the expression of this gene. Adult wings, where Su(H)-VP16 is expressed show ectopic margin structures, similar to these induced by Nintra (data not shown).

Altogether these results suggest that Su(H)-VP16 can mimic the activity of Nintra and further support our conclusion that Su(H) is required for the expression of all targets of Notch signalling during wing development. They also show that, in cases like wg and cut, Su(H) is not sufficient to activate expression and seems to require transactivation domain, probably Nintra, as a partner.

The results presented above suggest that Su(H) can activate gene expression in the absence of Notch activity. To test this further, we first overexpressed UAS-Su(H) in wing discs that lack Notch signalling, either because of the absence of ligands for Notch, as in an ap mutant (Couso et al., 1995; Klein and Martinez Arias, 1998b; Klein et al., 1998), or because of a failure to process Notch effectively, as in mutants for Presenilin (Psn) (Struhl and Greenwald, 1999; Ye et al., 1999). In both cases, expression of Su(H) under the control of dpp-Gal4 elicits the development of a wing pouch as defined by the growth of the tissue and the activity of the two enhancers of vg, vgQE and vgBE or vg expression itself (Fig. 5A-J). However, in contrast to the effects of Nintra, Ser or Dl, Su(H) does not induce expression of wg in these wings (Fig. 6I, data not shown). In Psn mutant discs, where Su(H) is activated, vg expression is induced in the wing area, a trait that is not observed in Psn mutant discs (Fig. 5J; Ye et al., 1999).

In agreement with Ye et al. (1999), we also find that expression of Nintra with dpp-Gal4 leads to a rescue of the wing pouch and margin in Psn mutant wing imaginal discs, but expression of UAS-Dl does not (Fig. 5H, data not shown).

A similar rescue is observed when Su(H) is expressed in Ser mutant wing discs where Notch signalling is initially weaker and decays during early wing development (data not shown, Klein et al., 1998b).

Although these experiments support the conclusion that Su(H) is able to activate gene expression in the absence of effective activation or processing of Notch, it is still possible that small amounts of Nintra might be present in the different mutants due to a spontaneous cleavage of Notch. These minute amounts could be sufficient to activate gene expression in association with abundant Su(H) and elicit the observed effects in the mutant situations. To rule out this possibility, we induced Notch mutant clones in the domain where UAS-Su(H) is expressed and monitored the expression of the vgBE (Fig. 6). The vgBE contain a functional Su(H)-binding site and represents a direct target of Notch/Su(H) signalling in the wing pouch (Kim et al., 1996). We find that Su(H) is able to ectopically activate the expression of the vgBE and induce strong proliferation in the absence of Notch as in wild-type cells (Fig. 6A-F). Antibody stainings with an antibody directed against the intracellular domain of Notch (antibody C17.9C6, Fehon et al., 1990) show that the clones lack Notch protein (data not shown). As a control, we induced Notch mutant clones in wild-type wing discs, which confirmed the requirement for Notch in the activity of the vgBE (data not shown).

In summary, these results clearly show that Su(H) is able to activate gene expression in the absence of Notch function and without the intracellular domain of Notch.

Since Su(H) is expressed ubiquitously and continuously (Gho et al., 1996), the fact that Su(H) can promote transcription without the presence of Nintra suggests that the activity of Su(H) must be suppressed in the absence of Notch activation during normal development. One possible mechanism for this is the binding of an inhibitory factor. A candidate for this function is the Hairless (H) protein, which has been shown to interact with Su(H) and antagonize its DNA-binding activity (Brou et al., 1994). Furthermore, several reports show that H antagonizes Notch signalling during adult PNS development (Bang et al., 1991; Bang and Posakony, 1992; reviewed in Posakony, 1994; Bang et al., 1995). To test whether H is an antagonist of Su(H) also during wing development, we first induced H mutant clones in the wing pouch and asked wether genes dependent only on Su(H) activity are expressed in these clones. If H regulates the activity of Su(H), the removal of H might lead to the activation of Su(H) and result in the expression of its targets, e.g. the vgBE and Ser. We find that both are ectopically activated in H mutant clones (Fig. 7A-G). The ectopic expression of the vgBE in the clones varies and is strongest near the DV boundary (Fig. 7B,C). This graded expression is possibly due to the requirement of a diffusible factor coming from the DV boundary. One candidate for this is Wg, which seems to be required for the proper expression of the vgBE (Klein and Martinez Arias, 1999; Zhang and Carthew, 1998). The cells in the H mutant clones do not express cut or wg, which are dependent on the presence of Nintra (Fig. 7H,I), suggesting that Notch is not activated in these clones. The loss of H function seems to elicit Su(H)-dependent target gene expression in the wing pouch, a region probably devoid of Notch activity. This suggests that the inactivation of H is sufficient to activate Su(H). To test further this conclusion, we looked at whether the activity of the vgBE is maintained in H mutant wing pouches if Notch is concomitantly removed. For this, we induced Notch mutant clones in H mutant wing discs (Fig. 8). In H mutant wing pouches, weak ubiquitous expression of the vgBE is observed throughout the whole area of the wing (Fig. 8A), confirming the clonal analysis described above. vgBE is also active in several Notch mutant clones near the DV and anteroposterior (AP) boundary (Fig. 8A-F), but the activity is not maintained in all clones. One explanation for this might be again the requirement of other so far unidentified factors enamating from the two compartment boundaries. In agreement with this, the vgBE enhancer has a late expression domain along the AP boundary, suggesting an input from these areas for its proper expression. However, as seen in Fig. 8G-I, this domain is also dependent on Notch during normal development. The removal of the Su(H)-binding site in the enhancer leads to the loss of all expression domains in the wing pouch, suggesting that Su(H) is required (Kim et al., 1996). Therefore, the fact that the cells of several mutant clones do express the vgBE suggest that the vgBE can be activated in the complete absence of Notch activity and that the inactivation of H is sufficient to activate Su(H). In agreement with Neumann and Cohen (1996), we never found any activation of the vgBE in Notch mutant clones induced in wild-type wing pouches (see Fig. 8G-I; insert in Fig. 8I), suggesting that during wild-type development, the activity of Notch is required to activate the vgBE. Hence, Notch probably activates Su(H) through inactivation of H. We next tested whether the degree of endogenous Su(H) activation that results from the removal of H is sufficient to elicit a biological effect. To assay this, we asked whether or not removal of H activity can induce Su(H)-dependent development of the pouch in wing discs in which Notch signalling is absent, such as ap and Psn mutant wing discs.

We find that loss of H function rescues the loss of wing development of ap mutants (Figs 9A-H, 11): whereas ap mutants have no wing pouch (Ng et al., 1996; Klein and Martinez Arias, 1998b, Fig. 11A), ap,H double mutants have large wing pouches with no margin structures (Figs 9D, 10D). The enlarged pouch of the double mutant discs expresses spalt (sal) and the two vg reporters, vgQE and vgBE, all of which are expressed specifically in the wing pouch in a Notch/ Su(H)-dependent manner and are not expressed in ap mutants (Fig. 9A-D,H). In contrast, no wg expression is induced in these double mutant discs (Fig. 9D), suggesting that the observed rescue is likely to be due to the activation of Su(H) in the double mutants. This is strongly supported by the fact that Su(H),H double mutants exhibit a small wing rudiment identical to that of Su(H) mutants (compare Fig. 10I with Fig. 4A). Expression of UAS-vg by dpp-Gal4 in ap mutant discs can recover the pouch-specific expression domain of sal (Fig. 9J, Klein et al., 1998a), suggesting that the activation of vg expression by Su(H) is responsible for the recovered sal expression in the ap,H double mutant wing discs. Similar to overexpression of UAS-Su(H) in ap mutant wing discs, the pouch in ap,H mutant discs develops near the residual wg expression in the remaining hinge (compare Figs 5E and 9C, arrowheads). As expected from the analysis of the wing discs, the pharate adult ap; H double mutants have large wing pouches, which are devoid of any margin like structure such as innervated bristles (Fig. 10D).

We further examined the effects of removing H on wing development in Psn mutants. As in the case of ap, loss of function of H effects a strong rescue of the wing pouch in the Psn,H mutant discs in comparison to the Psn mutant discs (Figs 5G, 9K,L). However, in this case, the morphology of the discs is more like wild type (Fig. 9L) and, in contrast to ap,H mutant discs, the pouch develops at its normal place (arrowhead in Fig. 9L). Closer monitoring of double mutant discs reveals some expression of wg and the vgBE along the DV boundary (Fig. 9K,L). This suggests that, in contrast to the situation of ap mutants, in Psn mutants, there is some activation of Notch and it seems that the lack of H activity can enhance this residual signalling of Notch at the DV boundary. This is remarkable considering that the wing phenotype caused by the loss of Psn is stronger than that caused by loss of Su(H) function (compare Fig. 4A and 5G).

Taken together, our results provide further evidence for a positive transcriptional activity of Su(H). They further show that H is an antagonist of Su(H) during early wing development and that it suppresses the activity of Su(H) in the absence of Notch signalling. The results also suggest that the inactivation of H is sufficient to activate Su(H) and that the activity of Notch is required to inactivate H during normal development.

Here we have assessed the role of Su(H) during early wing development and shown that it is required for all aspects of Notch signalling in this process. This is important to know in the light of an increasing number of reports of the existence of an Su(H)-independent Notch signalling event (Nofziger et al., 1999; Matsuno et al., 1997; Wang et al., 1997; Brennan et al., 1999; Zecchini et al., 1999). Our results show that this pathway does not operate during early development of the wing. We have observed that Su(H) can activate expression of some Notch target genes in the absence of Notch signalling and Nintra, and that the activity of Su(H) is antagonized by H during wing development as during lateral inhibition. Furthermore, we have shown that expression of Su(H) suppresses some of the Notch target genes in absence of Notch activation, providing evidence that Su(H), like its mammalian counterpart, can act as a repressor of transcription.

Our results shed a new light on some aspects of Notch signalling, which we discuss below.

Our observation that Su(H) can activate transcription of target genes in the absence of Notch suggests the existence of a mechanism that impedes this activity in vivo. Analysis of the interactions between H and Su(H) suggests that H is part of this mechanism. We observe that clones of H mutant cells express genes that can be activated by Su(H) without Notch activity and that the loss of H activity in ap mutant discs leads to a formation of a wing pouch, in a Su(H)-dependent manner. The activation of some Notch target genes in H clones is important, since the clonal analysis presented here shows that it occurs even in the absence of Notch activity. Therefore, the loss of H activity seems to be sufficient to activate Su(H). Our clonal analysis shows that removal of H leads to the activation of Su(H)-dependent gene expression in the absence of Notch. This indicates that the inactivation of H leads to the activation of Su(H). Furthermore, our analysis reveals that the activation of the vgBE requires the activity of Notch during wild-type development (Neumann and Cohen, 1996; see above), suggesting that Notch activation causes the inactivation of H and that this inactivation result in the activation of Su(H) (Fig. 11). The behavior of H mutant cells is not compatible with a simple function of H in defining a threshold for Notch signalling activity as has been proposed (Bang et al., 1995). If H only defines a simple threshold, one would expect that its removal in regions devoid of Notch activity would not lead to activation of Notch target genes. H can associate with Su(H) and this association prevents Su(H) from binding to its DNA target site in vitro experiments (Brou et al., 1995). This suggests that H might inactivate Su(H) through a physical association in vivo and that the activation of Su(H) occurs through its release from this inhibitory complex. The above discussed results further strongly suggest that this release might be mediated by Nintra, since it can activate the expression of all Notch-dependent genes. How Nintra achieves this is at present not clear and will require further biochemical experiments. Two observations, however, support our conclusion. First, it has been reported that Nintra can physically associate with H in vitro (Wang et al., 1997). Secondly, we have shown here that, in Psn mutants, loss of function of H, as well as expression of Nintra but not overexpression of Su(H), leads to the expression of wg, a target of Notch signalling that requires Nintra and Su(H). This suggests that the small amounts of Nintra that exist in a Psn mutant cannot interact with Su(H), even when there is abundant Su(H). In contrast, elimination of H can stimulate the not so abundant endogenous Su(H) to interact with these small amounts. Both results are in agreement with the suggested mode of Su(H) activation by Notch signalling through inactivation of H by Nintra and raise the possibility that Nintra inactivates H by direct physical interactions.

Our results show that overexpression of Su(H) leads to three different responses: (1) activation, as is the case for vg, some E(spl) genes, Dl and Ser; (2) inactivation, as shown for cut and E(spl)m8; or (3) no effect, as is the case for wg. This differential behavior is, at least in some cases, a consequence of direct binding of Su(H) to the promoters: The vgBE as well as the E(spl) genes contain Su(H)-binding sites to which Su(H) binds in vitro experiments and which are necessary for their activation in vivo (Bailey and Posakony, 1995; Lecortouis and Schweisguth, 1995; Kim et al., 1996; Nellesen et al., 1999). Despite that, they react differently towards Su(H) overexpression. Since E(spl)m8, which is suppressed by Su(H) overexpression, can be activated by expression of Su(H)VP16 or Nintra, we conclude that Nintra is required in addition to Su(H) to activate E(spl)m8 expression. Our results suggest that, in this case, Nintra probably acts as an activation domain of a dimeric transcription factor containing Su(H), as has been proposed (Jarriault et al., 1995; Hsieh et al., 1996; Wettstein et al., 1997; Lecourtois and Schweisguth, 1998; Kidd et al., 1998; Struhl and Adachi, 1998; Schroeter et al., 1998). From this, it follows that Nintra might have two function during a Notch signalling event: first it inactivates H, which leads to the release of Su(H) and then, in some instances, it provides the transactivation domain for free Su(H) to activate the expression of target genes (Fig. 11).

Flies carrying reporter lacZ constructs with up to 12 Su(H)-binding sites do not display any activity in the wing disc (Goet al., 1998). This suggests that Su(H) (even in association with Nintra) is not sufficient to activate transcription and requires other collaborating factors. It further suggests that, even in promoters that can be activated by Su(H) in the absence of Nintra, Su(H) probably interacts with other factors to promote gene expression. This is confirmed by a study of the vgBE. Although the Su(H)-binding site is absolutely necessary for its activity, other sites are equally important (Kim et al., 1997; Klein and Martinez Arias, 1999). So far the factors that bind to these sites are not identified. The dependence of Su(H) on these others factors is probably the reason for the differential expression of Notch target genes in H and H/N mutant clones that we have observed.

Recently it has been shown that Su(H) acts as a suppressor of sim-transcription during the formation of the midline cells in the embryonic central nervous system of Drosophila (Morel and Schweisguth, 2000). This observation provided the first evidence that Su(H), like its mammalian counterpart CBF1, can act as a suppressor of transcription. The inactivation of the cut and E(spl)m8 expression in absence of Nintra suggests that Su(H) can act as a suppressor of gene expression also during adult development and provides further evidence for a suppressing activity of Su(H). However, we show here that this suppression is context dependent and not a general feature of Su(H). This context dependency might also exist for CBF1, since only the reaction of a small number of genes towards its activity has been tested so far and it is possible that some target genes can be activated by CBF1 in the absence of Nintra in a similar way, as we have here shown for Su(H). In summary, these results suggest that the consequence of the binding of Su(H) to a promoter is dependent on its local architecture and, therefore, Su(H) can at the same time activate and suppress gene expression, like many other transcription factors, such as the mammalian WT1 gene (see e.g. Little et al., 1999).

The removal of both the maternal and zygotic expression of H during embryogenesis seems to have no consequence for the embryo (Schweisguth and Lecourtois, 1998). Since the overactivation of Notch/Su(H) signalling during embryogenesis has deleterious consequences (Struhl et al., 1993, Rebay et al., 1993; Lieber et al., 1993), this observation contradicts our conclusion that H is required to inactivate Su(H). However, the context dependency and differential reaction of the target genes observed during wing development offer several explanations for this discrepancy, without having to postulate an unknown factor, which can functionally replace H. First, it is likely that the interacting factors, which are required for gene expression in concert with Su(H), are different during embryogenesis and this could modulate the responsiveness of the target promoters. This conclusion is supported by the observation that the genes of E(spl)C, although probably all require Su(H) for their expression, are all very similar expressed in the embryo, but their expression pattern in the wing imaginal disc is very different (Campos-Ortega, 1993; deCelis et al., 1996a; Nellesen et al., 1999). Another explanation is that the target promoters of Su(H) during embryogenesis might be all of the type, which require the additional activity of Nintra. Therefore they would stay inactive even in the presence of free Su(H) until Notch is activated.

Thanks to A. Lüschnig, S. Cohen, S. Carroll, J. deCelis, E. Knust, T. Lieber, M. Young, S. Bray, F. Schweisguth, K. Tamura, M. Fortini, N. Perrimon and J. A. Campos-Ortega for supplying material crucial for this work. We would further like to thank F. Schweisguth for his critical comments on the manuscript. This work was supported by the Wellcome Trust (T. K. and A. M. A.), The Deutsche Forschungs-Gemeinschaft (T. K.) and Ligue Nationale contre le Cancer, Association pour la Recherche contre le Cancer (ARC) (L. S. and M. H.).