Ectopic expression of individual E(spl) genes has differential effects on different cell fate decisions and underscores the biphasic requirement for Notch activity in wing margin establishment in Drosophila

Notch signalling plays an important role in a large number of cell fate decisions in all metazoans (Kimble and Simpson, 1997). Activation of the Notch receptor by an extracellular ligand sets off an intracellular response that culminates in the expression of a number of target genes. Transcriptional activation in response to Notch is usually dependent on the Suppressor of Hairless [Su(H)] transcription factor, which seems to be directly activated by Notch (Bailey and Posakony, 1995; Lecourtois and Schweisguth, 1995; Schroeter et al., 1998; Struhl and Adachi, 1998). The most frequent target genes of Notch in Drosophila are seven Enhancer of split [E(spl)] genes, E(spl)m8, E(spl)m7, E(spl)m5, E(spl)m3, E(spl)mβ, E(spl)mγ and E(spl)mδ, which encode closely related proteins of the basic-helix-loop-helix family of transcription factors (Delidakis and Artavanis-Tsakonas, 1992; Jennings et al., 1994). E(spl)bHLH proteins act as transcriptional repressors in a complex with the co-repressor protein Groucho (Fisher and Caudy, 1998). Genetic and molecular analyses have so far detected little specificity amongst the seven E(spl)^* proteins, proposing instead that they act in a redundant fashion (Delidakis et al., 1991; Schrons et al., 1992). Yet, considering the number of Notch-dependent processes and the context dependence of the outcome of Notch signalling, it is reasonable to hypothesize that a different subset of nuclear effectors, including perhaps a different subset of E(spl) proteins, might be acting in different instances. To date, suggestions of functionally distinct roles of the E(spl)bHLH proteins have come from their selective dimerization ability with other bHLH proteins and their distinct, albeit overlapping, patterns of expression in imaginal development (Alifragis et al., 1997; de Celis et al., 1996).

In an attempt to discern possible functional specialization among E(spl) proteins, we have chosen to focus on the wing disk, where E(spl) genes display distinct expression domains. During the third larval instar (L3) and pupal periods, Notch activity is required for the specification and development of sensory organs in both notum and wing pouch domains of the wing disk, as well as in wing vein formation (de Celis et al., 1997; Huppert et al., 1997; Muskavitch, 1994). In both of these processes, Notch executes inhibitory signalling, diverting disk epithelial cells away from the neural and vein fates, respectively. Earlier in development, the basic coordinates of the wing disk are laid down by signalling from two organizing centres at the anteroposterior (AP) and dorsoventral (DV) boundaries, mediated by the secreted morphogens Decapentaplegic (Dpp) and Wingless (Wg), respectively (reviewed in Neumann and Cohen, 1997b). Establishment of the DV boundary takes place during L2 in a process that depends on Notch signalling. Activation of Notch at the future DV boundary directs the localized expression of vestigial (vg), wg and members of the E(spl) family (Diaz-Benjumea and Cohen, 1995; Jennings et al., 1995; Kim et al., 1996). E(spl) and vg, which encodes an unrelated nuclear protein, are direct targets of Notch activation via Su(H); the molecular mechanism of wg transcriptional activation is still unclear. Signalling by Notch and Wg subsequently interact in a complex process, which eventually leads to cut expression in DV boundary cells during late L3 (Micchelli et al., 1997; Neumann and Cohen, 1996). The combination of Wg and Dpp signalling is responsible for initiating a later round of vg expression from a ‘quadrant’ enhancer, which drives expression throughout the wing pouch except for the DV boundary (Kim et al., 1996, 1997; Neumann and Cohen, 1997a).

Among these Notch-dependent processes that take place in the developing wing disk, the role of E(spl) proteins is best understood in sensory organ precursor (SOP) selection. Neural precursor fate in the Drosophila wing depends on the expression of the proneural class of activator bHLH proteins encoded by the genes of the achaete-scute Complex (ASC) (Modolell, 1997). E(spl)bHLH proteins have been shown to directly repress the achaete gene, as well as to interact with proneural proteins making it likely that E(spl) downregulates proneural activity at many levels (Alifragis et al., 1997; de Celis et al., 1996; Giebel and Campos-Ortega, 1997; Jiménez and Ish-Horowicz, 1997; Nakao and Campos-Ortega, 1996). Of the seven members of the family, E(spl)m8 and E(spl)m7 are expressed in most proneural clusters of the wing disk, whereas E(spl)mγ and E(spl)mδ are only expressed in a subset (de Celis et al., 1996). A similar repressive action of E(spl) proteins probably accounts for Notch’s vein suppression effect, where the candidate target gene is veinlet/rhomboid (de Celis et al., 1997). E(spl)mβ is strongly expressed in intervein regions during L3 and flanks vein domains, where ve is expressed. The role of E(spl) genes at the DV boundary is less clear. Four genes are expressed there, namely E(spl)m8, E(spl)m7, E(spl)mβ and E(spl)mγ, yet they appear to have little or no function, as clones mutant for the whole E(spl) locus do not display the dramatic wing scalloping caused by mutant clones for N or Su(H) (de Celis et al., 1996). Expression of the remaining two E(spl) genes, E(spl)m5 and E(spl)m3, is not detectable in wing disks.

By ectopically expressing individual E(spl) genes in various domains of the wing disk, we have sought to identify differential effects in different processes: SOP selection, vein determination and wing patterning. We have indeed noted differential susceptibility of SOPs and veins to expression of E(spl) genes, which correlates with the wild-type expression pattern of the genes in question. Furthermore, we have documented repression of the vg quadrant enhancer by a subset of the E(spl) proteins, pointing towards a possible role of these proteins at the DV boundary. Finally, we have detected a positive role of some E(spl) proteins in the later phase of DV boundary formation, namely during expression of cut. Our results are consistent with two distinct episodes of Notch signalling occurring sequentially at the DV boundary.

Notch alleles (N^54l9, nd³ and N^ts1) and wg-lacZ (wg^en11) are described in FlyBase (http://flybase.harvard.edu:7081/). vg-lacZ enhancers (boundary and quadrant) are described in Kim et al. (1996). m8-lacZ is described in Lecourtois and Schweisguth (1995). The FRT chromosomes bearing P[πMyc], gro^E48, Df(3R)gro^b32.2 and N^54l9 are described in Xu and Rubin (1993), Heitzler et al. (1996) and Baker and Yu (1997). The GAL4 lines used were 32B, ptc^G559.1, ap^md544, omb-GAL4, h^1J3 and 455.2 (all described in FlyBase). We generated all UAS lines, except for UAS-m8 (Nakao and Campos-Ortega, 1996).

Cloning into the pUAST vector (Brand and Perrimon, 1993) and generation of UAS transformant lines for the E(spl)mβ and E(spl)mδ genes is described in de Celis et al. (1996). The same procedure was followed to generate the UAS-m5, UAS-m7, UAS-m3 and UAS-mγ transformant lines. Briefly, PCR fragments consisting of the coding regions only of the respective genes (sequence confirmed) were placed in a modified pUAST vector downstream of a synthetic oligonucleotide bearing an optimized translation start site. 5′ and 3′ UTRs are provided from the vector. Construction details are available on request. Transformants were obtained in a yw^67c23 background.

Crosses for studying the effects of UAS-E(spl) in a N^ts1 background were as follows:

For third chromosome lines, e.g. UAS-mδ:

N^ts1 × 32B-GAL4 UAS-mδ / TM6B

For second chromosome lines, e.g. UAS-mβ:

N^ts1 × UAS-mβ ; 32B-GAL4 / T(2;3) SM5, TM6B

For m8-lacZ:

N^ts1/ FM7c; 32B-GAL4 UAS-mδ / TM6B × m8-lacZ

Crosses were kept at 18°C and changed to new vials daily. Vials were placed at 29°C or 30°C at different developmental stages and for different durations as described in the Results. Developmental time is given as hours after egg laying (AEL); for consistency, we use the equivalent time at 25°C – development at 18°C takes approximately twice as long. For monitoring the expression of cut, wg, Dll and m8-lacZ, we compared side by side Tb⁺ (N^ts1/Y; 32B-GAL4 UAS-mδ/+) male larvae and control Tb (N^ts1/Y; TM6B/+) males from the same vial.

Clones were induced by heat shocking larvae (1 hour, 38°C) 48-96 hours AEL of the following genotypes:

y w^a N^54l9 FRT18A / P[πMyc]5A10D FRT18A; hsFLP/ 32B-GAL4

UAS-mδ

hsFLP/+; FRT82B P[πMyc] 87E97E/ FRT82B kar² ry⁵⁰⁶

P[gro⁺ry⁺] Df(3R)gro^b32.2

hsFLP/+; FRT82B P[πMyc] 87E97E/ FRT82B kar² ry⁵⁰⁶ gro^E48

For the Myc-marked clones, larvae were picked as wandering third instar, heat shocked again for 90 minutes (38°C) to induce πMyc expression and then allowed to recover for 90 minutes before dissection.

In situ hybridization with digoxigenin-labeled wg DNA and histochemical detection of β-galactosidase were done as described (Cubas et al., 1991). For immunocytochemistry, larvae were dissected in phosphate buffer and fixed as described in Xu and Rubin (1993). Antibody dilutions were as follows: Anti-Cut monoclonal antibody (kindly provided by Karen Blochlinger) 1/100; anti-Dll monoclonal antibody (kindly provided by Stephen Cohen) 1/1000; anti-N monoclonal antibody C17.9C6 (kindly provided by Spyros Artavanis-Tsakonas) 1/1000; rabbit anti-c-Myc (Santa Cruz Biotechnology) 1/1000. Horseradish-peroxidase-coupled secondary antibodies were from Jackson Immunological Laboratories (used at 1/250); diaminobenzidine was used for development. For mitotic clones, fixed tissues were reacted first with the anti-Cut antibody, developed with DAB+NiCl / CoCl₂ (black product) and subsequently with anti-cMyc or anti-Notch, developed with DAB alone (brown product).

Wings were mounted in Aquamount mountant modified (BDH Laboratory Supplies). Nota were mounted in Hoyers (Wieschaus and Nüsslein-Volhard, 1986) and incubated overnight at 60°C.

In order to obtain in vivo data on potential specific functions of E(spl) proteins, we ectopically expressed each one in mesothoracic imaginal disks using the GAL4-UAS system (Brand and Perrimon, 1993). The most prominent phenotypes observed were loss of sensory organs and loss of wing veins, in accordance with the gain-of-function phenotypes previously reported for E(spl)m8, E(spl)m5 and E(spl)mβ (de Celis et al., 1996; Nakao and Campos-Ortega, 1996; Tata and Hartley, 1995). Qualitatively, all seven E(spl) genes produced these same overexpression phenotypes; however, the severity of these phenotypes depended on the individual E(spl) protein. E(spl)m5 stood out among the seven members as the most inactive protein, as it produced mild gain-of-function phenotypes only when two copies of the UAS transgene were used (six lines tested). We therefore concentrated our comparative analysis on the remaining six UAS-E(spl) bHLH transgenes.

We initially focussed our attention on the specification of wing veins and the specification of three types of external sensory organs: notum macrochaetae, notum microchaetae and anterior wing margin bristles. Between three and seven lines of each UAS transgene were tested for adult phenotypes and only quantitative variations were observed. More impressive were the phenotypic differences seen between different transgenes, representative examples of which are shown in Figs 1 and 2. E(spl)mβ, which is normally expressed in intervein regions (de Celis et al., 1997), produced the most dramatic loss of vein (Fig. 1D) when driven by 32B-GAL4, which expresses uniformly in the wing pouch (Fig. 1O). UAS-E(spl)mγ had an equally severe effect on veins, while the remaining four had milder effects (Fig. 1A-F). Loss of notum microchaetae and wing margin bristles was also seen in some 32B-E(spl) combinations, strongest with E(spl)m8 and E(spl)m7 (Fig. 1G-N). Interestingly, whereas E(spl)m8 is more effective in abolishing the notum microchaeta fate, E(spl)m7 is most active against wing margin bristles. We observed macrochaeta loss using the h^1J3-GAL4 line (Fig. 2A-C). The phenotypes ranged from more than half of the macrochaetae deleted (by UAS-m8) to rare loss of one scutellar macrochaeta only (by UAS-mγ and -m3). A summary of the relative efficiency with which the different E(spl) genes affect each process studied is shown in Table 1. Some of these effects are evident from Figs 1 and 2, while others were obtained by comparing flies raised at higher temperatures, where the GAL4 system’s expressivity increases, and/or flies bearing two copies of the responder transgene (not shown).

The differences in E(spl) activity described in Table 1 were observed with GAL4 lines that exhibit moderate levels of expression. When stronger expressing GAL4 lines were used, the differences were lessened. For example, with ap-GAL4 (Fig. 2D-I) most E(spl) transgenes gave complete loss of macrochaetae and microchaetae, except for the very anterior edge of the notum; the sole exception (besides E(spl)m5) was E(spl)mδ, which allowed a few central microchaetae to form (Fig. 2F). With the scutellum-specific 455.2-GAL4, all UAS-E(spl) (except E(spl)m5) resulted in severe loss of scutellar bristles (not shown). This observation argues in favour of functional interchangeability of the six proteins, but does not detract from the fact that at levels closer to physiological each protein has varying efficacy according to developmental context. We have excluded the possibility that E(spl) genes displaying low activity might be an artifact of not having recovered a strongly enough expressing transformed line: each transgene shows dramatic effects in some process, in which other transgenes are less active (see Table 1); in the case of E(spl)mδ, which appears less active than other proteins in Table 1, we have shown that it affects R8 cell specification in the developing eye much more strongly than any of the other transgenes (Ligoxygakis et al., 1998).

We extended our analysis of the specificity of UAS-E(spl) transgenes using a number of molecular markers pertaining to wing patterning. Expression was driven in a domain centered around the AP boundary of the wing pouch by the omb-GAL4 driver. Neither wg-lacZ (not shown) nor the boundary enhancer of vg (Fig. 2L) were affected by ectopic expression of E(spl) genes. In contrast, the vestigial (vg) quadrant enhancer was strongly repressed by E(spl)m3 and E(spl)mγ, but was unaffected by the other E(spl) transgenes (Fig. 2J,K). In addition to revealing the activity of our UAS-m3 transgene, which was less active in other contexts, this finding suggests a role for the normal E(spl)mγ expression at the DV boundary, namely to repress vg(quad) expression. This is in agreement with the published observation that this enhancer contains a potential DNA binding site for Hairy/E(spl) proteins (Kim et al., 1996) and could account for the effects of N alleles on this enhancer reported by Go et al. (1998).

Although ectopic expression of E(spl) genes phenocopied activation of the Notch pathway in abolishing SOPs and veins, it did not generate ectopic wing margin in the same way as was observed with expression of Notch ligands or activated Notch itself (e.g. Diaz-Benjumea and Cohen, 1995). We were able to detect effects of E(spl) on the wing margin by studying their ectopic expression in genetic backgrounds with reduced Notch function. Flies heterozygous for a N null allele or hemizygous for the hypomorph nd³ have distally nicked wings, as well as thickened wing veins, in accordance with the need for Notch in establishment of the DV organizer and restriction of vein fate. When E(spl) transgenes were expressed in these backgrounds under 32B-GAL4, we observed an unexpected effect. E(spl)mδ and E(spl)mγ restored the missing distal margin and adjacent wing blade tissue, while E(spl)m8 and E(spl)m7 enhanced wing notching (Fig. 3). E(spl)mβ had a weaker effect, being able to rescue only nd³ (not shown); finally, E(spl)m5 had no effect and E(spl)m3 was not tested due to pupal lethality with the 32B-GAL4 line. The transgenes that showed strong vein suppression effects in wild-type backgrounds still displayed this phenotype (Fig. 3B-E), presumably by overcoming the reduction in Notch activity.

We then tested the 32B-E(spl) combinations in a more severe N loss-of-function background, by using the temperature-sensitive allele N^ts1. First, a series of temperature shifts was performed to define Notch requirements during different time periods. Flies were reared at 18°C and shifted to the non-permissive temperature (29°C) for 14 hours at various times AEL (see Materials and Methods). A broad temperature-sensitive period was observed for wing notching spanning from 60-108 hours AEL, i.e. from roughly mid-second instar to late third instar, in agreement with Shellenbarger and Mohler (1978). We compared the wing phenotypes of N^ts1/Y flies in the presence or absence of ectopically expressed E(spl) after shifting to the non-permissive temperature for 14 hours starting at early (72-84 hours AEL) or mid- (96-108 hours AEL) third instar larval stage. Early shifts of control N^ts1 flies resulted in long paddle-shaped wings with extensive anterior and posterior scalloping (Fig. 4A). Late shifts produced wing nicking around the margin, with less extensive loss of adjacent tissue (Fig. 4B). In the presence of 32B-GAL4 and UAS-mδ or UAS-mγ, the late upshift regimen gave wings with no or slight margin nicking (Fig. 4D), similar to the rescue observed for N/+ or nd³/Y genotypes, while UAS-mβ, UAS-m7 or UAS-m8 had no effect. Surprisingly, in the early shifts all five E(spl) transgenes gave enhancement of the N^ts1 phenotype resulting in miniscule wings (Fig. 4C) that resemble the vestigial wing phenotype.

These data suggest that the broad temporal requirement observed for Notch during the third larval instar reflects at least two distinct phases of Notch activity in the process of wing margin specification with opposite response to overexpression of E(spl)mδ or E(spl)mγ. The effects of the various transgenes on wing notching are summarized in Table 1.

Loss of function of the genes wg and cut results in wing scalloping phenotypes, such as those that we observed in our N^ts1 experiment (Blochlinger et al., 1993; Couso et al., 1994). Because Notch signalling at the DV boundary is needed for the expression of wg and cut, we sought to determine whether the wing margin rescue produced by E(spl)mδ and E(spl)mγ might reflect restoration of expression of either of these genes.

We assayed wg RNA expression in wing imaginal disks from N^ts1 animals in the presence or absence of 32B-GAL4, UAS-E(spl)mδ. Larvae were shifted to the restrictive temperature (30°C) at 96-108 hours (mid-late L3) for 16 or 22 hours and then immediately dissected. As reported by Diaz-Benjumea and Cohen (1995), these regimens resulted in specific wg loss from the DV boundary, but not from other regions of the wing disk. No difference in wg expression was observed between N^ts1 control disks and N^ts1 disks with ectopically expressed E(spl)mδ (not shown). To conclusively prove that no Wg activity is induced by ectopic E(spl), we studied expression of the Wg downstream gene Dll (Neumann and Cohen, 1997a) using immunocytochemistry as a more sensitive assay. Although Dll expression was not completely abolished after the 22 hours temperature shift, it was significantly reduced and no increase was seen with ectopic expression of E(spl)mδ (Fig. 5A-C). These results suggest that rescue of wing margin loss by E(spl)mδ ectopic expression is not due to restoration of wg expression at the DV boundary. We further confirmed that ectopic expression of E(spl)mδ did not restore expression of Notch-responsive genes, as the m8-lacZ boundary stripe was lost after heat inactivation of N^ts1 and was not restored by concomitant expression of E(spl)mδ (Fig. 5D-F).

We used similar temperature-shift regimens to assay cut expression in N^ts1 flies with or without 32B-GAL4, UAS-mδ or UAS-mγ. When we shifted up at approximately 96 hours AEL, Cut protein accumulation in N^ts1 disks was abolished after 22 hours (Fig. 5G). Reduction in Notch activity affected the DV stripe specifically and not the expression of Cut in adepithelial cells of the notum. In contrast, wing disks from N^ts1 flies ectopically expressing UAS-mδ or UAS-mγ displayed an almost wild-type Cut stripe (Fig. 5H,I). In another experiment, the upshift was performed at approximately 72 hours AEL and the flies were maintained at the restrictive temperature until the late third instar (wandering stage), i.e. about 48 hours. In this case, both control and E(spl)-expressing animals lacked the boundary Cut stripe (Fig. 5J,K). To address whether the difference observed was due to the extended duration of the heat pulse, we upshifted larvae at ∼72 hours AEL and kept them at the restrictive temperature only for 24 hours. After a further 2 days at the permissive temperature (i.e. ∼one day equivalent at 25°C), they reached the wandering stage and were dissected. Again no margin Cut expression was observed regardless of whether they expressed ectopic E(spl) or not (Fig. 5L).

We therefore conclude that expression of E(spl)mγ or E(spl)mδ can promote or maintain cut expression in the absence of Notch activity during late L3. Yet if Notch activity is compromised during early L3, E(spl)mγ or E(spl)mδ cannot show this effect, suggesting two sequential Notch-dependent events in the activation of cut. In agreement with this hypothesis, N null mitotic clones induced during late L2 produced autonomous loss of Cut even in the background of 32B-GAL4, UAS-E(spl)mδ (Fig. 6D). It is worth noting that, although in the 96 hour (late) N^ts1 upshift the shape of the disks was normal, the disks shifted at 72 hours (early) had severely atrophic wing pouches regardless of the duration of the heat pulse (Fig. 5J-L). It thus appears that early requirements for N in the proliferation of the wing pouch, just as in DV boundary-specific gene expression, are not rescued by ectopic expression of E(spl)mδ or E(spl)mγ.

To test the hypothesis that E(spl) might be involved in cut expression, we analysed mitotic clones null for the E(spl)-C, using the deficiency of the locus Df(3R)gro^b32.2, which deletes all E(spl)bHLH genes but does not affect gro. Although such clones were not completely devoid of Cut protein, they frequently decreased the levels of Cut in an autonomous fashion (Fig. 6A,B). This result was unexpected given the inability of the same mutant clones to produce wing margin notching (de Celis et al., 1996); the lower levels of Cut must be sufficient for wing margin integrity. The interesting possibility therefore arises that some protein encoded by the E(spl)-C, most likely E(spl)mγ, is involved in cut expression, but acts redundantly with some other factor. If an anti-repression mechanism is active in cut expression (see Discussion), we would predict that clones null for the Gro co-repressor might also show defects in the activation of cut, since Gro appears to be a necessary co-factor for the activity of a number of repressors including E(spl) (Fisher and Caudy, 1998). We tested this by analysing clones homozygous for the null allele gro^E48. Interestingly, all clones impinging on the cut expression domain did not just decrease, but autonomously abolished cut expression (Fig. 6C). We, therefore, conclude that gro⁺ is absolutely required for cut activation and that it is recruited not only by E(spl)mγ, but also via (an)other unidentified partner(s).

E(spl) genes are the most common transcriptional targets of Notch signalling known. One or more of these seven genes is turned on in each instance of Notch signalling, yet, when viewed individually, they display distinct expression patterns, especially during postembryonic stages (Bailey and Posakony, 1995; de Celis et al., 1996; Jennings et al., 1994; Lecourtois and Schweisguth, 1995). Even though functional analysis to date has suggested that the functions of the seven proteins are interchangeable, we have herein presented data showing that individual E(spl) proteins are most effective at influencing the cell fate determination events that occur within the domains of their highest respective expression. E(spl)m7 and E(spl)m8, which are most strongly expressed in proneural regions, are most effective at suppressing sensory organ fates. In contrast, E(spl)mβ is most effective at antagonizing vein fate, conforming with its expression in intervein regions. E(spl)mγ is active at the wing DV boundary to repress the vg quadrant enhancer and to activate cut. Finally, E(spl)mδ, which is strongly expressed in the morphogenetic furrow of the eye, is best at suppressing photoreceptor R8 fate (Ligoxygakis et al., 1998). What the functional specialization of E(spl)m5 and E(spl)m3 might be still remains unclear, as our analyses so far have focussed on the wing and eye disks, where these genes are normally not expressed.

In SOP singularization as well as vein specification, the observed ectopic phenotypes of E(spl) are congruent with what would be expected by overactivation of Notch signalling, which suppresses both SOP and vein fates (Muskavitch, 1994). In the wing margin, we have detected two phases of Notch signalling: the early one is curiously antagonized by E(spl) activity, whereas the late one is assisted by E(spl)mγ and E(spl)mδ, but not by others. How can E(spl) have two opposing activities during sequential phases of wing margin specification? Our observation that some E(spl) proteins can repress vg(quad)-lacZ suggests a possible explanation. This enhancer ultimately depends on Notch signalling, as it requires prior expression of DV boundary-specific genes, like wg, that are turned on by Notch (Kim et al., 1996, 1997; Neumann and Cohen, 1997a). Perhaps inappropriate expression of E(spl) in the wing pouch enhances the tissue loss caused by heat-inactivated N^ts1, by decreasing the levels of Vg protein, which is needed for wing cell viability (Kim et al., 1996). This model is not fully consistent with the fact that we observed repression of vg(quad) only with E(spl)m3 and E(spl)mγ, whereas all E(spl) proteins tested were equally able to enhance the early N^ts1-induced tissue loss. The different genetic backgrounds of the two experiments may account for this. It is possible that in the sensitized N^ts1 background mild repression of vg(quad) by any E(spl) protein could manifest itself as enhancement of wing blade loss. Ectopic expression of E(spl) proteins is not sufficient to have the same effect in a N⁺ background (Fig. 1), thus accounting for the fact that when the N^ts1 upshift is done at a later time – after the establishment of vg expression – the same driver (32B-GAL4)-responder (UAS-mγ or UAS-mδ) combination gives wing blade rescue (Fig. 4). It should also be pointed out here that GAL4-driven expression is much more robust at 30°C; perhaps the effects of the early upshifts are due to concomitant N activity reduction and very high E(spl) accumulation at a time most sensitive for vg(quad) expression.

What the above model does not account for is the stronger effect of E(spl)m8 and E(spl)m7 on wing margin loss in the N/+ background (Fig. 3). This cannot be due to vg repression, as we have seen that these two proteins are not good repressors of either of the two known enhancers of this gene (Table 1). It appears that this effect stems from yet a different activity of E(spl) proteins. In fact, we have preliminary data (not shown) that high level UAS-m8 expression in a wild-type background results in repression of cut during late larval/ pupal stages.

Notch signalling has been shown to be necessary and sufficient for cut [as well as for vg, wg and E(spl)] expression within the wing pouch (Diaz-Benjumea and Cohen, 1995; Jennings et al., 1995; Micchelli et al., 1997; Neumann and Cohen, 1996). Whereas vg and E(spl) are direct targets of the N-responsive transcription factor Su(H) (Bailey and Posakony, 1995; Kim et al., 1996; Lecourtois and Schweisguth, 1995; Neumann and Cohen, 1996), there is no evidence to date that cut is activated by a similar mechanism. Instead, the present work suggests that cut expression requires Gro and partly also depends on E(spl)bHLH factors. One possibility is that E(spl) (at least E(spl)mγ and E(spl)mδ), like a number of other transcription factors (Fisher and Caudy, 1998), might have a dual function as either a transcriptional repressor or an activator depending on context. Such an activation role has never been suggested before for either E(spl) or Gro. Alternatively, two models can be envisaged that reconcile a repressor activity of E(spl) proteins with their role in cut activation. In one, E(spl) could act by repressing a negative regulator of cut transcription. In the other, they could repress a negative regulator of Notch signalling. In the latter case, E(spl) expression would promote a positive feedback loop to enhance Notch signalling, thus increasing the signalling output from the severely compromised N^ts1 receptor at the restrictive temperature. The fact that no restoration is observed in the expression of two other Notch targets, namely wg and E(spl)m8-lacZ (Fig. 5A-F), argues against this hypothesis. We therefore favour a direct role of E(spl)mγ and E(spl)mδ in cut expression, either as activators or as repressors of a repressor, but not as general positive regulators of Notch signalling. Detailed study of transcriptional regulators that bind the cut wing margin enhancer will provide leads to unravel this rather complex control mechanism.

Ectopic expression of E(spl)mγ/E(spl)mδ is not sufficient for cut expression in a wild-type background. Rather, it appears that the ability of ectopic E(spl)mγ/E(spl)mδ to induce cut is spatially restricted to the normal domain of cut expression (Fig. 5G-I). As activated Notch is sufficient to ectopically turn on cut (Neumann and Cohen, 1996), it follows that some other Notch-responsive event, besides E(spl) expression, must also contribute to cut expression. This is consistent with our findings that early reduction of Notch activity abolishes cut expression despite concomitant ectopic expression of E(spl)mδ (Figs 5J-L, 6D). Molecular analysis has shown that cut expression requires the transcription factor Scalloped (Sd) (Morcillo et al., 1996). sd is a candidate target gene of Vg (Williams et al., 1993), which in turn is initially activated by Notch independently of E(spl) (Kim et al., 1996). It is possible that expression of vg and sd at the wing margin during early L3 could make these cells competent for cut expression. This would only be initiated later, when a second pulse of Notch signalling during mid-L3 activates (or relieves repression of) cut via E(spl)mγ or another Gro-interacting protein.

The existence of genes with overlapping functions is quite common in mammals, yet it is rare in Drosophila, where many single gene mutations produce phenotypic consequences. Even when structurally similar genes are encountered in Drosophila, many of the individual members have evolved discrete functions as evidenced by point mutant phenotypes – e.g., engrailed and ase (Dominguez and Campuzano, 1993; Simmonds et al., 1995). Unlike these, the E(spl) locus comprises a cluster of seven apparently redundant genes (no individual loss-of-function phenotypes characterized) interspersed with structurally unrelated genes. This organization is not a quirk of D. melanogaster, but is also found in the distantly related species D. hydei (Maier et al., 1993), raising the question of its importance through evolution.

In the case of another complex locus, the achaete-scute Complex, the presence of multiple genes with (partially) overlapping functions has been proposed to have been selected for, in order to allow the use of extensive lengths of transcriptional regulatory regions that would be difficult to act on a single gene (Modolell, 1997). It is clear that besides Notch activation, each of the seven E(spl) genes requires additional distinct transcriptional regulation, as their patterns of expression are different, yet their cis-regulatory regions do not appear to be very long (Bailey and Posakony, 1995; Lecourtois and Schweisguth, 1995). Thus, it is unclear why these regulatory regions have been allocated to different coding regions instead of being used as different enhancers for the same gene. A possible hypothesis stems from our work that suggests that individual members of the E(spl) family do indeed possess some functional specialization. Perhaps the pleiotropy of Notch function, namely the number of target genes to be regulated in different instances of signalling, would be too much to be handled by one individual effector protein, and gene duplications were favoured to produce similar effectors with different specializations. It appears that having multiple E(spl) genes confers a selective advantage to the organism, but individual gene knock-outs might yield effects of such low expressivity and penetrance that they would be undetectable by the currently practiced gross phenotypic analysis. Study of the genomic organization of homologous genes in more phylogenetically distant arthropods should shed light on the evolution of this complex locus.

The authors are grateful to Karen Blochlinger, Steve Cohen, Nick Baker, Seth Blair and Spyros Artavanis-Tsakonas for providing antibodies, clones and fly stocks. Nick Baker, Kitsos Louis and Michalis Averof provided critical comments for the manuscript. We also thank Pavlos Alifragis and Rena Skoula for help with plasmid constructions and embryo injections. This work was funded by EU SCIENCE Programme grant SCI*CT92-0819 to CD and SJB and by HFSP grant RG 490/94-M to CD.