A Notch-independent function of Suppressor of Hairless during the development of the bristle sensory organ precursor cell ofDrosophila

Notch signalling plays a fundamental role during a great number of developmental processes in multicellular animals(Artavanis-Tsakonas et al.,1999; Mumm and Kopan,2000). It mediates communication between adjacent cells and is often employed for binary fate decisions. Interactions between Notch and its ligands trigger the proteolytic cleavage of the Notch receptor, releasing the intracellular domain that travels to the nucleus. In the nucleus, the intracellular domain binds to the Su(H)/CBF transcription factor to activate the expression of target genes (Barolo et al., 2002; Furriols and Bray,2001; Morel and Schweisguth,2000). During its activation, Notch is cleaved twice. The first cleavage is ligand dependent and is performed by an ADAM metalloprotease encoded in Drosophila by the kuzbanian gene (kuz)(Klein, 2002;Lieber et al., 2002;Pan and Rubin, 1997;Sotillos et al., 1997;Wen et al., 1997). Kuz cleaves Notch in the extracellular part adjacent to the transmembrane domain. The resulting fragment within the membrane is then cleaved a second time by a protease encoded by the Presenilin gene (Psn). The cleavage occurs in the transmembrane region and releases the intracellular domain into the cell (Brou et al., 2000;Lecourtois and Schweisguth,1998; Qi et al.,1999; Schroeter et al.,1998; Struhl and Adachi,1998; Struhl and Greenwald,1999).

In many developmental processes Su(H) seems to act as a repressor of the expression of target genes in the absence of the Notch signal(Barolo et al., 2002;Furriols and Bray, 2000;Morel and Schweisguth, 2000). For this `default repression' it requires the Hairless protein (H), which acts as a bridge between Su(H) and its co-repressors CtBP and Groucho(Barolo et al., 2002;Morel et al., 2001).

One intensely studied process in which the Notch pathway plays an important role is the development of the bristle sense organ of the adult peripheral nervous system (PNS) of Drosophila(Modolell and Campuzano,1998). These bristles are simple mechanosensory organs that consist of only four cells. All four cells are generated by a single precursor, referred to as the sensory organ precursor cell (SOP)(Fig. 1A). The SOP is selected from a cluster of cells that are defined by the expression patterns of the proneural genes, such as the genes of the achaete-scute complex(AS-C) (Fig. 1B,C). The activity of the proneural genes enables all cells of a cluster (proneural cluster) to develop as SOPs. The selection of the SOP in the proneural cluster occurs through a process called lateral or mutual inhibition and is mediated by the Notch signalling pathway. Lateral inhibition ensures that only a defined number of cells of a proneural cluster develop as SOP, whereas the rest switch fate and develop to epidermoblasts. During lateral inhibition, the SOP sends an inhibitory signal via the Notch ligand encoded by Delta(Dl) to its neighbours to activate the expression of the genes of theEnhancer of split-complex [E(spl)-C] in these cells(Bailey and Posakony, 1995;Hinz et al., 1994;Jennings et al., 1994;Lecourtois and Schweisguth,1995; deCelis et al.,1996; Lai et al.,2000a; Lai et al.,2000b). The activity of the genes of the E(spl)-Cantagonizes that of the proneural genes(Knust et al., 1992;Nakao and Campos-Ortega,1996). As a result the neighbouring cells switch fate and develop as epidermal precursors. The expression of the genes of the E(spl)-Cis directly activated by a transcription complex that includes Su(H) as the DNA-binding part (Bailey and Posakony,1995; Hinz et al.,1994; Jennings et al.,1994; Lecourtois and Schweisguth, 1995; deCelis et al., 1996; Lai et al.,2000a; Lai et al.,2000b).

During normal development, the SOP arises in a distinct position in each proneural cluster, indicating that the cells at these positions have a bias to develop the SOP fate (Cubas et al.,1991; Cubas and Modollel, 1992). It is thought that this slight bias is amplified during lateral inhibition through a feedback loop between the activity of Notch and the expression of Dl(Heitzler et al., 1996;Schweisguth, 1995): the more Notch is active in a given cell, the less it expresses Dl. Hence, the ability of that cell to inhibit their neighbours decreases over time. Conversely, a cell where Notch is less active expresses higher levels of Dl and can inhibit its neighbours more efficiently, and its inhibiting ability increases over time.

Once a SOP is selected through lateral inhibition, it starts to express genes such as asense (ase), neuralized(neur), senseless (sens; Ly —FlyBase) and hindsight (hnt; pub — FlyBase). The expression of these markers is important for the correct development of the SOP. In particular, sens appears to be essential for the normal development of the sensillum (Nolo et al.,2000). It encodes a zinc-finger transcription factor and is activated in the SOP by the proneural proteins Achaete (Ac) and Scute (Sc)(Nolo et al., 2000). Upon ectopic expression, Sens is able to induce supernumerary SOPs, indicating that it is sufficient to initiate the development of sensory organs(Nolo et al., 2000). At the time when sens and neur expression is initiated, the expression of the proneural genes is switched off in the SOP(Cubas et al., 1991).

The SOP subsequently divides to generate two second-order precursors, pIIa and pIIb (Hartenstein and Posakony,1990) (Fig. 1C). pIIa divides once more and generates the bristle and socket cell. pIIb divides to give rise to a pIIIb precursor cell and a glia cell(Gho et al., 1999)(Fig. 1C). The glia cell migrates away from the developing sense organ. The pIIIb divides another time to give rise to the neurone and a sheath cell(Fig. 1C).

The Notch pathway is required repeatedly during the further development of the bristle sensillum(Hartenstein and Posakony,1990) (Fig. 1C,D):It sends an inhibitory signal from the second order precursor cell pIIb to pIIa that prevents pIIa from choosing the pIIb fate. Later, the Notchpathway is again required to send a signal from the bristle to the socket cell and from the neuron to the sheath cell to prevent the receiving cells from choosing the same fates as the sending cell(Fig. 1C). Thus, loss ofNotch function results in the development of all cells of a proneural cluster into SOP cells. These SOP cells then generate an excess of neurones at the expense of the other fates (Fig. 1D). This scenario implies that loss of function mutants of all genes that are involved in the Notch pathway should display an excess of SOPs that subsequently generated an excess of neurones(Fig. 1D), a phenotype named neurogenic.

Schweisguth and Posakony (Schweisguth and Posakony, 1992) have reported that in Su(H) mutant wing imaginal discs not all proneural clusters can be detected with the SOP-specific neur^A101-lacZ marker(neur^A101). By contrast, cells of all clusters express this marker in kuz mutant wing discs(Sotillos et al., 1997). The differences in the mutant phenotype of kuz and Su(H) raise the possibility that Su(H) might have a function during SOP development that is independent from its function during Notch signalling. To test this possibility, we compared the consequences of loss-of-function mutations of genes that are involved in the Notch pathway on the development of the SOP, and the differentiation of neurones. We found that in Su(H)mutants, the development of the SOPs arrests during an early phase. This is not observed in Psn, Notch and kuz mutants, and suggests that Su(H) is required for SOP development in a Notch independent fashion. We provide evidence that this arrest is caused by the loss of the activity of the gene senseless (sens), which is crucial for SOP development. Our results suggest that Su(H) acts as a repressor of one or more members of the E(spl)-complex that in turn repress the expression of sens.

We further find that Su(H) mutant cells are unable to prevent the SOP fate in normal neighbours that are located at the position of the proneural cluster, where the SOP normally forms. It appears that cells at this position are insensitive to the Dl signal. This observation suggests that the positional information within a proneural cluster is more precise than anticipated and that the position of the SOP is strongly determined.

The following alleles were used in this work:Su(H)^Δ47 P(B)FRT40A (a null mutant)(Morel and Schweisguth, 2000),Su(H)^AR9, Su(H)^SF8(Schweisguth and Posakony,1992), Psn^C1 [null mutant described by Struhl and Greenwald (Struhl and Greenwald,1999)], Psn^I2 [strong mutation described by Ye et al. (Ye et al., 1999)],kuz¹⁴⁰⁵, kuz¹⁴⁰³ [strong mutants, see Sotillos et al. (Sotillos et al.,1997)], Df(1)N^81K FRT101 (null mutation)(Brennan et al., 1997) andneur^A101 (Huang et al., 1991). The Df(3R)E(spl)^b32.2 is described by Schrons et al. (Schrons et al.,1992).

Reporter strains were E(spl)m8-lacZ(Lecourtois and Schweisguth,1995; Nakao and Campos-Ortega,1996), E(spl)mβ CD2(de Celis et al., 1998), SOP-E(Culi and Modolell, 1998) and Gbe+Su(H) (Furriols and Bray,2001).

UAS stocks were UASsens (Nolo et al., 2000), UASSu(H)(Klein et al., 2000), UASSu(H)ΔH (Furriols and Bray,2000) and UAS GFP [a gift from S. Bahri and Yeh et al.(Yeh et al., 1995)].

Gal4 drivers wre scaGal4 (Hinz et al., 1994) and dppGal4 (a gift from S. Carroll).

Antibody staining was performed according to standard protocols. The anti Dl, anti Wg and anti Hnt antibodies were 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. The anti Sens antibody was a gift of H. Bellen(Nolo et al., 2000). The anti 22C10 and anti Elav antibody were a gift of C. Klämbt.

Fluorochrome-conjugated antibodies were purchased from Molecular Probes.

The wing imaginal disc of Drosophila gives rise to the wing proper and one half of the mesothorax (notum) of the fly. In a wing imaginal disc of the late third larval instar stage, the proneural clusters in the notum have fully developed and the first SOPs are recognizable(Fig. 1B). These SOP and their corresponding proneural clusters are arranged in a stereotyped pattern and form the large bristles, which are called machrochaete. The development of the machrochaete provides a classical model for sensillum development in insects.

To find out whether there is a difference in the phenotypes ofSu(H) and mutants of other genes involved in the Notchpathway, we have compared the development of the SOPs of the machrochaete in wing imaginal discs that were mutant for of Su(H), Notch, Psn andkuz. We used two other markers, besidesneur^A101-lacZ, that specifically label SOPs of the wing imaginal disc (sens and hnt). Expression of both genes is restricted to the SOPs in the notum (Fig. 2A,E,J) (Nolo et al.,2000; Pickup et al.,2002).

We found that in Notch and kuz mutant wing imaginal discs all proneural clusters of the notum could be detected withneur^A101-lacZ (Fig. 2A,B,D). As previously reported, the situation is different inSu(H) mutants (Schweisguth and Posakony, 1992). For example, the cells of the dorsocentral cluster (arrow in Fig. 2A-D;see Fig. 1B for the naming of the clusters) strongly express neur^A101 in kuzdiscs and Notch mutant clones(Fig. 2B,D; data not shown),whereas expression of this marker is strongly reduced or absent in the cells of the corresponding cluster in Su(H) mutants(Fig. 2C). Likewise, the proneural clusters for the ANP and PNP (arrowheads in C,D) areneur^A101 positive in Notch mutant clones and inkuz, but not in Su(H) mutant discs(Fig. 2C,D). The difference between Su(H) mutants and mutants of other genes of theNotch pathway were even clearer if we looked at the expression ofsens and hnt. All proneural clusters are detectable with these two markers in notae of Notch, Psn and kuz mutant wing imaginal discs (Fig. 2E-G,I,J-L,N highlighted by arrows). However, expression ofsens and hnt was strongly reduced or absent in the cells of the clusters of Su(H) mutant wing discs(Fig. 2H,M).

During the course of our experiments, we noticed that the loss ofPsn function causes the strongest phenotype. We found that the proneural clusters are larger than in other mutants and often fused (seeFig. 2G,L). This observation is in agreement with the results of the analysis of Psn mutants during wing development, where loss of its function also causes the strongest phenotype (Klein et al.,2000).

To look at the consequences of loss of the SOP markers for neural differentiation of the progenies of Su(H) mutant SOPs, we monitored the expression of 22C10 (futsch) and elav, which are specific for mature neurones (Fig. 3). Although all cells of Psn and Notch mutant proneural clusters express 22C10 and some in addition expresselav (Fig. 3A,B,D),the expression of these markers was not detectable in cells of Su(H)mutant clusters (Fig. 3C,E). This indicates that the loss of expression of SOP marker such as sens,hnt and neur^A101 was accompanied by a lack of neural differentiation in Su(H) mutant wing imaginal discs.

The results presented so far reveal a qualitative difference between the phenotypes of Su(H) mutants and mutants of other genes of theNotch pathway, including Notch. This suggests that Su(H) has a function during the development of the SOP of the machrochaete that is independent from the Notch pathway. To confirm this conclusion, we analysed the phenotype of Su(H); Psn double mutant proneural clusters. We found that Su(H); Psn double mutant discs display a phenotype similar to the Su(H) mutant one: they express early SOP markers such as the SOP-E, but fail to express sens or22C10 in cells of most proneural clusters(Fig. 4A-D). Furthermore,expression of UAS Su(H) in these double mutant clusters withscaGal4, re-establishes the expression of sens and22C10 (Fig. 4E). These results indicate that the mutant phenotype of Su(H) is epistatic over that of Psn. They further confirm the conclusion that Su(H) is required for the development of the SOP in a Notch-independent manner.

The lack of expression of SOP markers in proneural clusters inSu(H) mutants could be caused by the loss of the cells of a cluster or by an arrest in their early development as a SOP. To discriminate between these possibilities, we monitored the expression of genes that are expressed in proneural clusters, such as E(spl)m8, and the activity of the earliest marker for the SOP, the achaete-scute SOP enhancer (SOP-E)(Culi and Modolell, 1998), inSu(H) mutant wing imaginal discs. We further used the expression ofDl, which is, as we have found, strongly expressed in cells of proneural clusters of Psn mutants, (data not shown). All these markers were expressed in cells of proneural clusters in the notum ofSu(H) mutant wing discs (Fig. 5A-C). Likewise, many cells of the proneural clusters ofSu(H); Psn double mutant wing discs also express the SOP-E(Fig. 4B). In summary, the cells of proneural clusters of Su(H) mutant notae are present and many of them express early SOP markers, such as the SOP-E andneur^A101, but they fail to express later markers such assens and hnt. Therefore, we conclude that in Su(H)mutants, the cells of the proneural clusters arrest their development during an early phase of SOP development (see Fig. 5F).

The strong expression of E(spl)m8 in SOP in Su(H) mutant wing imaginal discs is surprising because expression of this gene is switched of in the SOP during normal development(Nolo et al., 2000)(Fig. 5D,E). Thus, it appears that Su(H) is also required to switch off expression of this gene in the SOP during normal development.

sens encodes a zinc-finger containing transcription factor that is essential for SOP development (Nolo et al., 2000). Thus, we wondered whether it is the loss ofsens activity that causes the arrest in SOP development inSu(H) mutant wing imaginal discs. To test this hypothesis, we looked to see if forced expression of sens could restore expression of neural markers such as 22c10 and elav and SOP markers such as hnt and the SOP-E in Su(H) mutant discs(Fig. 6). In the first series of experiments, we expressed UAS sens with dppGal4(Fig. 6A-G). We found that most cells that express Sens were able to activate the expression of the SOP-E, as well as hnt (Fig. 6A,C,F,G). Most sens-expressing cells of the notum also initiated 22C10 expression (Fig. 6B,C). The ability of Sens to activate these genes ectopically was restricted to the notum.

Sens was also able to activate the expression of elav. However,activation of elav expression was more restricted and appeared in clusters of cells that mapped to regions of the proneural clusters(Fig. 6D,E). This indicates that the activity of Sens is required for the expression of elav but,in contrast to 22C10, hnt and the SOP-E, is not sufficient. Similar results were found when UAS sens was expressed with scaGal4(data not shown).

In summary, the results show that Sens is able to activate the expression of those genes that are normally absent in Su(H) mutant proneural clusters and suggests that the loss of sens activity causes the arrest in SOP development in Su(H) mutants. Hence, Su(H) appears to be required in the SOP to activate the expression of sens in aNotch-independent manner.

We also analysed the function of Su(H) during SOP development by inducing mutant cell clones in the notum during first larval instar [24-48 hours after egg laying (ael)] (Fig. 7). Using this type of analysis, we found a spectrum of phenotypes. In several cases, the mutant cells express early markers such asneur^A101 or the SOP-E, but fail to express hnt. The arrowhead in Fig. 7A-Epoints to such an example. In this case one Su(H) positive cell lies in the cluster. It is this cell that expresses SOP-E and hnt. By contrast, the mutant cells of the cluster express only the SOP-E. In other cases, we found a varying fraction of Su(H) mutant cells that express the early marker and also hnt, suggesting that some of the mutant cells do not arrest their development (arrows inFig. 7B-E;Fig. 7F-I). However, many of the cells of this class expressed lower levels of hnt than normal(see Fig. 7H,I). Altogether,these observations confirm our conclusion that Su(H) is required for the development of the SOP.

We observed that the fraction of cells that expresses hnt varies among the clusters of the notum. We concentrated on the SC and DC clusters: in six out of seven cases, we found hnt-positive cells in a fraction ofSu(H) mutant cells of the SC cluster. By contrast, we found only one case out of 15, where one cell weakly expresses hnt in clones that include parts of the DC cluster (Fig. 7F,G). In this case, the hnt-expressing cell is at a position where the pDC would normally arise. At the position of the aDC, two cells express hnt (Fig. 7F), but do not express the SOP-E (inset inFig. 7F). This suggests that the aDC has already divided to give rise to pIIa and pIIb. By contrast, the pDC still express the SOP-E and weakly hnt and has not divided,indicating that its development is delayed in comparison with that of the aDC. During normal development, the aDC develops later than the pDC and switches off the SOP-E before dividing (Huang et al., 1991). Thus, it appears that the pDC is strongly delayed in its development. This observation further supports our conclusion from the analysis of the homozygous mutants, that Su(H) mutant proneural cells arrest SOP development at an early stage.

For the other clusters, where we have less cases examined, we found that in four out of six cases of clones, including parts of the pPA/tr2/aPA+tr1 cluster we found a fraction cells weakly expressing hnt. In one out of the four observed cases of the PSA cluster, we found weak hntexpression in one cell. In the two clones we found for the ANP/PNP cluster no expression was observed.

As expected, kuz mutant cell clones that include regions of proneural clusters contain big clusters of hnt-expressing cells,indicating that cells of kuz mutant proneural clusters can progress in their development as SOP (Fig. 7J).

During the course of the clonal analysis of Su(H), we very often found that a single enlarged and Su(H)-positive cell adjacent or nearly surrounded by mutant cells that express the SOP-E(Fig. 7B-E;Fig. 8I-M). Invariantly, thisSu(H)-positive cell expressed hnt(Fig. 7B-E) and was located at the position of the cluster, where a SOP would normally arise(Fig. 7B-G). This observation contradicts the lateral inhibition model, which predicts that cells in which Notch is least active have the highest potential to inhibit their neighbours from adopting the SOP fate. Hence, cells in a cluster that are defective inNotch signal reception should be very potent to inhibit their wild-type neighbours and a Su(H)-positive cell should therefore never adopt the SOP fate, if adjacent to Su(H) mutant cells.

To investigate this paradox further, we examined if the Su(H)mutant cells express high levels of Dl as we have observed in homozygous mutant discs (see Fig. 5A). Indeed, we found that Su(H) mutant cells in proneural clusters do express high levels of Dl (Fig. 8A-D). This raises the possibility that D1 protein inSu(H) mutant cells might not be active. We therefore looked to see whether the Su(H) mutant cells of a proneural cluster are able to activate Notch in adjacent Su(H)-positive cells. As a measure ofNotch activity, we used the activity of the Gbe+Su(H)-lacZconstruct [Gbe+Su(H)] (Furriols and Bray,2001). The activity of this construct is dependent on Su(H) and the presence of a functional Notch receptor(Furriols and Bray, 2001). The expression pattern of Gbe+Su(H) faithfully reveals the activity domains of Notch in the wing imaginal disc (Furriols and Bray, 2001). In the notal region of the disc, Gbe+Su(H) is active in a pattern that is similar to that of Dl(Fig. 8E-G). We observed that expression of Gbe+Su(H) is switched off in SOPs. As an example the posterior SOP of the DC cluster is shown in Fig. 8H-J. At the stage when the SOP initiates expression ofhnt, an upregulation of the expression of Gbe+Su(H) can be observed in the adjacent cells (Fig. 8H-J). The SOP itself does not express the construct(Fig. 8H-J). The expression of Gbe+Su(H), including the ring of higher expression of around the SOP, is strictly dependent on the activity of Notch and Su(H) (data not shown, Fig. 8M-O). This observation is in agreement with the model of lateral inhibition and suggests that the SOP sends a signal that activates the Notch pathway in its immediate neighbours.

We found that Su(H) mutant cells of a proneural cluster can activate the Gbe+Su(H) construct in their Su(H)-positive neighbours. This is indicated by the upregulation of the construct in these cells(Fig. 8K-O, arrowhead in N, O). Hence, the cells of Su(H) mutant proneural cluster seem to express an active form of Dl. However, one exception was observed: Gbe+Su(H) is not activated in the hnt expressing, Su(H) positive SOPs that are located next to mutant cells (arrows in Fig. 8K,M,O). Expression of Notch itself in Su(H) mutant cells was normal (data not shown). These observations indicate that Su(H)mutant cells, although expressing active Dl, cannot activate theNotch pathway in cells that are located at positions where the SOP develops. Thus, it appears that the position of the SOP within a proneural cluster is strongly determined. Cells at this position appear to be insensitive to lateral inhibition.

The conclusion that the position of the SOP within a proneural cluster is pre-determined is further supported by another observation: We found three cases where a Su(H) mutant clone includes almost all cells of a proneural cluster. In these cluster, one or two cells of the cluster weakly express Hnt (arrows in Fig. 7B,E). These cells are located at positions, where the SOP would be expected to arise. Thus, it appears that cells at certain positions in a cluster are strongly biased towards the SOP fate.

The results above suggest that Su(H) is required for the proper expression of sens in the SOP. Su(H) could activate the expression ofsens by binding directly to its promoter. Such aNotch-independent activation of target genes by Su(H) has recently been discovered (Barolo et al.,2000; Klein et al.,2000). Alternatively, Su(H) could act as a repressor that switches off the expression of a factor that in turn represses the expression ofsens. Default repression by Su(H) in absence ofNotch-signalling seems to be a common mechanism to silence expression of Notch target genes in the absence of Notch activity(Barolo et al., 2002). We have performed the following experiments to discriminate between the two possibilities. First, we made use of a construct in which a VP16 transactivation domain is fused to Su(H) (Su(H)VP16)(Kidd et al., 1998). UASSu(H)VP16 acts exclusively as an transcriptional activator that activates all Notch target genes in the embryo and wing, similar to activated forms of Notch (UAS Nintra)(Kidd et al., 1998;Klein et al., 2000). If Su(H)is a direct activator of sens transcription, expression of UASSu(H)VP16 might activate sens in cells where it is expressed. However, if Su(H) represses the expression of a repressor, UASSu(H)VP16 should activate expression of this repressor and thussens expression should be lost. We observed that expression of UASSu(H)VP16 with dppGal4 did not induce expression ofsens in the notum. By contrast, Su(H)VP16 appears to suppress its expression in most parts of the notum(Fig. 9D,E). This result favours the possibility that the repressor function of Su(H) is required forsens expression. However, we cannot rule out that, in this experiment, the loss of sens expression is caused indirectly through the suppression of the determination of the SOP by UAS Su(H)VP16.

To explore the possibility that Su(H) might act as a repressor further, we have analysed SOP development in a Hairless (H) mutant background. H is a crucial part of the repressor complex that assembles around Su(H) to repress target gene expression(Klein et al., 2000;Furriols and Bray, 2000;Morel at al., 2001;Barolo et al., 2002). If Su(H)acts as a repressor, H should also be required. Hence, removal of Hfunction in Psn mutant wing imaginal discs should lead to the loss ofsens expression in a similar way to that observed for Su(H)or Su(H); Psn double mutants (see above). We found that this is indeed the case (Fig. 9A-C). InPsn H double mutant wing discs, expression of sens was dramatically reduced or lost in proneural clusters(Fig. 9A,C). However, the cells of the clusters still express the SOP-E(Fig. 9B,C). This observation shows that the double mutant cells have not died, but fail to develop. It appears that the loss of H function in Psn mutants leads to an arrest of SOP development in a similar manner as in Su(H) and inSu(H); Psn double mutants.

We further expressed a form of Su(H) that cannot bind H because it lacks the H-binding domain (Furriols and Bray, 2000). Overexpression of this UAS Su(H)ΔHconstruct by scaGal4 abolished hnt expression inPsn mutant wing imaginal discs(Fig. 9F). As in the H Psn double mutants, the cells of the proneural clusters were present, as visualized by the expression of UAS GFP driven by scaGal4. Thus,expression of UAS Su(H)ΔH in cells of the proneural clusters appears to cause an arrest of SOP development as observed in Su(H)mutants. In summary, these results support the conclusion that Su(H) requires H for its function in SOP development (seeFig. 9H).

Su(H) is directly required for the activation of the genes of theE(spl)-C (Bailey and Posakony,1995; Hinz et al.,1994; Jennings et al.,1994; Lecourtois and Schweisguth, 1995; deCelis et al., 1996; Lai et al.,2000a; Lai et al.,2000b). Seven of the genes of this complex encode bHLH repressor proteins that are required for the suppression of the SOP fate in the cells of the proneural clusters during the process of lateral inhibition(Knust et al., 1992;Nakao and Campos-Ortega, 1996)(Fig. 10A). Four other genes of the complex encode members of the bearded protein family(Lai et al., 2000a;Lai et al., 2000b)(Fig. 10A). The data raise the possibility that the repressor of SOP development is encoded by one or more genes of the E(spl)-C.

Initially, we tested this possibility by monitoring the expression ofsens and hnt in Su(H) mutant wing imaginal discs that carried a deletion for the whole E(spl)-C, Df(3R)E(spl)^b32.2 (gro+) (Fig. 10B,C). As a result these discs had only half the number of the genes of the complex. To our surprise, we found that in theseSu(H)^Δ47; Df(3R)^b32.2(gro+)/+discs, the cells of all proneural clusters expressed high levels ofsens and hnt (Fig. 10B,C), indicating that reducing the activity of the genes of theE(spl)-C by half is sufficient to remove the block in SOP development in Su(H) mutant wing discs. Hence, one or more genes of the complex seem to encode the repressors that prevent the expression of sens and the development of the SOPs in absence of Su(H).

We then tried to determine, which of the genes of the complex encode the repressor. From previous work, it is known that only one of the bH1H proteins,E(spl)m8, and two of the bearded-like proteins, Mα and M4, are expressed in cells of Su(H) mutant proneural clusters(Bailey and Posakony, 1995;deCelis et al., 1996;Lai et al., 2000a;Lai et al., 2000b)(Fig. 10A). The other members of the complex are either not expressed in the notal region of the wing imaginal disc, or the expression is lost in Su(H) mutant cells(Bailey and Posakony, 1995;deCelis et al., 1996;Lai et al., 2000a;Lai et al., 2000b). Hence, it is likely that persistent expression of one of the three proteins expressed in cells of Su(H) mutant proneural clusters causes the arrest of SOP development. However, forced expression of UAS m4 or UAS mα in cells of the proneural clusters with scaGal4 results in the formation of supernumerary bristles (Lai et al.,2000b). This suggests that these proteins stimulate rather than preventing SOP development.

By contrast, the expression of E(spl)m8 is switched off in the SOP during normal development (Nolo et al.,2000) (Fig. 5D,E). Thus, Su(H) appears to be required to switch off the expression ofE(spl)m8. Furthermore, expression of UAS E(spl)m8 withscaGal4 prevents SOP development in Su(H) or Psnmutant proneural clusters (Klein et al.,2000) (data not shown). These facts suggest that the abnormal persistent expression of E(spl)m8 in Su(H) mutant proneural clusters might cause the arrest in SOP development. One prediction for this hypothesis is that E(spl)m8 should not be abnormally expressed in SOPs of mutants that are involved in Notch signalling, but do not affect the formation of the Su(H)/H repressor complex. One such an example is the mutants of Psn. Thus, we monitored the expression ofE(spl)m8 in Psn mutant wing discs. In addition we looked at the expression of the E(spl)mβ gene, which is expressed in a broader domain and seems to include all regions of the wing imaginal discs where Notch is active (de Celis et al.,1996) (Fig. 10E). We found that E(spl)m8 is strongly expressed in cells of Psnmutant proneural clusters in a similar way to expression in Su(H)mutants (compare Fig. 10D withFig. 5C). Hence, it is unlikely that the abnormal expression of E(spl)m8 alone causes the arrest in SOP development observed in Su(H) mutants. As expected, expression ofE(spl)mβ was reduced in both mutants in a similar manner and is not elevated in cells of the proneural clusters(Fig. 10E,F; data not shown).

Altogether, the results indicate that the repressor of SOP development inSu(H) mutants is encoded by one or more genes of theE(spl)-C. At the moment, it is difficult to determine whether the repressor activity is encoded by one member of the complex or by a combination of E(spl)m8 with one or both of the beaded-like proteins.

The Notch pathway is one of the fundamental signalling systems that are conserved throughout the animal kingdom. Therefore, it is important to gain information about the function of its core members. We here have identified a new role of Su(H) during a fundamental differentiation process of one Drosophila cell type, the development of the SOP of the mechanosensory bristles. Our results suggest that Su(H) is required to promote SOP development. This is based on the fact that most cells of proneural clusters in the notum that lack Su(H) function do not express SOP markers such as Sens, Hnt and partially neur^A101-lacZ. Loss of neur^A101-lacZ expression in some proneural clusters of Su(H) mutant discs has been observed before(Schweisguth and Posakony,1992). This loss has been attributed to a `general sickness' of the mutant discs, as the lack of neur^A101-lacZ expression was only observed in the late developing proneural clusters(Schweisguth and Posakony,1992). Our data argue against such an explanation: Psnmutant wing imaginal discs exhibit a stronger neurogenic phenotype than doSu(H) mutants. Similar to Su(H) mutants, homozygousPsn mutant animals also die during the early pupal phase. Nevertheless, the cells of the proneural clusters of these mutants express all tested markers, indicating that SOP development is not affected. The same is true for kuz mutants, whose mutant phenotype is comparable with that of Su(H) mutants. Hence, general sickness of the wing imaginal disc cells is not likely to explain the arrest of SOP development in Su(H)mutants.

A role of Su(H) in development of the SOP is surprising, because it is a core element of the Notch signalling pathway and the activity of this pathway is required to prevent SOP development in cells of the proneural clusters (Schweisguth and Posakony,1992; de Celis et al.,1996; Heitzler et al.,1996; Nakao and Campos-Ortega,1996; Klein et al.,2000). Importantly, in this new role, Su(H) seems to function independently of the Notch signalling pathway. This is indicated by the finding that the Su(H) mutant phenotype is epistatic over that ofPsn mutants.

The data presented here indicate that Su(H) appears to be required to suppress the activity of one or more members of the E(spl)-C, that in turn suppress the expression of genes such as hnt and sens. This conclusion is based on: (1) the failure of Su(H)VP16 to activatesens, (2) the fact that Psn H double mutants display a similar loss or reduction of sens expression as Su(H) andSu(H); Psn double mutants and (3) the fact that expression of aSu(H) construct that is unable to bind H (UAS Su(H)ΔH)leads to an arrest of SOP development in Psn mutant wing imaginal discs. Several reports show that H is involved in Su(H)-related suppression of gene expression in the absence of Notch signalling(Furriols and Bray, 2001;Klein et al., 2000; Morel and Schweisguth, 2001; Barolo et al.,2002). Recently, it has been shown that H acts as a bridge between Su(H) and the general co-repressors CtBP and Gro(Barolo et al., 2002; Morel and Schweisguth, 2001). It is therefore likely, that this Su(H)/H/Gro/dCtBP complex mediates the repressor function required during SOP development.

Repression by Su(H) is not strictly required in all proneural clusters to allow expression of sens and other late SOP markers. Examples are the clusters in the wing region, such as the clusters of the dorsal radius. However, even in these clusters, sens and hnt are not expressed in all cells that express early markers, such asneur^A101 (e.g. compare wing imaginal discs inFig. 2C with 2H,M; clonal data not shown). Therefore, it appears that the activity of Su(H) promotes SOP development also in these clusters. The clusters of the dorsal radius give rise to other types of sense organs, such as companiforme sensilla, and it is possible that there are different requirements for the activity ofSu(H) for the development of the different types of sense organs

We show that the removal of one copy of the E(spl)-C is already sufficient to relieve the block in SOP development in Su(H) mutants,indicating that the arrest is probably caused by the abnormal expression of one or more members of the complex. Although the complex encodes for several well-characterized repressors of neural development, we were not able to pinpoint the repressor function to any particular gene. Many studies by various groups have studied the regulation of the genes of theE(spl)-C (Bailey and Posakony,1995; deCelis et al.,1996; Lai et al.,2000a; Lai et al.,2000b). From these studies, it is clear that only three genes of the complex are expressed in the cells of Su(H) mutant proneural clusters. All other members are either not expressed in the notal region of the wing imaginal disc or their expression is lost in the mutant cells. Previous studies have shown that both bearded-like proteins that are expressed in Su(H) mutant proneural clusters promote SOP development(Lai et al., 2000a;Lai et al., 2000b). Hence, it is unlikely that the abnormal expression of these genes causes the observed arrest in SOP development. To our surprise, we found that the strongest candidate, the bHLH repressor encoded by E(spl)m8, is also abnormally expressed in Psn mutants, where SOP development proceeds and the Su(H)/H-containing complex is intact. The observation is interesting, because it suggests that the activity of the whole Notch pathway is required to switch off the expression of E(spl)m8 in the SOP, but it also indicates that abnormal expression of the gene cannot be the reason for the arrest in SOP development in Su(H) mutant cells. Thus, the repressor activity might not be encoded by a specific member of theE(spl)-C.

One possibility is that the combination of the three abnormally expressed genes of the complex generates the repressing activity. An alternative is that Su(H) controls the expression of other genes that act in combination with the upregulated members of the complex to suppress SOP development. Another possibility is that more genes of the complex are de-repressed inSu(H) mutants at a level not detectable by the currently available methods. In this scenario, the weak expression of several bHLH-encoding genes will sum up to a level of repressor activity sufficient to stop SOP development. Using currently available techniques, it is very difficult to discriminate between these possibilities.

We found that in Su(H) mutant cell clones induced during the first larval instar stage, hnt is expressed in a fraction of cells of specific proneural clusters, such as the scutellar cluster, but absent or strongly reduced in other clusters. We further found that in Su(H)mutant wing imaginal discs, expression of sens and hnt is lost or stronger reduced than in mutant cell clones induced during the first larval instar.

A high stability of the Su(H) protein is a possible explanation for this discrepancy. In favour of this explanation is the observation that the maternal component of Su(H) is sufficient to allow the development of homozygous animals until early pupal stages(Lecourtois and Schweisguth,1995). Furthermore, we found that vestigial (vg), a target gene of the Notch/Su(H) pathway during wing development(Kim et al., 1996) is expressed in Su(H) mutant wing imaginal disc of the early third larval instar stage (S.K. and T.K., unpublished). This indicates the presence of Su(H) activity at this stage. This residual activity of Su(H) must be provided by the maternal component. Both observations suggest that the Su(H) protein is degraded slowly and thus persists in mutant cells for a long time. It is therefore likely that Su(H) mutant cells, induced at the first larval instar, contain residual amount of Su(H). This residual amount of Su(H) might be sufficient to activate expression of late SOP marker in cells of certain proneural clusters.

An alternative explanation for the milder phenotype observed in theSu(H) mutant clones is that it requires time to accumulate a sufficient level of activity of the repressor(s) of the E(spl)-C to stop SOP development. Hence, in the case of the clonal analysis, the loss ofSu(H) activity occurs later than in homozygous mutant wing imaginal discs and lower levels of repressor activity would be present in cells of the proneural clusters of the machrochaete.

The development of the machrochaete is a paradigm for the assignment of different fates to initially equivalent cells. Proneural genes are expressed in clusters of cells and confer on these cells the potential to become SOPs. Careful examination has revealed that the SOPs of the machrochaete arise at the same positions within the proneural cluster (Cubas and Modollel, 1992),indicating that the selection of the SOP is not random. It is thought that other factors, such as Extramachrochaete, Pannier and Wingless introduce small differences in proneural activity that favour cells at specific positions within the cluster to become the SOP (Cubas and Modollel, 1991) (reviewed bySimpson, 1997). These small differences in proneural activity are enhanced through the activity of theNotch pathway: a cell with high proneural activity expresses high levels of Dl and is therefore potent to inhibit its neighbours(Heitzler and Simpson, 1991;Hinz et al., 1994). Cells with a high activity of Notch have less proneural activity and Dl. Thus, they are less potent to inhibit its neighbours (reviewed bySimpson, 1997). In this scenario, the Notch pathway is required to amplify initially small differences in proneural activity in cells among a cluster. This amplification eventually results in the accumulation of high levels of activity in the SOP and loss of activity in the neighbours. In this way, the pathway acts to resolve a crude pre-pattern to the level of a single cell. According to this model, cells defective in Notch signal reception should be very potent in lateral inhibition. However, we here found the opposite: A cell that is located at the position where the SOP arises is able to adopt the SOP fate,even if surrounded by Su(H) mutant cells. It can do so despite the fact that the mutant neighbours accumulate high levels of proneural activity(indicated by the expression of the SOP-E), as well as Dl. We here show that Dl, expressed in the Su(H) mutant cells at high level, is active and can activate Notch signalling in wild-type cells with the exception of the SOP. Thus, although the SOP was adjacent to cells with an extremely high potency for lateral inhibition during the whole live of a proneural cluster, it has succeeded in adopting the SOP fate. It appears that the SOP is determined by positional cues that are much more precise than anticipated. These cues render the cell at the correct position in the cluster insensitive to lateral inhibition. This suggests that small differences in proneural activity are not the crucial bias imposed on cells within a proneural cluster and that lateral inhibition might not be required to resolve a crude pre-pattern.

Nevertheless, the big clusters of SOPs observed in other neurogenic mutants, indicate that the Notch pathway has a function in preventing the SOP fate in all cells of a proneural cluster and also in cells that are located further away from the SOP.

Furthermore, we observed a ring of high expression of the Gbe+Su(H)construct around the SOP during normal development, suggesting that the SOP sends a signal that activates the Notch pathway in its immediate neighbours. This lateral inhibition is relatively late, as we observe it only around SOPs that already express hnt. It also occurs only in the cells adjacent to the SOP.

Altogether, these observations suggest that the Notch pathway might have two separable functions during SOP development. During early phases of a proneural cluster, the activity of the pathway keeps the cells of the cluster undecided, perhaps by mutual repression. Owing to positional cues, one cell becomes insensitive to the inhibitory signal and adopts the SOP fate. Subsequently the SOP inhibits its immediate neighbours by sending an inhibitory signal through Dl. A similar function of the Notch pathway has recently been proposed for the segregation of the embryonic neuroblasts ofDrosophila (Seugnet et al.,1997).

Schweisguth (Schweisguth,1995) reported that, during michrochaete development, cells can adopt the SOP fate, even if they are located adjacent to Su(H) mutant cells, suggesting that, also during development of this bristle type, the mutant cells cannot inhibit its normal neighbours. It was suggested that theSu(H) mutant cells might loose contact with neighbouring wild-type cells and, because Dl/Notch signalling depends on cell contact, this could prevent the activation of the Notch pathway in the wild-type cells. Our data suggest that this is not the case: the mutant cells can activate the pathway in adjacent wild-type cells. Hence, the results of Schweisguth (Schweisguth,1995) suggest that also during michrochaete development positional cues might be important for the determination of the SOP.

We thank S. Bray, S. Campuzano, J. Campos-Ortega, J. Modolell, A. Martinez-Arias, H. Bellen, N. Baker, G. Struhl, M. Fortini. C. Klaembt and F. Schweisguth for sending stocks and reagents. We also thank R. Wilson, M. Kaspar and K. Reiners for critical comments on the manuscript. This work was supported by the Deutsche Forschungsgemeinschaft through SFB 572 and the ARC program of the DAAD.