The EGF receptor and N signalling pathways act antagonistically in Drosophila mesothorax bristle patterning

During development, the epidermis of Drosophila generates over one thousand bristles and other types of sensory organs (SOs) (Lindsley and Zimm, 1992). Many of these bristles appear at stereotyped positions, such as the conspicuous large bristles (macrochaetae) that arise on the head and the dorsal mesothorax (notum). This arrangement of macrochaetae provides a classical model to study pattern formation (Lindsley and Zimm, 1992). Each macrochaetae derives from a single SO mother cell (SMC) which undergoes two differential divisions (Bodmer et al., 1989; Hartenstein and Posakony, 1989). The four progeny cells subsequently differentiate into the components of the SO. During the third instar larva and early pupa stages, SMCs appear in precise positions of the imaginal discs, the larval epithelia that will give rise to a large part of the adult epidermis (Cubas et al., 1991; Huang et al., 1991). Thus, the accurate position of macrochaetae is largely due to the emergence of their corresponding SMCs at specific sites of the imaginal discs.

Formation of this pattern of SMCs require the participation of genes collectively known as the proneural genes and of cell to cell signalling systems. Proneural genes confer on cells the ability to become SMCs. Two of them, achaete (ac) and scute (sc), members of the ac-sc complex (AS-C) (Campuzano and Modolell, 1992), are the most important for the development of macrochaetae. They encode transcriptional regulators of the basic region-helix-loop-helix (bHLH) family (Garrell and Campuzano, 1991; Jan and Jan, 1993) and probably commit cells into becoming SMCs by activating downstream genes that participate in the neural differentiation program. ac and sc are coexpressed in relatively small groups of cells, the proneural clusters, which prefigure the pattern of macrochaetae (Cubas et al., 1991; Skeath and Carroll, 1991). A fixed number of SMCs arise from each cluster, usually one or two. In the imaginal wing disc, a typical cluster that gives rise to one bristle may consist of 20-30 cells, but the SMC is selected from a smaller subgroup of cells that accumulate higher levels of Ac-Sc proteins than their neighbours (the proneural field, Cubas et al., 1991; Cubas and Modolell, 1992; Skeath and Carroll, 1991). This subgroup and the SMC, which accumulates the highest levels of Ac-Sc, always occupy the same position within the cluster. The SMC also accumulates Asense, another bHLH protein encoded in the AS-C (Brand et al., 1993; Domínguez and Campuzano, 1993; Jarman et al., 1993). Recently, an enhancer that mediates the increased accumulation of proneural protein in the SMC has been characterized (Culí and Modolell, 1998). It promotes proneural gene self-stimulation specifically in this cell and this activation is an early and essential step of SMC commitment. Additional, as yet uncharacterized factors are required for the action of this enhancer. Thus, SMC commitment is at present a poorly understood process.

The Notch (N) cell to cell signalling pathway prevents additional cells of a proneural cluster from becoming SMCs, and therefore the development of many macrochaetae from a single cluster. Indeed, in the absence of N signalling many proneural cluster cells become SMCs (Artavanis-Tsakonas et al., 1995; review). It is currently thought that the more proneural protein a cell accumulates, the stronger is its ability to signal and the less inhibited it will be by their neighbours. Thus, in a proneural cluster, the cells of the proneural field (which have the highest levels of proneural protein) tend to escape from the inhibition. When a cell does so, it becomes an SMC, it signals maximally and prevents its neighbours from acquiring the same fate (lateral inhibition; Heitzler and Simpson, 1991; Simpson, 1990; Simpson, 1997). The reception of this strong signal maintains the SMC-specific enhancer in these neighbouring cells in an ‘off’ state (Culí and Modolell, 1998).

The Epidermal growth factor receptor (EGFR) signalling pathway has also been implicated in macrochaetae development (Clifford and Schüpbach, 1989). EGFR signalling is transduced by the Ras/Raf/MAP kinase cascade. It participates in diverse processes of Drosophila development, like embryonic ventral ectoderm fate, head development, wing and haltere development, notum differentiation, ommatidial cell recruitment and differentiation, induction of dorsal follicle cell fate in the ovary, etc. (Freeman, 1998; Schweitzer and Shilo, 1997; reviews). How activation of the same Ras/Raf/MAP kinase pathway causes cells to adopt different fates is a very active area of research, specifically motivated by the fact that misactivation of the pathway in humans is associated with many kinds of tumours (Li and Perrimon, 1997; Moghal and Sternberg, 1999; reviews). Contrary to the inhibitory signals mediated by the N pathway, in Drosophila EGFR signalling seems to promote macrochaetae development. Indeed, hypomorphic mutations at the Egfr gene were found to remove several notum macrochaetae with different frequencies (Clifford and Schüpbach, 1989), although some macrochaetae were sometimes duplicated. Absence and duplications of macrochaetae have also been observed in nota mosaic for hypomorphic Egfr alleles (Díaz-Benjumea and García-Bellido, 1990). Microchaetae develop normally in these clones, although with a higher density probably due to the reduced size of the mutant cells. Moreover, clones of cells with amorphic Egfr alleles in the tergites can autonomously develop bristles and attract and incorporate neighbouring wild-type bristles. These results suggest that different groups of bristles have distinct Egfr function requirements.

We have analysed the role of EGFR signalling in the determination of the notum macrochaetae. While distinct proneural clusters show different requirements in the level of EGFR signalling for wild-type levels of ac-sc expression, SMCs generally fail to be determined in the absence of this signal. Our data indicate that reception of the EGFR signal is necessary for the triggering of the self-stimulatory loop of sc that is characteristic of and a requisite for SMC determination. We also show that the levels of EGFR signalling have to be regulated, as excess signalling leads to too many cells from each proneural cluster becoming SMCs. This regulation may be accomplished in part by the N-mediated interactions that occur among cells of the proneural cluster.

In(1)sc^10.1 and N^ts have been described previously (Lindsley and Zimm, 1992). A temperature sensitive condition for the Egfr gene was obtained using the heteroallelic combination Egfr^tsla/Egfr^CO, where Egfr^CO is a deficiency of the locus (Kumar et al., 1998). Larvae were raised at 18°C and incubated at 30°C for 12-15 hours before dissection. Lines carrying the transgenes UAS-ras^V12 (Karim and Rubin, 1998), UAS-argos (Freeman, 1994), UAS-raf^DN2.1 (Martín- Blanco et al., 1999), UAS-Egfr and UAS-Egfr^DN (Buff et al., 1998) and UAS-Spitz^soluble (Schweitzer et al., 1995b) were used in combination with the Gal4 drivers C253 (Culí and Modolell, 1998), C765 (Gómez-Skarmeta et al., 1996), pnr^MD237 (Heitzler et al., 1996), 179b (Brand and Perrimon, 1993), ap-Gal4 (Calleja et al., 1996), sca- Gal4 (Nakao and Campos-Ortega, 1996) and dpp^disk-Gal4 (Staehling- Hampton et al., 1994). Both sca-Gal4 and 253-Gal4 drive expression in proneural clusters (our unpublished data, and Culí, 1998), as they are presumably activated by ac-sc (Culí, 1998; Mlodzik et al., 1990). To generate clones of cells expressing the Gal4 protein and the green fluorescent protein (GFP) marker, females of the genotype yFLP122; P{Act5C>y+>Gal4}, UAS-GFP (Ito et al., 1997) were crossed with males harbouring the required UAS line. Larvae (36-60 hours after egg laying; AEL) were incubated at 36°C for 2-6 minutes and raised at 25°C until dissection. The SMC-specific lacZ reporter transgenes used were neuralized (neu)-lacZ A101.IF3 (Huang et al., 1991) and SRV-lacZ (Culí and Modolell, 1998).

lacZ expression was analysed in wing imaginal discs by X-gal staining (Gomez-Skarmeta et al., 1995). Anti-Sc and anti-Sens (Nolo et al., 2000) antibody staining was performed as described by Cubas et al. (Cubas et al., 1991) for conventional microscopy or using lissamine rhodamine-conjugated anti-rabbit and CY5-conjugated anti-guinea pig secondary antibodies for confocal microscopy.

In situ hybridization to detect rho/ve mRNA was performed as described by González-Crespo and Levine (González-Crespo and Levine, 1993) using an antisense DIG-labeled RNA probe.

Weak hypomorphic Egfr alleles cause the partial removal of several notum macrochaetae (Clifford and Schüpbach, 1989). The effect of stronger loss-of-function Egfr mutations has not been determined since these mutations drastically reduce the size of the imaginal wing discs and cause lethality. Moreover, clones of cells homozygous for amorphic or nearly amorphic Egfr mutations do not survive in the prospective notum (Díaz- Benjumea and García-Bellido, 1990; and our unpublished results). Consequently, we have reexamined these findings using the temperature sensitive combination Egfr^tsla/Egfr^CO (Kumar et al., 1998; Table 1). At a permissive temperature (18°C), three bristles (ASA, PSA and PPA) were often missing and the anterior postalar (APA) and anterior dorsocentral (ADC) were frequently duplicated. When late third instar larvae were placed at a non-permissive temperature (30°C) for 15 hours (pupation took place during this interval) and completed development at 18°C, the presence of all notum macrochaetae was affected to different extents, excepting the scutellars and the APA, a bristle that was sometimes duplicated (Table 1). Stronger phenotypes were obtained by overexpressing a dominant negative form of EGFR (UAS- Egfr^DN) with either the drivers sca-Gal4 (expressed in proneural clusters; Mlodzik et al., 1990) or ap-Gal4 (expressed in the dorsal compartment of the disc; Table 1). With ap-Gal4 at 29°C most notum macrochaetae were removed, although microchaetae were unaffected. UAS-aos, which encodes the Argos protein (an EGFR inhibitory ligand; Schweitzer et al., 1995a), driven in proneural clusters by C253-Gal4 suppressed both macro and microchaetae and only a few bristle sockets remained (Fig. 1H). These results suggest that EGFR signalling is essential for bristle development. Consistent with this conclusion, increasing the levels of the wild-type receptor (ap-Gal4/UAS-Egfr) promoted development of extra macrochaetae, but only in the vicinity of the extant ones (Fig. 1I). This suggests that the excess signalling causes extra macrochaetae precursors to arise within the extant proneural clusters.

We next examined which stage(s) of macrochaetae development is/are affected by the decreased function of EGFR. One of the earliest events, the establishment of the proneural clusters, was analyzed by examining the accumulation of Sc protein. Heat-treated Egfr^tsla/Egfr^CO discs showed that the proneural clusters located at the central region of the prospective notum, the dorsal radius and the pleura had reduced levels of Sc, while those at the scutellar and ANP clusters were increased (compare Fig. 1A and B). These effects, which were not observed in discs from non heat-treated larvae (not shown), suggest different requirements for Egfr function to accomplish wild-type levels of sc expression in different regions of the imaginal disc. Note, however, that for many proneural clusters the modified levels of Sc protein still permitted the development of the corresponding macrochaetae with near wild-type frequency (Table 1).

The next step in bristle development, SMC emergence from proneural clusters, was also sensitive to the loss of EGFR signalling. Accumulation of the Aos inhibitory ligand in proneural clusters (C253-Gal4/UAS-aos) was almost not enough to modify the Sc protein levels (Fig. 1C, compare with 1A). However, SMCs, distinguishable by their enhanced accumulation of Sc, were not identifiable. Their absence was verified by the lack of expression of the SMC-specific marker neu-lacZ (A101.IF3 enhancer trap line; Huang et al., 1991; Fig. 1F). Consistent with these findings, in the heat-treated Egfr^tsla/Egfr^CO discs, expression of neu-lacZ was clearly reduced and/or delayed (Fig. 1E). These results suggest that EGFR signalling is generally required for SMC emergence, although the necessary levels may vary for different SMCs. This may explain the suppression of essentially all notum bristles by Aos overexpression (Fig. 1H).

EGFR signalling is mediated by the activation of the Ras/Raf signal transduction pathway (Freeman, 1998; Schweitzer and Shilo, 1997; reviews). We verified that bristle development also required the activation of this pathway by overexpressing a dominant negative form of Raf (UAS-raf^DN2.1). Relatively mild but ubiquitous expression in the notum (C765-Gal4 driver) eliminated (95-100%) the ASA and PSA macrochaetae and promoted (45% of heminota) the generation of one extra DC. Relatively late overexpression in proneural clusters (C253-Gal4 driver) only eliminated the APA (83%) and occasionally generated an extra DC (7% of heminota). With the pnr^MD237-Gal4 driver, which promotes early expression at the dorsal-most part of the presumptive notum, the DC proneural cluster was sharply reduced and the DC bristles were absent (100% of ADC and 91% of PDC, 33 thoraces examined, not shown). A generalized and stronger expression of UAS-raf^DN2.1 (179b-Gal4) was lethal and the few pharate adults recovered lacked 50-80% of notum macrochaetae, although in approximately half of these cases the socket of the missing bristle was present. This socket phenotype, as well as that shown in Fig. 1H, suggests that the Ras/Raf signalling pathway is also required for the late process of bristle differentiation.

We also examined the effect of UAS-raf^DN2.1 on the establishment of proneural clusters and SMC emergence by analyzing Sc accumulation in clones of cells overexpressing this transgene. In the SC and PNP clusters, this accumulation was not overtly modified (Fig. 2A,B). In contrast, cells within the APA, DC or vein L3 proneural clusters that expressed UAS- raf^DN2.1 autonomously lost or had reduced expression of sc (Fig. 2C,D). These results confirm that cells of proneural clusters near the central region of the prospective notum require the EGFR signal to optimally express sc, while those located further away, like the NP and SC clusters, do not show this requirement (Fig. 1A,B). SMCs, recognized by their increased accumulation of Sc or their senseless (sens) expression (Nolo et al., 2000) were found only outside the UAS-raf^DN2.1-overexpressing clones. Moreover, these cells can emerge in abnormal positions when the clones occupied the sites where SMCs normally appear (Fig. 2E-H).

Overactivity of EGFR signalling was mimicked by overexpressing a constitutively activated form of Ras by means of the UAS-ras1^V12 transgene (Karim and Rubin, 1998). With either ap-Gal4, pnr^MD237-Gal4 and dpp^disk-Gal4, which drive expression in subregions of the wing disc, or 179b-Gal4 and C765-Gal4, which promote ubiquitous expression, sc was ectopically activated (Fig. 3A,B and results not shown). In the notum territory, high levels of ectopic sc expression occurred mostly in single cells. Many of them were SMCs, as they expressed lacZ under the control of an sc SMC-specific enhancer (Culí and Modolell, 1998) (Fig. 3D,E). We could not determine whether these SMCs gave rise to bristles since the overexpression of activated Ras was lethal.

Activated Ras expressed within proneural clusters (C253- Gal4/UAS-ras1^V12) did not overtly affect Sc accumulation in the clusters (Fig. 3C). However, extra SMCs did appear within the clusters, since the neu-lacZ SMC-specific marker often revealed several neighbouring neu-lacZ-positive cells (Fig. 3F,G) and several macrochaetae were generated near the extant ones (Fig. 4E). This again indicated that EGFR signalling promotes SMC determination.

When the expression of UAS-ras1^V12 was restricted to clones of cells, it was clear that sc activation was cell-autonomous, although the levels of expression varied from site to site within the disc (Fig. 4B). At the central part of the prospective notum and pleura (not shown) activation was maximal, while it could not be detected in areas like the proximal-most notum. Moreover, the accumulation of Sc also varied among the cells of a clone, and many of the ones with the highest levels were probably SMCs, since they expressed the SRV-lacZ SMC- specific marker (Fig. 4C,D). Moreover, flies in which few clones of UAS-ras1^V12-expressing cells were induced (2 versus 6 minute induction) survived to adulthood and showed clusters of adjacent macrochaetae (Fig. 4F,G). Taken together, these results indicate that reception of the Ras- mediated signal can induce accumulation of Ac-Sc and SMC commitment. As this can occur in adjacent cells (Figs 3F,G, 4C,F,G) excess Ras signalling appears to overrule lateral inhibition promoted by N signalling (Heitzler and Simpson, 1991; Simpson, 1990; Simpson, 1997).

The Ras/Raf signalling cassette is downstream of several receptor tyrosine kinases (Perrimon, 1994). Hence, we verified that the sc activation observed by overexpressing Ras1^V12 could also be accomplished by activating EGFR itself. Expression in cell clones of the soluble form of Spitz, an activating ligand of EGFR (Golembo et al., 1996; Schweitzer et al., 1995b), promoted ectopic expression of sc in cells both within and outside of the clones (Fig. 4A), as expected of a diffusible ligand. The effect was maximal within and near the soluble Spitz-producing cells and, again, in the more central regions of the prospective notum.

The transmembrane protein Rhomboid/veinlet (Rho/ve) is known to activate EGFR signalling (reviewed by Wasserman and Freeman, 1997) by presenting and helping solubilize the ubiquitous, membrane-bound form of Spitz (Bang and Kintner, 2000). In the wing imaginal disc, rho/ve is strongly expressed in the presumptive wing veins, the wing lack Ac and Sc proneural proteins (Fig. 5B). This suggests that the proneural proteins activate rho/ve in proneural clusters and this should help promote EGFR signalling within at least part of their cells. Note however, that a low level of EGFR activation must also occur in a rather generalized way in the prospective notum, since cell proliferation is impaired in the complete absence of EGFR activity (Clifford and Schüpbach, 1989; Simcox et al., 1996).

We attempted to monitor in wild-type nota the activity of the EGFR pathway by examining the accumulation of the doubly phosphorylated mitogen-activated protein (MAP) kinase (dp- ERK; Gabay et al., 1997). However, the low levels of dp-ERK in the presumptive notum precluded consistent detection with currently available antibodies.

The previous results indicate that EGFR signalling occurs among the cells of proneural clusters and that it promotes SMC emergence. In contrast, Dl-N signalling among cells of a margin, part of the dorsal radius and the nascent trachea (Sturtevant et al., 1993), but expression at the prospective notum has not been appropriately described. Although weak, rho/ve mRNA accumulation was detected in most proneural clusters, the DC showing the highest levels (Fig. 5A). Expression was not detectable in all the cells constituting each proneural cluster and the clusters in which expression occurred varied from disc to disc. This is compatible with a dynamic and short-lived expression. rho/ve transcription was dependent on ac- sc, as it was undetectable in In(1)sc^10.1 discs, which

In a reciprocal experiment, we found that a large decrease in EGFR activity (C253-Gal4/UAS-aos) did not significantly modify the levels of E(spl)-m8 mRNA in proneural clusters (not shown). This suggests that the EGFR pathway does not affect N signalling. We also found that under conditions of sharply reduced N signalling (overexpression of a dominant negative form of the ligand Dl in proneural clusters), EGFR signalling was still required for macrochaetae development, since in this genetic background the overexpression of UAS- aos eliminated these sensory organs (Fig. 5F,G). This is consistent with the N and EGFR pathways acting antagonistically and in parallel on the SMC-specific enhancers (see Discussion).

The development of the Drosophila mesothoracic macrochaetae requires the activity of the EGFR pathway. We have analyzed this requirement and found that EGFR signalling is involved in at least three stages of the development of these sensory organs, namely, formation of proneural clusters, emergence of SMCs from these clusters, and SO differentiation. This last aspect is suggested by the ‘sockets without bristle’ phenotype observed in flies in which EGFR activity is impaired by the expression of the inhibitory ligand Aos or the dominant negative form of Raf. We have not further studied this role of EGFR.

The earliest stage in macrochaetae development is the formation of the proneural clusters of ac-sc expression. We find that accumulation of Sc in cells of proneural clusters located at the more central positions of the wing disc decreases upon reduction of the level of EGFR signalling. The effect is cell- autonomous, which indicates that reception of the signal is important for cells to express sc properly. In contrast, more marginally located clusters, like the notopleural or scutellar, were unmodified or slightly enhanced under conditions of insufficient EGFR signalling. It is known that expression of ac-sc in different proneural clusters depends on separate, functionally independent enhancers which are thought to respond to local, specific combinations of transcription factors (prepattern) (Gómez-Skarmeta et al., 1995). The different, spatially restricted effects of the insufficiency of EGFR function may thus be due to interference in the deployment or function of particular factors expressed in the affected area. Interestingly, the expression of the homeobox genes of the iroquois complex, necessary for the expression of ac-sc in many notum proneural clusters (Leyns et al., 1996), is especially sensitive to the expression of the Vein EGFR ligand in the central region of the notum (Wang et al., 2000). Alternatively, since EGFR function is a well known requisite for growth and patterning of imaginal discs (Clifford and Schüpbach, 1989; Díaz-Benjumea and García-Bellido, 1990; Díaz-Benjumea and Hafen, 1994; Nagaraj et al., 1999; Simcox et al., 1996; Sturtevant et al., 1993; Wang et al., 2000), the reduced expression of sc may be due to a more general impairment of the patterning of the central area of the disc.

Our data support a key role for EGFR signalling in the emergence of the notum macrochaetae SMCs from proneural clusters. Indeed, expression of the EGFR inhibitory ligand Aos exclusively in proneural clusters, a condition that permits essentially wild-type Sc accumulation in these clusters, almost completely suppressed the appearance of SMCs and SOs. SMC emergence was also impaired in discs from heat-treated Egfr^tsla/Egfr^CO larvae and in clones of cells expressing UAS- raf^DN2.1. Moreover, when the cells that accumulated Raf^DN2.1 occupied positions where SMCs normally appear, wild-type neighbouring cells could give rise to displaced SMCs. This phenomenon is reminiscent of and in accordance with the observation, made with mosaic individuals, that when the position of a dorsocentral bristle is in ac⁻ territory, this bristle does not develop, but a nearby ac⁺ cell can give rise to a dorsocentral bristle displaced from its wild-type position (Stern, 1954). The cell-autonomous effect of Raf^DN2.1 indicates that reception of the EGFR signal, mediated by the Ras/Raf/MAP kinase cassette, is essential for notum macrochaetae SMC determination. This was further substantiated by the cell autonomous induction of SMCs and bristles in clones of cells overexpressing a constitutively activated form of Ras. Taken together, these results indicate that reception of the EGFR signal promotes sc expression and SMC determination.

We found that in the notum anlagen the expression of rho/ve occurred mainly in proneural clusters and that this expression was dependent on ac-sc. Rho/ve facilitates the processing of Spitz, an activating ligand of EGFR (Bang and Kintner, 2000). We also found that the soluble, active form of Spitz promoted ectopic sc expression and SMC emergence. Hence, these data suggest that, in proneural clusters, Ac-Sc promote expression of rho/ve, which by activating Spitz, would stimulate EGFR signalling in the cells of the cluster (Fig. 6). (The Vein EGFR ligand probably does not specifically act in proneural clusters, as many of these lie outside of its expression domain; Simcox et al., 1996; F. Cavodeassi, personal communication.) We thus propose that EGFR mediates a mutual positive signalling among cells of the proneural cluster, which promotes SMC emergence by probably reinforcing ac-sc expression. We call this positive signalling lateral cooperation. Evidently, this does not exclude an autocrine activation of the EGFR pathway in the cells that express AS-C proteins, but we favor the lateral cooperation hypothesis since it is well established in other systems that the EGFR pathway is used mainly for intercellular communication (Freeman, 1998; Schweitzer and Shilo, 1997; reviews). As discussed below, this signalling should facilitate the acquisition of the SMC state by one or a few cells of a proneural cluster.

The SMC state is associated with greatly increased levels of proneural protein (Brand et al., 1993; Cubas et al., 1991; Culí and Modolell, 1998; Domínguez and Campuzano, 1993; Jarman et al., 1993; Skeath and Carroll, 1991). These are accomplished by the self-stimulation of ac, sc and ase mediated by AS-C enhancers that activate these genes specifically in the cells that become SMCs (Culí and Modolell, 1998). As we have shown that Ras1^V12 elicits the expression of both sc and SRV-lacZ, we propose that, in the extant proneural clusters, the SMC-specific enhancers are targets of EGFR signalling. Unidentified effector(s) of the EGFR/Ras pathway should facilitate the self-stimulation of the proneural genes mediated by the SMC-specific enhancers by, possibly, binding to these enhancers. Conclusive evidence in support of this role requires the identification of the signalling effector(s) and of their interaction with the enhancer. Interestingly, overexpression of the effector Pointed P1 promotes development of many extra macrochaetae on the notum (J. Culí, unpublished) and we have detected putative Ets-domain binding sites in the sc and ase SMC enhancers (GTGGAAAT and ACGGAAAC, respectively, Culí and Modolell, 1998).

EGFR-mediated lateral cooperation should tend to activate the SMC-specific enhancers in many cells of the proneural clusters (Fig. 6). This, however, is prevented by N signalling, which is activated by Ac and Sc in the cells of the cluster (Simpson, 1997, review). This signalling, by means of the bHLH proteins of the E(spl)-C (Artavanis-Tsakonas et al., 1995; review), blocks the ac-sc-ase self-stimulatory loop promoted by the SMC-specific enhancers (Culí and Modolell, 1998) (Fig. 6A). However, within a proneural cluster the cells of the proneural field accumulate greater amounts of Ac-Sc proteins (Cubas et al., 1991; Skeath and Carroll, 1991). As it has been hypothesized that cells that signal the most are the least inhibited by their neighbours, eventually, a cell of the proneural field will be released from the inhibitory loop and its levels of E(spl)-C bHLH protein will become minimal (Jennings et al., 1995). This cell will turn on the ac-sc-ase self-stimulation and become an SMC (Fig. 6B). The SMC signals maximally to its neighbours and prevents them from following the same fate (lateral inhibition).

Our results add to this scenario the requirement for EGFR- mediated signalling for one cell of the proneural field to turn on the ac-sc-ase self-stimulatory loops and become an SMC (Fig. 6). According to this model, Ac-Sc activate both the N- and EGFR-mediated signalling pathways, with their SMC- suppressing and SMC-promoting abilities, respectively, and both signalling systems appear to act on the same SMC- specific enhancers. Since an excess signalling by the N or the EGFR pathway will either prevent SMC determination or promote emergence of ectopic SMCs, the respective levels of signalling should balance each other so that only one SMC is determined at a time from each proneural cluster. How is this balance accomplished? This is at present unclear. The large enhancement of rho/ve mRNA in proneural clusters under conditions of insufficient N signalling suggests that this pathway may prevent the Rho/Ve-promoted activation of EGFR from rising to excessively high levels. In contrast, the insensitivity of the levels of E(spl)-m8 protein to the overexpression of UAS-aos in proneural clusters suggests that the EGFR pathway does not affect N signalling. Antagonistic interactions between the N and the EGFR pathways are found in other developing systems, as in the wing preveins (de Celis et al., 1997) and in the reiterative recruitment, from a long-lived atonal proneural cluster, of the precursors of the 70-80 scolopidia of the femoral chordotonal organs (zur Lange and Jarman, 1999). In this later case, EGFR signalling promotes commitment of neural precursors and the Dl-N interaction prevents too many cells from being committed.

In the presumptive notum, the inability of available antibodies to reliably detect dp-ERK and, in proneural clusters, the low levels of rho/ve mRNA (compared to those in the wing preveins; Sturtevant et al., 1993; Gómez-Skarmeta et al., 1996) suggest that low levels of EGFR activity are sufficient to ensure the emergence of the macrochaetae precursor cells. This may explain the failure of the Egfr hypomorphic alleles compatible with cell or adult viability to completely eliminate notum macrochaetae (Clifford and Schüpbach, 1989; Díaz-Benjumea and García-Bellido, 1990; and this paper). The notum microchaetae appear to be even more resistant to the lowering of EGFR signalling. Perhaps, they do not directly require it for development, similarly to the terguite bristles that can arise within Egfr amorphic clones (Díaz- Benjumea and García-Bellido, 1990). An essential difference between notum macrochaetae, on the one hand, and notum microchaetae and terguite bristles, on the other, is that the first appear in fixed positions while the others do not do so, being instead organized in density patterns. We speculate that EGFR signalling among the cells of the proneural field may make the selection of the SMC less ambiguous and, therefore, spatially more precise. A cell centrally located within this subset would receive the strongest signalling from their neighbours and would become a SMC in preference to more marginally located neighbours (Fig. 6). The observation that slight reduction in the level of EGFR signalling causes duplications of some notum macrochaetae (Clifford and Schüpbach, 1989; Díaz-Benjumea and García-Bellido, 1990; and Table 1), that is, it makes the decision of which cell becomes an SMC less precise and it allows two SMCs to arise from presumably the same proneural cluster, may be consistent with this interpretation.

The overexpression of UAS-aos in proneural clusters removes essentially all bristles, including those of the tergites (not shown). This may indicate that all SOs require some level of EGFR signalling to develop. However, the fact that in the tergite clones homozygous for amorphic Egfr⁻ alleles still develop bristles (Díaz-Benjumea and García-Bellido, 1990) suggests that the Aos overexpression may be interferring with additional tyrosine kinase receptors that would be redundant with EGFR in the development of these bristles.

We thank A. Martínez Arias, J. F. de Celis and colleagues of our laboratory for helpful comments during the course of this work and constructive criticisms on the manuscript. We are grateful to J. F. de Celis for flies overexpressing UAS-Egfr and UAS-Egfr^DNand H. Bellen for anti-Sens antibody. A contract from Ministerio de Educación y Cultura to E. M. B., and a University of Cambridge (Department of Zoology) visiting professorship from Fundación Banco Bilbao Vizcaya to J. M., are acknowledged. This work was supported by grants from Comunidad Autónoma de Madrid (07B/0033/1997), Dirección General de Investigación Científica y Técnica (PB93-0181 and PB98-0682) and an institutional grant from Fundación Ramón Areces to the Centro de Biología Molecular Severo Ochoa.