The Pax-homeobox gene eyegone is involved in the subdivision of the thorax of Drosophila

In Drosophila, much is known about genetic subdivisions of the body. An early and fundamental one is the segregation of anterior (A) and posterior (P) compartments along the anteroposterior (AP) axis of the body,which establishes the parasegmental trunk (reviewed by Lawrence and Morata, 1994; Lawrence and Struhl, 1996; Mann and Morata, 2000). The various Hox genes become active in specific positions, thus generating the morphological diversity along this axis.

However, within each segment, the morphological diversity does not depend on the Hox genes but on other developmental genes that are expressed in specific regions within segments. For example, engrailed (en)determines the difference between A and P compartments(Morata and Lawrence, 1975), apterous (ap) is involved in distinguishing the dorsal and ventral regions of the wing (Diaz-Benjumea and Cohen, 1993), and pannier (pnr) discriminates the dorsal-medial and dorsal-lateral regions of the thorax and abdomen(Calleja et al., 2000; Herranz and Morata, 2001).

The second thoracic segment develops in the dorsal region of the wing and the corresponding part of the thorax, known as mesothorax. These structures derive from the wing imaginal disc, which is subdivided from the beginning into A and P compartments (Lawrence and Morata, 1977). This subdivision affects both the wing and the thoracic region. In the mesothorax, the A compartment forms the greater part,the notum, whereas the P compartment is a small featureless region known as the postnotum.

The notum is made up of approximately 15,000 cells that develop a complex and highly stereotyped bristle pattern in which no lineage restriction has been found (Calleja et al.,2000). Thus, its genetic/morphological diversity has to be generated by non-lineage subdivisions. The notum has been shown to be subdivided into two major regions defined by the activities of the pnr and Iroquois (Iro) genes(Calleja et al., 2000; Gomez-Skarmeta et al., 1996). The Iro genes are expressed originally in all the notum cells and specify the development of the entire notum (Diez del Corral et al., 1999). Later, pnr restricts Iro gene activity to the lateral region and also specifies dorsal-medial development(Calleja et al., 2000).

However, pnr and the Iro genes are expressed in comparable domains in head, thoracic and abdominal segments, suggesting that they encode general properties such as dorsal-medial or dorsal-lateral, which apply to all segments. That is, they do not determine the development of one specific segment, but are probably involved in a general combinatorial mechanism together with other general factors, such as the Hox genes or engrailed (reviewed by Mann and Morata, 2000). It follows that there should exist other genes,regulated by the combinations of the above selector genes, which would be responsible for the morphological diversity within the different segments.

Using the `yellow' method (Calleja et al., 1996), we have isolated a number of Gal4 lines conferring expression in various parts of the thorax. One of them (EM461) yielded Gal4 activity in most of the scutum, the part of the notum from the anterior border to the suture with the scutellum (Fig. 1). No activity was observed in the abdomen. We report a functional study of a gene whose expression directs Gal4 activity in the EM461 line. The gene was discovered to be eyegone (eyg), which encodes a homeodomain Pax protein (Jun and Desplan,1996). eyg is one of the elements of the genetic network activated during eye development (Hazelett et al., 1998; Hunt,1970), and it is also involved in the development of the salivary gland duct (Jones et al.,1998). Our results demonstrate that eyg, and its related and adjacent gene twin of eyegone (toe), play a role in the genetic subdivision of the thorax. The eyg/toe function is necessary for scutum formation, and its ectopic activity in the scutellum transforms this into a scutum-like tissue. eyg expression in the scutum is regulated by a positive input of the Iro genes and pnr, and by the repressor activity of the Hh and Dpp signalling pathways.

By using plasmid rescue, we cloned and sequenced the flanking genomic sequence 3′ of the pGalW element(Brand and Perrimon, 1993). The insert is located at 157,094 bp within the scaffold AE 003541 (FlyBase), 102 bp upstream of the eyg transcription start. toe is located 30 kb downstream of eyg (Flybase).

The following Drosophila strains were used in this work to generate loss of function clones: Df(3L)iro(Leyns et al., 1996); eyg^20MD1, Gal4-eyg^gv5 (generous gift of Maria Dominguez); FRT80 iro^DfM1/TM6(Gomez-Skarmeta et al., 1996); FRT42ptc¹⁶/CyO, FRT40Atkv^a12/Cyo(Nellen et al., 1996); FRT40A Mad^B1/Cyo(Wiersdorff et al., 1996); FRT40Asmo³/Cyo (Strutt and Mlodzik, 1997); FRT2A eyg^SA2/TM6B; and FRT2A mwh/TM3 Sb. The FLP/FRT technique(Xu and Rubin, 1993) was used to generate loss of function clones. Larvae of the appropriate genotype were heat shocked for 1 hour at 37°C, at different larval stages. The clones were visualised in discs by either loss of GFP or of β-galactosidase expression.

For gain-of-function experiments, the following Gal4 lines were used: ywf^36a hsFLP122; abx FRT f⁺ FRT Gal4-UAS-lacZ/Cyo (de Celis and Bray,1997); ywflp122; act-FRT y⁺ FRT Gal4 UASGFP/SM5 Tb (Ito et al., 1997); ap-Gal4 (Rincon-Limas et al.,2000); 455-Gal4(Martin-Blanco, 1998); ush-Gal4 (Calleja et al.,2002); 248-Gal4(Sanchez et al., 1997); pnr-Gal4 (Heitzler et al.,1996); and 638-Gal4 (M. Calleja and G.M., unpublished). The gain-of-function clones were generated by recombination at the FRT sequences. The gain of Gal4 and lacZ activity can be detected by β-Gal staining. In the adult cuticle, the clones can be scored because they are mutant for f^36a and contain y⁺ activity. The UAS-ara(Diez del Corral et al., 1999)and UAS-tkv^QD(Hoodless et al., 1996) lines used have been described previously. UAS-eyg and UAS-toelines were generated by cloning the whole ORF into the pUAST vector, using the following cloning sites: NotI/XhoI for eyg, and EcoRI/XhoI for toe. The constructs were injected into yw embryos and stable lines were selected by rescue of the white phenotype. eyg cDNA was kindly provided by C. Desplan,and the toe cDNA by the BDGP. For the apoptosis experiments we used an UAS-P35 line (Hay et al.,1994).

The lacZ reporter lines used were: en-LacZ(Simcox et al., 1991), neu-LacZ (Flybase) and esg-lacZ(Whiteley et al., 1992).

Males homozygous for the pGal4 insertion (EM461) were crossed to females carrying the hop transposase. Excisions of the pGal4 transposon were selected by the loss of the w⁺ eye in the F1 progeny. Individual revertants were crossed to TM3/TM6 flies and balanced. PCR analysis was carried out with individual stocks with strong phenotypes over Df(3L)iro. We used one primer located 2.5 kb downstream of the P-element insertion site and two primers located upstream of the insertion site, one 3 kb upstream and the other 11.5 kb upstream.

Downstream primer, 5′-CCGGTGGACTATGGCGCGAACGGACGCG-3′;

Upstream 1, 5′-CGGCGTGGCCACCTTGGGCTTTGAGCC-3′; and

Upstream 2, 5′-CGGCGAGGGGAGTGGGGCCTGATGGG-3′.

The complete eyg ORF was cloned in the pQE vectors (Qiagen) and the recombinant protein purified by following the QIA express protocols. The protein was injected into guinea pigs and the serum obtained was used as a polyclonal antibody.

Discs were dissected in PBS and fixed in 4% paraformaldehyde for 20 minutes at room temperature. Discs were subsequently washed in PBS, blocked in blocking buffer (PBS, 0.3% Triton, 1% BSA), and incubated overnight with the primary antibody: anti-β-Gal (1:2000; rabbit), anti-Eyg (1:200; guinea pig) or anti-Ara (1:200; rat) diluted in blocking buffer at 4°C. Washes were performed in blocking buffer, and the appropriate fluorescent secondary antibody was added for 1 hour at room temperature. Following further washes in blocking buffer, the discs were mounted in Vectashield. Anti-Ara antibody was kindly by S. Campuzano, anti-β-Gal (rabbit) and anti-caspase 3 were purchased from Cappel and from Cell Signalling, respectively. Images were taken in a laser MicroRadiance microscope (Bio-Rad) and subsequently processed using Adobe Photoshop.

In situ hybridisation and antibody/in situ hybridisation-double labelling were performed as described (Azpiazu and Frasch, 1993), and embryos were mounted in Permount (Fisher Scientific). Digoxigenin-labelled RNA probes were synthesised as described(Tautz and Pfeifle, 1989). eyg-specific antisense RNA probe was synthesised from a 570 bp SacI/KpnI fragment of the cDNA provided by C. Desplan. The toe-specific antisense RNA probe was synthesised from a plasmid provided by the BDGP and consists of an 475 bp EarI/XhoI fragment.

Adult flies were prepared by the standard methods for microscopic inspection. Soft parts were digested with 10% KOH, washed with alcohol and mounted in Euparal. Embryos were collected overnight and aged an additional 12 hours. First instar larvae were dechorionated in commercial bleach for 3 minutes and the vitelline membrane removed using heptano-methanol 1:1. Then,after washing with methanol and 0.1% Triton X-100, larvae were mounted in Hoyer's lactic acid (1:1) and allowed to clear at 65°C for at least 24 hours.

The EM461 Gal4 line directs expression in most of the scutum, antennae and the genitalia. EM461>UAS-y adults show no y⁺rescue either in the legs or in the abdomen. The insertion was located in the third chromosome and is homozygous viable with no visible phenotype.

By using plasmid rescue, we cloned sequences at the insertion point of the P-element (see Materials and Methods), and found that it was inserted 102 bp 5′ of the transcription start site of eyg (FlyBase). The expression pattern found in EM461>UAS-lacZ embryos was similar to that of eyg (see below). Also, the imaginal expressions of EM461 and of eyg were largely coincident.

The analysis of the genomic region flanking the insertion point revealed the presence of another transcription unit, toe, and one predicted gene (CG32102; Fig. 1). toe is located 35.7 kb upstream from the P-element insertion, and CG32102 1.1 kb downstream from the insertion point. No cDNA has been reported for the latter gene. By contrast, toe has been previously cloned and shows high sequence identity with eyg (FlyBase). Their sequence conservation and location indicates that eyg and toe derive from a duplication of a primordial eyg-like gene, and raises the possibility that they may have redundant functions.

We have analysed the expression of eyg and toe during embryonic and imaginal disc development using the UAS-lacZ construct, eyg and toe specific RNA probes, and an anti-Eyg polyclonal antibody generated in our laboratory. The distribution of the eyg and toe transcripts appears to be very similar during embryogenesis, and also in the wing imaginal disc, although the expression levels of toeare consistently lower than those of eyg. The β-Gal distribution in eyg-Gal4/UAS-lacZ embryos and imaginal discs matches that observed for the eyg and toe transcripts, and that observed with the anti-Eyg antibody. Thus, we shall use the anti-Eyg antibody as indicative of the expression of both genes.

Our results on the embryonic expression of eyg/toe are in agreement with those already reported for eyg using in situ hybridisation (Jones et al.,1998) (see also Fig. 3), and will not be considered further. We checked the possibility that eyg might be expressed in the embryonic primordia of dorsal discs by double-label experiments with escargot (esg)(Whiteley et al., 1992) and found that none of the cells expressing eyg show esgactivity (data not shown). This result suggests that at the beginning of the wing disc development eyg is not yet active. eyg/toe is expressed in the eye, wing and haltere discs. The expression and function of eyg in the eye disc has been reported(Hazelett et al., 1998);expression in the wing disc is shown in Fig. 2. The first sign of activity is observed at the beginning of the third instar(Fig. 2B), and, by that stage,the Eyg product already is restricted to a part of the thoracic region of the disc. In mature discs there are, in addition to the major domain in the thorax, two small expression domains in the hinge(Fig. 2B,D) and the pleura.

We have delimited the eyg/toe domain in the thorax by double labelling wing discs with anti-Eyg antibody and probes from genes also expressed in the thorax: the Iro genes, pnr and en. The expression domain does not overlap with that of en(Fig. 2B), indicating it is only active in the anterior compartment, but it overlaps partially with the pnr and Iro gene domains in the anterior notum(Fig. 2D,E). The eyg/toe domain does not extend to the whole of the lateral region of the notum, as indicated by the X-Gal staining of eyg-Gal4/UAS-lacZ flies (Fig. 2A), where some of the macrochaetes are not labelled, and also by double labelling with neu-lacZ(Fig. 2C), which marks all the macrochaetes; several of them are outside the eyg/toe domain.

The general conclusion from these experiments is that in the wing disc eyg/toe expression is activated approximately at the onset of the third larval period. Its expression is restricted from the beginning to a subgroup of thoracic cells that will form the anterior central portion of the mesothorax.

We have generated a large number of revertants of the eyg-Gal4 line that lose the w⁺ gene. Of these, the eyg^SA1 and eyg^SA2revertants fail to complement and their trans combination shows a clear notum phenotype, lacking most of the bristles. They also fail to complement eyg^20MD1, a P-lacZ insertion located 3′ of the eyg gene. The eyg^SA2 and eyg^20MD1 alleles show a stronger phenotype in their trans combinations, and over Df(3L)iro, a deletion of eyg, toe and several other genes (see Materials and Methods). eyg^SA2 is a 9 kb deletion including the entire eyg transcription unit, but not that of toe. eyg^SA1 is a 3.8 kb deletion that removes part of the eyg transcription unit. Most of the phenotypic analysis has been carried out using the eyg^SA2 allele and the Df(3L)iro. There are no individual mutants for the toegene.

The phenotype of eyg^SA2/Df(3L)iro is shown in Fig. 3A. The notum is much reduced owing to the lack of practically the entire eyg/toedomain. The majority of the microchaetes of the central region and the dorsocentral bristles are missing, but the scutellar and lateral bristles are present. The zones not affected by the mutations are those that in wild-type flies do not possess eyg/toe activity.

As most of the normal eyg domain is lacking in eyg^SA2/Df(3L)iro flies, there was the possibility that the loss of eyg function may cause apoptosis in cells normally expressing eyg. It has been reported that eyg mutants produce apoptosis in the eye cells anterior to the morphogenetic furrow(Jang et al., 2003). However,in our experiments, we do not find apoptosis in the mutant wing discs and there is no detectable caspase activity in the notum region (using an anti-caspase antibody; data not shown). Moreover, we carried out an experiment designed to assay the contribution of apoptosis to the eyg-mutant phenotype. The mutation eyg^gv5 is a Gal4 insertion at eyg with a strong phenotype (Fig. 3H), which can be used to drive the activity of the baculovirus gene P35, a general inhibitor of apoptosis(Hay et al., 1994). Flies of the genotype eyg^gv5/Df(3L)iro; UAS-P35 show the same phenotype as their siblings eyg^gv5/Df(3L)iro, indicating that apoptosis is not a major factor contributing to the eyg syndrome in the notum.

We have also induced eyg-mutant clones cells in the notum. As expected, these clones behaved normally outside the eyg domain, but those inside the domain show loss of bristles and abnormal patterning, which is often associated with alterations in the normal polarity of surrounding bristles and trichomes. In the disc they adopt a round shape, suggesting that they tend to sort out from surrounding eyg-expressing cells. These results indicate that the activity of eyg is connected with the acquisition of specific cell affinity properties.

The interpretation of the phenotype of eyg mutations may be complicated by the possibility that although the eyg^SA2 is a complete loss-of-eyg transcription, there could be some rescue by toe, which is expressed in the same domain. However, we note that, in eyg-mutant flies, most or all the eyg domain is lost(Fig. 3A,H), suggesting that if there is rescue by toe it has to be weak. The ability of the Toe product to rescue the eyg-mutant phenotype was tested in an experiment in which we used the eyg^gv5 Gal4 line to drive the Toe product. As shown in Fig. 3H,I, flies of the genotype eyg^gv5 >UAS-toe show partial rescue of the eyg^gv5 phenotype in the notum. This experiment indicates that the Toe product has a developmental potential similar to that of Eyg (see also below and Fig. 4). However, toetranscript levels appear normal in eyg^SA2 mutants(Fig. 3G), and still are not able to rescue the strong phenotype of eyg^SA2, suggesting that the normal toe levels are not sufficient to substitute for the loss of eyg activity.

We have used eyg and toe specific RNA probes to examine the transcriptional activity of the two genes in eyg mutants. In eyg^SA2 embryos there is no eyg transcription(Fig. 3E,), whereas that of toe appears normal (Fig. 3F). In wing discs of the same genotype we detect, using the anti-Eyg antibody, a small but consistent amount of cross-reacting product(Fig. 3C). As this suggested that the anti-Eyg antibody recognises both the Eyg and Toe products, we carried out an experiment expressing toe under the control of an Ubx-Gal4 driver line. In Ubx-Gal4 > UAS-toe embryos, eyg transcription is not altered, but anti-Eyg antibody staining depicts the ectopic activity of the toe gene (data not shown).

Altogether, the preceding results indicate that normal eygfunction is required for the appearance of the anterior central part of the notum. There is little, if any, rescue of the eyg-mutant phenotype by toe, in spite of the fact that it is expressed in the same domain. The results also demonstrate that toe is not regulated by eyg, as the loss of eyg activity does not affect toe expression.

To study the effects of ectopic eyg and toe expression by the Brand and Perrimon method (Brand and Perrimon, 1993), we generated UAS-eyg and UAS-toe transformants, which then were crossed to various Gal4 lines. We also used the flip-out method (Chou and Perrimon, 1992) to induce marked clones of eyg-expressing cells.

The effect of misexpressing eyg and toe in the notum was analysed using the 248-Gal4, pnr-Gal4, 638-Gal4, ap-Gal4,455-Gal4 and ush-Gal4 driver lines. The combinations of the Gal4 lines with UAS-eyg, and with UAS-toe, give essentially the same results (Fig. 4A-C),indicating that the two gene products have a similar, or the same,developmental capacity.

In the combinations with 248-Gal4, 455-Gal4, 638-Gal4, ush-Gal4 and pnr-Gal4 the scutellum is modified and develops a scutum-like pattern, as indicated by shape alterations and the appearance of microchaetes(Fig. 4A-E). We also observe that there is an abrupt change in polarity of the bristles and trichomae close to the border of the transformation, suggesting that eyg and toe might have a role in defining polarity of the epidermal elements.

In contrast to the preceding combinations, in ap-Gal4/UAS-eygflies, the scutellum is absent but there is no indication of scutellum to scutum transformation (Fig. 4F). As ap is expressed in all notal cells after the second larval period (Diaz-Benjumea and Cohen, 1993; Rincon-Limas et al., 2000) and precedes eyg expression, we believe that inducing the activity of eyg early in notum development programmes the whole notum to develop as the eyg/toe domain, thus preventing the formation of scutellum (see Discussion).

The transformations observed in clones of eyg-expressing cells(data not shown) fit well with those described with the Gal4 lines. In the normal eyg territory these clones develop normally, indicating that increased levels of eyg activity do not appear to have a major effect. In the scutellum and lateral region they tend to sort out from surrounding tissue and also exhibit a change of fate; this is especially clear in the scutellum, where they develop scutum-like tissue.

These results suggest that eyg and toe are involved in the determination of the specific development of the anterior central region of the notum. This function is also reflected by the acquisition of specific cell affinities, which may have a role in the partitioning of the thorax (see Discussion). eyg- expressing clones in the disc are of normal shape in the eyg domain but have a round shape in the scutellum region,again indicating that eyg affects cell affinities.

One significant result of the misexpression experiments is that whereas ectopic eyg or toe expression in the thorax induces scutum-like tissue, their activity in the wing does not produce a comparable transformation. The expression of eyg, or of toe, alone produces defects in growth and differentiation in the wing, but there is no clear indication of transformation towards notum(Fig. 5B). This behaviour is clearly illustrated by the 638-Gal4 line which drives expression in the posterior notum as well as in the wing pouch(Fig. 5A). In 638-Gal4 >UAS-eyg flies the scutellum, but not the wing, is transformed towards scutum. Only in a small region in the centre of the wing is there is some tissue resembling notum (Fig. 5B).

This suggested that eyg may require a developmental cofactor normally present in the thorax but not in the wing. As the obvious candidates were the Iro genes, which we show to be activators of eyg (see below), we have compared the effect of eyg in the wing in either the presence of the absence of the ara gene. Of the three genes of the Iro-complex, araucan (ara), caupolican (cau), and mirror(mirr), the first two are expressed in the same domain of the wing disc and appear to have the same developmental role(Gomez-Skarmeta et al.,1996).

Flies of genotype 638-Gal4 > UAS-ara exhibit normal notum development and in addition the wing also becomes transformed towards notum(Fig. 5C). Thus the Ara product has the capacity to induce notum development, as expected from previous studies (Diez del Corral et al.,1999). In the favourable cases, it is possible to discriminate the presence of scutellum structures in the ectopic notum(Fig. 5C). In 638-Gal4 >UAS-ara UAS-eyg flies, ara and eyg are co-expressed in the same cells of the wing. In these flies, the ectopic notum produced in place of the wing displays, additionally, a scutellum to scutum transformation(Fig. 5D). This result strongly suggests that ara (and probably cau) is providing the necessary `thoracic' context for the normal function of eyg/toe. We believe that the notum-like tissue observed in the centre of the wing of 638-Gal4 > UAS-eyg flies(Fig. 5B) is caused by the Iro gene activity normally present in that region(Gomez-Skarmeta et al.,1996).

Altogether the results obtained about the developmental consequences of eliminating or misexpressing eyg and toe indicate that these genes have a role in the subdivision of the notum into an anterior-central and a posterior-lateral region. To achieve this function they require the contribution of the Iro genes.

The eyg/toe expression domain in the thorax covers only part of the notum; as illustrated in Fig. 2, eyg/toe-expressing cells in the mature disc are restricted to the anterior and central region of the disc. The misexpression experiments show that this spatial restriction is important for the normal patterning of the thorax. Thus, the factors controlling eyg/toeexpression are significant components of the pattern formation process of the thorax.

We first checked the regulatory role of the Iro genes, for although in mature discs eyg and the Iro genes are not co-extensive(Fig. 2D), there is evidence that in early disc development Iro gene expression covers much or all the mesothorax (Diez del Corral et al.,1999). Thus the eyg domain would be a subset of the initial Iro gene domain, which suggests that the Iro genes might function as early activators of eyg. Moreover, cells lacking Iro gene activity are not able to differentiate thoracic structures(Diez del Corral et al.,1999), suggesting that they also lack eyg activity.

We find that, in absence of Iro gene activity, eyg expression is lost: clones of Iro gene mutant cells induced in the first instar abolish eyg expression even in regions in which the Iro genes are not active in mature discs (Fig. 6A-C). An intriguing feature of these clones is that their effect on eygextends to the zone outside the clone. Similar non-autonomous effect of the Iro genes have been reported (Diez del Corral et al., 1999).

However, in the third larval period the eyg and Iro gene domains are not co-extensive (Fig. 2). Clones of cells lacking Iro gene function induced at this stage lack eygactivity, but only if they localise in the Iro gene domain; Iro^-clones in the pnr domain retain eyg activity(Fig. 6D-F). This indicates that there should be other factor(s) responsible for the maintenance of eyg expression in the part of the domain where the Iro genes are not active at this stage. The obvious candidate is pnr, which in late disc development is expressed in a notum domain complementary to that of the Iro genes (Calleja et al.,2000). We have induced pnr-mutant clones in third instar discs (72-96 hours of development) and examined the effect on eygexpression. These clones are small, and because they have differential cell adhesion properties (Calleja et al.,2000), tend to be round in shape and sort out from surrounding cells. In these clones, eyg expression is not modified, probably because the loss of pnr function results in gain of Iro gene activity(Calleja et al., 2000), which upregulates eyg. Nevertheless, ectopic pnr expression induces ectopic activation of eyg(Fig. 6I), just like the Iro genes do (Fig. 6H). This effect is not mediated by concomitant Iro gene expression as pnr does not activate (but rather repress) the Iro genes(Calleja et al., 2000). These results strongly suggest that both the Iro genes and pnr act as positive regulators of eyg.

Another important aspect of eyg regulation is how its expression is inhibited in the scutellum and lateral notum. These regions are close to the AP compartment border, suggesting that the Hh and Dpp pathways might have a repressing role. There is evidence(Mullor et al., 1997) that the scutellum and the zone close to the AP border is patterned by Hh. This region is also expected to contain high levels of Dpp.

The role of the Hh pathway was tested by examining eyg expression in clones of cells that either are defective in [by eliminating the activity of the smoothened (smo) transducer] or express high levels of hh activity (by removing ptc activity or misexpressing hh). As illustrated in Fig. 7A-C, smo^- clones located in the proximity of the AP border gain eyg activity. This gain of eyg activity is also reflected in the cuticular differentiation of these clones, which form scutum-like structures (Fig. 7D). On the contrary, ptc^- clones localised in the eyg domain show loss of eyg expression(Fig. 7E-G). Similarly, ectopic expression of hh in the eyg domain results in loss of eyg expression (data not shown).

The regulatory role of Dpp was assessed by inducing clones of cells mutant for Mothers against Dpp (Mad), a gene necessary for the transduction of the Dpp signal (Newfeld et al.,1996). The results are shown in Fig. 8A-F: the behaviour of the mutant clones regarding eyg expression depends on their position in the disc. In those clones located inside the eyg domain, eygexpression remains normal, although in some cases there is a noticeable increase of expression (Fig. 8D-F). Mad^- clones outside but close to the eyg domain exhibit gain of expression(Fig. 8A-F), but in those located further away eyg is not de-repressed. This effect of position can be observed in the larger Mad^- clones, like the one depicted in Fig. 8A-C; the zone of the clone close to the border of the domain expresses eygectopically whereas the zone opposite does not. We believe the reason for this differential behaviour of the clones is due to the various levels of Hh signalling in this region (see Discussion). We have found several Mad^- clones in the posterior compartment, as judged by their position, one of which is shown in Fig. 8D-F. None of these clones show gain of eyg expression. Because hh does not play a role in the posterior compartment (Basler and Struhl, 1994), the failure of Mad^- clones to de-repress eyg expression suggests that the latter is regulated by a different mechanism in the posterior compartment. The repressor role of Dpp upon eyg expression is further supported by the results obtained from forcing high levels of Dpp activity in the eyg domain: in cells mutant for a constitutive form of the Dpp receptor, thick veins(Nellen et al., 1996), eyg activity is abolished (Fig. 8G-I).

In conclusion, the Iro and pnr genes appear to act as activators of eyg, whereas the Hh and Dpp pathways repress eygactivity. The combination of these two antagonistic activities yield the normal eyg expression pattern.

The homeobox gene eyg is known to play a role in eye development(Hazelett et al., 1998; Hunt, 1970; Jang et al., 2003), where it acts in combination with other eye forming genes. We describe a new, hitherto unknown, function of eyg in the development of the thorax. We also show that eyg and its twin gene, toe, perform similar and redundant roles, although eyg exerts the principal role. Given the similarity in expression and function of the two genes, we will refer to eyg/toe as a single functional unit.

As are the other segmental structures of Drosophila, the mesothorax is subdivided into anterior (notum) and posterior (postnotum)compartments by the activity of the en gene. Subsequently, these compartments are further subdivided(Calleja et al., 2000) into medial and lateral regions by the activities of the pnr and Iro genes. These regions are not defined by restricted lineages, but are kept developmentally segregated by the differential affinities of the pnr-and Iro gene-expressing cells (reviewed by Mann and Morata, 2000; Calleja et al., 2000). We now find that the notum, but not the postnotum, is further subdivided by the activity of eyg/toe. Its role in this subdivision is clearly demonstrated both by the experiments in which its function is eliminated and when it is ectopically expressed. The lack of eyg results in flies in which the central anterior region, corresponding to the eyg/toe notum domain, is missing(Fig. 3A). Conversely, ectopic expression of either eyg or toe in the scutellum transforms it towards scutum (Fig. 4). The conclusion is that eyg/toe is involved in a new step in the subdivision of notum, which is superimposed over the previous one established by the Iro genes and pnr. The result is the appearance of four distinct regions, according to the active or inactive state of eyg/toe in the pnr and the Iro gene domains. The specific activation of eyg/toe in the notum, the anterior mesothoracic compartment, but not in the postnotum, the posterior mesothoracic compartment,appears to be an important factor in the generation of diversity in the thorax.

A significant functional feature of eyg/toe is that it is unable to induce notum structures by itself, but requires co-expression of its activator the Iro gene (Fig. 5), and probably pnr. For example, whereas ectopic eyg/toe activity induces scutum-like structures in the scutellum(which is also part of the notum and which expresses pnr), it fails to do so in most of the wing. Interestingly, it only induces notal structures in the middle of the wing (Fig. 5B), precisely the place where there is Iro gene activity in normal development (Gomez-Skarmeta et al.,1996). This mode of action is unlike that of selector or selector-like genes, such as the Hox genes, en, Dll, pnr or the Iro genes (reviewed by Mann and Morata,2000), which are able to induce, out of context, the formation of the patterns they specify. This indicates that eyg/toe is not of the same rank, but that it is developmentally downstream of the Iro genes and pnr, and appears to mediate their `thoracic' function. The restriction of eyg/toe activity to the thorax, unlike the Iro genes and pnr, which are also expressed in the abdomen, is fully consistent with this role. We note that eyg/toe is also expressed in a similar domain in the metathorax, suggesting that it may perform a parallel role in this segment.

The eyg/toe expression domain occupies the larger part of the notum, extending from the anterior border to the suture between the scutum and the scutellum. This domain coincides with the region affected in the eyg mutations, and is consistent with the gain-of-function experiments. Thus the eyg/toe expression domain corresponds to the zone where eyg/toe function is required. As it is only a part of the notum, a question of interest is to find out how eyg/toeexpression is restricted to this zone. This restriction is necessary for the appearance of distinct anterior-central and posterior-lateral subdomains, for if eyg/toe are expressed uniformly, as in ap-Gal4 >UAS-eyg flies (Fig. 4F),the entire notum develops as the anterior-central domain.

Our experiments indicate that the localised expression of eyg/toeis achieved by the activity of two antagonistic factors: the promoting activity induced by the Iro and pnr genes, and the repressing activities exerted by the Hh and the Dpp pathways. The latter are probably mediated by Hh and Dpp target genes that are yet to be identified.

Both Iro genes and pnr act as activators of eyg/toeexpression. In Iro gene-mutant clones eyg is abolished, and ectopic Iro gene activity results in ectopic eyg expression(Fig. 6). Although pnr^- clones do not lose eyg activity, the probable explanation is that they show Iro gene activity(Calleja et al., 2000), which upregulates eyg. However, ectopic pnr expression induces eyg activity (Fig. 6). Because the Iro gene and pnr expression domains cover the entire notum, in the absence of any other regulation they would induce eygactivity in the whole structure.

The elimination of eyg activity from the scutellum and lateral notum is caused by the Hh and Dpp pathways. Because the AP compartment border is displaced posteriorly in the notum, these two pathways are active at high levels in the posterior region of the mesothorax. Assuming that the two signals behave as in the wing (Lecuit et al., 1996; Nellen et al.,1996), Hh activity will be higher in the region close to the AP border, whereas the effect of Dpp will extend further anteriorly. Thus the repressive role of Hh will be greater in the proximity of the AP border and that of Dpp will be greater in more anterior positions. This is precisely what our results indicate. In the region close to the AP border the high Hh levels alone are sufficient to repress eyg(Fig. 8C). However. in more anterior positions, close to the border of the eyg domain, Hh levels are lower and, although Hh is still necessary, it is not sufficient to repress eyg. Here there is an additional requirement for Dpp activity(Fig. 8A-F).

Thus, the part of the notum that does not express eyg can be subdivided into two distinct zones according to the mode of eygregulation: a region close to the AP border that requires only Hh, and a more anterior region that requires both Hh and Dpp. In the posterior compartment,the repression of eyg has to be achieved by a different mechanism because neither the inactivation of the Hh pathway in smo^-clones, nor the inactivation of the Dpp pathway(Fig. 8D-F), induces ectopic eyg activity. A probable possibility is that en itself may act as repressor.

The result of the antagonistic activities of the Iro genes and pnron the one hand, and of the Hh and Dpp signalling pathways on the other,subdivides the notum into an eyg/toe expressing domain and a non-expressing domain. The localised expression of eyg/toecontributes to the morphological diversity of the thorax, as it distinguishes between an anterior-central region and a posterior-lateral one. It provides another example of a genetic subdivision of the body that is not based on lineage. It also provides an example of a patterning gene acting downstream of the combinatorial code of selector and selector-like genes. Its mode of action supports a model in which the genetic specification of complex patterns, such as the notum, is achieved by a stepwise process involving the activation of a cascade of regulatory genes.

We thank Eduardo Moreno for providing the EM461 insert and the initial characterisation of the Gal4 line; Maria Dominguez for the gift of the eyg mutants; Claude Desplan for providing us with the eygcDNA; and Sonsoles Campuzano for the UAS-ara flies. We also thank Manuel Calleja and Ernesto Sanchez-Herrero for discussions and comments on the work and all the Morata lab members for their support. Many thanks to Rosa Gonzalez and Angélica Cantarero for their technical help. The experimental work has been supported by a grant from the Dirección General de Investigación Científica y Técnica(BMC2001-2131 and PB1998-0511). S. Aldaz is a recipient of a fellowship from the Ministerio de Ciencia y Tecnología and N. Azpiazu is supported by the Ramón y Cajal Programme.