Antagonistic and cooperative actions of the EGFR and Dpp pathways on the iroquois genes regulate Drosophila mesothorax specification and patterning

Drosophila imaginal discs provide an excellent model system to investigate how genetic information specifies the different body regions of an organism and how these regions are patterned. A pair of wing imaginal discs gives rise to the notum, pleura, wings and wing hinges of the insect,structures that display characteristic patterns of sensory organs and veins. Morphologically, cells from third instar imaginal discs are essentially indistinguishable one from another. However, classical genetic analyses and molecular data have uncovered the existence of a plethora of genes expressed in the discs with specific spatial and temporal patterns whose products provide a vast amount of positional information(Mann and Morata, 2000). The genes that actually carry out the specification of the different body regions and fates, for instance the specification of body wall versus appendage or the implementation of neural versus epidermal fate, decode this positional information. Their cis-regulatory elements (REs) act as integrating devices receiving inputs present in the cells, which consist of combinations of transcription factors, and eliciting responses in the form of the restricted patterns of expression of the genes they are controlling(Modolell and Campuzano,1998).

The araucan (ara), caupolican (caup)and mirror (mirr) genes of the Iro-C(Gomez-Skarmeta et al., 1996; McNeill et al., 1997) belong to the class of genes whose expression specifies territories. Thus, the homeoproteins they encode are essential for, among other functions (reviewed by Cavodeassi et al., 2001),the specification of the notum territory and its subsequent patterning(Diez del Corral et al., 1999; Gomez-Skarmeta et al., 1996). These diverse functions rely on their spatially and temporally restricted expression patterns. Thus, at the second larval instar, these genes are expressed in the most proximal region of the wing disc. There, they confer on cells the ability to form notum since, in their absence, these cells give rise to structures of the dorsal wing hinge(Diez del Corral et al.,1999). Furthermore, the confrontation of Iro-C-expressing and nonexpressing cells generates an organizing border similar to those found between compartments (Diez del Corral et al., 1999; Villa-Cuesta and Modolell, 2005). Afterwards, expression of Iro-C is restricted to the prospective lateral notum where the Iro proteins function as components of the prepattern that governs the expression of the achaete-scute complex (AS-C)genes and, accordingly, help define the pattern of adult sensory organs(Gomez-Skarmeta et al.,1996).

To understand the development of the notum it is thus necessary to clarify how expression of the Iro-C is controlled. Previous reports have shown that,in the prospective notum region of the wing disc, Iro-C is under the control of the Epidermal growth factor receptor (EGFR) and Decapentaplegic (Dpp)signalling pathways. During the second larval instar, EGFR signalling is necessary for notum development and expression of Iro-C(Wang et al., 2000; Zecca and Struhl, 2002a; Zecca and Struhl, 2002b),whereas the Dpp pathway contributes to the confinement of the expression of Iro-C to the notum region (Cavodeassi et al., 2002). In the early third larval instar, this pathway further restricts expression of the Iro-C genes to the prospective lateral notum(Cavodeassi et al., 2002). How both pathways converge on the regulation of Iro-C genes is still unknown. To address this point, we have characterized the regulatory sequences of the Iro-C, a gene complex that spans approximately 150 kb of DNA(Gomez-Skarmeta et al., 1996; McNeill et al., 1997). Expression of Iro-C genes is thought to be controlled by enhancer regulatory sequences that would act jointly on, at least, ara and caup(Gomez-Skarmeta et al., 1996). We have identified five partially redundant wing disc-specific cis-regulatory elements within the Iro-C (IroREs) and demonstrated the ability of some of them to mediate regulation by the EGFR and Dpp signalling pathways. We show that the transcription factor Pointed (Pnt) mediates the activation of Iro-C by the EGFR pathway and the involvement of Pannier (Pnr), U-shaped (Ush) and Mothers against Dpp (Mad) transcription factors in Dpp-dependent repression. In addition, we propose a mechanism for the coexistence of Iro-C expression and the activity of the Dpp pathway at the prospective lateral notum near the anterior-posterior (AP) compartment boundary. Our results help clarify how the territorial specification and patterning of the Drosophila thorax is effected by the antagonistic/cooperative action of EGFR and Dpp pathways on Iro-C expression.

All the mutant alleles and transgenes are described in FlyBase(http://flybase.org). Marked clones of mutant cells were generated by FLP-mediated mitotic recombination (Xu and Rubin,1993; Lee and Luo,1999) in larvae of the following genotypes:

y w hsFLP1.22; ubi-nlsGFP FRT40A/ush^VX22FRT40A (or tkv^a12 FRT40A, or yan⁴⁴³FRT 40A or yan⁸⁸⁴ FRT 40A); IroRE-lacZ/+

y w hsFLP1.22; IroRE-lacZ/+; FRT82B ubi-nlsGFP/FRT82B grn^7L12 (or FRT82B pnt^Δ88 or FRT82B pnr^VX6)

y w hsFLP1.22; IroRE-lacZ/+; FRT82B M ubi-nlsGFP/FRT82B ttk^1e11

yw hsFLP tubα1-Gal4, UAS-GFP; tubα 1-Gal80 FRT40A/ush^VX22 FRT40A; UAS-tkv^QD/IroRE-lacZ

Second instar (48-72 hours after egg laying; AEL) or early third instar(72-96 hours AEL) larvae were heat shocked for 1 hour at 37°C.

Clones of cells overexpressing different genes were obtained as described by Ito et al. (Ito et al.,1997). Actin5C>yellow⁺>Gal4, UAS-GFP;IroRE-lacZ males were mated with females carrying hsFLP1.22 and different UAS transgenes (UAS-tkv^QD, UAS-Mad, UAS-pnr, UAS-ush,UAS-vein, UASRas^V12, UAS-Raf^DN, UAS-Raf^Act,UAS-pntP1). Larvae were treated 24-48 hours AEL or 72-96 hours AEL, for 10 minutes at 37°C. The clones were revealed by GFP expression.

EcoRI or HindIII fragments of genomic Iro-C DNA from the lambda phage and P1 walk described by Gomez-Skarmeta et al.(Gomez-Skarmeta et al., 1996),encompassing 110 kb of the DNA proximal to the iro^DFM2breakpoint, were subcloned in the C4PLZ enhancer tester plasmid that contains a weak P-element promoter (Wharton, Jr and Crews, 1993). Fragments encompassing IroRE¹ and IroRE² were ligated and subcloned into C4PLZ to obtain the transgenic flies IroRE¹-IroRE²-lacZ. IroRE^2-B1 to IroRE^2-B5 fragments were obtained by PCR amplification using 1.6 kb IroRE^2-B as a template and appropriate primer sets. (Primers sequences are available upon request.) PCR-amplified fragments cloned into pGEM Teasy vector (Promega), were excised by EcoRI digestion and subcloned into C4PLZ. The lacZ reporter plasmids were introduced into y w¹¹¹⁸ embryos by standard P-element transformation (Ashburner,1989). Three to six independent transgenic lines were established and examined for each construct.

Imaginal discs were dissected and stained as described previously(Gomez-Skarmeta et al., 1996). Primary antibodies were: rabbit and mouse anti-β-galactosidase (Cappel and Promega), rat anti-Caup, an antibody that recognizes both Ara and Caup(Diez del Corral et al., 1999)and mouse anti-Wingless (DSHB). Secondary antibodies were from Jackson Immunoresearch Laboratories and Molecular Probes. Confocal images were acquired using Bio-Rad Microradiance and Zeiss LSM510 Meta confocal microscopes. Images were imported and assembled using Adobe Photoshop 7.0 software.

Flyenhancer(http://flyenhancer.org/Main)(Markstein and Levine, 2002)and cis-analyst(http://rana.lbl.gov/cis-analyst/)(Berman et al., 2002) programs were used in the search of putative transcription factor binding sites. Cross-species sequence conservation was monitored using the VISTA program(http://genome.lbl.gov/vista/index.shtml)(Couronne et al., 2003).

Mutagenesis was performed by the site directed QuikChange system(Stratagene). To obtain the IroRE^2B-2-GATA-mut,IroRE^2B-2 DNA subcloned into pGEM Teasy, was used as template to mutate the putative GATA binding site GATAAG into CTGAAG with the following primers: forward 5′-GATGGCGATGGCAGCctgAAGCCCCATGATTTTG-3′ and reverse 5′-CAAAATCATGGGGCTTcagGCTGCCATCGCCATC-3′. Iro-RE^2B-2DNA was similarly used to obtain the Iro-RE^2B-2-ETS-mut, with the putative conserved ETS binding site CGGGATG changed into CGCATTG, using the following primers: forward 5′-CCGGGGATTGGGAAATGGGTTCGcatTGGCCAGTTTAGTCG-3′ and reverse 5′-CGACTAAACTGGCCAatgCGAACCCATTTCCCAATCCCCGG-3′. Altered bases are shown in lowercase. Mutations were confirmed by sequencing.

The Iro-C genes ara and caup show similar patterns of expression in the wing disc. In early second instar larvae, they are expressed in the whole prospective mesothorax region(Cavodeassi et al., 2002; Gomez-Skarmeta et al., 1996). Later, in the third instar, their expression is restricted to the lateral notum (Fig. 1A). In addition,at this developmental stage, novel domains of expression appear in the prospective regions of the L1, L3 and L5 veins, tegula, dorsal radius, dorsal and ventral pleura and alula (Diez del Corral et al., 1999; Gomez-Skarmeta et al., 1996). The expression of mirr is slightly different, being absent from the L3, L5 and tegula domains but present at the other domains(Kehl et al., 1998). The Iro-C harbours two additional transcription units, lincoyan(linc), whose pattern of expression at the notum is identical to that of ara and/or caup and quilapan (quil),which is ubiquitously expressed (R. Diez del Corral, PhD thesis, Universidad Autónoma de Madrid, Spain, 1998). Previous genetic analysis suggested the existence of enhancer-like REs that would drive the coincident expression of ara and caup in the wing disc(Gomez-Skarmeta et al., 1996). Thus, In(3L)iro^DFM2, associated with a breakpoint within the ara transcription unit (Fig. 1B), removes ara expression in the wing disc except in the L3 vein domain, in contrast to caup expression which is only lost from that domain (Gomez-Skarmeta et al.,1996). This suggests the existence of vein L3-specific RE(s)distal to the In(3L)iro^DFM2 breakpoint and other RE(s),specific for the remaining domains of Iro-C expression, located proximal to such breakpoint. To identify notum-specific REs, we analyzed the regulatory potential of 31 different genomic fragments, spanning approximately 110 kb of genomic Iro-C DNA (Fig. 1B).

Only five of those fragments drove lacZ expression at specific regions of the imaginal wing disc (Fig. 1B). One of them, 3.3 kb in length and named Iro regulatory element² (IroRE²), reproduced most of the expression pattern of Iro-C in the prospective notum(Fig. 1D,D′,H,H′). Thus, IroRE²-lacZ was expressed in the proximal region of early third instar wing discs (the presumptive notum region; Fig. 1H,H′) and at the presumptive lateral notum in third instar wing discs(Fig. 1D,D′). Note,however, that the pattern of IroRE²-mediated lacZexpression did not exactly coincide with that of ara/caup. Thus,β-gal was not detected in a triangular area, located near the notum/hinge border and centred around the AP compartment boundary, where expression of ara/caup is enhanced (Fig. 1D, arrowhead, compare with D″). This is precisely the region where expression of lacZ was driven by another Iro-C genomic fragment of 3.9 kb, IroRE¹ (Fig. 1C,C′). Accordingly, an IroRE¹-IroRE²composite RE was found to drive lacZ expression in a pattern very similar, albeit not identical, to that of the endogenous ara/caupgenes (see Fig. S1 in the supplementary material).

Two other genomic fragments, IroRE³ and IroRE⁴ (3.4 and 3.7 kb), adjacent to each other (Fig. 1B), drove lacZ expression in a stripe of cells located at the proximal region of the presumptive lateral notum, which partially overlapped with the caup expression domain(Fig. 1E-F′). Finally,IroRE⁵ (2.8 kb) drove expression mainly in the prospective alula and peripodial membrane (Fig. 1G).

The identified REs also drove expression of lacZ in other imaginal discs and in the embryo, in patterns resembling that of the Iro-C genes (see Figs S2, S3 in the supplementary material). We did not identify any RE driving expression in the L1, L3 or L5 veins.

In the proximal region of the wing disc, activation of EGFR by its ligand vein leads to the expression of the Iro-C genes(Wang et al., 2000; Zecca and Struhl, 2002b). We thus examined whether EGFR signalling similarly regulates expression driven by IroRE², by monitoring β-gal accumulation in clones of cells with altered EGFR signalling. Excess of signalling conditions(UAS-Ras^V12 or UAS-Raf^Actoverexpression clones), promoted upregulation of lacZ expression,similar to that observed with the endogenous ara/caup genes and with similar topological restrictions (Fig. 2A-B′) (Zecca and Struhl, 2002b). Namely, in clones induced at the second instar, lacZ activation occurred at the hinge and pleura domains(Fig. 2A,A′, arrows) but not at the central wing pouch or proximal notum(Fig. 2A,A′, arrowheads and not shown). In later induced clones, lacZ expression occurred within the wing pouch, albeit at lower levels than at the hinge(Fig. 2B,B′, insets), and in the anterior region of the medial notum(Fig. 2B,B′, arrow). Conversely, lacZ expression was eliminated in UAS-Raf^DN overexpressing clones(Fig. 2C,C′). As described for the endogenous ara/caup genes, overexpression of vein had no effect (Zecca and Struhl, 2002a). We conclude that EGFR signalling positively controls IroRE².

Dpp signalling represses Iro-C expression(Cavodeassi et al., 2002). Thus, we next investigated the ability of this pathway to modulate the expression of IroRE²-lacZ. Inactivation of the pathway(tkv^a12 clones) caused strong ectopic expression of lacZ in the proximal notum (Fig. 3A,A′) and a weaker one at the hinge(Fig. 3C,C′). Furthermore, lacZ expression was upregulated in the extant domain of IroRE² (Fig. 3B,B′). Conversely, IroRE²-lacZ was repressed when the Dpp pathway was overactivated in UAS-tkv^QD (a constitutively activated form of Tkv)(Nellen et al., 1996) or UASMad overexpressing clones (Fig. 3D,D′ and not shown). Thus, Dpp signalling negatively regulated IroRE². Taken together the above results indicate that the EGFR and Dpp pathways regulate ara/caup expression in the notum mainly through IroRE².

We next whished to identify the sites of IroRE² that mediate its response to the EGFR and Dpp pathways. We reduced the IroRE² to a 1.6 kb subfragment (IroRE^2-B; Fig. 4B), which maintained enhancer activity in the notum (Fig. 4C) and was activated by EGFR and repressed by Dpp signalling (not shown). Next, we examined its sequence for putative binding sites of effectors of these pathways using Flyenhancer(Markstein and Levine, 2002)and cis-analyst (Berman et al.,2002) programs.

Regulation of gene expression by the EGFR pathway is mediated by Pointed(Pnt; an ETS activator transcription factor) and/or by removal of repression by the ETS factor Yan and the zinc-finger factor Tramtrack (Ttk) (reviewed in Rebay, 2002). As shown in Fig. 4B (and see Fig. S4 in the supplementary material), IroRE^2-B contains nine putative ETS binding sites (Xu et al.,2000) and five putative Ttk binding sites(Brown and Wu, 1993; Read et al., 1990). With respect to the Dpp pathway, heteroliogomers of Mad and Medea act as intracellular signal transducers (Affolter et al., 2001). In accordance with its Dpp-dependent regulation, we identified a putative binding site for Mad(Kim et al., 1997) within IroRE^2-B (Fig. 4B;see Fig. S4 in the supplementary material).

Previous genetic evidence has shown repression of ara/caupexpression in the prospective medial notum by the GATA transcription factor Pannier (Pnr) (Calleja et al.,2000), which is a target of Dpp signalling in the wing disc(Sato and Saigo, 2000; Tomoyasu et al., 2000). Consistent with a direct regulation, in IroRE^2-B we also found nine putative GATA binding sites (Haenlin et al., 1997; Martin and Orkin,1990) (Fig. 4B; see Fig. S4 in the supplementary material).

A high proportion of the putative binding sites found in our analysis were located within conserved regions between D. melanogaster, D. pseudoobscura, D. virilis and D. mojavensis as shown by comparison with the VISTA program(Couronne et al., 2003)(Fig. 4A; see Fig. S4 in the supplementary material).

We identified putative binding sites for Pnt, Yan and Ttk in the IroRE^2-B sequence. Accordingly, we investigated the function of these transcription factors in the regulation of IroRE²-lacZ and ara/caup genes.

We first examined the function of pnt. Gain-of-function pnt clones (induced at 72-96 hours AEL) caused ectopic accumulation of β-gal in the wing pouch, wing hinge and proximal notum and enhanced accumulation (above the extant levels) at the lateral notum(Fig. 5A,A′, arrowheads).(The central hinge and wing pouch were somehow refractory to such activation when clones were induced at 48-72 hours AEL; not shown.) Consistently, loss of function of pnt (pnt^Δ88clones) reduced or totally removed lacZ expression driven by IroRE², especially in the central and posterior regions of the lateral notum (Fig. 5B,B′). Moreover, expression of ara/caup was similarly abolished in pnt clones(Fig. 5C,C′,arrowhead).

The expression of IroRE²-lacZ and ara/caupgenes was unaffected either in cells devoid of Yan (yan⁴⁴³or yan⁸⁸⁴ clones) or Ttk(ttk^1e11clones) (not shown). This suggested that Yan and Ttk do not regulate IroRE²-lacZ or ara/caupexpression. Thus, Pnt appears to be the main effector of the EGFR pathway in the regulation of Iro-C through IroRE².

Pnr is a GATA DNA binding protein that acts on its own as a transcriptional activator but switches to repressor when in a complex with the Friend of GATA(FOG) protein U-shaped (Ush) (Cubadda et al., 1997; Haenlin et al.,1997; Ramain et al.,1993). Since Pnr represses ara/caup in the proximal region of third instar wing discs (Calleja et al., 2000), we assayed the effect of pnr or ush loss-of-function on the transcription of IroRE²-lacZ. pnr^VX6 or ush^VX22 mutant clones, located at the most proximal region of the disc, exhibited ectopic lacZ expression(Fig. 6A,A′ and Fig. 7A,A′, arrows),indicating a repressor function of these proteins on transcription mediated by IroRE². Similarly, lack of pnr or ush caused ectopic ara/caup expression (Fig. 7A,A″, arrows) (Calleja et al., 2000). Consistent with a repressor role for Pnr/Ush, UAS-ush was able to repress expression of IroRE²-lacZ and ara/caup in the proximal most domain of the lateral notum (Fig. 7B-C′, arrows), most likely through the formation of Pnr/Ush heterodimers.

Surprisingly, the effect of lack of pnr on transcription mediated by IroRE² differed along the proximodistal axis of the disc. Thus,β-gal was absent from pnr^VX6 clones situated in a region marked by wg expression(Fig. 6B,B′, arrows). Since Pnr activates expression of wg whereas the heterodimer Pnr/Ush represses it (Calleja et al.,1996; Sato and Saigo,2000; Tomoyasu et al.,2000), expression of wg reveals cells containing Pnr but devoid of Ush. These data suggested that Pnr activates expression of IroRE²-lacZ in the absence of Ush and, indeed, activation of this reporter gene was observed in pnr overexpression clones(Fig. 6C,C′). On the contrary, overexpression of pnr inhibited ara/caupexpression in the notum and did not ectopically activate it (not shown). Thus,regulation of ara/caup and IroRE²-lacZ by Pnr appears to be slightly different, most likely due to the fact that additional regulatory elements, other than the IroRE², cooperate to establish the pattern of the endogenous ara/caup genes (see Discussion).

Considering that the expression of pnr and ush in partially overlapping domains in the medial region of the wing disc depends on Dpp activity (Sato and Saigo,2000; Tomoyasu et al.,2000), the above results suggest that Pnr/Ush dimers may mediate Iro-C repression by the Dpp pathway at the medial notum through IroRE². However, these observations do not rule out a possible direct repression of IroRE²-lacZ by Mad binding to its putative binding sites present in this RE. To address this point, we made use of the MARCM technique (Lee and Luo,1999) to overexpress UAS-tkv^QD in ush^VX22 clones. We reasoned that if the repressor effect of UAS-tkv^QD was solely due to activation of pnrand ush (Sato and Saigo,2000) and formation of the Pnr/Ush repressor heterodimer,repression should not occur in the absence of Ush. However, absence of ush did not relieve lacZ repression by the Dpp pathway(Fig. 7D). Reduced expression of IroRE²-lacZ was also found in UAS-Mad, ush^VX22 clones (not shown). These results suggest a role for Mad, in addition to Pnr/Ush, in Dpp-dependent repression through IroRE² in the medial notum (see below).

We also found that expression of IroRE²-lacZ was upregulated in tkv^a12 clones located at the wing hinge(Fig. 3C,C′). Interestingly, ush and the GATA family protein Grain (Grn) are expressed at the hinge region abutting the Iro-C expression domain(Brown and Castelli-Gair Hombria,2000; Haenlin et al.,1997). This opened up the possibility that an Ush/Grn dimer might mediate the Dpp-dependent Iro-C repression at the hinge. However, there was no upregulation of ara/caup or IroRE²-lacZexpression in ush or grn mutant clones at the hinge region(Fig. 7A-A″, arrowheads,and not shown), suggesting that this was not the case. Accordingly,elimination of Mad-dependent repression could account for the enhanced expression of lacZ found in tkv^a12 clones in the wing hinge. A similar situation may apply to the lateral notum(Fig. 3B,B′), a region devoid of Pnr and Ush, where Mad could be the Dpp pathway effector.

IroRE^2-B contains putative binding sites for effectors of the EGFR and Dpp pathways. To determine the in vivo functionality of these binding sites, we first set to define the minimal functional length of this RE by assaying the enhancer potential of five overlapping 500 bp subfragments. Three of them (IroRE^2-B1, IroRE^2-B2 and IroRE^2-B4) drove lacZ expression in the wing disc in overlapping regions contained within the Iro-expressing domain but also outside of it (Fig. 4D-F). Interestingly, putative ETS and GATA binding sites are not scattered within the IroRE^2-B sequence but appear clustered within these subfragments (Fig. 4B). Clustering of binding sites might contribute to synergistic regulation by the transcription factors (Arnosti et al.,1996). Removal of the 200 initial bp of the IroRE^2-B,which harbours several putative binding sites, did not affect its regulatory potential (not shown), indicating the dispensability of these sites. Furthermore, a lacZ construct containing exclusively the 200 initial bp of the IroRE^2-B was not expressed at the wing disc (not shown). In sum, these data pointed to the 240 bp sequence shared by IroRE^2-B1 and IroRE^2-B2 and the region covered by IroRE^2-B4 as the most relevant for the control of lacZexpression. Note that the former region harbours a highly conserved sequence among the Drosophila species analyzed(Fig. 4A, asterisk) and putative ETS and GATA binding sites (Fig. 4B).

Neither the IroRE^2-B2 fragment, which contains a Mad binding site, nor the IroRE^2-B1 and IroRE^2-B4 fragment, devoid of such a site, drove expression in the prospective medial notum(Fig. 4D-F). Since the three fragments, on the contrary, contain GATA binding sites, this suggests that the GATA protein Pnr should be the main repressor of Iro-C at the medial notum through the IroRE^2-B. A mutation was created in the unique GATA site of IroRE^2-B2 to assess the relative contribution of Pnr to the establishment of the proximal border of Iro-C expression. The resulting IroRE^2-B2GATAm-lacZ construct showed expansion of lacZ expression into the medial notum(Fig. 4G,G′), thus stressing the main contribution of GATA proteins. The potential contribution of Mad to IroRE^2-B2-lacZ regulation was assayed in tkv^a12 clones generated in flies harbouring the IroRE^2-B2GATAm-lacZ construct. Interestingly, further upregulation of lacZ expression was found in the clones(Fig. 4I-I″). These results, in agreement with those of Fig. 7D-D″, indicate that Mad, in addition to Pnr/Ush, should contribute to Dpp-dependent repression in the medial notum.

It is noteworthy that removal of the ETS binding site of IroRE^2-B2 that most closely matches the consensus strongly reduced lacZ expression in the wing disc, although it did not affect leg and eye disc expression (Fig. 4H,H′ and not shown). This indicates that EGFR signalling activates IroRE^2-B2-mediated transcription in the wing disc by using the conserved Pnt binding site.

Although the Dpp pathway represses Iro-C expression(Cavodeassi et al., 2002) and,that mediated by IroRE², this does not prevent the spatial coincidence, in the lateral notum of third instar discs, of Iro-C proteins and elevated levels of Dpp activity (Fig. 9C). Moreover, the overexpression of UAS-tkv^QDis unable to repress Iro-C expression in a domain of the lateral notum near the Dpp source (Cavodeassi et al.,2002). Interestingly, IroRE¹ drives lacZexpression in that domain (Fig. 1C,C′). Thus, we examined the effect of the Dpp pathway on IroRE¹-lacZ regulation. Overexpression of UAS-tkv^QD did not repress lacZ expression but instead activated it, although exclusively in a domain restricted to the lateral notum (Fig. 8A,A′, arrows). Accordingly, loss of Dpp signalling(tkv^a12 clones) significantly reduced β-gal accumulation (Fig. 8B,B′,arrows). IroRE¹-mediated transcription was also activated by the EGFR pathway, as shown by ectopic lacZ expression in UAS-Raf^Act clones (Fig. 8C,C′, arrows) and its converse reduction in UAS-Raf^DN clones (Fig. 8D,D′, arrow). Thus, Iro-RE¹ may be instrumental in allowing expression of Iro-C genes in the posterior lateral notum despite the repressor effect of the Dpp pathway through other regulatory elements.

A common theme in development is the convergence of different signalling pathways to implement a given developmental program. For instance during embryonic development, the antagonistic activity of the EGFR and Dpp pathways sets the limits between the neuroectoderm and the dorsal ectoderm(von Ohlen and Doe, 2000). A similar situation applies to the specification of prospective body regions within the wing imaginal disc. Here, during the early second instar, EGFR and Dpp pathways act antagonistically on the regulation of the Iro-C(Cavodeassi et al., 2002; Wang et al., 2000; Zecca and Struhl, 2002b)restricting its expression to the prospective notum region where it specifies notum development rather than hinge (Diez del Corral et al., 1999). Later, at the early third instar, again the concomitant activity of EGFR and Dpp signals (the latter now also emanating form the most proximal region of the wing disc) partition the prospective notum into two different subdomains, the medial and the lateral notum, the latter being specified by ara/caup expression(Cavodeassi et al., 2002; Diez del Corral et al., 1999). Thus, to understand how regionalization of the adult fly body is achieved it is important to elucidate the mechanisms responsible for the joint interpretation of both signalling pathways.

The opposing effects of the EGFR and Dpp pathways on Iro-C expression may have resulted from direct cross talk between these pathways. For instance,both in vertebrates and invertebrates the ability of EGFR to antagonize BMP signalling by phosphorylation and inhibition of Smad proteins has been reported (Kretzschmar et al.,1997; Kubota et al.,2000). Although not previously described, Dpp signalling may similarly interfere with the activity of the EGFR pathway. However, this appears not to be the case in the wing disc since expression of argos, a readout of the EGFR pathway(Rebay, 2002) can be detected in the proximal-most notum of the wing disc (not shown), a region with high levels of Dpp signal. Furthermore, overexpression of activated tkvdoes not repress expression of kekkon, another EGFR downstream gene(Rebay, 2002) (our unpublished results).

Here we show that the opposing effects of the EGFR and Dpp pathways on Iro-C expression result from the convergence of both pathways on at least two distinct Iro-C regulatory elements, IroRE¹ and IroRE². These two REs drive gene expression in two complementary domains of the prospective notum region of the wing disc, and appear to mediate most of the regulation of the Iro-C genes by the Dpp and EGFR pathways in this region of the wing disc. Furthermore, IroRE¹ provides a regulatory mechanism for the coexistence at the prospective lateral notum of Iro-C expression and Dpp pathway activity, notwithstanding the negative regulation of Iro-C by such pathway (Fig. 9).

The transcriptional regulation of the Iro-C genes is modular. Thus, the non-coding Iro-C genomic DNA contains a series of five separate enhancers that control the expression of a reporter gene in sub-domains within the realm of Iro-C expression in the prospective notum region of the wing disc. None of the identified fragments reproduces on its own the entire pattern of expression of Iro-C in the prospective notum. However, IroRE¹ and IroRE² promote expression in complementary domains that entirely cover the territory of the presumptive lateral notum. Furthermore,IroRE²-mediated transcription recapitulates expression of Iro-C at the whole prospective notum at the second larval instar. We hypothesize that the combined activity of both REs would be responsible for a great part of the regulation of Iro-C expression in the notum territory. Moreover, although IroRE³, IroRE⁴ and IroRE⁵ mediate lacZ expression in patterns only partly related to that of the Iro-C genes, these REs probably contribute to the complex regulation of the Iro-C. In addition, we cannot exclude the possibility of other RE(s) located outside the tested region that would help to establish the final pattern of Iro-C expression. Indeed, Iro^DFM3, a deficiency obtained by imprecise excision of the iro^rF209 P-lacZ element that extends up to the mirr promoter(Gomez-Skarmeta et al., 1996),maintains some lacZ expression in part of the central notum (not shown).

The identified REs might act simultaneously on ara and caup expression to give rise to their almost coincident patterns of expression. As previously proposed (R. Diez del Corral, PhD thesis,Universidad Autónoma de Madrid, Spain, 1998), such coincidence cannot be attributed to cross-regulation between ara and caup since in iro^rF209 mutant discs (iro^rF209 is an ara null allele, Fig. 1B) expression of caup is unmodified. Regulation of ara/caup would be, accordingly, similar to that of the achaete-scute genes of the AS-C, which show identical patterns of expression due to the use of shared enhancers (reviewed in Modolell and Campuzano, 1998). Expression of the vertebrate Iroquois (Irx) genes appears to be similarly regulated. Thus, the analysis of the regulatory potential of highly and ultra conserved non-coding regions present in the intergenic regions of the Irx clusters suggests these genes to be regulated by partially redundant enhancers shared by the components of each cluster(de la Calle-Mustienes et al.,2005).

Expression of mirr in the notum region of the wing disc largely coincides with that of ara/caup and most likely is under the control of the same REs. Thus, activity of the IroRE² may account for the unmodified expression of mirr in iro¹ imaginal discs (associated with an inversion breakpoint located within the caup transcription unit; Fig. 1B) (R. Diez del Corral, PhD thesis, Universidad Autónoma de Madrid, Spain, 1998). In addition, differences in the expression of ara/caup and mirr might be due to the presence of repressor RE(s) or insulator sequences (Gerasimova and Corces, 1996) that would prevent the action of the RE(s)controlling ara/caup on the mirr promoter. This is consistent with the previous observation of ectopic expression of mirr in Mob1 mutants, a regulatory mutation mapped within the Iro-C (Kehl et al.,1998).

Our identification of REs present in the Iro-C has allowed us to the unveil some of the molecular mechanisms of its transcriptional regulation at the level of DNA-protein interaction and to analyse the interplay of positive and negative inputs from convergent signalling pathways.

EGFR activation in the proximal region of the wing disc leads to expression of Iro-C (Wang et al., 2000; Zecca and Struhl, 2002a; Zecca and Struhl, 2002b). Here we demonstrate that both IroRE¹ and IroRE² mediate positive regulation by the EGFR pathway(Fig. 9A,B). We show that Pnt mediates activation of IroRE²-lacZ by the EGFR pathway. Furthermore, as previously shown for the Iro-C genes(Zecca and Struhl, 2002a; Zecca and Struhl, 2002b),EGFR-dependent activation is cell context dependent. This suggests the existence, in the cells receiving EGFR signalling, of presently unknown factors that would contribute to ara/caup activation and/or the presence of counteracting repressing mechanisms, which should prevent their activation. Clearly, the Dpp pathway is so far the best candidate, since it has been shown that it can repress Iro-C(Cavodeassi et al., 2002) and the IroRE²-lacZ transgene (this work).

The molecular mechanism of Dpp-dependent regulation of Iro-C expression appears to be more complex. The Dpp pathway can repress or activate Iro-C through different REs and different effector proteins. IroRE²appears to mediate Dpp-dependent repression at the medial notum (most probably through direct binding of the heterodimer Pnr/Ush and Mad) and at the hinge and lateral notum (independently of Pnr, Ush and the GATA factor Grn in these domains). Dpp-dependent repression of Iro-C may be mediated, in addition,through a different RE, namely, through a brk silencer element(brkSE), shown to mediate Dpp-dependent repression of brk by binding of a Medea/Mad/Schnurri repressor complex(Pyrowolakis et al., 2004),which is present at the Iro-C within IroRE⁵ (A.L. and S.C.,unpublished observations).

Despite the Dpp-mediated repression through IroRE², a high level of Iro-C proteins accumulates in the lateral region of the notum, near the strong source of Dpp at the AP border (Fig. 9C). Furthermore, in this region of the wing disc Iro-C expression is refractory to Dpp-dependent repression(Cavodeassi et al., 2002). It is noteworthy that, IroRE¹ mediates lacZ expression exclusively in that region of the wing disc and it appears to provide a regulatory mechanism for the co-existence of Iro-C expression and Dpp pathway activity, since the Dpp pathway does not repress but, on the contrary,activates IroRE¹-mediated lacZ expression(Fig. 9B). Activation is restricted to the lateral notum, most likely because of the presence, in the hinge and medial notum territories, of repressors [Muscle segment homeobox,Msh; also known as Drop (Villa-Cuesta and Modolell, 2005) and Pnr/Ush, respectively] that would counteract activation. Putative binding sites for both Msh [consensus sequence G/C TTAATTG (Kuzuoka et al.,1994)] and GATA proteins are indeed present in IroRE¹(not shown). Thus, IroRE¹ and IroRE² represent two different REs in the same gene that respond in opposite ways to the same positional information, i.e. Dpp signalling. A similar complex level of regulation has been found in the case of segmentation genes such as even skipped (eve) where the same transcription factor (Hunchback) is able to activate or repress transcription mediated through different eve position-specific enhancers(Small et al., 1992; Small et al., 1996).

In addition a Dpp-independent mechanism based in the mutual repression between Iro-C and the homeoprotein Msh helps to maintain the distal border of Iro-C expression (Villa-Cuesta and Modolell, 2005). This repression could be mediated by direct binding of Msh to one putative Msh binding site present in the Iro-RE^2-B sequence.

Further analysis of IroRE³, IroRE⁵, other putative RE(s) located outside of the tested region and putative repressor RE(s) that could not have been identified in our study, would help to establish the final pattern of Iro-C expression, which is essential for the specification of the notum and its subsequent patterning.

We are grateful to J. Modolell, F. Cavodeassi, J. Culí, J. F. de Celis, J. de Navascués, R. Diez del Corral, E. González-Pérez, J. L. Gómez-Skarmeta, M. Ruiz-Gómez, D. Shaye, S. Sotillos and E. Villa-Cuesta for constructive criticisms on the manuscript, to E. Caminero for her invaluable help in the generation of the transgenic lines, to N. Barrios for her help, to N. Cisneros for technical assistance, to F. Cavodeassi for Fig. 9C, to M. Affolter, P. Badenhorst, A. Baonza, and the Bloomington Stock Center for materials and stocks. A predoctoral fellowship from the MEC is acknowledged. This work was supported by grants from MCYT (BMC2002-00411) to J. Modolell and (MEC BFU2005-02888/BMC) to S. Campuzano. An institutional grant from Fundación Ramón Areces to the Centro de Biología Molecular Severo Ochoa is acknowledged.