Capicua DNA-binding sites are general response elements for RTK signaling in Drosophila

Receptor tyrosine kinase (RTK) signaling pathways control a broad spectrum of developmental decisions, including cell proliferation, differentiation, morphogenesis and survival (Schlessinger, 2000; Simon, 2000). Many RTK pathways signal through the conserved Ras/MAPK cassette, which then leads to phosphorylation of nuclear transcription factors and other cellular proteins. At the transcriptional level, RTK signals induce a wide variety of target gene responses in different contexts, but the molecular mechanisms underlying these responses are not well understood. In Drosophila, in vivo validated RTK effectors include the Ets factors Pointed and Yan (Simon, 2000; Tootle and Rebay, 2005), the HMG-box repressor Capicua (Cic) (Jiménez et al., 2000; Goff et al., 2001; Roch et al., 2002; Astigarraga et al., 2007; Tseng et al., 2007) and the Groucho (Gro) co-repressor (Hasson et al., 2005; Cinnamon et al., 2008; Cinnamon and Paroush, 2008; Jennings and Ish-Horowicz, 2008). Consequently, the analysis of these effectors can provide general insights into the regulatory mechanisms by which RTK signals control gene expression and development.

The Drosophila Torso RTK pathway represents an excellent model of transcriptional regulation in response to RTK activation (Furriols and Casanova, 2003). In this system, localized activation of the Torso receptor at each pole (termini) of the early blastoderm embryo controls the specification of terminal body structures by inducing the expression of two zygotic gap genes: tailless (tll) and huckebein (hkb) (Pignoni et al., 1990; Brönner and Jäckle, 1991). This induction involves a mechanism of derepression: both genes are normally repressed in medial regions of the embryo and the Torso signal relieves this repression at the poles (Liaw et al., 1995; Paroush et al., 1997; Jiménez et al., 2000). Repression of tll and hkb requires several nuclear factors, including Cic and Gro, which are both downregulated by the Torso signal (Paroush et al., 1997; Häder et al., 2000; Jiménez et al., 2000; Goff et al., 2001; Astigarraga et al., 2007; Cinnamon et al., 2008). Thus, loss of Cic or Gro function causes derepression of tll and hkb in medial regions of the embryo, which then leads to repression of central gap genes such as knirps (kni) and Krüppel (Kr) (Paroush et al., 1997; Jiménez et al., 2000; Goff et al., 2001; Löhr et al., 2009) (see Fig. S1 in the supplementary material). Conversely, mutations that render Cic or Gro insensitive to MAPK phosphorylation cause inappropriate repression of tll and hkb at the poles (Astigarraga et al., 2007; Cinnamon et al., 2008). Additionally, various studies have implicated other factors, such as GAGA/Trx-like, Dorsal, Retained (Retn; also known as Dead-ringer) or Tramtrack, in tll and/or hkb regulation (Liaw et al., 1995; Häder et al., 2000; Chen et al., 2002).

It is currently assumed that terminal gap genes contain complex enhancer regions that are bound by several, perhaps redundantly acting, transcription factors. However, how these activities converge to regulate Torso-dependent expression of tll or hkb is not understood. For example, analysis of a hkb enhancer indicated a role of Dorsal, Retn and Gro in Torso-mediated regulation of this enhancer (Häder et al., 2000). Cic is also required for hkb repression, but it has not yet been possible to demonstrate direct binding of Cic to hkb cis-regulatory regions (Jiménez et al., 2000). Recently, a DNA-binding motif for the Cic protein has been identified in humans (Kawamura-Saito et al., 2006), and it has been noted that this motif resembles a short regulatory element in the tll upstream region, the torso response element (tor-RE), which restricts tll expression to the posterior pole of the embryo (Liaw et al., 1995; Löhr et al., 2009). Consequently, it is possible that Cic represses hkb expression by binding to tor-RE-like elements, thus contributing to Torso-dependent regulation of this target.

Here, we report that tor-RE-like octameric sequences present in the hkb enhancer region function as binding sites for Cic and play a central role in the response of this target to Torso regulation. We also show that these Cic-binding motifs are essential for recruitment of the Gro co-repressor to hkb enhancer sequences in vivo. We then show that similar elements control the restricted expression of the intermediate neuroblasts defective (ind) gene in the neuroectodermal region of the embryo. This regulation occurs downstream of the EGFR RTK signaling pathway, indicating that Cic-binding sites function downstream of different RTK signals. Identical sites mediate Cic-dependent regulation of another EGFR target, argos, in the developing wing. Using synthetic enhancer constructs, we find that Cic octamers are sufficient to provide the regulatory information necessary to translate RTK signaling inputs into precise transcriptional responses in different tissues. We conclude that Cic octameric sites are general response elements for RTK signaling in Drosophila.

A GST-Cic^HMG expression construct was generated by amplifying a fragment encoding the Drosophila Cic HMG-box region (corresponding to residues 481-580) with primers hmg1 (5′ AAT GAA TTC CCG CAG CTG GGC AGC 3′) and hmg2 (5′ TAT CCC GGG TCC GCT CGC CTT TCC 3′), and subcloning the resulting fragment into pGEX-6P-2. This construct (pGEX-6P-2-Cic^HMG) is structurally equivalent to the pGEX6P-2-Cic-HMG construct from human Cic made by Kawamura-Saito et al. (Kawamura-Saito et al., 2006).

To generate hkb^0.4-lacZ, the hkb^0.4 fragment was amplified using primers hkb1 (5′ AAT GAA TTC ACG TTC GCT GGC CGA G 3′) and hkb2 (5′ GAA GGA TCC ATA AAA CGC GGT CCG 3′), digested with EcoRI and BamHI, and subcloned in EcoRI/BamHI-digested pCaSpeR-hs43-lacZ. hkb^0.4mut-lacZ was made similarly but using a pUC57-hkb^0.4mut plasmid template in which the two TGAATGAA sites had been mutated to CACACGCA by recombinant PCR.

hb-lacZ was generated by amplifying a 270 bp hb enhancer with primers hb1 (5′ ATG AAT TCG CTA GCT GCC TAC TCC 3′) and hb2 (5′ AAT GCG GCC GCA CGC GTC AAG GGA 3′) and digesting the resulting product with EcoRI and NotI for cloning into pCaSpeR-hs43-lacZ. hbC-lacZ was made by inserting two TGAATGAA sites as NotI-SpeI and SpeI-BamHI adaptors downstream of the hb sequence.

Bcd-lacZ was made by amplifying a synthetic array of four Bcd-binding sites separated by scrambled spacers (Hanes et al., 1994), digesting the PCR product with EcoRI and BamHI, and subcloning the resulting fragment in pCaSpeR-hs43-lacZ. To generate CBcdC-lacZ, we first joined a 45 bp module from hkb^0.4 containing two TGAATGAA sites with the above Bcd-binding site fragment using recombinant PCR. This fragment was subcloned upstream of a second copy of the above 45 bp element to create a CBcdC module, which was then inserted as an EcoRI-BamHI fragment into pCaSpeR-hs43-lacZ. CBcdC^TRE-lacZ and CBcdC^mut-lacZ were made similarly, using versions of the hkb 45 bp module mutated to TCAATGAA or CACACGCA, respectively.

ind^0.5-lacZ was created by amplifying the ind^0.5 fragment with primers ind1 (5′ AAT GAA TTC AAA CGT TTT GTT ATA ATC 3′) and ind2 (5′ GAA GGA TCC GGA AGA CAC TTC ATG 3′), and subcloning the resulting fragment in pUC57. The 0.5 kb ind^0.5 fragment was then recovered by digesting the pUC57-ind^0.5 plasmid with BamHI and (partially) with EcoRI, and ligated to EcoRI/BamHI-digested pCaSpeR-hs43-lacZ. ind^0.5mut-lacZ was made similarly using a pUC57-ind^0.5mut plasmid template in which the TGAATGAA sites had been mutated to CACACGCA by recombinant PCR.

argos^1.0-lacZ was generated using the argos^1.0 enhancer fragment amplified with primers argos1 (5′ ATG AAT TCG AGA TGA AAG TTT ATA G 3′) and argos2 (5′ CAT TTT CAC ACC TGA CTG CAG 3′), and subcloning the resulting fragment in T-overhang pUC57. argos^1.0 was then recovered as an EcoRI-BamHI fragment and subcloned into pC4PLZ. argos^1.0mut-lacZ was made similarly using the corresponding argos^1.0mut fragment carrying mutated Cic sites (CACACGCA).

CUASC-lacZ was made by first joining five tandem Gal4-binding sites to the 45 bp module from hkb^0.4 containing two Cic sites. This fragment was then inserted upstream of a second copy of the Cic-site module to create a CUASC enhancer, which was then subcloned as an EcoRI-BamHI fragment in pC4PLZ.

GST-HMG-box fusion proteins were expressed and purified as described previously (Paroush et al., 1994). In vitro binding assays were carried out as described by Kawamura-Saito et al. (Kawamura-Saito et al., 2006). Briefly, incubations were performed in a 15 μl volume containing 0.1-0.2 μg of GST-HMG-box protein, 10 mM Tris-HCl (pH 7.5), 50 mM NaCl, 1 mM DTT, 6% glycerol, 0.5% Triton-X100, 10 μg BSA, 1-2,5 μg poly(dI-dC) and 1 μg of single-stranded DNA. After 15 minutes of preincubation at 4°C, ~0.05 pmol of ³²P-labeled DNA probe was added and the incubation was continued for another 45 minutes at the same temperature. Reactions were resolved on 5% nondenaturing polyacrylamide gels at 4°C in 0.5× TBE.

The cic¹, cic², cic^fetE11 and tor⁴⁰²¹ alleles have been described before (Jiménez et al., 2000; Goff et al., 2001; Roch et al., 2002; Klinger et al., 1988). cic^ΔC2 embryos were obtained from transheterozygous females carrying two different cic^ΔC2 insertions (Astigarraga et al., 2007). Embryos devoid of maternal gro activity were obtained using the gro^MB36 allele (Jennings et al., 2008) in combination with the ovo^D-FLP-FRT system (Chou et al., 1993). Embryos lacking maternal Ras function, alone or in combination with cic, were generated similarly using the Ras^ΔC40b and cic^Q474X alleles (Tseng et al., 2007). dorsal (dl) mutant embryos were derived from dl¹/dl⁴ mothers (FlyBase). Other transgenic insertions and mutants used were cic-HA construct (Astigarraga et al., 2007), argos^w11 (Freeman et al., 1992), the rho^ve vn¹ combination (Diaz-Benjumea and García-Bellido, 1990), UAS-cic (Lam et al., 2006) and UAS-λtop (Queenan et al., 1997). Transgenic lines were obtained by standard P-element transformation and several independent lines were analyzed for each reporter construct.

Embryos were fixed in 4% formaldehyde-PBS-heptane for 20 minutes. In situ hybridizations were carried out using digoxigenin-UTP labeled antisense RNA probes, and anti-digoxygenin antibodies conjugated to alkaline phosphatase (Roche). Immunostainings were performed using the following primary antibodies: anti-dpErk (Cell Signaling; 1:50 dilution), anti-HA (12CA5, Roche; 1:400 dilution) and anti-β-galactosidase (40-1a, Developmental Studies Hybridoma Bank; 1:250 dilution). Signals were detected using secondary fluorochrome-conjugated antibodies (Molecular Probes). Embryos were mounted in Permount (in situ hybridizations) or Fluoromount-G (immunostainings). Wing discs were fixed in 4% paraformaldehyde-PBS for 20 minutes, processed for immunostaining using anti-HA and anti-β-galactosidase (anti-β-Gal) antibodies, and mounted in Fluoromount-G.

ChIP assays were performed using staged embryo collections from homozygous lines containing the hkb^0.4-lacZ or hkb^0.4mut-lacZ transgenes. Embryos were dechorionated in 100% bleach and subsequently fixed for 20 minutes in 10 ml crosslinking buffer (3% formaldehyde, 50 mM HEPES [pH 7.6], 1 mM EDTA, 0.5 mM EGTA, 100 mM NaCl) and 30 ml heptane. Crosslinking was stopped with 125 mM glycine. Crosslinked chromatin was sheared by sonication to an average size of 500 bp and immunoprecipitated using anti-Gro antibodies (two different rabbit polyclonal antisera raised against the N-terminal region of the protein). Control experiments using pre-immune serum or no antibody resulted in signals below 0.05% of input. Immunoprecipitated complexes were sequentially washed with low salt buffer [50 mM HEPES (pH 7.9), 1 mM EDTA, 1% Triton X-100, 0.1% SDS, 140 mM NaCl, 0.1% deoxycholate], high salt buffer [50 mM HEPES (pH 7.9), 1 mM EDTA, 1% Triton X-100, 0.1% SDS, 500 mM NaCl, 0.1% deoxycholate], LiCl buffer [20 mM Tris-HCl (pH 8.0), 1 mM EDTA, 250 mM LiCl, 0.5% deoxycholate, 0.5% NP-40] and TE. The chromatin was eluted with TE containing 1% SDS and 0.1 M NaHCO₃, and cross-linking was reversed by incubating at 65°C overnight. The resulting DNA was purified by chloroform extraction and ethanol precipitation, and quantified by qPCR using the FastStart SYBR Green Master Mix (Roche) on an Opticon Monitor 2 system (Bio-Rad). Three to five independent biological replicates (in which independent embryo collections were subjected to separate crosslinking and IP before separate qPCR) were analyzed for each amplicon. As reported in other Gro ChIP studies (Martinez and Arnosti, 2008), these replicates produced some variability reflected in the standard deviation (s.d.) of the data. Nevertheless, the results were highly consistent over multiple experiments using the two anti-Gro antibodies. A P-value was calculated by comparing hkb^0.4 with hkb^0.4mut for all data points from amplicons within hkb^0.4 (amplicons C-F) using a two-tailed t-test.

The human Cic protein binds the octameric sequence TGAATG(G/A)A (Kawamura-Saito et al., 2006). This element exhibits a single-nucleotide mismatch when compared with the core sequence of the tor-RE, TGCTCAATGAA (Liaw et al., 1995; Löhr et al., 2009). We assayed the ability of human and Drosophila Cic to bind to TGAATGAA and TCAATGAA sequences in gel-shift assays and observed similar interactions with both sites, indicating that T(G/C)AATGAA motifs are recognized by Drosophila Cic in vitro (see Fig. S2 in the supplementary material). We have also analyzed the role of Cic in repression of tll via the tor-RE. Using transgenes that contain tll enhancer sequences (Liaw et al., 1995), we provide evidence that Cic represses tll by binding to the tor-RE, and that this repression is inhibited by Torso signaling at the posterior pole (see Fig. S3 in the supplementary material).

We then searched for T(G/C)AATGAA motifs in the hkb upstream region and identified several TGAATGAA elements that are well conserved among Drosophila species (Fig. 1A; data not shown). Two such conserved sites are included in the hkb enhancer region identified by Häder et al. (Häder et al., 2000). To test whether Cic represses hkb through these motifs, we first defined a minimal hkb enhancer fragment that accurately reproduces the endogenous hkb pattern (Fig. 1A; data not shown). This 0.4 kb enhancer (designated hkb^0.4) directs highly restricted expression at both poles of the embryo (Fig. 1B). This pattern depends on Cic repression because it expands in cic¹ embryos lacking maternal cic function (Fig. 1C) (Jiménez et al., 2000). Conversely, cic^ΔC2 embryos expressing a Cic derivative insensitive to Torso-mediated downregulation (Astigarraga et al., 2007) show diminished hkb^0.4-lacZ expression at both poles (Fig. 1D). Mutagenesis of the two TGAATGAA sites in hkb^0.4-lacZ causes expanded reporter expression that resembles the pattern of hkb^0.4-lacZ in cic¹ embryos (Fig. 1E). We conclude that regulation of hkb expression requires direct binding of Cic to conserved TGAATGAA cis-acting octamers.

For comparison, we also analyzed hkb^0.4-lacZ expression in embryos devoid of maternal Gro function. Gro activity is essential for restricting tll and hkb expression to the embryonic poles, although the mechanism of Gro action in this context remains uncertain (Paroush et al., 1997; Jiménez et al., 2000; Häder et al., 2000; Cinnamon et al., 2008) (see below). As shown in Fig. 1F, there is significant hkb^0.4-lacZ derepression in gro^MB36 mutant embryos, similar to the effect seen in cic¹ embryos. Thus, both Cic and Gro play similar roles in repressing the hkb^0.4 enhancer. By contrast, embryos lacking Dorsal activity, another maternal regulator which functions as both an activator and repressor and is implicated in hkb regulation (Häder et al., 2000; Hong et al., 2008), displayed reduced hkb^0.4-lacZ expression at the posterior pole (Fig. 1G), indicating that Dorsal is required for activating hkb expression in posterior regions (see below).

Although binding of Cic to hkb^0.4 is essential for repressing this enhancer, the response to Torso regulation might involve additional factors bound to the enhancer. To address this issue, we asked whether Cic octamers are sufficient for Torso-dependent regulation of synthetic enhancers. We first tested whether Cic-binding sites (TGAATGAA) linked to a heterologous enhancer would make it responsive to Torso regulation. We selected a 270 bp promoter fragment from the hunchback (hb) gene, which normally drives intense staining in the anterior third of the embryo (construct hb-lacZ; Fig. 2A,B) (Struhl et al., 1989). Linking the same fragment to a single pair of Cic-binding motifs (construct hbC-lacZ) caused restricted expression from ~91 to 100% embryo length (EL; 0% being the posterior tip of the embryo; Fig. 2C). This pattern resembles the anterior domain of hkb expression and precisely corresponds to the area of Cic downregulation by the Torso pathway (Jiménez et al., 2000; Kim et al., 2010). Furthermore, this pattern depends on Cic because it expands posteriorly in cic¹ embryos (Fig. 2D). Thus, the addition of Cic repressor sites confers Torso-dependent expression to the hb enhancer.

The hb enhancer is activated by the anteriorly expressed Bicoid (Bcd) factor (Struhl et al., 1989; Driever and Nüsslein-Volhard, 1989). Therefore, we tested whether a simple combination of Bcd and Cic sites would also respond to Torso regulation. A construct containing four multimerized Bcd sites drive anterior expression from ~73 to 100% EL (construct Bcd-lacZ; Fig. 2A,E). By contrast, a transgene in which the Bcd sites are flanked by two Cic sites on either side is expressed in a restricted pattern from 92 to 100% EL (construct CBcdC-lacZ; Fig. 2A,F). In cic¹ embryos, CBcdC-lacZ expression expands posteriorly up to ~74% EL (Fig. 2G), whereas it almost disappears in cic^ΔC2 embryos (Fig. 2H). A similar construct containing TCAATGAA sites corresponding to the tor-RE (CBcdC^TRE-lacZ) also showed highly restricted expression in the Torso signaling domain (Fig. 2I). Finally, mutation of the four Cic sites to CACACGCA caused derepressed reporter expression similar to the Bcd-lacZ pattern (construct CBcdC^mut-lacZ; Fig. 2J). These results indicate that Cic repressor sites combined with Bcd activator sequences are sufficient to provide a direct highly localized readout of Torso signaling activity at the anterior pole.

Because the Gro co-repressor does not bind DNA, it is believed to be recruited to terminal enhancers by one or more DNA-bound repressors (Paroush et al., 1997; Häder et al., 2000; Jiménez et al., 2000; Cinnamon et al., 2008; Jennings and Ish-Horowicz, 2008). We and others have proposed different mechanisms by which Gro could interact with terminal repressors such as Dorsal, Retn or Cic to silence tll and hkb expression (Häder et al., 2000; Jiménez et al., 2000). Given that hkb^0.4-lacZ expression depends on both Gro activity and intact Cic regulatory sites (Fig. 1), we analyzed whether such sites are required for recruitment of Gro to the hkb^0.4 enhancer. To this end, we first monitored association of Gro to the hkb^0.4-lacZ transgene by chromatin immunoprecipitation (ChIP) assays using anti-Gro antibodies and qPCR. These experiments were performed using staged embryo collections (90-180 minutes after egg laying) carrying two copies of hkb^0.4-lacZ. We designed a set of amplicons that span the hkb^0.4 enhancer and the flanking sequences present in the reporter construct (Fig. 3A). Some of these amplicons (A, B, G, H and I) are specific for the reporter and do not amplify endogenous genomic sequences, whereas amplicons C-F potentially amplify both the homozygous transgenic and endogenous hkb^0.4 enhancers. As shown in Fig. 3B, we found association of Gro with most of the intact hkb^0.4 enhancer (blue bars for amplicons C-F), but not with regions flanking the enhancer (amplicons A, B, G and I), although a small peak is observed at the transcriptional start site (amplicon H). Interestingly, within the enhancer, Gro levels were somewhat higher upstream of the Cic sites (amplicons D and E). This upstream region includes binding sites for Dorsal and Retn (Fig. 3A), two factors that have been implicated in hkb regulation and are known to bind Gro directly (Dubnicoff et al., 1997; Valentine et al., 1998; Häder et al., 2000).

We then used the same approach to assay binding of Gro to the hkb^0.4mut enhancer containing mutant Cic sites. In this case, Gro was detected at significantly lower levels compared with the wild-type enhancer (red bars in Fig. 3B). This decrease is observed throughout the enhancer, including the region upstream of the mutant Cic sites. Averaging across the four amplicons within the enhancer, we find that mutagenesis of the Cic sites reduces Gro association with the enhancer by 4.5-fold (P<0.01). As amplicons C-F should detect signals derived from both transgenic and endogenous enhancers, the larger than twofold decrease observed for the mutant sample suggest a higher efficiency of Gro immunoprecipitation from transgenic versus endogenous hkb^0.4 sequences, perhaps owing to a more open conformation of the transgene. Similar data were obtained using two different anti-Gro antibodies (data not shown). Taken together, these results indicate that binding of Cic to specific sites in hkb^0.4 is essential for recruitment of Gro to this enhancer.

We searched for potential Cic-binding sites in genes that might be targets of other RTK signaling pathways. One gene identified in these analyses, ind, functions as a target of the EGFR RTK pathway in the neuroectodermal region of the embryo (Skeath, 1998; Weiss et al., 1998; von Ohlen and Doe, 2000; Hong et al., 2008). ind expression begins at mid stage 5, forming sharp four- or five-cell wide longitudinal stripes on either side of the embryo (Weiss et al., 1998) (Fig. 4C). This pattern requires activator inputs from the Dorsal morphogen, as well as from the EGFR pathway, which is active in lateral domains overlapping the ind stripes (Skeath, 1998; von Ohlen and Doe, 2000; Hong et al., 2008) (Fig. 4B,D). The ind stripes are limited ventrally by the Ventral neuroblasts defective (Vnd) repressor expressed in the ventral neuroectoderm, and dorsally by the dorsal limit of the EGFR signaling domain (Weiss et al., 1998; von Ohlen and Doe, 2000). Previous analyses of ind regulation have identified a repressor element (the A-box motif) that controls the dorsal limit of ind expression through an unknown factor (Stathopoulos and Levine, 2005). We noted that the A-box sequence (WTTCATTCATRA) matches the complementary sequence of the Cic-binding motif. This, together with the requirement of EGFR signaling for ind expression, prompted us to study the role of Cic in ind regulation.

We reasoned that activation of ind expression by the EGFR pathway could involve downregulation of Cic protein in the lateral ectoderm. Indeed, EGFR activity in the neuroectoderm (as visualized by immunostaining against double-phosphorylated Erk/MAPK) precisely correlates with a sharp decline in Cic protein levels in this region (Fig. 4B). This suggests that EGFR signaling controls the dorsal limit of ind expression by defining a corresponding limit of Cic downregulation. We then monitored ind expression in cic¹ embryos and observed altered ind stripes that appeared both dorsally expanded at the anterior and retracted from the posterior (Fig. 4E). In embryos derived from cic¹/cic² females, which contain even lower Cic activity (Roch et al., 2002), the ind stripes are shorter and further expanded towards the dorsal side, resulting in eight- to 10-cell wide staining at late stage 5 (Fig. 4F). Thus, Cic has two effects on ind expression: it defines the dorsal limit of the ind stripes and maintains their expression in the abdominal region. The latter effect is reminiscent of the indirect positive role of Cic on kni and Kr expression (see Fig. S1 in the supplementary material), and it could thus reflect repression of ind by the terminal gap genes extending from the posterior pole in the absence of Cic. Consistent with this idea, we fail to detect ind expression in torso⁴⁰²¹ gain-of-function mutant embryos where the terminal gap genes are severely derepressed (Fig. 4G) (Klinger et al., 1988; Brönner and Jäckle, 1991).

To test whether Cic represses ind through A-box elements, we generated a lacZ reporter driven by a 0.5 kb ind enhancer fragment encompassing two A-box motifs and a Dorsal-binding site (Fig. 4A) (Stathopoulos and Levine, 2005). This reporter (ind^0.5-lacZ) forms weak, discontinuous lateral stripes of expression in late blastoderm embryos (Fig. 4H). As in the case of endogenous ind stripes, the ind^0.5-lacZ stripes are abolished in embryos lacking EGFR/Ras signaling activity (derived from females carrying Ras^ΔC40b germ-line clones; Fig. 4D,I). In cic¹ embryos, the ind^0.5-lacZ stripes are more uniform and expand by three to five cell diameters dorsally (Fig. 4J), whereas cic¹/cic² embryos show ectopic expression almost up to the dorsal midline (Fig. 4K). Additionally, this expanded ind^0.5-lacZ expression persists in embryos derived from Ras^ΔC40b cic^Q474X double mutant germline clones (Fig. 4L), implying that EGFR/Ras signaling normally induces ind^0.5-lacZ expression by downregulating Cic. Finally, mutation of the two A-box motifs in ind^0.5 caused derepression throughout lateral and dorsal regions of the embryo (Fig. 4M), and this pattern was unaffected in embryos lacking EGFR/Ras activity (Fig. 4N), consistent with removal of Cic being sufficient for EGFR-dependent induction of the ind^0.5 enhancer. Together, these results indicate that the A-box motifs in ind are binding sites for Cic protein that respond to EGFR regulation via Cic derepression.

Previous analyses showed that Cic behaves as a repressor of argos expression in the wing imaginal disc (Roch et al., 2002). argos is an EGFR signaling target that encodes a feedback inhibitor of this pathway (Freeman et al., 1992; Golembo et al., 1996). During wing development, EGFR activity defines the position of wing veins and leads to downregulation of Cic in presumptive vein cells, particularly in two rows of cells running along the future wing margin and in prospective veins L3, L4 and L5 (Fig. 5A) (Roch et al., 2002). This pattern of Cic downregulation is markedly complementary to the expression of argos, as visualized with the argos^w11 enhancer trap reporter (compare Fig. 5A with 5B) (Gabay et al., 1997). Reduced Cic function in cic²/cic^fetE11 discs causes argos^w11 derepression in intervein cells at levels similar to those of endogenous wing margin and L5 stripes (Fig. 5C; see also Fig. S4 in the supplementary material) (Roch et al., 2002). This ectopic expression is weaker than in stripes L3 and L4, suggesting that these stripes are subject, at least in part, to Cic-independent regulation, an idea supported by the relatively normal development of the L3-L4 intervein region in cic mutant adults (Fig. S4 in the supplementary material). We also analyzed argos^w11 expression in discs lacking both Cic and EGFR signaling activities, using the cic²/cic^fetE11 background in combination with rhomboid (rho) and vein (vn) alleles that eliminate EGFR signaling in the wing disc (Martín-Blanco et al., 1999). This caused generalized argos^w11 expression throughout the wing pouch without enhancement in stripes L3 and L4 (Fig. 5D), suggesting that EGFR signaling induces argos^w11 expression in prospective veins by relieving Cic repression, and that an additional EGFR-dependent input reinforces this expression in stripes L3 and L4.

To investigate whether Cic represses argos directly through Cic octameric sites, we first identified several conserved TGAATG(G/A)A motifs within the first intron of argos (Fig. 5E; data not shown). Next, we selected a 1.0 kb intron fragment containing five such sites (four TGAATGAA and one TGAATGGA motifs) and other conserved sequences. When placed upstream of a lacZ reporter, this fragment (designated argos^1.0) directs restricted expression in presumptive veins L3 and L4 (Fig. 5E,F), indicating that it mediates partial aspects of argos regulation. By contrast, the same fragment carrying mutated Cic sites drives widespread expression in the wing pouch and peripheral regions of the disc (Fig. 5E,G; data not shown). Thus, conserved Cic-binding sites in argos restrict its expression to prospective wing vein cells of the disc.

To test whether Cic-binding sites are sufficient to mediate EGFR-dependent regulation in the wing, we assayed an artificial enhancer containing five GAL4-binding sites flanked on either side by two tandem TGAATGAA motifs (construct CUASC-lacZ; Fig. 5E). Indeed, inducing ubiquitous GAL4 expression under the control of the tubulin or hsp-70 promoters leads to localized activation of the CUASC enhancer in prospective veins (Fig. 5H and data not shown). This restricted pattern depends on Cic, as it becomes significantly derepressed in cic mutant discs (Fig. 5I). We also monitored CUASC-lacZ expression driven by the C5-GAL4 line, which is active in the presumptive wing pouch (Fig. 5J) (Yeh et al., 1995). As shown in Fig. 5K, C5-GAL4 activates CUASC-lacZ expression only in presumptive vein cells of the wing pouch. Co-expression of CUASC-lacZ and Cic [using an UAS-cic construct (Lam et al., 2006)] with the same driver resulted in loss of lacZ expression in presumptive vein L5 (Fig. 5L), which correlated with loss of vein L5 in adult wings (Fig. S4 in the supplementary material). Conversely, co-expressing CUASC-lacZ together with UAS-λtop, which encodes a constitutively active form of EGFR (Queenan et al., 1997), caused severe lacZ derepression throughout the presumptive wing pouch (Fig. 5M; see also Fig. S4 in the supplementary material). This pattern recapitulates C5-GAL4-mediated activation of the standard UAS enhancer lacking Cic sites (Fig. 5J), and is therefore consistent with generalized downregulation of Cic in the C5-GAL4>UAS-λtop background. Thus, our results indicate that EGFR signaling controls argos expression through octameric Cic sites, and that such sites are sufficient to define a complex pattern of EGFR-mediated activation in the developing wing.

RTK signaling pathways play key functions in metazoan development, but the molecular mechanisms underlying RTK-initiated responses are not well understood. Until recently, it was generally assumed that Pointed and Yan were the only nuclear effectors of all RTK pathways in the fly [see, for example, Simon (Simon, 2000)]. However, several studies have identified the Cic repressor as an important sensor of some of these pathways (Jiménez et al., 2000; Goff et al., 2001; Astigarraga et al., 2007; Tseng et al., 2007). Here, we have shown that Cic regulatory functions downstream of Torso and EGFR signals depend on common TGAATGAA DNA octamers and that, at least in certain assays, these octamers are sufficient to induce localized RTK responses in vivo.

Our results show that regulation of hkb expression in response to Torso signaling crucially depends on conserved TGAATGAA elements recognized by Cic (Fig. 1). We also find that these elements combined with Bcd activator sequences are sufficient to establish localized reporter expression in the anterior pole of the embryo (Fig. 2). It thus appears that binding of Cic to specific sites in hkb is the key step for delimiting hkb expression in response to Torso activation. Therefore, although we cannot rule out that other (possibly redundant) Torso-dependent factors contribute to hkb regulation, we propose that this regulation largely depends on broadly distributed activators such as Bcd, Dorsal and Lilliputian (Reuter and Leptin, 1994; Häder et al., 2000; Tang et al., 2001) (Fig. 1G), and localized Cic repression.

We also find that association of Gro to the hkb enhancer requires the presence of intact Cic octamers in the enhancer. How does this association occur? Although Cic and Gro proteins interact in vitro, we have not yet demonstrated a direct correlation between such binding and Cic repressor activity in vivo (Jiménez et al., 2000; Astigarraga et al., 2007) (C.N. and G.J., unpublished). Our finding that Gro associates with sequences containing Dorsal and Retn sites is consistent with a role of these factors in recruiting Gro to the hkb enhancer, possibly through cooperative interactions with Cic. However, mutations in dorsal or retn do not cause clear derepression of hkb^0.4-lacZ or hkb expression (Fig. 1G) (Häder et al., 2000). It is also possible that local recruitment of Gro by Cic results in subsequent spreading of the co-repressor along the entire hkb^0.4 enhancer, a mechanism that may involve oligomerization of Gro and binding to hypoacetylated histones (Courey and Jia, 2001; Song et al., 2004; Martinez and Arnosti, 2008).

Our results indicate that patterning of the dorsal-ventral (DV) embryonic axis requires a mechanism of EGFR-mediated derepression that is similar to the role of Torso signaling in the anterior-posterior (AP) terminal system. In both cases, a local source of RTK activation downregulates the Cic repressor, thus inducing expression of Cic targets in restricted patterns (Fig. 6A). During DV patterning, the Dorsal morphogen activates the expression of several targets in ventral and lateral regions of the embryo, and it is believed that decreasing amounts of Dorsal protein help establish the dorsal limits of those expression domains. However, Dorsal nuclear levels appear rather uniform across the ind expression domain (Kanodia et al., 2009; Liberman et al., 2009), indicating that other mechanisms define the dorsal limit of ind expression. Indeed, previous studies have shown that EGFR signaling plays a key role in setting this border (Weiss et al., 1998; von Ohlen and Doe, 2000), and suggested the existence of unknown repressors restricting ind expression in dorsal regions (Stathopoulos and Levine, 2005). Our results indicate that these two events are linked through a mechanism of EGFR-mediated downregulation of Cic repressor activity.

During wing vein specification, there is a precise correlation between EGFR/MAPK signaling, Cic downregulation and argos transcription in prospective wing vein cells (Fig. 5A,B) (Gabay et al., 1997). Furthermore, our data show that Cic represses argos directly (Fig. 5E-G), and that Cic octamers alone are sufficient to interpret the EGFR activation signal to produce an argos-like response (Fig. 5H-M, Fig. 6B). However, the CUASC-lacZ reporter does not recapitulate all aspects of argos transcription, because only endogenous argos shows elevated expression in presumptive veins L3 and L4. This difference probably depends on localized determinants that regulate gene expression in the L3-L4 region (Blair, 2007), and do not affect CUASC-lacZ. Still, argos regulation during wing development appears largely dependent on EGFR-mediated downregulation of Cic as well as on positive input(s) by localized or ubiquitous activators, which may include the Osa/Eyelid factor (Terriente-Félix and de Celis, 2009). In addition, both loss- and gain-of-function experiments show strong correlation between Cic-dependent activity through TGAATGAA elements and differentiation of wing veins in the adult (Fig. 5 and see Fig. S4 in the supplementary material) (Goff et al., 2001; Roch et al., 2002), suggesting that Cic is an important sensor of EGFR signaling in this system. Cic probably controls additional EGFR targets involved in wing vein specification and other EGFR-regulated processes such as cell proliferation in imaginal discs (Tseng et al., 2007). Future studies will probably reveal new roles of Cic and its binding sites downstream of RTK signaling cascades.

In summary, Cic regulates multiple RTK signaling responses by binding to conserved octameric sites in target enhancers, indicating that conservation between these RTK pathways extends to specific response elements in cis-regulatory regions. Notably, these octamers are sufficient to translate RTK signaling inputs into localized transcriptional responses in different tissues: RTK signals produce complementary gradients (Torso) or boundaries (EGFR) of Cic downregulation that are then translated into complementary patterns of target gene expression through relief of Cic repression. This mechanism represents a particular case of ‘default repression’, a general strategy of developmental control whereby target genes induced by signaling pathways are maintained repressed in the absence of signaling (Barolo and Posaknony, 2002). For example, a similar derepression switch occurs during TGF-β/Dpp-mediated induction of optomotor-blind transcription via relief of Brinker repression (Sivasankaran et al., 2000; Barolo and Posaknony, 2002).

Finally, the human Cic protein binds octameric sequences related to those characterized here (Kawamura-Saito et al., 2006). The best-characterized targets of Cic in human cells are ETS genes of the pea3 family (Kawamura-Saito et al., 2006), which are known to respond to FGF RTK activation in different vertebrate systems (e.g. Roehl and Nüsslein-Volhard, 2001; Raible and Brand, 2001). Therefore, it will be interesting to ascertain whether Cic octamers also mediate RTK responses in those systems.

We thank A. Olza for assistance with Drosophila injections, L. Bardia for support with confocal analyses, I. Becam, J. Bernués, M. Martínez-Balbás, M. Mannervik, M. Milán, F. Roch and S. Shvartsman for scientific advice, and J. Botas, J. Casanova, M. Grillo, I. Hariharan, B. Jennings, T. Nakamura, S. Hanes, T. Schüpbach, F. Serras and the Bloomington Drosophila Research Center for reagents and fly stocks. This work was funded by grants from the Spanish Ministerio de Ciencia e Innovación (BFU2005-02673 and BFU2008-01875/BMC to G.J.), the Generalitat de Catalunya (2009SGR-1075 to G.J.), the National Institutes of Health (GM44522 to A.J.C.), the Israel Science Foundation (Center of Excellence 180/09 to Z.P.) and the Król Charitable Foundation (to Z.P.). G.J. is an ICREA Investigator. Deposited in PMC for release after 12 months.