Inverse regulation of target genes at the brink of the BMP morphogen activity gradient

During development of the Drosophila wing imaginal disc, Dpp, a member of the BMP superfamily, spreads in a graded manner from its expression domain along the A–P compartment boundary. Dpp functions together with another BMP family member, Glass bottom boat (Gbb) that is uniformly distributed across the A–P patterning field (Haerry et al., 1998). Consequently, a medial to lateral gradient of BMP signalling is generated that is ultimately responsible for proper patterning and growth of Drosophila adult wing (Entchev et al., 2000; Lecuit et al., 1996; Nellen et al., 1996; Schwank et al., 2008; Teleman and Cohen, 2000; Wartlick et al., 2011). The BMP activity gradient leads to differential activation of the Punt–Thickveins (Tkv) kinase receptor complex, which propagates the signal via phosphorylation of Mothers against dpp (Mad) (Kim et al., 1997; Newfeld et al., 1997). Phosphorylated Mad (pMad) along with its binding partner Medea (Med) translocates to the nucleus and regulates transcription. In fact, most BMP-induced target genes are not directly regulated by the pMad-Med complex, but are indirectly induced through repression of a transcriptional repressor Brk (Campbell and Tomlinson, 1999; Jaźwińska et al., 1999; Minami et al., 1999). At the molecular level, upon BMP signalling, the pMad–Med complex binds to silencer elements in the brk locus and recruit the zinc-finger protein Schnurri (Shn), which confers repressive activity to the complex (Marty et al., 2000; Torres-Vazquez et al., 2000). As a result, an inverse, i.e. lateral to medial, nuclear gradient of Brk emerges, which in turn negatively regulates the expression of different BMP targets above distinct threshold concentrations (Müller et al., 2003). Brk represses BMP target genes via direct binding to the sequence GGCGYY in their regulatory regions. These targets include spalt (sal), optomotor-blind (omb) and vestigial (vg) (Campbell and Tomlinson, 1999; Jaźwińska et al., 1999; Kim et al., 1997; Minami et al., 1999), and have nested expression domains centred around the high point of the BMP activity gradient at the centre of the wing disc. This nested expression pattern, which forms the basis of the threshold model of BMP/Brk target gene regulation, is explained by a differential sensitivity to Brk, with sal being repressed by very low levels of Brk, while repression of vg requiring higher levels. However, the fact that a high threshold BMP target such as sal is also expressed in lateral regions (Fig. 1A), where Brk is at its maximal levels, prompted us to re-examine the morphogen gradient model. Unexpectedly, our data revealed that in lateral regions of the wing disc both sal and vg are negatively regulated by BMP signalling and positively by Brk. By studying the regulation of sal expression in different regions of the wing disc, we identify a new mechanism where Brk induces the expression of a classic BMP target through an enhancer that contains neither Brk nor pMad-Med-Shn complex binding sites. Brk appears to induce the lateral expression of sal indirectly through repression of a negative regulator. Our analysis provides a working model explaining how the activities and mutual interactions of pMad, Brk and the new negative regulator of sal (NRS) differentially and inversely regulate sal expression along the anterior-posterior axis of the wing disc. Based on these position-specific, distinct and inverse outcomes of BMP/Brk-dependent patterning system, we subdivide the wing disc along the A–P axis into four regions (Fig. 1A).

Spreading of Dpp into both compartments establishes, with the help of uniformly expressed Gbb, a gradient of BMP signalling a long the A–P axis with peak levels in the centre of the wing pouch that exponentially decline towards the periphery. Consequently, the BMP activity, as reflected in the pMad concentration, continually decreases from the centre of the wing pouch establishing a parallel exponentially decaying gradient (Bollenbach et al., 2008; Entchev et al., 2000; Teleman and Cohen, 2000). It is proposed that BMP signal can attribute precise positional information only within the range of the wing pouch (Bollenbach et al., 2008). Consequently, at the periphery of the wing disc, expression of the target genes, like sal, was attributed to BMP-independent regulation (Campbell and Tomlinson, 1999; de Celis et al., 1999).

We wished to examine whether BMP activity is indeed inconsequential at the periphery as proposed. We therefore generated clones of cells mutant for the BMP type I receptor, tkv (which is required for signal transduction of both Dpp and Gbb) (Bangi and Wharton, 2006; Khalsa et al., 1998) and monitored Sal expression across the entire patterning field. Consistent with previous studies, we found that Sal expression was abolished in the mutant clones located in the centre of the wing pouch (region I) (Fig. 1C, white arrowheads). Strikingly, however, Sal expression was ectopically upregulated in lateral regions of the wing pouch (region III) and even in the periphery of the wing disc (region IV) (Fig. 1C, yellow arrowheads). As another read-out for BMP signalling, we monitored Brk expression in the same tkv mutant clones, as BMP signalling activates genes by attenuating the brk repressor. Although low, upregulation of Brk expression was apparent in many of the tkv mutant clones, located at the periphery of the wing disc (region IV) (Fig. 1D, yellow arrowheads), indicating that BMP signalling is active all across the patterning field including at the periphery.

Similar results were obtained when we overexpressed Dad (Fig. 1E–G), an inhibitory Smad that interferes with BMP signal transduction at the level of Mad (Tsuneizumi et al., 1997). These results show that BMP signalling exerts opposite influence on sal expression in a region-specific fashion. In the medial regions of the wing disc BMP signalling positively regulates sal expression, whereas in lateral regions, where the BMP activity gradient declines (regions III & IV); it appears to actively repress Sal expression apparently via pMad activity.

It is important to note that both ligands Gbb and Dpp signal via the BMP downstream machinery components including the Tkv receptor and RSMAD family of transcription factor(s) (Bangi and Wharton, 2006; Khalsa et al., 1998). Thus, our experiments cannot distinguish between the relative contributions of Gbb and Dpp to BMP signalling and hence to regulation of target genes.

sal expression is regulated not only by Brk repression but also through direct activation by pMad-Med complex (Moser and Campbell, 2005). In order to reassess the effect of activation of BMP-dependent pMAD signalling on sal expression, we generated clones overexpressing Tkv^QD, a constitutively active receptor that phosphorylates Mad in a ligand-independent manner (Lecuit et al., 1996; Nellen et al., 1996). We found, as expected, that Sal expression was ectopically upregulated all over the wing pouch (regions I, II and III; Fig. 2C,E,I). However, in most of Tkv^QD overexpressing clones located in the periphery of the wing disc (region IV) the normal expression of Sal was strongly repressed (Fig. 2C,E), similar to Brk expression (Fig. 2K,L). By contrast, the expression of nab, a newly identified positive BMP target in the wing disc (Ziv et al., 2009), was ectopically upregulated in the same lateral clones (Fig. 2F) where Sal expression was repressed (Fig. 2E). Thus, in lateral regions of the wing disc (as opposed to medial regions) the pMad-Med complex appears to actively repress sal expression. It should be noted, however, that at times, Sal expression was either not reduced or even upregulated (Fig. 2A–C,K) in Tkv^QD overexpressing clones located in region IV (see below).

The unexpected observation that loss of tkv in lateral regions of the wing disc (region III and IV) resulted in upregulation of both brk and sal raises the possibility that in these regions sal, like brk, is directly regulated by the pMad-Med-Shn repression complex. Alternatively, sal expression could be positively regulated by Brk. To address this, we generated loss of function (LOF) clones of brk and specifically compared the accumulation of Sal in clones in different regions of the wing disc. This strategy allowed us to assess the influence of Brk on sal expression without interfering with BMP signalling at the level of Mad phosphorylation. As expected, near the centre of the wing pouch (region II), complete removal of Brk function, resulted in ectopic expression of Sal (Fig. 3C, white arrowheads). Surprisingly, however, ectopic upregulation of Sal is not observed in brk clones located more laterally (region III) (Fig. 3F). Even more strikingly, Sal expression in lateral regions, where it normally coincides with high levels of Brk (region IV), was lost in brk mutant clones (Fig. 3C,F,I, yellow arrowheads). In contrast to sal, nab expression was ectopically upregulated even in the lateral regions upon removal of brk (Fig. 3H, yellow arrowheads). These results suggest that the lateral expression of sal is not regulated directly by the pMad-Med-Shn repression complex. More importantly, while Brk activity restricts the expression of positive BMP target genes (Campbell and Tomlinson, 1999; Jaźwińska et al., 1999; Minami et al., 1999), it appears to be required for maintaining sal expression in lateral regions of the wing disc.

The loss of function analysis suggested that in different regions of the wing disc, Brk regulates sal expression in a qualitatively distinct fashion. We thus wondered if overexpression of Brk across the wing disc would also have distinct and reciprocal outcomes. To test the idea, we generated Brk overexpressing clones and specifically compared sal expression in clones situated in different regions of the wing disc. As previously reported, Sal expression in the centre of the wing pouch was abolished in such clones (Fig. 3L, white arrowhead). Importantly, in lateral regions of the wing pouch (region III), overexpression of Brk resulted in ectopic expression of Sal (Fig. 3L, yellow arrowheads). In summary, these results demonstrate that in lateral regions of the wing disc (regions III and IV) Brk positively regulates sal expression.

Our results contradict the notion that Brk functions as a default repressor of BMP targets. Rather the results presented above suggest that Brk may drive lateral expression of sal indirectly by inhibiting a negative regulator of sal expression. Alternatively, Brk may act as a direct activator of sal in lateral regions of the wing disc. To distinguish between these two possibilities it was necessary to decipher how Brk actually regulates transcription of sal. To analyse this, we employed a Brk-VP16 chimeric fusion protein, in which the repression domain of the Brk repressor is replaced with the activation domain of the herpes simplex virus protein VP16 (Weiss et al., 2010). This recombinant protein is expected to display Brk's cognate DNA binding specificity. However, instead of repressing Brk target genes, it should be able to activate them. If in the lateral regions of the wing disc, endogenous Brk acts directly as an activator of sal, then overexpression of Brk-VP16 should also activate sal expression in lateral regions of the wing pouch (region III), similar to overexpression of Brk. By contrast, if in lateral regions Brk activates sal indirectly, via repression of another negative regulator, then Brk-VP16 will upregulate that negative regulator, ultimately leading to downregulation of sal expression (in region IV). As expected, in medial regions of the wing disc (region II) Brk-VP16 was able to ectopically upregulate Sal expression (Fig. 3O, white arrowhead). In lateral regions, however, overexpression of Brk-VP16 resulted in downregulation of Sal (Fig. 3O, yellow arrowheads), indicating that in order to activate sal expression in lateral regions of the wing disc, Brk acts as a repressor: It downregulates a (yet unidentified) negative regulator of sal (which we refer to as NRS hereafter), and in an indirect manner de-represses sal expression.

Previous studies have identified several independent regulatory regions which are responsible for distinct spatial aspects of sal expression in the wing disc (de Celis et al., 1999). Characterisation of the medial enhancer that appears to control sal expression domain only in the centre of the wing pouch (Barrio and de Celis, 2004) (Fig. 1A) revealed that it contains Brk binding sequences. Consistent with the presence of a repressor binding sites, this particular enhancer fusion construct is not expressed in the lateral regions of the wing disc where substantial levels of Brk protein are present. Another enhancer, namely AK, recapitulates sal expression in the lateral regions of the wing disc when fused to a lacZ reporter (AK-lacZ) (de Celis et al., 1999) (Fig. 1A). We wished to determine using bioinformatics if this lateral enhancer of sal could be regulated directly by either pMad-Medea-Shn complex or Brk. We sequenced the AK enhancer (753 bp; supplementary material Fig. S1) and found (using fuzznuc/EMBOSS) no binding sites for pMad-Med-Shn complex, indicating that sal expression in lateral regions is not subjected to direct regulation by this repressive complex. Consistent with the results presented above, we found no Brk binding sites, indicating that Brk does not directly bind the regulatory region responsible for sal expression in lateral regions, but rather acts in an indirect manner by repressing the negative regulator of sal (NRS) that subsequently acts by directly binding to the AK enhancer. Interestingly, the AK enhancer is highly conserved in other Drosophila species and contains several putative binding sites for known transcriptional repressors (supplementary material Fig. S1).

According to our model, the expression of the AK-lacZ reporter is restricted to the periphery of the wing disc presumably due to the repressing activity of the negative regulator of sal (NRS) in more medial regions. If this assumption is correct then increase in Brk dose will result in repression of NRS expression and subsequent activation of the AK-lacZ reporter in more medial regions. We tested this prediction by analysing the expression of the AK-lacZ reporter in Brk overexpressing clones. Indeed, we found that the lacZ reporter was ectopically induced in clones generated in lateral as well as medial regions of the wing pouch (Fig. 4C). This ‘in vivo activity assay’ indicates that the negative regulator of sal (NRS), which regulates sal expression via the AK enhancer, is normally repressed by Brk in lateral regions. However, it is expressed and active all along the wing pouch where Brk levels decline (regions I, II and III). Importantly, endogenous sal expression was downregulated in Brk overexpressing clones located in the centre of the wing disc and upregulated in lateral clones (region III) (Fig. 4B, white and yellow arrowheads respectively). This implies that in contrast to the isolated lateral (AK) enhancer, the full promoter of sal responds to NRS repressing activity in a position-dependent manner, presumably due to integration of additional inputs likely mediated by other regulatory regions.

Our model further predicts that activating BMP signalling or removing Brk activity in the periphery of the wing disc (region IV) should lead to upregulation of NRS and consequently to downregulation of the AK-lacZ reporter. Indeed, when we either generated tkv^QD overexpressing clones or clones mutant for brk, expression of the AK-lacZ reporter was abolished within the clones (Fig. 4F,I, respectively). Consistent with our model when we antagonized Brk-repressing activity by overexpressing Brk-VP16 in clones, we repressed the AK-lacZ reporter (Fig. 4L), presumably because the recombinant protein activated the expression of NRS. In summary, these results indicate that Brk drives endogenous sal expression in lateral regions of the wing disc (region IV) by repressing NRS, a negative regulator of sal, which acts through the AK enhancer. Moreover, the absence of sal expression in lateral regions of the wild type wing pouch (region III) is not a result of Brk repression but rather is engineered by NRS activity.

We wondered whether this unusual influence exerted by BMP/Brk signalling in lateral regions is unique to sal regulation or if it is a general mechanism that is more broadly utilised to regulate other target genes as well. In the wing disc, vg is a low-threshold BMP target that is expressed, like sal, also in lateral regions where Brk is in its maximal levels. To test if expression of vg in the lateral region of wing disc is also regulated by Brk, we generated brk LOF clones and analysed Vg expression by antibody staining. As in the case of sal, Vg was downregulated upon removal of brk in region IV (Fig. 5), suggesting that vg expression in the wing disc periphery is also positively regulated by Brk. Interestingly, the enhancer fragment responsible for vg expression in the wing pouch (vgQ) (Kim et al., 1997), is ectopically induced in the periphery of the wing disc upon removal of brk (Campbell and Tomlinson, 1999). Taken together with our analysis these data indicate that in the context of the full promoter Brk cannot repress vg expression through the vgQ enhancer in the periphery of the wing disc. Moreover, similar to sal, vg expression in the wing disc periphery is positively regulated by Brk.

During Drosophila wing disc development the BMP morphogenetic gradient established by the collective actions of the two BMP ligands, Dpp and Gbb, patterns the cellular field by modulating gene expression in a concentration-dependent manner. How is the BMP concentration gradient translated into coordinated target gene expression? Current model relies upon two opposing activity gradients of the transcriptional regulators pMad and Brk, established in response to the BMP gradient. This model has been particularly successful in elucidating the regulatory influence exerted by BMP signalling although most of the attention has been focused on wing pouch, a region proximal to the peak of the gradient. Moreover, it is assumed that the target gene expression, in lateral regions where the BMP activity gradient decline, is independent of the signalling influence. Data presented here demonstrate that this supposition is incorrect. Alterations in pMAD signalling even in the lateral regions of the wing disc indeed lead to changes in positive target gene expression such as vg and sal albeit in an unexpected manner. In a classic ‘role reversal’ mode, the expression of the same targets is positively regulated by Brk and negatively by BMP signalling. Thus, while the classical morphogen model assumes that morphogens pattern a homogeneous field of responding cells, we show that in the developing wing disc interpretation and response to the BMP morphogenetic signal qualitatively differ along the anterior-posterior axis. The fact that sal is already expressed in both medial and lateral regions of the wing disc during early second larval stage (Grieder et al., 2009), implies that this subdivision occurs early in development.

By comparing sal expression in the different regions of the wing disc, we have uncovered a novel circuitry underlying inverse regulation of an archetypal BMP target gene in distant regions where the morphogen levels decline. How is this counter regulation of sal achieved? Using the enhancer fragment that drives sal expression just in the lateral regions of the wing disc, we provide evidence that Brk induces expression of sal at the periphery of the wing disc indirectly through repression of a negative regulator (NRS). On one hand, NRS represses sal expression by binding to a cis-regulatory element that contains neither Brk nor pMad-Med-Shn complex binding sites. On the other, NRS is itself negatively regulated by Brk. The experiments described in this paper (summarised in Table 1) provide an initial working model (Fig. 6) to explain how the expression pattern of an established BMP target like sal is in fact differentially and inversely regulated in different regions along the anterior-posterior axis of the wing disc.

Phenotypic consequences of the loss of brk on sal expression are qualitatively distinct and the effects vary in a position-dependent manner. Compromising brk function results in upregulation of sal in the central region (I and II); shows no effect towards the edge of the wing pouch (region III), and leads to a loss of sal expression in the periphery of the disc (region IV) (Fig. 3A–I; Table 1).

Brk is known to repress lateral expression of classic BMP responsive genes such as omb and dad through direct binding to specific sequences within their enhancers (Sivasankaran et al., 2000; Weiss et al., 2010), indicating that it is highly active in lateral regions. Since the medial enhancer of sal also contains two Brk consensus-binding sequences (Barrio and de Celis, 2004), this raises the question as to why the medial enhancer does not function to repress sal in the lateral zone? While we don't know the precise mechanism underlying this position-dependent-repression, the fact that a P-lacZ reporter of the sal medial enhancer is induced in lateral regions upon removal of Brk (Campbell and Tomlinson, 1999), indicates that in isolation the medial enhancer responds to Brk repressive activity also in the periphery of the wing disc. Similarly, Barrio and de Celis found that mutating the relevant Brk binding sites in the isolated medial enhancer of sal expanded the expression to lateral regions (Barrio and de Celis, 2004). Combined together, these observations suggest that in the context of the full-length endogenous promoter of sal, (yet unknown) trans-factors are differentially distributed along the A–P axis of the developing wing disc to prevent repression by Brk (via the medial enhancer) in lateral regions and thus to confer position-dependent expression.

Our analysis implies that Brk drives endogenous sal expression in region IV by repressing a negative regulator of sal (NRS), which targets the lateral enhancer. Brk levels, which decline medially, enable NRS to be active which in turn represses sal expression in lateral regions of the wing pouch (region III) (Fig. 6). By manipulating Brk levels and monitoring the activity of the AK-lacZ reporter, we provide evidence that NRS is normally expressed and active all along the wing pouch (regions I, II and III). This raises the question as to how endogenous sal in the centre of wing pouch escapes repression mediated by NRS. In principle, high pMad activity in medial regions (I and II) could overcome the repressive function of NRS. While plausible, this is an unlikely scenario as in the absence of both brk and mad, sal is ectopically expressed in medial regions (Campbell and Tomlinson, 1999; Jaźwińska et al., 1999). We therefore propose that in the context of the endogenous promoter, the activity of NRS is antagonized specifically in centre of the wing pouch (regions I and II) with the assistance from the localised, non-uniform distribution of trans-acting factors along the A–P axis of the wing disc to confer a position-dependent transcriptional response.

In our experiments sal is ectopically induced near the edge of the wing pouch (region III) in tkv mutant clones (Fig. 1F). Even more perplexingly tkv^QD overexpressing clones also behave in a similar manner (Fig. 2I). Both of these outcomes are difficult to reconcile with the current model describing how the BMP morphogen gradient is linearly interpreted as both extreme situations; either complete loss (tkv) or substantial gain (tkv^QD) of endogenous pMad activity, results in ectopic expression of sal. However, this conundrum can be partially resolved by taking into account the activity of the newly invoked, additional component NRS into the BMP-dependent patterning system.

In the wild-type wing disc, pMad activity in region III, although low, is still sufficient to downregulate Brk levels just enough to allow for concomitant rise in NRS levels ultimately resulting in repression of sal expression. Importantly, low levels of pMad in region III (acting through the medial enhancer; ME) are inadequate to antagonize the repressor activity of NRS (mediated through the lateral enhancer; LE) (Fig. 6A, region III). The absence of pMad activity in tkv mutant clones in region III increases the levels of Brk, which in turn represses NRS and thus de-represses sal expression. In the case of tkv^QD overexpressing clones, the substantially elevated pMad activity represses brk expression leading to elevated NRS activity (Fig. 6B). In region III clones (periphery of the wing pouch) the high activity of pMad (acting from the ME) overcomes the repressing activity of NRS (acting from the LE), ultimately resulting in activation of sal. By contrast, in the vast majority of tkv^QD overexpressing clones (57 out of 63 clones) located in region IV (periphery of the wing disc) endogenous expression of sal is either lost or diminished. Why in region IV pMad inducing activity does not have an edge over NRS repressing activity, as is the case in region III? We propose that in the context of the endogenous promoter, the activity of pMad is antagonized specifically in region IV due to the activity of unevenly distributed trans-acting factors. The rare occasions, where in the tkv^QD overexpressing clones located in region IV sal expression was upregulated could be due to a rare event leading to acquisition of wing pouch like identity by the cells at the wing disc periphery, presumably due to early exposure to high pMad activity. Nevertheless, differential transcriptional response behaviour exhibited by the cells from regions III v/s IV supports the subdivision of the developing wing disc on the basis of distinct regional competence.

How an exponentially decaying morphogen gradient of BMP gives rise to computable changes in gene expression ultimately leading to discreet morphological structures is a fascinating question. A steep slope of the BMP activity gradient near the peak allows sharp expression domains of target genes to be defined within the wing pouch area (Bollenbach et al., 2008; Entchev et al., 2000; Teleman and Cohen, 2000; Wartlick et al., 2011). However in the lateral regions of wing disc, the activity gradient of BMP dips considerably raising the question as to how small differences in signal strength provide discrete threshold responses. Indeed, it is believed that BMP/Brk patterning system does not regulate the lateral expression of sal (Affolter and Basler, 2007; Campbell and Tomlinson, 1999; Jaźwińska et al., 1999). Our data argue that not only the signalling is active in the lateral regions but the inverse regulatory mode adopted by the signalling circuitry is in fact responsible for generating distinct threshold responses.

We generated overexpressing Flp-out clones using the act >CD2 >Gal4 cassette, recombined to a UAS-GFP construct for the detection of the clones. Larvae were subjected to a 37°C heat shock for 10 minutes. Genotypes of dissected larvae were as follows: Tkv^Q235D-overexpressing clones: yw hsp70-flp; AK; act >CD2 >Gal4 UAS-GFP/UAS-tkv^Q235D. Brk-overexpressing clones: yw hsp70-flp; UAS-brk/AK; act >CD2 >Gal4 UAS-GFP. Brk-VP16 clones: yw hsp70-flp; UAS-brk-vp16/AK; act >CD2 >Gal4 UAS-GFP. Dad-overexpressing clones: yw hsp70-flp; UAS-dad; act >CD2 >Gal4 UAS-GFP.

We generated mutant clones using Flp-mediated mitotic recombination and identified them by the loss of the GFP marker. Clones were induced by heat shock (60 minutes at 37°C). Genotypes of dissected larvae were as follows. brk loss-of-function clones: Ubi-GFP FRT18A/yw brk^M68 FRT 18A; hs-flp. tkv loss-of-function clones: yw hsp70-flp; tkv^a12FRT40/Ubi-GFP FRT40.

Imaginal discs from third instar larvae were fixed and stained by standard techniques. The specific primary antibodies used were: mouse anti-β-gal (1∶1000; Promega), rabbit anti-Spalt [1∶1000; a gift from A. Salzberg (Halachmi et al., 2007)], rat anti-Nab [1∶1000 (Suissa et al.</citref>, 2011)], rat anti-Brk (1∶1000; a gift from F. A. Martín and G. Morata, CDBM University Autonoma De Madrid, Madrid, Spain), rabbit anti-cleaved Caspase 3 (1∶40; Cell Signaling) and Rabbit anti-Vg (1∶20; a gift from Sean Carroll). Secondary antibodies used: RRX- or Cy5-conjugated (1∶400; Jackson Laboratories). Images were taken on a TE2000-E confocal microscope (Nikon) using a 20× objective. Figures were edited using Adobe Photoshop 7.0.

DNA contains the first 173 aa of the Brk repressor, includes the DNA-binding domain (DBD) site, nuclear localisation sequence (NLS), but not the repression domain (construct A2 from (Winter and Campbell, 2004)) cloned into pUAST vector that contains VP16 activation domain (a gift from Dr Adi Zalsberg – Technion). 50 µg of Qiagen-purified DNA sent to Genetic Services Inc. (GSI) where they injected the DNA into mutant embryos lacking the gene W (for red eyes). Adult flies were single-crossed to yw, and F1 containing red eyes were selected. Transgenic flies were kept as balanced stocks.

We thank R. Barrio, J. F. de Celis, G. Campbell, A. Salzberg, M. Affolter, G. Morata, S. Carroll and The Bloomington Drosophila Stock Center for fly stocks and antibodies; T. Kahan for bioinformatics analysis; and B. Shilo for comments on the manuscript. The authors declare that they have no conflict of interest.