A conserved transcriptional network regulates lamina development in the Drosophila visual system

The insect visual system comprises two separate structures: the compound eye (often called ‘retina’), which contains the photoreceptors, and the underlying optic lobes (OLs), which are part of the brain hemispheres (Meinertzhagen and Hanson, 1993; Sanes and Zipursky, 2010). Despite their functional importance as primary visual processing centers, our knowledge of the genetic mechanisms that control the specification and early development of the Drosophila OLs is poor compared with the eye. In addition, knowledge gained in Drosophila might be of general importance for understanding the mechanisms that specify the vertebrate visual neuroepithelium (reviewed by Erclik et al., 2009; Sanes and Zipursky, 2010).

The compound eye is formed by ∼800 ommatidia, or unit eyes. Each comprises six outer (R1-6) and two inner (R7 and R8) photoreceptor (PR) neurons, together with pigment and lens-secreting cone cells. Outer PRs are in charge of motion and brightness detection and spatial vision, whereas inner PRs function as color, UV and polarized light sensors (Hardie, 1985). According to their differentiated function, R1-6 innervate the first OL neuropil, which is the lamina, whereas R7 and R8 axons run across the lamina to innervate the neuropil beneath, which is the medulla. Next, the lobula and lobula plate (together called the ‘lobula complex’) receive signals from medulla neurons and send projections to the higher-order visual centers of the brain (Meinertzhagen and Hanson, 1993; Sanes and Zipursky, 2010). This tetralayered organization is conserved within insects and shared with malacostracan crustaceans (Strausfeld, 2009).

The lamina, medulla and lobula complex are derived from the embryonic OL primordium (OP) (Green et al., 1993), which segregates into the outer and inner OL anlagen. After embryonic invagination, and during larval life, these two anlagen proliferate extensively and are termed the outer and inner proliferative centers (OPC and IPC), respectively. The OPC-derived neurons have their cell bodies in the lamina and medulla, whereas the IPC-derived neurons have their cell bodies mostly in the lobula complex (Hofbauer and Campos-Ortega, 1990; Meinertzhagen and Hanson, 1993).

During late third larval stage (late L3) the OPC neuroepithelium is characterized by densely packed cells expressing DE-cadherin (DE-cad; Shotgun – FlyBase) that give rise to medulla neuroblasts medially and to lamina neurons laterally (Egger et al., 2007). Neuroepithelial and lamina cells are separated by an indentation called the lamina furrow (LF) (Selleck and Steller, 1991; Meinertzhagen and Hanson, 1993). Lamina cells are characterized by the expression of the transcription factor dachshund (dac) (Mardon et al., 1994; Huang and Kunes, 1996). As cells exit the LF, they are contacted by incoming retinal PR axons in what is termed the preassembly domain (Umetsu et al., 2006). Hedgehog (Hh), delivered by PR axons, triggers the lamina differentiation program along with a final cell division (Selleck and Steller, 1991; Huang and Kunes, 1996; Huang et al., 1998). Next, lamina cells and PR axons reorganize and assemble into lamina columns (Meinertzhagen and Hanson, 1993), a process that requires the expression of the bHLH-PAS transcription factor single-minded (sim) (Umetsu et al., 2006).

In recent years, the mechanisms of specification and early development of the eye primordium have been extensively studied [reviewed by Amore and Casares (2010)], in contrast to other components of the visual system. Eye specification is dependent on the expression of two Pax6 genes, twin of eyeless (toy) and eyeless (ey), with toy being required for the initiation of ey expression (Halder et al., 1995; Czerny et al., 1999). Then, ey-expressing cells are maintained as proliferative and undifferentiated progenitors as long as they express the TALE class homeodomain gene homothorax (hth) (Bessa et al., 2002; Peng et al., 2009; Lopes and Casares, 2010). Repression of hth by Decapentaplegic (Dpp; Drosophila BMP2/4) allows the upregulation of a class of genes collectively known as retinal determination (RD) genes (Bessa et al., 2002; Lopes and Casares, 2010). These include sine oculis (so), a Six2-type homeodomain transcription factor, its partner eyes absent (eya), and the transcription factor dac (Silver and Rebay, 2005). The RD genes, which also include the Zn-finger genes teashirt (tsh) and tiptop (tio), are knitted together in the RD gene regulatory network that, through extensive feedbacks, locks in the eye fate [reviewed by Kumar (2010)].

RD genes are also expressed in the OLs. However, their putative role during development has remained elusive. The RD genes so and eya are expressed in the embryonic OP, where they are required for its invagination (Cheyette et al., 1994; Serikaku and O'Tousa, 1994; Bonini et al., 1998; Daniel et al., 1999). After OP invagination, transcription of so and eya is reinitiated only during late larval stages (Cheyette et al., 1994; Serikaku and O'Tousa, 1994; Bonini et al., 1998). In addition to small or absent eyes, so and eya alleles result in reduced OLs, the lamina being specially affected (Bonini et al., 1993; Cheyette et al., 1994; Serikaku and O'Tousa, 1994). However, experiments (Fischbach and Technau, 1984) using so chimeras have indicated that the lamina and medulla reduction seen in so flies is likely to be the consequence of a lack of innervation from the reduced eyes, raising the possibility that the OL expression of at least so has no major relevance for OL development. However, a cell-autonomous requirement for so or eya has not been tested to date. As mentioned above, the expression of dac, another RD gene, is detected in lamina cells, where it is required for the ensuing differentiation of the lamina (Huang and Kunes, 1996; Chotard et al., 2005). In the compound eye dac expression requires eya and so input (Chen et al., 1997; Pignoni et al., 1997), but it has not been established whether this is also the case during lamina development.

Here, we investigate the specification mechanisms of the lamina, as the first neuropil of the OLs, focusing on the expression and function of RD genes. We show that the expression of the RD genes eya and so is mutually dependent. These genes are required cell-autonomously for dac expression, which additionally requires the Hh signaling provided by incoming retinal PR axons. dac is instrumental for the differentiation of lamina neurons as it is required to repress hth, which would otherwise impair sim expression. In addition, we identify a role for hth in the neuroepithelium (NE), where it is normally co-expressed with eya/so. hth is required for normal NE growth and to control the extent and levels of expression of eya. Therefore, an hth-eya-so-dac core network is shared by the lamina and the compound eye. However, major differences exist in the roles played by these genes, including a role of dac in lamina differentiation as a necessary hth repressor.

The development of the lamina and of the compound eye show a number of important similarities. The differentiation of lamina neurons is triggered by Hh. Differentiation progresses as a fan-shaped wave marked by an indentation, the LF (Selleck and Steller, 1991), which moves from lateral to medial as incoming axons are sent in by newly formed rows of ommatidia (Umetsu et al., 2006). This is reminiscent of the Hh-driven movement of the morphogenetic furrow in the eye (Wolff and Ready, 1991). In addition, expression of the RD genes dac, eya and so during L3 lamina development has been reported (Cheyette et al., 1994; Serikaku and O'Tousa, 1994; Bonini et al., 1998). Therefore, and in order to analyze these similarities in more detail, we first mapped the expression domains of eya, so, dac, sim and hth in the developing OL.

During L3, NE progenitors, which express high levels of DE-cad (Egger et al., 2007), give rise to dac-expressing lamina cells laterally and to medulla neuroblasts medially (Fig. 1A-D). The expression of lethal of scute [l(1)sc] in a narrow band, two to three cells in width, marks the transition zone after which medulla neuroblasts are generated medially (Yasugi et al., 2008) (Fig. 1E), while the LF marks the transition between the NE progenitor cells and the lamina (Hofbauer and Campos-Ortega, 1990; Huang and Kunes, 1996). Therefore, l(1)sc expression medially and the LF laterally define the NE progenitor domain (Fig. 1E,G). The onset of dac expression starts posterior to the LF, in its posterior slope, and is maintained throughout lamina development (Huang and Kunes, 1996; Chotard et al., 2005; Umetsu et al., 2006). Differentiating lamina cells expressing sim are located posterior to the LF (Umetsu et al., 2006). Expression of sim is detected weakly and uniformly in the preassembly domain, but clear nuclear staining can be seen only in the assembly domain (Umetsu et al., 2006) (Fig. 1C,D). hth expression is detected in medulla neuroblasts and neurons, in the NE progenitors (Reddy et al., 2010; Hasegawa et al., 2011; Morante et al., 2011; Li et al., 2013; Suzuki et al., 2013) and also within the posterior slope of the LF (Fig. 1D,E,F). However, hth expression is absent in more internal lamina regions (i.e. the preassembly and assembly domains). Interestingly, the expression of hth and sim abut each other within the lamina (Fig. 1D). eya expression does not extend into the medulla, being detected in the transition zone, overlapping l(1)sc, the NE progenitors and through the furrow into the lamina (Fig. 1E). eya levels increase just posterior to the furrow, and these levels are sustained throughout the lamina. The expression of so follows exactly that of eya (supplementary material Fig. S1).

Thus, three domains can be distinguished along the lamina differentiation path: (1) NE progenitor cells that express hth and low levels of eya/so anterior to the LF; (2) lamina precursor cells (LPCs), located in the posterior slope of the furrow, are characterized by the expression of hth and high levels of eya/so and dac; and (3) the preassembly and assembly domains of the lamina, where hth is no longer expressed and which accumulate increasing Sim. These patterns are summarized in Fig. 1G.

The similar expression of eya and so suggested that these genes might regulate the expression of each other. To test this, we induced clones expressing RNAi transgenes specific for either eya or so. In these clones, the levels of the targeted proteins, as detected by specific antibodies, were reduced to background levels, indicating severe gene expression loss (Fig. 2). When eya was knocked down So expression was lost within the clones (Fig. 2A), and when so was knocked down Eya expression was strongly reduced (Fig. 2B). Therefore, the expression of eya and so is mutually dependent, although this dependence for eya might be only partial. A feedback between eya and so has also been described during eye and ocelli development (Pignoni et al., 1997; Brockmann et al., 2011).

dac expression is known to lie downstream of eya and so in the compound eye (Chen et al., 1997). We tested whether this was also the case in the lamina by inducing clones in which eya or so function had been lost (eya^E8) or knocked down (so RNAi). In both experiments, dac expression was not activated within the clones and the effects were cell-autonomous (Fig. 3A,B). Interestingly, loss of eya or so resulted in hth expression beyond its normal limit, raising the possibility that dac loss was due to its repression by hth. To distinguish between a direct activating role by Eya/So versus a repression by hth, we compared the effects on dac of removing so alone or together with hth. dac expression was cell-autonomously lost in both so-RNAi clones (in which hth is maintained) as in so-RNAi+hth-RNAi clones (Fig. 3B,C). Therefore, dac regulation requires the positive input of the RD genes and does not seem to be negatively regulated by hth (see below). To test this specifically, we induced hth-RNAi clones and checked for changes in Dac levels. We compared the Dac immunofluorescence signal within and outside hth-RNAi clones (Fig. 3D) and found no difference (ten clones from seven OLs were analyzed). Therefore, hth is not acting as a dac repressor behind the LF. In addition, the fact that dac expression did not extend medially into the eya/so-expressing domain in hth^– clones indicated that these RD genes are not sufficient to induce dac expression. This is consistent with the hh pathway being additionally required for dac expression (Huang and Kunes, 1996).

Previous work by Pappu and co-workers identified two enhancers in the dac gene (Pappu et al., 2005), one of which, dac3EE, was noted to drive strong reporter gene expression in the lamina. We confirmed very strong expression of this enhancer in the lamina, plus weaker expression in the lobula (Fig. 3E; data not shown). eya knockdown clones, in which dac expression was lost, also lost dac3EE-Z activity (Fig. 3E). This suggests that eya acts via dac3EE to regulate dac lamina expression.

The expression of dac builds up in the posterior slope of the LF, preceding the shutting off of hth in the lamina preassembly domain. This observation raised the possibility of dac being involved in hth repression. In fact, blocking the Hh signaling pathway by removing the Hh signal transducer smoothened (in smo³clones) results not only in a loss of dac (Fig. 4A) (Huang and Kunes, 1996) but also in an expansion of hth expression into the lamina (Fig. 4B). Interestingly, the expression of eya does not depend on hh, as its expression remains unaltered in smo^– clones spanning the lamina (Fig. 4B) in spite of hth expression. In test if hth repression is mediated by dac, we examined the effect of removing dac on hth expression. In dac^– clones, hth is upregulated in internal regions of the lamina whereas eya expression is unaffected (Fig. 4E), suggesting that Eya/So cannot repress hth in the absence of Dac. To prove that this effect was not due to a regulatory feedback of dac on the Hh signaling pathway, we checked dac requirement for Hh signaling activity. Whereas in smo^– clones the expression of the downstream signaling component cubitus interruptus (ci) (Motzny and Holmgren, 1995; Alexandre et al., 1996) is reduced (Fig. 4C), in dac^– clones ci expression was unaltered (Fig. 4D), indicating that dac is not generally required for Hh signal transduction. Altogether, these results indicate that dac is required downstream of eya/so and the hh pathway to repress hth. As smo^– cells cannot differentiate as lamina neurons, these smo^–, eya-expressing, dac-nonexpressing cells are likely to remain in a lamina precursor state. When we performed the converse experiment and expressed dac ectopically in the NE, hth expression was downregulated cell-autonomously (Fig. 4F), indicating that dac is not only required but also sufficient to repress hth.

Molecularly, Dac proteins have been shown to work, at least in some developmental contexts, in a complex with So/Six and Eya proteins (Chen et al., 1997; Pignoni et al., 1997; Ikeda et al., 2002). In order to gain insight into whether such a complex might be involved in sim activation and hth repression in the developing lamina, we examined the expression of both genes in clones expressing dac but lacking eya (and, therefore, also so) using the MARCM technique. eya⁻ dac⁺ clones, as is also the case for clones mutant for eya only, abutted the lamina. In these clones (N=5), Hth expression was maintained at levels similar to those in adjacent, non-mutant cells (supplementary material Fig. S2A). Only in two cases (out of more than 40 clones examined) the eya⁻ dac⁺ clone clearly spanned the preassembly domain, allowing the analysis of clones in the Sim-expressing region. In one case, Sim expression was reduced within the clone (supplementary material Fig. S2B,C), whereas no noticeable alteration in Sim levels was observed in the second clone (supplementary material Fig. S2D,E). eya⁻ dac⁺ clones falling entirely within the lamina were extremely rare and small (1-3 cells). In one such example, Sim expression was absent from the eya⁻ dac⁺ cells (supplementary material Fig. S2D,E, inset). These results suggest that Dac is unable to regulate hth in the absence of its partner Eya; with the limited evidence at hand, this might also be the case for sim. The low recovery of eya⁻ dac⁺ cells in the OLs might be due to high apoptosis rates induced by dac, as we detect increased levels of the apoptosis marker activated Caspase 3 in OLs from brains containing dac⁺ clones (data not shown).

Finally, we examined the impact of misregulation of hth, eya/so and dac on sim expression, as this gene marks further differentiation steps in the lamina (Umetsu et al., 2006). In so^– clones, sim expression is reduced (Fig. 5A). However, in these clones hth is derepressed (Fig. 5A and see Fig. 3B). Therefore, the loss of sim could be due to either the loss of a positive input (RD) or the presence of a repressor (Hth). To distinguish between these two possibilities, we simultaneously knocked down so and hth. In so^– hth^– clones, sim expression is still reduced or absent (Fig. 5B), pointing to the need for a positive sim regulator. Since eya and so are required for dac expression, we next checked the effect of only removing dac (which does not affect the expression of its upstream regulators eya/so; Fig. 4E). Again, sim expression was reduced cell-autonomously in dac^– clones (Fig. 5C). However, and as we showed previously, loss of dac is accompanied by hth upregulation (Fig. 5B, see also Fig. 4E). In order to determine if hth was directly responsible for sim repression in dac^– cells, we induced hth-expressing clones in the lamina (Fig. 5D,E). In these clones sim is repressed (Fig. 5D) without detectable changes in dac expression (Fig. 5E).

We conclude that the RD genes are required for full sim expression, acting both as sim activators and hth repressors. A crucial role in this regulation is played by dac, which is necessary to repress hth.

Our results so far indicate that the repression of the transcription factor hth within the lamina is necessary to allow its proper differentiation. However, hth is expressed at earlier stages in the NE progenitors, where it overlaps with eya and so (Fig. 1E,F; supplementary material Fig. S1). To establish whether hth plays any positive role in the NE, we first analyzed the impact of loss of hth on tissue growth. hth^– clones were recovered throughout the OL, although they were 30-40% smaller than their wild-type twin clones (Fig. 6A,B), suggesting a role of hth in proliferation. Next, we investigated whether eya expression would be affected by removing hth. Owing to the smaller size of hth^– clones, we induced hth loss-of-function mosaics using the Minute technique in order to recover larger clones and make the analysis of the effects of hth removal on eya expression easier. hth, M⁺ clones showed a slight expansion of eya expression (Fig. 6C). In addition, when the intensity of the eya signal was compared between hth mutant and adjacent control tissue, we observed an increase of ∼20% in eya levels in the hth^– cells (12 hth-RNAi clones from seven OLs analyzed).

These results indicate that hth modulates the extent and levels of eya expression, suggesting that the levels of hth might be subject to tight regulation. Indeed, this does seem to be the case, as clones overexpressing hth block eya expression in the NE (Fig. 6D). We noted, however, that this repression was only detectable in the NE. Neither in smo^– nor in dac-mutant clones, where hth expands a few rows into the lamina, did we observe any significant change in eya expression posterior to the LF. This might indicate that another, as yet unidentified, factor aids hth in modulating eya expression in the NE. Alternatively, dac, the expression of which is turned on just after the LF, could enhance eya expression, making it insensitive to hth. Loss of hth, in hth-RNAi clones, does not result in premature Sim expression (Fig. 5F), consistent with the requirement of multiple inputs for sim activation.

In this study we have uncovered a gene regulatory network that operates cell-autonomously during the specification of lamina neurons in the OLs. The regulatory model that emerges from our work is summarized in Fig. 7.

We have shown that the expression of the RD genes eya and so is required cell-autonomously for the specification of lamina neurons through the activation of at least dac and sim. Their expression is initiated in the NE cells at low levels, but this expression increases after the LF. The fact that eya and so positively regulate each other would, above a certain expression threshold, cause eya and so to lock in their transcription to maximal levels. hth might regulate that threshold, since hth acts in vivo as a repressor of eya in NE progenitors. Regulation of eya expression might be further required for the spatial segregation of cells in different states along the lamina differentiation pathway. Thus, eya^– clones are seldom recovered in internal regions of the lamina (Fig. 3A), which might be indicative of a segregation of the eya-mutant tissue or the elimination of these cells from within the lamina. Something similar happens with smo and sim mutant clones, which do not appear in internal regions of the lamina either (Umetsu et al., 2006). In addition, hth is required to sustain normal NE cell proliferation, as hth-mutant clones grow, on average, 30% less than wild-type clones.

The recruitment of lamina precursors from NE progenitors is driven by incoming waves of R1-6 axons, thereby coupling it to retinal differentiation (Selleck and Steller, 1991; Huang and Kunes, 1996). Axon-delivered Hh activates its downstream pathway and, as a consequence, dac expression is upregulated in eya/so-expressing cells (Huang and Kunes, 1996; Chotard et al., 2005). We show that the net result of this activation is an efficient repression of hth to allow the full expression of sim and, thereby, normal lamina differentiation. Dac, and its vertebrate Dach homologs, have been shown to form a protein complex with So/Six and Eya family proteins (Chen et al., 1997; Pignoni et al., 1997; Ikeda et al., 2002) and to synergize with them in ectopic eye induction assays (Chen et al., 1997; Pignoni et al., 1997). In addition, Dac possesses a DNA-binding domain (Kim et al., 2002). The fact that ectopic dac expression does not seem capable of repressing hth without eya suggests that this function would require the formation of a trimeric complex with So and Eya. hth, eya, so and dac are transiently co-expressed in some cells of the posterior slope of the LF, before hth is shut off. This posterior slope might represent a transition zone in which these regulatory processes are taking place (Fig. 1G). Regarding the activation of sim expression, we have not been able to clearly determine whether it can be carried out by Dac alone or by a So-Eya-Dac complex, although the limited evidence that we have gathered is compatible with Dac requiring Eya/So cooperation.

In the compound eye, progenitors are characterized by hth expression, whereas the precursor population (contained in the so-called pre-proneural domain) expresses high levels of eya, so and dac, and no hth (Bessa et al., 2002). However, detailed inspection of the eye progenitor domain shows that the hth-expressing cells also co-express RD genes such as eya, although at low levels (supplementary material Fig. S3), as we have shown to be the case in NE progenitors in the OL. Therefore, eye and lamina derive from progenitors that show many similarities. However, there are several profound differences. Neither ey nor toy, the two Pax6 genes positioned at the top of the genetic hierarchy of eye development, is expressed during lamina development (Callaerts et al., 2001; Morante et al., 2011; Southall et al., 2013) (supplementary material Fig. S4), and all available information on ey function, including that from ey mutants (Callaerts et al., 2001), expression of dominant-negative ey forms (Morante et al., 2011) or RNAi-mediated ey and toy knockdowns (our unpublished data), is compatible with these genes not playing a direct role in lamina development, at least during larval stages. Also, neither tsh nor its paralog tio is expressed in the OPC NE (Southall et al., 2013; data not shown), although cells of the OPC NE are ready to respond to tsh expression by proliferating and blocking lamina differentiation (supplementary material Fig. S5). Since, in the eye, tsh and hth have been shown to directly interact with Yki to activate Hippo-regulated target genes (Peng et al., 2009), and since the proliferation in the OPC is controlled by the Hippo pathway (Reddy et al., 2010), providing tsh is likely to engage hth and yki in maintaining the progenitor state of neuroepithelial cells as well. Another significant difference between lamina and eye development is that, during eye development, hh does not regulate hth expression, at least not directly, because blocking the hh pathway in smo^– clones does not affect the hth expression pattern (Firth and Baker, 2009; Lopes and Casares, 2010). In the eye, hth repression is carried out mostly by the BMP2 homolog Dpp (Lopes and Casares, 2010), which is itself an Hh target (Heberlein et al., 1993; Greenwood and Struhl, 1999; Fu and Baker, 2003). In the OLs, Dpp has been shown to play a different role: the specification of lamina glia (Yoshida et al., 2005).

Perhaps the most striking difference, though, lies in the role played by dac. In the eye, dac is required for the initiation of retinal differentiation; however, once the differentiation wave is progressing, removal of dac has little effect on the process (Mardon et al., 1994). Accordingly, dac^– clones do not derepress hth (C. Bras-Pereira and F.C., unpublished). By contrast, dac is necessary for the correct differentiation of the lamina, where it is required for hth repression. Our study shows that sim expression is reduced in dac-mutant cells. In a previous study, Umetsu et al. (2006) found no effect on sim upon dac removal. We note that our clones were induced using a heat shock-inducible flipase (hsFLP), whereas the clones used by Umetsu and co-workers were induced with the NP6099-GAL4 line (Yoshida et al., 2005) driving UAS-FLP, which may cause a milder phenotype due to delayed timing of clone induction. This could have resulted in only a subtle reduction in Sim expression. The fact that in the absence of dac the expression of sim is reduced but not absent argues for the existence of a dac-independent sim activator, probably Hh (Fig. 7).

The use of the hth-eya-so-dac cassette seems to be evolutionarily conserved, as genes of the Pax, Six, Eya, Dach and, in some instances, Meis gene families have been shown to be co-expressed and functional during the development of many different organ types in vertebrates: from eyes, sensory placodes or brain regions to muscle, kidney or pancreas (Ikeda et al., 2002; Zhang et al., 2002,, 2006; Li et al., 2003; Bessarab et al., 2004; Purcell et al., 2005; Bumsted-O'Brien et al., 2007; Kaiser et al., 2007; Erickson et al., 2010; Santos et al., 2011). Therefore, further study of the early development of the Drosophila lamina might shed light on the general mechanisms governed by this multipurpose genetic cassette in vertebrates.

Larvae were raised at 25°C, unless otherwise indicated. w¹¹¹⁸was used as control strain. P{PZ}dac P(ry⁺)/Cyo; ry⁵⁰⁶ was used as a reporter for dac expression. For targeted misexpression we used the UAS/GAL4 system (Brand and Perrimon, 1993). UAS strains used were: UAS-eya-RNAi (VDRC, 43911), UAS-so-RNAi (VDRC, 104386), UAS-hth-RNAi (VDRC, 12764), UAS-GFP-hth (Casares and Mann, 2000), UAS-Flag-HA-eyaB2 and UAS-Flag-HA-tsh (both kindly provided by C. M. Luque, Universidad Autónoma, Madrid, Spain), UAS-dac and UAS-GFP (Bessa and Casares, 2005).

Clones of cells mutant for hth (hth^P2), smo (smo³), eya (eya^E8) and dac (dac³) were generated through mitotic recombination (Xu and Rubin, 1993). These alleles are described in FlyBase. Clones were induced by a 30 min heat shock at 37°C between 48 h and 72 h after egg-laying (AEL) in larvae of the following genotypes: yw, hsFLP;; FRT82B hth^P2/FRT82B Ubi-GFP, yw, hsFLP; eya^E8 FRT 40A/Ubi-GFP FRT40A, yw, hsFLP, smo³ FRT40/Ubi-GFP FRT40A and yw, hsFLP; dac³ FRT40/Ubi-GFP FRT40A. In all these cases, mutant cells were marked by the absence of GFP. In order to give hth^P2 mutant cells a growth advantage, hth^P2 clones were induced using the Minute technique (Morata and Ripoll, 1975). Clones were induced in yw, hsFLP; FRT82B hth^P2/FRT82B arm-lacZ, M(3)w¹²⁴ larvae by a 30 min heat shock at 37°C between 48 h and 72 h AEL. Clones were detected by the absence of β-galactosidase. The MARCM technique (Lee and Luo, 2001) was used to ectopically express Dac in the absence of eya. yw, hsFLP, tub-Gal4,UAS-GFP; FRT40A,tub-Gal80/CyO females were crossed to eya^E8 FRT40A/CyO; UAS-Dac/TM6B males. Larvae were heat shocked between 48 h and 72 h AEL for 45 min at 37°C. Non-Tb, GFP-positive larvae were selected for analysis. Mutant tissue was positively marked with GFP.

Ectopic expression clones were generated randomly using the flip-out method (Struhl and Basler, 1993). yw, hsFLP, act>y+>Gal4;; UAS-GFP/TM6B,Tb females were crossed to males carrying the UAS transgenes (either homozygous or balanced over TM6B, Tb). Clones were induced between 48 h and 72 h AEL by a 10 min heat shock at 35.5°C. For UAS-RNAi lines only, after heat shock larvae were grown at 29°C to maximize transgene expression; otherwise, cultures were maintained at 25°C. To induce clones within the lamina region, heat shock was performed between 72 h and 96 h AEL. Non-Tb larvae were selected for dissection and analysis. Clones were positively marked with GFP.

The genomic region containing the dac3EE enhancer (Pappu et al., 2005) was PCR amplified, cloned into PCR8/GW/TOPO (Invitrogen) and transferred into attB-pRVV-lacZ vector (kindly provided by R. S. Mann, Columbia University, New York). The attB construct was inserted in the second chromosome at the 22A attP site via phi-C31-mediated transgenesis (Bischof et al., 2007). The primers used were: 5′-GATCCCAAAAG-GACATCTTCAA-3′ and 5′-TCGAATGCAATTTTAACAGAAAAA-3′. Standard genetic techniques were used to introduce the dac3EE-Z line into appropriate genetic backgrounds.

Eye imaginal discs and brains were dissected and fixed according to standard protocols. Primary antibodies used were: guinea-pig anti-Hth, 1/2000 (Casares and Mann, 1998); rabbit anti-β-galactosidase, 1/1000 (Cappel, 55976); rabbit anti-GFP, 1/1000 (Molecular Probes, A11122); guinea-pig anti-So, 1/1000 (Jemc and Rebay, 2007); rabbit anti-Toy, 1/50 (Jacobsson et al., 2009); rat anti-Ey, 1/100 (Halder et al., 1995); rat anti-L(1)sc, 1/100 (gift from M. D. Martín-Bermudo, CABD, Seville); and mouse anti-Sim, 1/50 (gift from A. Baonza, CBMSO, Madrid). Mouse anti-Eya (eya10H6; 1/100), rat anti-DE-cad (DCAD2; 1/100), rat anti-Elav (7EBA10; 1:1000), mouse anti-Dac (mAbdac 1-1; 1/100) and mouse anti-Ci (mAb2A1; 1/5) were from Developmental Studies Hybridoma Bank. Primary antibodies were incubated in PBS with 0.2% Triton X-100. Fluorescently labeled secondary antibodies (anti-mouse-488 and -568; anti-rabbit-488 and -568; anti-rat-488 and -647 and anti-guinea pig-567 and -647) were from Molecular Probes. Images were obtained on SPE or SP2 Leica confocal systems and processed with Adobe Photoshop.

In order to detect quantitative changes in gene expression in gain- or loss-of-function cell clones, we measured the immunofluorescence signal within the clones and in neighboring (control) areas as an expression level correlate. Analysis was performed with ImageJ (NIH). Each clone was compared with a neighboring patch of tissue of similar area.

We thank Carlos M. Luque for the generation of the UAS-Flag-HA-eyaB2 and UAS-Flag-HA-tsh transgenes; Max Sánchez for help with statistical analysis; R. S. Mann, I. Rebay, Å. Rasmuson-Lestander, A. Baonza, M. D. Martín-Bermudo and P. Callaerts for antibodies and DNAs; G. Mardon, the Vienna Drosophila Resource Center (VDRC) and Bloomington Stock Center for fly strains; the Developmental Studies Hybridoma Bank (Iowa University) for antibodies; and the CABD Advanced Microscopy facility.