Retinal determination genes as targets and possible effectors of extracellular signals

Abstract

Retinal determination genes are sufficient to specify eyes in ectopic locations, raising the question of how these master regulatory genes define an eye developmental field. Genetic mosaic studies establish that expression of the retinal determination genes eyeless, teashirt, homothorax, eyes absent, sine oculis, and dachshund are each regulated by combinations of Dpp, Hh, N, Wg, and Ras signals in Drosophila. Dpp and Hh control eyeless, teashirt, sine oculis, and dachshund expression, Dpp and Ras control homothorax, and all the signaling pathways affect eyes absent expression. These results suggest that eye-specific development uses retinal determination gene expression to relay positional information to eye target genes, because the distinct, overlapping patterns of retinal determination gene expression reflect the activities of the extracellular signaling pathways.

Introduction

Pax6/Eyeless was described as a ‘Master Regulatory gene’ when it was found that its ectopic expression could direct development of an ectopic eye (Halder et al., 1995). A small set of other genes have also been found capable of directing ectopic eye development, including another Pax6 homolog twin of eyeless (toy), sine oculis (so), eyes absent (eya) and dacshund (dac) (reviewed in (Pappu and Mardon, 2002; Silver and Rebay, 2005). These retinal determination (RD) genes have roles in mammalian eye development, suggesting conserved master-regulatory gene mechanisms (reviewed in(Nilsson, 2004; Treisman, 2004).

It has been suggested that RD gene products define the eye field by binding to many genes that are expressed in the eye, and interacting with the various transcription factors that are direct effectors of positional signals throughout the body, so that the RD protein/positional information combination defines eye-specific programs of gene expression in response to extracellular signals (Curtiss et al., 2002; Mann and Carroll, 2002). This combinatorial mechanism might be illustrated by an eye-specific enhancer of the hh gene, which is activated by the Ras pathway through a Ras-dependent transcription factor Pnt, in combination with the RD protein So that provides eye specificity (Rogers et al., 2005). The model that RD gene products provide for eye-specific interpretation of positional information could explain why extracellular signals have such an influence on the locations where RD gene expression can form ectopic eyes, something that is harder to understand if RD genes simply activate eye programs irrespective of the cell's position (Chen et al., 1999; Pappu et al., 2003; Bessa and Casares, 2005; Weasner et al., 2006).

The notion that RD genes define the eye response to positional signals is complicated by the fact that RD gene expression is itself dynamic in space and time (Bessa et al., 2002; Pappu and Mardon, 2002). Dynamic RD gene expression is partly attributable to regulatory interactions between the RD genes themselves, but these are not sufficient to explain all the spatial and temporal aspects, which also depend on Dpp signaling to cue changes in RD gene expression, as summarized below (Bessa et al., 2002). One explanation could be that there is a distinction between extracellular signaling pathways. As there are several examples of RD gene regulation by Dpp signaling, it could be that Dpp plays a unique role in coordinating RD gene expression. By contrast, Wnt, Hh, Notch, and Ras pathways might provide the positional information that induces specific eye target genes and patterns eye-specific fates within the eye field, according to the status of RD gene expression at the time.

To investigate the control of RD gene expression in more detail, we made a systematic study of the contribution of extracellular signaling pathways during the third larval instar, when patterning, specification, and differentiation of individual retinal cells occurs. We used clonal analysis of a battery of mutations affecting signal reception and transduction to explain the spatial and temporal dynamics of RD gene expression in terms of specific roles for Hh, Wg, Notch and Ras signals, in addition to Dpp. We found that all the extracellular signals regulate RD gene expression. The results suggest that Dpp does not play a unique role, and lead to a new discussion of the relationship between RD genes as master regulators and the role of extracellular signals and positional information.

Specification and differentiation of individual retinal cells occur within a domain of the eye imaginal disc that co-expresses So, Eya and Dac. This co-expression domain spreads anteriorly across the eye imaginal disc, progressively replacing expression of three other genes, Ey, Tsh and Hth (Figure 1A-1E) (Bessa et al., 2002). Ey, Tsh and Hth are all DNA-binding transcription factors that interact directly, and promote each other's expression (Bessa et al., 2002). The Ey/Tsh/Hth combination represses expression of Eya and Dac (Bessa et al., 2002). Eya and Dac interact with the DNA-binding protein So, and there is evidence that Dac can also bind DNA (Bonini et al., 1997; Chen et al., 1997; Pignoni et al., 1997; Kim et al., 2002). The combination of So/Eya/Dac promotes expression of each of their own genes, and represses expression of Ey/Tsh/Hth genes (Bessa et al., 2002). Cells in the eye disc may be poised to switch between these two gene expression combinations by the cell- autonomous antagonism between them. However, Hth shuts off earlier than Ey and Tsh do, and So and Eya are expressed earlier in most eye cells than Dac is, suggesting that the notion of a single switch between only two expression states is oversimplified.

Figure 1.

(A) A cartoon of the third instar eye imaginal disc relating proliferation, G1 arrest, the morphogenetic furrow (MF) and differentiation to gene expression domains. Anterior is to the left in this and other figures. Hth is expressed most anteriorly and overlaps with more posterior expression of Ey and Tsh. Ey and Tsh overlap in turn with Eya, So and Dac, which begin expression 5-8 cell diameters anterior to where cell cycle arrest occurs in G1. 6-10 cell diameters posterior to cell cycle arrest, Ey and Tsh disappear, leaving Eya, So and Dac to persist during differentiation. Yellow bar shows the initial stripe of atonal expression in response to Dpp and Hh signaling. (B) Wild type eye disc epithelium showing the expression of Hth (green), Ey (red) and Eya (blue). Arrow indicates the position of the morphogenetic furrow. Individual channels are shown in panels (C) Hth, (D) Ey and (E) Eya. (F-J) The effect of ectopic Dpp signaling (act>tkv*, Lac Z) on (F) Hth, (G) Ey, (H) Tsh, (I) Eya, and (J) Dac levels. Flip-out clones expressing activated Tkv* and β-Galactosidase are labeled green. Arrows in panel (J) indicate anterior clones where Dac is induced poorly.

The dynamic gene expression patterns correlate with a morphogenetic furrow that moves anteriorly across the eye disc and marks the progression of retinal differentiation (Figures 1A-1E) (Bonini et al., 1993; Pichaud and Casares, 2000; Bessa et al., 2002; Kenyon et al., 2003). The morphogenetic furrow moves across the eye disc in response to Hh and Dpp signaling, antagonized by Wg signaling, and with a contribution from N (Ma et al., 1993; Ma and Moses, 1995; Treisman and Rubin, 1995; Greenwood and Struhl, 1999; Curtiss and Mlodzik, 2000; Baonza and Freeman, 2001; Fu and Baker, 2003). Hh, Dpp, and N ligands are all expressed in various cells in or near the posterior, differentiating parts of the retina as it expands; Wg is expressed ahead of the furrow by undifferentiated cells at the lateral and anterior margins. Dpp signaling is already known to contribute to activation of the so, eya, and dac genes, and to accelerate repression of the hth and ey genes(Curtiss and Mlodzik, 2000; Lee and Treisman, 2001; Bessa et al., 2002). These changes are unaffected or only delayed in cells deficient for Dpp signal transduction, suggesting that other signals also contribute to regulate RD genes.

Materials and Methods

Loss of function clones

Clones of cells mutant for genes were obtained by the FLP-mediated mitotic recombination technique with a heat shock inducible FLPase(Golic, 1991; Xu and Rubin, 1993). Homozygous mutant cells wwere identified through the absence of transgene-encoded arm-β-gal (Vincent et al., 1994). Apart from dac⁴, clones of the genotypes below were generated in a Minute heterozygous background (Morata and Ripoll, 1975). Minute heterozygous larvae were heat shocked at 48-96 hours AEL for 1 hour at 37°C.

The following alleles were used:

Mad¹² FRT40(Sekelsky et al., 1995); smo³ FRT40 and smo^D16 FRT40 (Chen and Struhl, 1998); Su(H) ^Δ47 FRT40[w⁺ l(2)35Bg⁺] (Morel and Schweisguth, 2000); dac⁴ FRT40 (Chen et al., 1997); FRT42 arr² (Wehrli et al., 2000); FRT42 mago³ (Boswell et al., 1991); FRT42 shn³(Chase and Baker, 1995); FRT42 shn¹ (Arora et al., 1995). Double and triple mutant clones were obtained using the following chromosomes:

• smo³ Su(H) ^Δ47FRT40 [w⁺ l(2)35Bg⁺];

• smo³ Mad¹² FRT40 [w⁺ l(2)35Bg⁺];

• Mad¹² Su(H) ^Δ47FRT40 [w⁺l(2)35Bg⁺];

• FRT42 mago³ shn³;

• FRT42 arr² shn¹;

• smo³ Mad¹² Su(H) ^Δ47 FRT40 [w⁺ l(2)35Bg⁺];

• FRT42 arr² shn¹.

Marked, Minute chromosomes used included:

• M21 [armLacZ] FRT40;

• FRT42 [armLacZ] M(2)56F;

• FRT82 M(3)96C [armLacZ].

Transgenes

UAS-Arm^S10 (Pai et al., 1997); UAS-Dac (Chen et al., 1997); UAS-Dac (Shen and Mardon, 1997); UAS-So^C.2 (Pignoni et al., 1997); UAS-Ras^V12 (Karim and Rubin, 1998); yw; p[act>CD2>Gal4]; UAS-LacZ (Pignoni and Zipursky, 1997); UAS-GFP^nls.

MARCM analysis

Dac and/or So were overexpressed in smo Mad cells using the MARCM technique (Lee and Luo, 2001). y w hsFLP, UAS-GFP/+; smo³ Mad¹² FRT40/ TubGal80 FRT40; TubGal4/UAS-Dac, y w hsFLP, UAS-GFP/+; smo³ Mad¹² FRT40/ TubGal80 FRT40; TubGal4/UAS-So^C2 and y w hsFLP, UAS-GFP/+; smo³ Mad¹² FRT40/ TubGal80 FRT40; TubGal4/UAS-Dac, UAS-So^C2 were dissected. Wingless signaling was activated in Mad and smo Mad Su(H) clones in y w hsFLP, UAS-GFP/UAS-arm*; Mad¹⁰ FRT40/TubGal80 FRT40; TubGal4/+ and y w hsFLP, UAS-GFP/UAS-arm*; smo³ Mad¹² Su(H)^Δ47 FRT40 [w+ l(2)35Bg+]/TubGal80 FRT40; TubGal4/+. Larvae were heat shocked 24-72 hours AEL. Mutant cells were positively marked and detected with an antibody against GFP or Dac ectopic expression.

Immunohistochemistry

Labeling of eye discs was performed as described (Firth et al., 2006). Preparations were examined on the BioRad Radiance2000 confocal microscope. Images were processed using Adobe Photoshop 6.0 and NIH Image J software. The following antibodies were used: rabbit anti-βGalactosidase (Cappel); mouse anti-βGgalactosidase (mAb40-1a); mouse anti-Dac (2-3) (Mardon et al., 1994); mouse anti-Elav (mAb9F8A9) (O'Neill et al., 1994); rabbit anti-Ey (Halder et al., 1998); mouse anti-Eya (10H6) (Bonini et al., 1993); guinea pig anti-Hth (Casares and Mann, 1998); guinea pig anti-So (Mutsuddi et al., 2005); rabbit anti-Tsh (Wu and Cohen, 2000);mouse anti-Wg (4D4) (Brook and Cohen, 1996); rabbit and mouse anti-GFP (Invitrogen).

Results

In this study we examined the cell-autonomous roles of extracellular molecules by generating clones of cells, located within the developing retinal field, that are mutant for downstream signaling components (Wassarman et al., 1995; Bray, 2006; Clevers, 2006; Affolter and Basler, 2007; Variosalo and Taipale, 2008). These studies aim to identify the extracellular signals to which retinal cells respond directly and cell-autonomously, and to distinguish signals that act indirectly and non-autonomously. Our study also focuses on the cells within the eye field that themselves differentiate to retinal fates; the initiation of the morphogenetic furrow at the posterior eye margin is genetically distinct(Lee and Treisman, 2002; Kenyon et al., 2003).

BMP signaling does not explain all transcriptional transitions

Dpp signaling regulates all the RD genes, and has been proposed as the chief source of their spatial patterning (Bessa et al., 2002). If this was so, we would expect that ectopic Dpp signaling would be sufficient to repress Ey, Tsh, and Hth, and to activate expression of Eya, So, and Dac. Ectopic expression of an activated Dpp receptor (act>tkv*) was used to test this hypothesis. Although ectopic Dpp signaling could repress or activate the expression of some of these genes, it was not sufficient for all. Specifically, Tkv* repressed Hth, but Ey was not repressed and repression of Tsh was variable (Figures 1F-1H)(Bessa et al., 2002). In addition, whereas Tkv* activated Eya expression (Figure 1I), Dac was induced only in cells just ahead of the MF (Figure 1J), corresponding to a modest acceleration of the normal program. Therefore, Dpp signaling was not sufficient to account for spatial patterning of retinal determination gene expression, and other spatial information within the eye disc must be important also.

Hh contributes to Ey and Tsh repression, but does not repress Hth

The most obvious suspect to collaborate with Dpp is Hh. Dpp and Hh together are required for the onset of differentiation and cell cycle arrest as the morphogenetic furrow progresses (Greenwood and Struhl, 1999; Curtiss and Mlodzik, 2000; Firth and Baker, 2005). The contribution of Hh was first assessed with regard to the Hth, Ey and Tsh genes that are repressed as the MF approaches. When the Dpp pathway was eliminated, in clones of cells mutant for Mothers against Dpp (Mad), repression of all three genes was delayed with respect to wild type cells, consistent with previous reports (Figures 2A, 2D and 2G) (Lee and Treisman, 2001; Bessa et al., 2002).

Figure 2. Combinatorial repression of Ey, Tsh and Hth by Dpp and other signals.

The effects of losing of Dpp and Hh signaling on Ey, Tsh and Hth expression were assessed by generating mitotic clones of Mad¹², smo^D16, and smo³ Mad¹² respectively (panels A-I). Hth expression was also examined in additional genotypes (panels J-N). Homozygous clones are outlined by absence of the β-Galactosidase in the magenta channel (except for panel N). Ey, Tsh or Hth proteins were labelled in the green channel and are shown as maximum projections of the z-axis of the disc epithelium. (A) Ey expression persisted in Mad clones but was eventually repressed (eg yellow arrows). (B) Ey expression persisted in smo clones but was eventualy repressed (eg yellow arrows). (C). Ey was never repressed in smo Mad clones (eg blue arrow). (D) Tsh expression persisted in Mad clones but was eventually repressed (eg yellow arrow). (E) Tsh expression persisted in smo clones but was eventually repressed (eg yellow arrows). (F) Tsh was never repressed in smo Mad clones (eg blue arrows). (G) Hth expression persisted in Mad clones but was eventually repressed (eg yellow arrows). In addition, the second phase of Hth expression that occurs in the posterior eye began early in Mad clones (eg orange arrows). Note that presence or absence of Hth depends on location in the anterior-posterior axis, not size of the clones. (H). Hth expression was repressed normally in smo clones (eg yellow arrow). (I-J) Hth expression persisted in smo Mad clones (I) or smo Mad Su(H) clones (J) but was downregulated after a delay (eg yellow arrows). Hth re-expression was accelerated (eg orange arrows). These phenotypes resemble the effect of Mad alone. Hth repression sometimes seems more complete in Mad clones (see panel G), which are expected to maintain higher Ras activity than smo Mad or smo Mad Su(H) clones. (K) Hth expression persisted in shn clones but was eventually repressed (eg yellow arrow). Hth re-expression was accelerated (eg orange arrow). Note that presence or absence of Hth depends on location in the anterior-posterior axis, not size of the clones. (L) Hth was repressed in mago clones (eg yellow arrow), after a slight delay. (M). Hth was never repressed in shn mago clones (blue arrows). (N) act>Ras^V12, GFP clones labeled for GFP in red. Differentiating neural cells are labelled with ElaV in blue. Ras activity repressed Hth in the anterior domain (eg blue arrows), even where there were no differentiating neural cells. Hth was also reduced by Ras activation and ectopic differentiation in the posterior domain (eg yellow arrows).

Cells mutant for smoothened (smo) were used to examine the role of Hh signaling. In smo^- clones near the MF, repression of Ey and Tsh was delayed (Figures 2B and 2E). Repression of Hth occured normally in smo^- clones, however (Figure 2H). These data show that Hh signaling contributes to the normal repression of Ey and Tsh.

The combined roles of Hh and Dpp were examined in clones of cells simultaneously mutant for both Mad and smo. The smo Mad^- clones completely failed to repress Ey or Tsh, and expression of these proteins was maintained in all eye disc cells (Figures 2C and 2F). By contrast, Hth repression was only delayed, similar to what was seen in Mad^-clones (Figure 2I). These findings indicate that ey and tsh expression are repressed by both Dpp and Hh. Each pathway contributes to timely repression, and neither is completely sufficient alone. Cells that cannot respond to Hh or Dpp maintain Ey and Tsh, suggesting that these two pathways together account for the spatial regulation of ey and tsh expression in the eye disc. Certainly, no other signal is present that is sufficient for their repression. By contrast, Hth must be regulated by Dpp and some other spatial signal or signals, distinct from Hh.

Hth repression is controlled by Dpp and Ras signals

The Notch (N) and Ras/MAPK signaling pathways are activated by ligand expression as eye development proceeds. We assessed the role of Notch by generating clones of cells mutant for the transcription factor that acts downstream of N, Suppressor of Hairless (Su(H)). Hth was normal in Su(H) or smo Su(H) clones (data not shown). Although Su(H) has both positive and negative roles in N signaling, (Morel and Schweisguth, 2000; Bray, 2006), Su(H) mutant cells cannot be influenced by N signaling so spatiotemporal differences between Su(H) mutant cells must be responses to other pathways. When Dpp and N signaling were both affected, in Mad Su(H) clones, Hth repression was delayed to the same extent as when only Dpp signaling was removed (data not shown). We also removed Dpp, N and Hh signaling simultaneously. Hth repression was similarly delayed in smo Mad Su(H)^-clones (Figure 2J). Thus, neither N nor Hh had any detectable role in repressing Hth, alone or in the absence of Dpp signaling.

We assessed the role of Ras/MAPK signaling in clones of cells mutant for a component of the Ras/MAPK signal transduction pathway, mago nashi (mago)(J.Y. Roignant and J. Treisman, personal communication). The mago gene is conveniently linked to schnurri (shn), which encodes a Mad co-repressor that is required for most effects of Dpp signaling(Affolter and Basler, 2007). Hth repression was delayed in shn^- clones as it was in Mad clones, confirming that this effect of Dpp required shn (Figure 2K). There was also a small delay in Hth repression in mago^- clones or egfr^- clones (Figure 2L and data not shown). In mago shn^- clones, Hth was not repressed, and Hth levels were maintained, throughout the eye disc (Figure 2M). The data indicate that Hth is repressed by Dpp and Ras/MAPK signaling.

To test whether Ras/MAPK pathway activity is sufficient to repress Hth, we examined clones of cells ectopically expressing an activated form of Ras (act> Ras^V12). As predicted, expression of activated Ras^V12 repressed Hth (Figure 2N). Ectopic Ras/MAPK activity can induce ectopic photoreceptor differentiation (Freeman, 1996), possibly leading to ectopic Dpp expression, but we found that act>Ras^V12 clones repressed Hth even when they lacked ectopic photoreceptors, and in addition that repression appeared cell-autonomous(Figure 2N). Ras^V12 also repressed Hth expression in the peripodial epithelium (data not shown).

These findings indicate that hth expression is repressed by both Dpp and Ras signaling. Each pathway contributes to timely repression at the proper location, and neither is completely sufficient alone. Cells that cannot respond to Dpp or Ras maintain Hth, suggesting that these two pathways together account for the spatial regulation of hth expression in the eye disc. No other signal is present that is sufficient for hth repression.

Maintaining Hth repression depends on Dpp

Hth repression is transient, and Hth is expressed again about 20 hours later in undifferentiated cells with basal nuclei that are fated to contributed to retinal pigment and sensory bristle cells(Pichaud and Casares, 2000). This second stage of hth expression posterior to the morphogenetic furrow began prematurely in clones of Mad^- or shn^- mutant cells unable to respond to Dpp (orange arrows Figures 2G and 2K). This suggests that Dpp signaling represses Hth continuously, and that Hth re-expressed posterior to the furrow is due to increasing distance from the morphogenetic furrow where Dpp is expressed in a stripe (Blackman et al., 1991). Roles of Dpp signaling anterior to the morphogenetic furrow are well-known, but to our knowledge initiation and maintenance of Hth repression provides the first indication that Dpp might diffuse and acts both anterior and posterior of the Dpp expression stripe in the morphogenetic furrow. This is intriguing as phosphorylation of the Mad protein is more apparent anterior to the furrow than in the differentiatin retina (unpublished observations).

Hth re-expression occurred with normal timing in mago clones, or egfr clones, indicating that a drop in MAPK signaling is not sufficient for Hth re-expression (Figure 2L, and data not shown). By contrast, elevated Ras signaling, which recruits ectopic photoreceptor cells, also prevented Hth re-expression posterior to the furrow (Figure 2N). It is possible that the role of Ras signaling in photoreceptor differentiation normally helps restrict Hth expression to undifferentiated cells posterior to the furrow.

Dpp and Hh are required for dac and so expression

The RD genes eya, so and dac are turned on as the MF approaches (Bonini et al., 1993; Cheyette et al., 1994; Mardon et al., 1994; Serikaku and O'Tousa, 1994). Because Dpp signaling was only sufficient to turn on Dac close to the furrow, and Dac expression occurs in cells unable to respond to Dpp, albeit after a delay, we examined the contribution of Hh to Dac regulation. As expected, Dac expression was delayed in Mad^- clones unable to respond to Dpp (Figure 3A). The timing of Dac expression was unaffected in smo^- clones that are unable to respond to Hh, but Dac levels appeared lower (Figure 3B). Dac was cell-autonomously absent (or severly reduced) in smo Mad clones that cannot respond to Hh or Dpp (Figure 3C). Thus, Dpp and Hh act redundantly to establish Dac expression.

Figure 3. Dpp and Hh regulation of Dac and So expression.

Homozygous clones lacking Dpp and/or Hh signaling are outlined by absence of the β-Galactosidase in the magenta channel. Dac or So proteins were labelled in the green channel. (A) In Mad clones the onset of Dac was delayed. (B) In smo clones the level of Dac was reduced. (C). Dac was never expressed in smo Mad clones. (D) In Mad clones So expression was reduced and sometimes delayed. (E) So expression was unaffected in smo clones. (F) So levels were greatly reduced in smo Mad clones (eg yellow arrow).

The regulation of So expression has not been examined previously. The level of So expression was subtly reduced in most internal Mad^- clones that are unable to respond to Dpp (Figure 3D). Sometimes this was accompanied by a short delay (data not shown). Expression was unaffected in smo^- clones that are unable to respond to Hh (Figure 3E). By contrast, So was almost absent in smo Mad clones that cannot respond to Hh or Dpp (Figure 3F). The effect was cell-autonomous. Sometimes a low level of So protein remained. This did not seem to correlate with the clone boundaries, so it was hard to conclude that such residual expression indicates a role for another signaling pathway, although this might be the case. Taken together, the data indicate that Dpp and Hh act redundantly to establish most So expression.

Multiple signals turn on Eya expression

Similar experiments led to a complex picture for regulation of Eya (Figure 4; higher magnification images in Supplementary Figure 1). Dpp signaling is sufficient to turn on eya expression (Figure 1I). Although Dpp signaling contributes to initiating eya expression promptly at the proper location, Eya is only delayed in the absence of Dpp signaling (Figure 4A) (Curtiss and Mlodzik, 2000). We found that although Eya was still expressed in smo Mad cells, there was a longer delay than in Mad clones alone (Figures 4B). This suggested that Hh also contributes to eya expression, but that additional cues can activate Eya more posteriorly.

Figure 4. Multiple signals contribute to Eya expression.

Homozygous clones are outlined by absence of the β-Galactosidase in the magenta channel. Eya protein labelled in green. (A) Mad¹²; (B) smo³ Mad¹²; (C) Mad¹² Su(H)^d47 ci⁹⁴; (D) mago³; (E) shn^TD4; (F) mago³ shn^TD4; (G) arr²; (H) arr² shn^TD4. Higher resolution images of these genotypes are shown at higher magnification in Supplementary Figure 1.

Eya expression was still further delayed in Mad Su(H) ci cells compared to Mad ci, indicating that N signaling contributes to Eya expression (Figure 4C, and data not shown). Eya expression was normal in Su(H) or Su(H) ci clones, showing that Eya was not very dependent on N when Dpp signaling was normal (data not shown). Although ci has both positive and negative roles in Hh signaling(Methot and Basler, 1999; Variosalo and Taipale, 2008), ci mutant cells cannot be influenced by Hh so spatiotemporal differences between ci mutant cells must indicate patterning by other pathways.

Eya expression was reduced and slightly delayed in mago clones that lack Ras signaling (Figure 4D). The absence of Dpp and Ras signaling in mago shn clones did not delay Eya expression more than in shn clones, although Eya levels were also reduced as in mago clones (Figures 4E and 4F). Thus, MAPK signaling appears to contibute to the level of Eya expression, Dpp signaling to the timely onset. These data suggest that in addition to Dpp signaling, Hh, N, and MAPK signaling each make contributions to eya expression. It is possible that still further spatial cues exist, because none of the genotypes yet examined abolish Eya expression.

Role of Wg in regulating RD gene expression

In addition to positive regulators, morphogenetic furrow movement and retinal differentiation are regulated negatively by Wingless (Wg) expression ahead of the furrow (Ma and Moses, 1995; Treisman and Rubin, 1995). The decay of Wg signaling below a threshold could contribute to Eya expression in posterior cells. Consistent with this notion, premature Eya ahead of the morphogenetic furrow has been reported for cells prevented from responding to Wg by over-expression of Axin or mutation of Frizzled-class receptors (Baonza and Freeman, 2002). However, while ectopic Wg signaling delays Eya expression in cells anterior to the MF, it does not prevent expression posterior to the MF (Baonza and Freeman, 2002). Thus, Wg contributes to the pattern of Eya expression, but is not sufficient to define it. We sought to determine whether repression by Wg and activation by other signals were together sufficient to account for the spatial regulation of Eya.

If Wg and Dpp provide the main signals that pattern Eya, increasing Wg signaling in the absence of Dpp signaling should repress Eya throughout the eye field. This model was tested using clones of Mad mutant cells expressing activated armadillo (arm*) (Pai et al., 1997). Although delayed, Eya was still expressed posterior to the morphogenetic furrow in clones expressing arm* that were mutant for Mad, similar to clones of cells expressing arm* that were wild type for Mad (Figures 5A and 5B). Since N and Hh appear to contribute to Eya expression in the absence of Dpp, we also examined smo Mad Su(H) mutant cells expressing arm*. These cells delayed Eya expression very substantially, more than smo Mad Su(H) clones, but Eya was often expressed by columns 7-10 posterior to the morphogenetic furrow (Figures 5C and 5D).

Figure 5. Wg signaling represses Eya expression.

MARCM experiments in which GFP-expressing clones are labelled in magenta. Eya protein labelled in green. All of the genotypes delayed Eya expression (eg yellow arrows), but Eya was expressed eventually (blue arrows). (A) GFP⁺, arm*; (B) GFP⁺, arm*, Mad¹⁰, (C) GFP⁺, smo³ Mad¹² Su(H)^Δ47 (D) GFP⁺, arm*, smo³ Mad¹² Su(H)^Δ47.

These findings were consistent with the notion that Wg signaling contributed to patterning Eya expression, in addition to signals that act positively. However, Wg might not act independently, but instead antagonize one or more positive signals. Conversely, one or more positive signals might contribute only indirectly, by interfering with Wg. These possibilities were investigated for the case of Dpp, using loss of function mutations. As expected, Eya was expressed prematurely in cells mutated for the Wg receptor Arrow (Arr) (Figure 4G). If the role of Wg were to prevent Dpp from inducing Eya, we would expect that Wg signaling would not affect cells unable to respond to Dpp, which would delay Eya expression. By contrast, Eya expression in arr shn double mutant cells began at an intermediate stage, close to that of wild type cells (Figure 4H). This suggests that Wg and Dpp contribute independently to Eya expression, in addition to an underlying pattern of other spatial information that includes the Hh, MAPK and N pathways, and perhaps other pathways not investigated here, which can turn on Eya almost normally in the absence of both Wg and Dpp signals.

Dpp and Hh repress Tsh, but not Ey, through Dac

Our studies identify extracellular sources of patterning for the hth, Ey, tsh, dac, and eya genes that are direct in the sense that cell-autonomous requirements rule out indirect effects mediated by other cell-cell signaling mechanisms. The cell-autonomous responses might be multi-layered, however. For example, regulation of ey, tsh, so and Dac expression by Hh and Dpp signaling could indicate direct binding of the Ci, Mad and Brk transcription factors to each of the Ey, tsh, so and Dac genes (Brk is a repressor controlled by Mad and Shn so that Brk targets are under Dpp regulation (Affolter and Basler, 2007)). Alternatively, Hh and Dpp might directly regulate one of these genes, or even some other transcription factor gene, which then regulates ey, tsh, and dac expression cell-autonomously.

There is already an extensive literature documenting cross-regulation between eye specification genes (Reviewed in (Pappu and Mardon, 2002)). We found evidence that Tsh is partly regulated though Dac, which is the more direct target of Dpp and Hh signaling. In dac^- clones, Tsh was not repressed posterior to the MF, and repression of Ey was delayed (Figures 6A and 6B). Reduced Ey expression was still present in more posterior dac^- clones however (arrow in figure 6B). If Hh and Dpp repress Ey and Tsh by establishing Dac, we would predict that Ey and Tsh would be repressed in smo Mad clones if Dac expression was restored. We found that smo Mad^- clones that expressed Dac repressed Tsh but not Ey (Figures 6C and 6D). This indicated that Dpp and Hh signaling were dispensable for Tsh repression if Dac was present, consistent with Dac expression mediating the repression of Tsh posterior to the furrow. Expression of Dac was not sufficient to repress Tsh prematurely ahead of the furrow, however, and the two proteins are coexpressed in some cells in normal development. Although Dac was partially required for Ey repression, ectopic Dac was not sufficient to repress Ey in smo Mad mutant cells (data not shown), indicating that there is another target of Hh and Dpp that is required.

Figure 6. Hh and Dpp repress Tsh through Dac.

(A) dac⁴ clones (absence of magenta) labeled for Tsh (green). Tsh repression required dac. (B) dac⁴ clones (absence of magenta) labeled for Ey (green). Ey repression was delayed but eventually occured without dac (eg yellow arrow). (C) smo³ Mad¹² mutant cells expressing ectopic Dac and GFP (magenta). Tsh (green) was repressed. (D) smo³ Mad¹² mutant cells expressing ectopic Dac and GFP (magenta). Ey(green) was not repressed. (E) smo³ Mad¹² mutant cells expressing ectopic So and GFP (magenta). Ey(green) was not repressed. (F). smo³ Mad¹² mutant cells expressing ectopic So and GFP (magenta). Tsh(green) was not repressed. (G) smo³ Mad¹² mutant cells expressing ectopic Dac and So (magenta). Ey(green) was not repressed. (H) smo³ Mad¹² mutant cells expressing ectopic Dac, So and GFP (magenta). Ato(green) was not expressed.

As there is evidence that So might repress Ey, (Pignoni et al., 1997), we tested whether Hh and Dpp repressed Ey through So by expressing So in smo Mad^- mutant cells. However, So was not sufficient to repress Ey (Figure 6E). So also did not restore Tsh repression in smo Mad^- mutant cells (Figure 6F). Co-over expression of Dac and So had only minor effects on Ey expression in smo Mad^- mutant cells (Figure 6G).

We also noted that expression of Dac and So failed to restore either Ato expression or ommatidial differentiation to smo Mad clones, even at stages where smo Mad cells express Eya (Figure 6H, and data not shown).

Discussion

Retinal Determination gene expression is regulated spatially and temporally by extracellular signals

We found that all the extracellular signaling pathways played roles in spatial and temporal regulation of RD genes. Although Dpp was important, none of the hth, tsh, ey, eya, dac, and so genes were regulated exclusively by Dpp. Instead, three distinct combinatorial ‘codes’, comprising Dpp+Hh, Dpp+Ras, and Dpp+Hh+Ras+Notch+Wg, each contributed to the expression of particular RD genes. Hth was repressed redundantly by Dpp and Ras signaling, Tsh, Ey, So and Dac were regulated redundantly by Dpp and Hh signals, and Eya regulation was affected by all five extracellular signaling pathways examined. Therefore, dynamic RD gene expression is a function of extracellular signaling by all the core signaling pathways, and perhaps by still further spatial cues not examined in our study (Figure 7).

Figure 7. Extracellular signals regulate and act through retinal determination genes.

A cartoon showing the spatial signals that affect retinal determination gene expression in the third instar eye disc, as the morphogenetic furrow induces differentiation within cells coexpressing Eya, So, and Dac, and extinguishes Hth, Ey and Tsh expression. Clonal analysis has defined cell-autonomous effects of extracellular signaling pathways, which must be ‘direct’ in that any intermediate steps have to be intracellular. The morphogenetic furrow is associated with and driven forwards by expression of Dpp and Hh; ligands for N and receptor tyrosine kinases are also expressed. Effects on growth suggest that EGFR and Ras function anterior to the furrow as well as posterior, although the ligands involved are uncertain. Non-autonomous interactions between these signals, which contribute to their pattern of expression, are not shown here. Morphogenetic furrow movement is antagonized by Wg, which is expressed in the anterior eye disc. The retinal determination genes are color-coded to indicate which extracellular signals are responsible for their spatial and temporal regulation. Roles of Wg signaling in the initiation of Hth, Ey and Tsh expression at earlier developmental stages are not shown. Potential indirect regulation of the Ey and Tsh genes is indicated by grey arrows. Our findings are consistent with repression of Tsh and to some extent Ey by Hh occurring indirectly via the activation of Dac. The data do not exclude additional, parallel regulation by Hh and Dpp directly, however. In addition, any of the regulatory relationships shown here may act through unidentified intermediates, so long as these function cell-autonomously. For example, Dac might regulate Tsh or Ey indirectly.

Two kinds of data show that signals other than Dpp were important in wild type flies, and did not just contribute to robustness or as back-up pathways that are mainly relevant in mutant genotypes. First, Dpp signaling alone was not sufficient to regulate all the genes in the third instar. Specifically, we found that activated Tkv was not sufficient for Ey repression, only partially repressed Tsh, and was not sufficient to turn on Dac throughout the eye field (Figure 1F-J). Secondly, cells that remain able to respond to Dpp signaling, but are deficient in responding to other signals, showed abnormal patterns of gene expression that indicated roles for other signals even when Dpp signaling was intact. Specifically, cells defective in Ras signaling delayed repression of Hth (Figure 2L) and had reduced levels of Eya (Figure 4D); cells defective in Hh signaling had reduced Dac levels and delayed repression of Ey and Tsh (Figures 2B, 2E, and 3B). Therefore, Dpp signaling is important but not sufficient by itself to specify the normal spatio-temporal pattern of retinal determination gene expression.

Retinal Determination genes may define eye-specific, indirect target genes of extracellular signals

One idea has been that retinal determination proteins define the eye field through eye-specific responses to core extracellular signaling pathways (Curtiss et al., 2002; Mann and Carroll, 2002). How can RD genes play such a high-level role if RD gene expression is itself changing dynamically in response to all the signaling pathways, like all the effector genes responsible for eye differentiation and morphogenesis? Rather than surrender the idea that RD genes, like other master regulatory genes, define how a particular organ interprets positional information, we suggest another possible explanation. If dynamic RD gene expression captures enough features of the extracellular signaling pattern, overlapping expression patterns of RD genes can themselves encapsulate the activity of extracellular signals, act as a relay of this positional information. In this view, target genes of RD proteins are being regulated by the extracellular positional information, but indirectly, through the dynamic, overlapping combinations of RD gene expression. This regulation would be eye-specific. For example, a gene that lacks any binding site for Ci could respond cell autonomously to Hh signaling through regulation by one or more retinal determination genes that are regulated by Hh, and this Hh-responsiveness would be eye specific. This model would only be plausible if RD gene expression reflected the activity of all the major extracellular signaling pathways, which we find is indeed the case.

Is there any evidence for eye genes that are regulated by extracellular signals indirectly through an RD gene relay? The atonal gene, which encodes the proneural bHLH protein for the founder R8 photoreceptor cells of the neural retina, may be an example(Tanaka-Matakatsu and Du, 2008). Hh and Dpp signaling induce ato transcription in a stripe just ahead of the morphogenetic furrow (Jarman et al., 1994; Greenwood and Struhl, 1999; Curtiss and Mlodzik, 2000). This ato expression depends an enhancer located 3′ to the gene, yet the 384bp that contain this enhancer lack any consensus binding site for the transcription factors Ci, Mad, or Brk, which mediate Hh and Dpp signaling (Sun et al., 1998; Zhang et al., 2006)(our unpublished observations). Instead, the enhancer activity depends on binding sites for Ey and So (Zhang et al., 2006). Our study now shows that the expression of both ey and so depends on Hh and Dpp signaling. Therefore the Hh- and Dpp-dependent transcription of ato may be explained by Hh- and Dpp-regulation of ey and so. There must be another Hh/Dpp-dependent factor in addition to Ey and So, however, because restoring So (and Dac) expression did not rescue ato expression to smo Mad mutant cells (these smo Mad cells already express Ey and Eya) (Figure 6H). We predict that the missing factor depends heavily on Hh signaling, because ato expression is more critically dependent on smo than on Dpp signaling (Dominguez, 1999; Fu and Baker, 2003).

It will require more than one example for the positional information relay model to prove useful. It will be interesting to determine whether Dpp regulates ey and tsh through Mad or Brk binding, or indirectly thorugh Hth and Dac (Figure 7). The Ras and Notch pathways have many roles in eye-specific cell fates, and it will be interesting to determine whether some of the target genes involved are regulated indirectly through Hth and Eya, which respond to Ras and Notch signals (Figure 7). The ‘positional-information relay’ model does not mean that RD gene products cannot act in combination with extracellular signals on the same promoters as has been proposed before. Indeed this is likely at some promoters (Curtiss et al., 2002; Mann and Carroll, 2002; Rogers et al., 2005; Hayashi et al., 2008) and probably necessary in activating expression of the RD genes themselves (Bessa and Casares, 2005; Pappu et al., 2005). We suggest, however, that evolution used RD genes and the relay mechanism to expand the repertoire of eye-specific gene expresion and recruit target genes that lack binding sites for direct transducers of extracellular signals.

Supplementary Material

01. Supplementary Figure 1. Many signals contribute to Eya expression.

This figure shows Eya expression in the same genotypes shown in Figure 4 of the main text, but at higher resolution and magnification. Homozygous clones are outlined by absence of the β-Galactosidase in the magenta channel. Eya protein labelled in green. (A) Mad¹²; (B) smo³ Mad¹²; (C) Mad¹² Su(H)^d47 ci⁹⁴; (D) mago³; (E) shn^TD4; (F) mago³ shn^TD4; (G) arr²; (H) arr² shn^TD4.