Signaling by the secreted hedgehog, decapentaplegic and wingless proteins organizes the pattern of photoreceptor differentiation within the Drosophila eye imaginal disc; hedgehog and decapentaplegic are required for differentiation to initiate at the posterior margin and progress across the disc, while wingless prevents it from initiating at the lateral margins. Our analysis of these interactions has shown that initiation requires both the presence of decapentaplegic and the absence of wingless, which inhibits photoreceptor differentiation downstream of the reception of the decapentaplegic signal. However, wingless is unable to inhibit differentiation driven by activation of the epidermal growth factor receptor pathway. The effect of wingless is subject to regional variations in control, as the anterior margin of the disc is insensitive to wingless inhibition. The eyes absent and eyegone genes encode members of a group of nuclear proteins required to specify the fate of the eye imaginal disc. We show that both eyes absent and eyegone are required for normal activation of decapentaplegic expression at the posterior and lateral margins of the disc, and repression of wingless expression in presumptive retinal tissue. The requirement for eyegone can be alleviated by inhibition of the wingless signaling pathway, suggesting that eyegone promotes eye development primarily by repressing wingless. These results provide a link between the early specification and later differentiation of the eye disc.
A small number of signaling pathways is used repeatedly to direct developmental processes in both Drosophila and vertebrates. Members of the hedgehog (Fietz et al., 1994), TGF-β (Kingsley, 1994) and Wnt (Cadigan and Nusse, 1997) families, typified by hedgehog (hh), decapentaplegic (dpp) and wingless (wg) in Drosophila, establish patterns of growth and differentiation at multiple stages of development. However, the interactions between these pathways can vary, allowing each tissue and appendage to acquire its characteristic properties. In Drosophila, hh can activate the expression of either wg or dpp (Basler and Struhl, 1994; DiNardo et al., 1994; Heberlein et al., 1993; Ma et al., 1993), while dpp and wg signaling can interact antagonistically (Brook and Cohen, 1996; Jiang and Struhl, 1996; Penton and Hoffmann, 1996; Theisen et al., 1996; Treisman and Rubin, 1995) or cooperate to activate specific targets (Campbell et al., 1993; Lecuit and Cohen, 1997; Riese et al., 1997). Both the patterns of expression of these genes, and the rules governing their interactions, must be established for each tissue by specific regulators.
The spatial control of differentiation in the eye imaginal disc requires the coordination and regulation of these three signals. During the third larval instar, differentiation of photoreceptor clusters begins at the posterior margin of the eye disc and gradually spreads anteriorly (Ready et al., 1976). The initiation of differentiation requires hh, which is expressed at the posterior margin of second instar larval discs and activates dpp expression there (Dominguez and Hafen, 1997; Royet and Finkelstein, 1997). dpp function is also required for initiation, which does not occur in cells unable to receive the dpp signal (Burke and Basler, 1996; Chanut and Heberlein, 1997; Wiersdorff et al., 1996). Although dpp is expressed at the lateral margins as well as the posterior margin (Masucci et al., 1990), initiation from the lateral margins is prevented by the presence of wg (Ma and Moses, 1995; Treisman and Rubin, 1995). wg expression is absent from the posterior margin due to dpp signaling (Wiersdorff et al., 1996), and its ectopic expression there can block initiation (Treisman and Rubin, 1995). However, it is not clear how the expression domains of dpp and wg are first established. Because hh pathway activity leads to dpp expression in posterior regions of the eye disc and wg expression in anterior regions (Dominguez and Hafen, 1997; Royet and Finkelstein, 1997; Heberlein et al., 1995), other factors must help to determine the specificity of the response to hh.
The anterior progression of differentiation is also driven by hh and dpp signaling, as inactivation of either protein using temperature-sensitive mutations arrests the process (Chanut and Heberlein, 1997; Ma et al., 1993). However, local loss of cell-autonomous downstream components of either signaling pathway does not have a dramatic effect on progression (Burke and Basler, 1996; Penton et al., 1997; Strutt and Mlodzik, 1997; Wiersdorff et al., 1996), suggesting that there is some redundancy between the two pathways, or that a third signal is involved. hh, now expressed in the differentiating photoreceptors, is required to activate a stripe of dpp expression in the morphogenetic furrow (Heberlein et al., 1993; Ma et al., 1993), the point at which cells undergo a shape change just prior to their differentiation (Ready et al., 1976). Ectopic expression of hh, but not dpp, is sufficient to trigger ectopic photoreceptor differentiation anterior to the normal furrow (Heberlein et al., 1995; Pignoni and Zipursky, 1997); however, ectopic dpp is effective in initiating a new morphogenetic furrow from the anterior margin of the disc, and can do so from a considerable distance (Pignoni and Zipursky, 1997; Chanut and Heberlein, 1997). The progression of differentiation through internal regions of the eye disc is still inhibited by ectopic wg (Treisman and Rubin, 1995).
The process by which hh or dpp signaling leads to photoreceptor differentiation is not entirely clear; it involves the expression of atonal (ato), a proneural gene encoding a helix-loop-helix (HLH) protein that is absolutely required for photoreceptor formation (Jarman et al., 1993, 1994, 1995), and the repression of hairy, encoding another HLH protein that inhibits premature photoreceptor formation in combination with extramacrochaetae (Brown et al., 1995; Heberlein et al., 1995). ato appears to specify R8, the first photoreceptor to form in each cluster; repeated activation of the epidermal growth factor (EGF) receptor signaling pathway by the spitz (spi) ligand then recruits the remaining cells of the cluster (Freeman, 1994, 1996; Tio et al., 1994; Tio and Moses, 1997).
Prior to the initiation of photoreceptor differentiation, a group of genes including eyeless (ey), eyes absent (eya), sine oculis (so), eyegone (eyg) and dachshund (dac) acts to determine the fate of cells in the eye disc, and is likely to control any differences in signaling mechanisms between the eye disc and other imaginal discs. All these genes are required for eye formation and have some ability to induce ectopic eye development in other imaginal discs (Bonini et al., 1993, 1997; Chen et al., 1997; Cheyette et al., 1994; Halder et al., 1995; Mardon et al., 1994; Pignoni et al., 1997; Quiring et al., 1994; Serikaku and O’Tousa, 1994; Shen and Mardon, 1997; C. Desplan and H. Sun, personal communication). They all encode nuclear proteins: ey encodes Pax-6 (Quiring et al., 1994) and eyg another Pax-like protein (C. Desplan and H. Sun, personal communication); so encodes a divergent homeodomain protein (Cheyette et al., 1994; Serikaku and O’Tousa, 1994), and eya and dac encode novel nuclear factors (Bonini et al., 1993; Mardon et al., 1994). The eya protein has been demonstrated to interact molecularly with both so and dac (Chen et al., 1997; Pignoni et al., 1997), suggesting that complexes between these molecules may activate target genes required for eye development. ey is first expressed in the embryonic eye disc primordium (Quiring et al., 1994), while so is expressed and required in the entire visual system (Cheyette et al., 1994; Serikaku and O’Tousa, 1994). Later expression of so and eya in the eye disc in the second and early third instar stages, in gradients with their highest levels at the posterior margin (Bonini et al., 1993; Mardon et al., 1994), is dependent on ey (Halder et al., 1998); expression of dac in a similar pattern requires eya (Chen et al., 1997). While ectopic ey expression efficiently induces eye development on the legs, wings and antennae (Halder et al., 1995), the other genes have a weaker ability to induce ectopic eyes and only do so appreciably when eya is expressed in combination with either so or dac (Bonini et al., 1997; Chen et al., 1997; Pignoni et al., 1997; Shen and Mardon, 1997).
We have examined the mechanism by which wg prevents photoreceptor development; we show that wg acts downstream of the dpp receptor thick veins and thus its inhibitory effect is not mediated by repression of the expression or activity of the dpp protein. One of the consequences of dpp function is the induction of photoreceptor development; this also requires activation of the GTP-binding protein ras by EGF receptor signaling (Simon et al., 1991; Xu and Rubin, 1993; Freeman, 1996). We show here that wg acts upstream of ras activation. To determine how the expression patterns of dpp and wg are regulated, we have tested the effects of eyg and eya, two genes essential for eye development. We show that both of these genes contribute to the activation of dpp expression and the inhibition of wg expression. The absence of photoreceptor development observed in eyg mutants can be rescued by inhibition of wg signaling, suggesting that repression of wg expression is a critical function of eyg.
MATERIALS AND METHODS
Alleles used were eyg1 (Lindsley and Zimm, 1992), Df(3L)iro2 (Gomez-Skarmeta et al. 1996), eya1 (Bonini et al., 1993), MadB1 (Wiersdorff et al., 1996), wgCX2 (Baker, 1987), puntl(3)10460 (Ruberte et al., 1995), sogY506 (Ferguson and Anderson, 1992), omb282 (Lecuit et al., 1996), and l(1)ombD4 (Grimm and Pflugfelder, 1996). The reporters were dpp-lacZ BS3.0 (Blackman et al., 1991) and wgP (Kassis et al., 1992). Transgenic lines used were UAS-tkvQD (Nellen et al., 1996), UAS-dTCFΔN (van de Wetering et al., 1997), UAS-dpp (Frasch, 1995), UAS-wg, UAS-hh (Azpiazu et al., 1996), UAS-s-spi (Schweitzer et al., 1995b), UAS-rasv12 (Karim and Rubin, 1998), UAS-fluΔarm (Zecca et al., 1996), UAS-omb (Grimm and Pflugfelder, 1996), and Act>CD2>GAL4 (Pignoni and Zipursky, 1997). dpp-GAL4 was constructed by cloning the 10 kb KpnI-XbaI-fragment of the dpp 3′ region used to make BS3.0 (Blackman et al., 1991) upstream of a NotI-KpnI fragment containing the minimal hsp70 promoter (Hiromi and Gehring, 1987) including 60 bp upstream and 200 bp downstream of the transcription start site, and placing this upstream of a KpnI-XbaI fragment containing the GAL4-coding region and the α-tubulin 3′UTR (Brand and Perrimon, 1993) in the pCasPeR4 vector (Pirrotta, 1988). Unlike the shorter enhancer previously used (Staehling-Hampton et al., 1994), this driver gives expression of lacZ at the posterior margin that is as strong as its expression at the lateral margins. However, expression does not move anteriorly with the morphogenetic furrow. UAS-sggact (also referred to as UAS-sggS9A) was constructed by site-directed mutagenesis of the cDNA coding for the most abundant sgg protein (SGG10) according to the protocols of the pALT system (Promega). The codon corresponding to ser-9 was changed into ala and verified by sequencing. A mutagenised 2.0 kb BglII-XbaI fragment was then cloned into the pUAST vector (Brand and Perrimon, 1993). ey-GAL4 was constructed by cloning a 3.6 kb EcoRI fragment containing the eye-specific enhancer of the ey gene into the vector p221-4 (a gift of E. Knust). p221-4 contains the GAL4 gene with a hsp70 minimal promotor in front of it.
To make loss-of-function clones, puntl(3)10460, eyg1, eya1, MadB1, sogYS04, omb282 and l(1)ombD4 were recombined with FRT elements at positions 82, 80, 40 and 18, respectively (Xu and Rubin, 1993). An FRT element at position 40 was recombined successively with wgCX2 and MadB1 to create a doubly mutant chromosome arm. Males of the resulting FRT lines were crossed with females carrying the same FRT element, an arm-lacZ (Vincent et al., 1994) P element on the same chromosome arm (except for the punt clone in Fig. 1B and the eya clone in Fig. 4B), and either hsFLP1 (Xu and Rubin, 1993) or eyFLP1 (a gift of B. Dickson). Crosses using hsFLP1 were heat shocked for 1 hour at 38°C in both first and second instar. To make gain-of-function clones, a stock carrying hsFLP1, Act>CD2>GAL4 and either UAS-lacZ or wg-lacZ was constructed and crossed to other UAS lines, either individually or in combinations. Larvae were heat shocked 30 minutes at 37°C in either second instar (for combinations including UAS-dpp) or first instar (for all other lines and combinations).
Eye discs were stained as described by Treisman and Rubin (1995), except that the fix used was 4% formaldehyde in PEM. Rat anti-elav (Robinow and White, 1991) was diluted 1:1, mouse anti-wg (Brook and Cohen, 1996) was diluted 1:10, mouse anti-omb (Grimm and Pflugfelder, 1996) was diluted 1:100, mouse anti-dac (Mardon et al., 1994) was diluted 1:5 and rabbit anti-ato (Jarman et al., 1995) was diluted 1:5000.
Furrow initiation requires dpp signaling even in the absence of the inhibitory wg signal
The initiation of photoreceptor development at the posterior margin of the eye disc requires dpp signaling (Burke and Basler, 1996; Chanut and Heberlein, 1997; Heberlein et al., 1993; Pignoni and Zipursky, 1997; Wiersdorff et al., 1996; Fig. 1A), and loss-of-function of components of the dpp pathway at the posterior margin results in the ectopic expression of wg in the mutant cells (Wiersdorff et al., 1996; Fig. 1B). Since the presence of wg at this position is sufficient to prevent morphogenetic furrow initiation (Treisman and Rubin, 1995), it is possible that the only requirement for dpp in initiation is to repress wg. This hypothesis has been proposed by Dominguez and Hafen (1997), based on their observation that clones of cells mutant for protein kinase A (PKA), in which the hh pathway is ectopically activated, can develop as photoreceptors in the anterior of the disc even when they lack both dpp and wg. However, it is not clear that the mechanism of normal furrow initiation can be inferred from the effects of loss of PKA in anterior regions. As a more direct test, we examined clones of cells mutant for both Mothers against dpp (Mad), which encodes an intracellular component required to transduce the dpp signal (Raftery et al., 1995; Sekelsky et al., 1995; Newfeld et al., 1996), and a null allele of wg. Cells in these clones are unable to respond to dpp, but are also unable to produce wg. When such clones of cells occur at the posterior margin of the eye disc, they autonomously fail to initiate photoreceptor development (Fig. 1C). Clones of cells singly mutant for Mad also fail to differentiate as photoreceptors, but often have an additional non-autonomous inhibitory effect on photoreceptor differentiation by surrounding cells, which is likely to be mediated by wg (Fig. 1A). Thus dpp signaling is required not only to repress wg expression, but also independently for morphogenetic furrow initiation.
wg inhibits photoreceptor formation downstream of the dpp receptors
wg is required to prevent ectopic morphogenetic furrow initiation from the lateral margins of the eye disc (Ma and Moses, 1995; Treisman and Rubin, 1995). However, the mechanism by which wg inhibits photoreceptor differentiation is not well understood. It has been suggested that wg acts by preventing dpp expression, as dpp expression is lost in clones of cells lacking the kinase encoded by shaggy/zeste-white 3 (sgg) (Heslip et al., 1997), which normally functions to inhibit the wg pathway. However, a low level of ectopic wg can inhibit photoreceptor differentiation without reducing dpp expression (Treisman and Rubin, 1995). As dpp positively autoregulates its own expression (Wiersdorff et al., 1996), inhibition of dpp function may result in a loss of dpp expression. The ability of wg to inhibit differentiation in the presence of dpp is illustrated in Fig. 2B, in which an eye-specific enhancer from the eyeless (ey) gene (Quiring et al., 1994) drives GAL4 to express a UAS-wg transgene throughout the eye disc beginning before the stage of furrow initiation (insets in Fig. 2B show the pattern of UAS-lacZ expression driven by ey-GAL4 in early and late third instar discs). If wg acted by inhibiting dpp expression, it should be possible to overcome its effects by expressing dpp from a heterologous promoter. However, co-expression of dpp and wg either under the control of the ey-GAL4 driver or using hsFLP-induced recombination to fuse GAL4 to the constitutive Actin5C promoter (flp-out-GAL4; Pignoni and Zipursky, 1997) did not allow initiation of photoreceptor development at the posterior margin (Fig. 2D,F).
Ectopic expression of dpp in the eye disc has been shown to specifically induce initiation of photoreceptor differentiation from the anterior margin of the disc in a non-autonomous fashion (Chanut and Heberlein, 1997; Pignoni and Zipursky, 1997; Fig. 2C,E). Surprisingly, we observed that this ectopic differentiation was not inhibited by wg signaling. Co-expression of dpp and wg throughout the disc under ey-GAL4 control resulted in initiation from the anterior margin at a much higher frequency than from the posterior margin (Fig. 2D). Furthermore, clones of cells co-expressing dpp and wg under the control of the flp-out-GAL4 system were still able to induce anterior morphogenetic furrow initiation, although clones at the posterior margin blocked photoreceptor differentiation (Fig. 2F). Thus initiation from the anterior margin must be able to overcome the inhibition normally caused by wg.
Another possible way for wg to inhibit dpp-induced photoreceptor differentiation would be by reducing the activity of the dpp protein. For example, wg might act on short gastrulation (sog), which encodes a secreted molecule thought, by analogy to its Xenopus homolog chordin, to bind the dpp protein in the ventral region of the embryo and prevent it from binding to its receptors (Holley et al., 1995; Piccolo et al., 1996; Schmidt et al., 1995). However, loss of sog function at the lateral margins of the eye disc, unlike loss of wg, did not induce premature photoreceptor differentiation (data not shown). To test directly whether wg acts on or downstream of the dpp protein, we co-expressed wg with a constitutively active form of the dpp type I receptor thick veins (tkvQD; Nellen et al., 1996; Lecuit et al., 1996) in the eye disc using ey-GAL4. Activated tkv was unable to overcome the inhibition caused by wg (Fig. 2H; compare to Fig. 2G for tkvQD alone), suggesting that wg acts downstream of or in parallel to this receptor.
wg inhibits photoreceptor differentiation upstream of ras activation
Rather than affecting the dpp pathway directly, wg might block photoreceptor differentiation at a stage subsequent to dpp signaling. Formation of all photoreceptors is known to depend on the EGF receptor and its downstream component ras (Simon et al., 1991; Xu and Rubin, 1993). Furthermore, wg has recently been shown to antagonize EGF receptor signaling during the specification of the cuticle pattern in the embryo (O’Keefe et al., 1997; Szuts et al., 1997). To determine whether wg also acts on this pathway in the eye, we tested whether a secreted and active form of the ligand spitz (s-spi; Schweitzer et al., 1995b) or a constitutively active form of ras (Fortini et al., 1992; Karim and Rubin, 1998) could bypass the block caused by wg. In discs expressing both wg and activated ras ubiquitously, we observed extensive photoreceptor differentiation and growth (Fig. 2L), as in discs expressing activated ras alone (Karim and Rubin, 1998; Fig. 2K). Thus wg must act upstream of ras activation to block differentiation. Expression of s-spi also rescues photoreceptor differentiation in discs expressing wg ectopically (Fig. 2J; compare to Fig. 2I for s-spi alone). The rescue is less robust than that caused by ras activation, either because expression of s-spi is not sufficient to fully activate ras, or because wg blocks stages both upstream and downstream of spi activity.
To determine whether the inhibition of photoreceptor differentiation is mediated by the conventional wg signal transduction pathway, we tested the ability of a constitutively active form of armadillo (Δarm; Zecca et al., 1996), the β-catenin homolog thought to participate in transcriptional activation of wg target genes (Peifer and Wieschaus, 1990; van de Wetering et al., 1997), to block photoreceptor differentiation. When we expressed activated arm in clones of cells using flp-out-GAL4, we indeed observed a block of photoreceptor differentiation within the clone (Fig. 2M). This block was not rescued by co-expression of activated tkv (Fig. 2N), but was overcome by co-expression of activated ras (Fig. 2P), consistent with the results for ectopic wg expression. We also tested whether a transcription factor known to be induced by wg at the lateral margins, optomotor-blind (omb; Pflugfelder et al., 1992; Zecca et al., 1996), could mediate the inhibition. We found, however, that ectopic expression of omb inhibited cell growth, making it difficult to evaluate its effect on differentiation (data not shown). Loss of omb function at the lateral margins did not lead to ectopic photoreceptor differentiation (data not shown), so it is likely that other target genes contribute to this effect of wg. Another known target gene in the eye disc, which we have not tested, is orthodenticle (Royet and Finkelstein, 1997).
The expression patterns of dpp and wg are regulated by eyes absent and eyegone
The normal restriction of morphogenetic furrow initiation to the posterior margin of the eye disc is due to the presence of hh and dpp and the absence of wg at this position (Dominguez and Hafen, 1997; Baker, 1988; Masucci et al., 1990). Although negative regulation of wg expression by dpp, and dpp function by wg, provides a mechanism for the maintenance of their expression domains, it does not explain how they are established. We therefore examined whether genes implicated in early events of eye development are involved in the regulation of dpp and wg expression. One such gene is eyes absent (eya), which encodes a novel nuclear protein required for eye formation (Bonini et al., 1993). No photoreceptors form in eya1 mutant eye discs and extensive cell death reduces the size of the third instar disc (Bonini et al., 1993). However, prevention of cell death does not appear to be the primary function of eya, as clones of eya mutant cells proliferate extensively prior to the third instar stage (Pignoni et al., 1997), although they are replaced by wild-type head cuticle in the adult eye (data not shown). eya mutant cells fail to differentiate as photoreceptors, resembling sgg mutant cells, in which the wg pathway is over-active (Heslip et al., 1997; Treisman and Rubin, 1995). We examined the expression of dpp and wg in eya mutant eye discs and in clones of eya mutant cells. dpp-lacZ expression was greatly reduced in early third instar eya mutant discs, prior to the initiation of the morphogenetic furrow (Fig. 3A,B), and was completely lost in eya mutant clones (Fig. 4B; Pignoni et al., 1997), suggesting that eya is required for dpp transcription. Although the initiation of wg expression in early eya mutant eye discs appeared to be normal (Fig. 3D,E), ectopic wg protein was observed in eya mutant clones in late third instar discs (Fig. 4C). This wg protein appears to be active, as omb, a target of wg in the eye disc (Zecca et al., 1996), was also expressed in eya clones (Fig. 4D).
The phenotypes of eye-specific mutations in eya and so are very similar, suggesting that these genes act at the same level in the hierarchy leading to eye disc specification (Pignoni et al., 1997). ey appears to act upstream of eya and so in both normal and ectopic eye development (Halder et al., 1998). We have therefore not examined the effects of mutations in so or ey. Another gene required for eye formation that has not been placed within this hierarchy is eyegone (eyg; Hunt, 1970); in its absence, no photoreceptors differentiate and the eye disc does not reach its normal size and shape (Figs 3C,F, 5A). We examined the expression patterns of dpp and wg in early third instar eyg mutant discs. dpp expression was restricted to the posterior margin of eyg mutant discs, in contrast to its expression around the posterior and lateral margins of wild-type discs (Fig. 3A,C). On the contrary, wg expression was expanded, especially on the dorsal side of the disc, where it extended to the posterior margin (Fig. 3D,F). eyg thus acts to delimit the domains of dpp and wg expression; since it encodes a Pax-like transcription factor (C. Desplan and H. Sun, personal communication), it is possible that this regulation is direct. Most clones of eyg mutant cells develop normally (Fig. 4E) and do not affect dpp or wg expression (data not shown). Possibly eyg mutant clones are rescued by dpp, hh or another secreted factor diffusing in from surrounding wild-type cells. Alternatively, eyg may act on a localized region or during a restricted time period in development.
The eyg phenotype results from its effect on wg expression
To determine whether these effects on dpp and wg expression were the basis for the effects of eya and eyg on photoreceptor development, we tested whether their mutant dpp itself at this position failed to rescue photoreceptor formation (Fig. 5C and data not shown). To inhibit the wg pathway, we expressed a constitutively active form of the protein kinase encoded by sgg, made by mutating serine-9, a site for inhibitory phosphorylation of the mammalian homolog glycogen synthase kinase-3β (Cross et al., 1995; Stambolic and Woodgett, 1994). sgg negatively regulates the activity of arm (Peifer et al., 1994), so a hyperactive form of sgg should block phenotypes could be rescued by restoration of dpp signaling or by inhibition of wg signaling. To specifically target initiation of the morphogenetic furrow in discs transheterozygous for eyg1 and a deficiency removing eyg, which completely lack photoreceptors (Fig. 5A), and to avoid early effects on growth of the eye disc, we used a dpp-GAL4 driver (see Materials and Methods), which directed expression at the posterior margin of wild-type and eyg mutant discs (Fig. 5B and data not shown). Activation of dpp signaling by expression of tkvact or transmission of the wg signal. Indeed, expression of this form of sgg at the wing margin prevented differentiation of the wg-dependent margin bristles (data not shown). We also used a dominant negative form of the wg-responsive dTCF transcription factor (van de Wetering et al., 1997). Expression of either of these molecules led to the initiation of a morphogenetic furrow in the eyg mutant discs (Fig. 5E,F) and formation of small adult eyes (data not shown). Since inhibition of the wg pathway at the posterior margin of eyg mutant discs is sufficient to allow photoreceptor formation, we conclude that the misexpression of wg observed at the posterior of the eyg mutant discs is a major cause of the absence of photoreceptor development. As expected since it can overcome the effect of ectopic wg (Fig. 2L), activated ras was also able to rescue photoreceptor differentiation in eyg mutant discs (Fig. 5G,H). Interestingly, expression of hh, which is sufficient to induce photoreceptor differentiation in the eye disc (Heberlein et al., 1995; Dominguez and Hafen, 1997), was unable to do this in the absence of eyg (Fig. 5D), showing that hh cannot overcome the inhibition of initiation caused by wg; the same conclusion was reached by co-expression of hh and wg in wild-type discs (data not shown).
eya has multiple functions in promoting eye development
We attempted to rescue the eya mutant phenotype by expressing the same molecules under the control of ey-GAL4, as dpp expression requires eya and ey expression does not (Figs 3B, 6A; Bonini et al., 1997; Halder et al., 1998). The ey-GAL4 driver induces sufficient target gene expression to rescue the eyg phenotype, although early effects on eye disc growth are also observed (Fig. 5H and data not shown). However, neither tkvact, sggact, dpp, hh, nor pairwise combinations of these factors were able to induce photoreceptor formation in eya mutant discs (Fig. 6B,C,E and data not shown). Expression of sggact did result in a reduction in size of the eya discs (Fig. 6C), as it does in wild-type discs (Fig. 6D); thus its effect on growth is not secondary to premature differentiation. The lack of dpp expression in eya mutant discs was not rescued by ectopic hh (Fig. 6E), suggesting that eya is required downstream of or in conjunction with hh to direct dpp expression; eya must also regulate additional factors required downstream of dpp for photoreceptor differentiation. Finally, we tested the ability of activated ras to rescue the eya phenotype when expressed under ey-GAL4 control, and found that it was also usually insufficient to allow photoreceptor differentiation (Fig. 6F); thus eya-regulated factors are still required downstream of activation of the EGF receptor pathway. The effects of eya and eyg on dpp and wg and the interactions between dpp and wg signaling are summarized in Fig. 7.
Inhibition of retinal differentiation by wg is not due to loss of dpp expression or activity
In the leg disc, dpp and wg have been shown to maintain their complementary domains of expression by mutual repression (Brook and Cohen, 1996; Heslip et al., 1997; Jiang and Struhl, 1996), and it has been suggested that this interaction also occurs in the eye disc (Heslip et al., 1997). dpp signaling is indeed required to repress wg expression at the posterior margin of the eye, as wg is ectopically expressed in cells unable to receive the dpp signal (Wiersdorff et al., 1996; Fig. 1B). However, this is not the only requirement for dpp signaling, as clones doubly mutant for Mad and wg still fail to initiate photoreceptor differentiation. Consistent with this finding, ectopic wg expression is only observed in clones mutant for strong loss-of-function Mad alleles, suggesting that wg repression requires only a low level of dpp signaling (Wiersdorff et al., 1996). The ability of dpp to induce anterior initiation, like its requirement for posterior initiation, cannot be attributed to its repression of wg, as wg does not prevent anterior initiation. Indeed, we have observed that dpp can induce the ectopic expression of wg at the anterior margin when it is misexpressed using either flp-out-GAL4 or ey-GAL4 (data not shown), suggesting that other factors determine the effect of dpp on wg transcription.
The primary effect of wg does not appear to be the regulation of dpp expression. Even though dpp expression is lost when the wg pathway is activated in sgg mutant clones (Heslip et al., 1997), we found that ectopic expression of wg can inhibit photoreceptor differentiation without reducing dpp expression (Treisman and Rubin, 1995; Fig. 2B). It is possible that a level of wg signaling too low to completely antagonize sgg or abolish dpp expression is still able to prevent photoreceptor formation. Our results show that expression of dpp (Fig. 2D,F) or constitutive activation of the dpp pathway using an activated tkv receptor (Fig. 2H,N) does not rescue the block caused by wg, as would have been expected if wg acted solely by altering the level of dpp expression or activity. Unlike dpp, the activated tkv construct is not sufficient to promote anterior initiation, although it does induce ectopic ventral initiation (Fig. 2C,G); this could be due to its lack of non-autonomous activity, as expression driven by ey-GAL4 is lost from the anterior margin during the third larval instar (Fig. 2B). Another explanation might be that, in this case, dpp signaling requires the combined action of tkv and the other type I receptor encoded by saxophone (sax; Brummel et al., 1994; Xie et al., 1994), although sax is clearly not sufficient for signaling as cells mutant for tkv fail to initiate a furrow (Burke and Basler, 1996). We have also found that ectopic wg does not prevent the expression of the proneural gene ato (Jarman et al., 1994; data not shown). It is not clear whether ato acts upstream or downstream of dpp, as mutations in either gene result in loss of expression of the other (Jarman et al., 1995; Dominguez and Hafen, 1997) and the two are probably involved in a feedback loop.
On the contrary, we have found that the phenotype caused by ectopic wg is rescued by expressing activated forms of spi or ras, raising the possibility that wg interferes with EGF receptor signaling upstream of ras. Recently, it has been shown that, in the embryonic segments, wg and secreted spi emanate from distinct sources and promote opposing cell fates. This led to the proposal that wg antagonizes signaling by spi through the EGF receptor and the ras/MAPK cascade (O’Keefe et al., 1997; Szuts et al., 1997). Since EGF receptor signaling is required for the formation of all photoreceptors (Freeman, 1996; Tio and Moses, 1997; Xu and Rubin, 1993), it is a possible target for wg inhibition in the eye disc. However, it does not appear that the effects of ectopic wg can be completely explained by antagonism of spi signaling, as mutations in spi allow the specification of R8 and the progression of the furrow (Freeman, 1994, 1996; Tio et al., 1994; Tio and Moses, 1997), while the presence of ectopic wg does not. It is possible that another ligand, such as vein (Simcox et al., 1996; Schnepp et al., 1996), normally activates the EGF receptor in R8 and that this ligand is also antagonized by wg. Another possibility is that ras activation in R8 is mediated by another tyrosine kinase receptor; one of the identified FGF receptors is expressed in the morphogenetic furrow (Emori and Saigo, 1993). The lower effectiveness of rescue by s-spi than by rasv12 could also suggest that wg has effects both upstream of spi expression or processing, and downstream of these events. Some factors known to be required between spi and ras that could be targets of wg inhibition are daughter of sevenless (Herbst et al., 1996; Raabe et al., 1996), downstream of receptor kinases (Olivier et al., 1993; Simon et al., 1993) and son of sevenless (Rogge et al., 1991; Simon et al., 1991). Alternatively, wg could act by stimulating the expression or function of argos, a secreted antagonist of spi (Schweitzer et al., 1995a).
Interestingly, expression of activated ras alone is sufficient to induce photoreceptor development in regions anterior to the morphogenetic furrow, as well as extra photoreceptor cells posterior to it (Fig. 2O). Such ectopic development appears to be restricted to a ‘zone of competence’ near the furrow, as the presence of activated ras in more anterior regions leads to dpp expression but not photoreceptor differentiation (data not shown). Such a zone has been described before as responsive to ectopic hh expression or to the loss of PKA (Heberlein et al., 1995; Pan and Rubin, 1995; Strutt et al., 1995). Thus, it appears that the effects of ras activation and hh expression are very similar, and most of the functions of hh could be achieved by activating ras.
dpp and wg can cooperate to induce ectopic furrow initiation at the anterior margin
Since wg appears to inhibit photoreceptor differentiation downstream of dpp signaling, our observation that it does not inhibit dpp-induced initiation from the anterior margin is surprising. This does not seem simply to be due to dpp diffusing further and stimulating differentiation beyond the range of wg inhibition, as co-expression of the cell-autonomous components tkvact and armact can also induce anterior initiation, even though tkvact alone does not (data not shown). One possible explanation is that a cofactor required for wg inhibition is absent from the anterior margin. The anterior margin is likely to have some molecular differences from the rest of the disc; for example, the homothorax gene is expressed at the anterior margin and is required there to inhibit ectopic furrow initiation (Pai et al., 1998). Loss of PKA also induces wg expression only in this region (Dominguez and Hafen, 1997), and it is the only part of the disc able to respond to ectopic dpp (Chanut and Heberlein, 1997; Pignoni and Zipursky, 1997). The anterior margin may require special characteristics because it forms the boundary between the eye disc and the antennal disc, two imaginal fields with quite different modes of development.
Another possibility is that initiation from the anterior margin actually results from a duplication of the eye field. Outgrowths and complete duplications of the eye disc are often seen when anterior initiation is induced by ectopic dpp (Pignoni and Zipursky, 1997). Proximal-distal growth of the leg requires adjacent cells expressing dpp and wg, and distally complete duplications of the leg can result from ectopic expression of one near the normal expression domain of the other (Basler and Struhl, 1994; Campbell et al., 1993; Diaz-Benjumea et al., 1994). Interestingly, one target gene activated by the combined presence of dpp and wg in the leg disc is dac (Lecuit and Cohen, 1997), which is sufficient to induce eye development in the antennal disc and other tissues (Shen and Mardon, 1997). It is possible that both dac expression and duplicating outgrowth can also be activated by the combination of dpp and wg in the eye disc; ectopic dac expression is indeed observed in conjunction with anterior initiation induced by clones of cells expressing dpp and wg (data not shown). Early expression of wg throughout the eye disc (Royet and Finkelstein, 1997) might cooperate with hh-induced dpp at the posterior margin to induce normal dac expression and eye development; our observation that expression of activated sgg dramatically reduces the size of the eye disc even in the absence of photoreceptor differentiation (Fig. 6C,D) supports a role for wg signaling in early growth of the disc.
Genes required to specify the eye disc regulate the expression patterns of dpp and wg
Regional differences in the response to dpp, wg and hh are likely to be due to preexisting differences in the distribution or function of earlier acting genes. In the case of the leg disc, wg expression in ventral anterior cells is already present when cells are recruited to form the primordium and imposes an asymmetric response to the hh signal, restricting dpp expression to the dorsal region (Cohen et al., 1993). However, in the eye disc primordium, neither dpp nor wg expression has been shown to be inherited from embryonic expression. Both eya and eyg appear to be required for the normal activation of dpp expression and repression of wg expression in the eye disc, although eya has a stronger effect on dpp and eyg on wg. As eya is first expressed in the eye disc in a gradient with its high point at the posterior margin (Bonini et al., 1993), it is a good candidate to promote posterior dpp expression and prevent wg expression upon hh induction. However, in eya mutants, a low level of dpp expression can be initiated (Fig. 3B; Pignoni et al., 1997) and early wg expression is restricted to its normal domain. The effects of loss of eya appear stronger in late third instar homozygous mutant discs or clones of cells (Fig. 4 and data not shown). eya might therefore be required for the maintenance of these expression patterns by the autoregulatory and wg-repressing functions of dpp, as well as for the induction of dpp by hh (Fig. 6E). The effects of ectopic eya expression have only been examined using a dpp-GAL4 driver (Bonini et al., 1997; Pignoni et al., 1997), making it difficult to evaluate whether eya is sufficient for dpp expression or wg repression.
Our results support those of Pignoni et al. (1997), who showed that even very late loss of eya function, posterior to the morphogenetic furrow, results in the absence of photoreceptors, and proposed that eya acts at multiple stages of photoreceptor development. Similarly, we show that the eya phenotype cannot be rescued by altering the activity of the dpp, wg or hh pathways; expression of the proneural gene ato is not restored to eya mutant discs by overexpression of activated tkv, activated sgg or hh (data not shown), confirming that eya blocks differentiation upstream of ato expression (Jarman et al., 1995). However, activation of ras is also insufficient to rescue eya mutant discs, although it does rescue the block caused by ectopic wg, which we show is downstream of ato expression. Thus eya must be required for photoreceptor development both upstream and downstream of ato expression. Although the expression pattern of eyg has not been described, its requirement for neuronal differentiation can be rescued by inhibiting wg signaling at the posterior margin of the eye disc, showing that an important part of its function must be to repress wg in this region. As neither hh, dpp nor activated tkv can rescue the eyg phenotype, eyg must act downstream of, or in conjuction with, hh and dpp activity. If eyg acted solely to repress wg, it would be difficult to understand how its ectopic expression could induce ectopic eye development (H. Sun, personal communication). However, ectopic eyes induced by eyg are only observed ventral to the normal eyes, and may thus result from ectopic initiation at the ventral margin of the eye disc, as opposed to induction of eye development in another imaginal disc. Interestingly, ectopic ey is not able to induce eye development in wg-expressing domains of other discs (Halder et al., 1998), and coexpression of eyg enhances ectopic eye formation by ey (H. Sun, personal communication), perhaps through its ability to repress wg.
Although we do not know whether eya and eyg directly regulate dpp or wg, this is a possibility as eyg contains two DNA-binding domains (Jun and Desplan, 1996) and eya a transcriptional activation domain (Pignoni et al., 1997). so affects dpp and may affect wg in the same way as eya, since the so and eya mutant phenotypes are very similar and their encoded proteins can form a complex (Pignoni et al., 1997). As ey appears to act upstream of these genes (Halder et al., 1995; Bonini et al., 1997), it may affect dpp and wg indirectly by regulating the expression of so and eya. In contrast, dac is not required for dpp expression (Mardon et al., 1994), although it is required to prevent wg expression at the posterior margin (Treisman and Rubin, 1995). dac is not required for normal growth of the eye disc, suggesting that it functions later than eyg. It would be interesting to test whether its mutant phenotype in the eye disc is solely due to ectopic wg expression.
In summary, our results show that wg inhibits normal photoreceptor differentiation in a manner independent of dpp expression or activation. The expression patterns of both dpp and wg, and perhaps their cross-regulatory interactions, are determined during early eye development by genes including eya and eyg. Such tissue-specific regulators may explain how very different processes can be controlled by the same signals.
We thank Ron Blackman, Steve Cohen, Barry Dickson, Manfred Frasch, Jose-Luis Gomez-Skarmeta, Ulrike Heberlein, Yuh-Nung Jan, Felix Karim, Elisabeth Knust, Graeme Mardon, Marek Mlodzik, Mark Peifer, Gerd Pflugfelder, Francesca Pignoni, Gerry Rubin, Benny Shilo, and Gary Struhl for their generous gifts of fly stocks and reagents. We also thank Tanja Eggert for assistance in generating the ey-GAL4 lines. We are grateful to the members of the Treisman and Lehmann labs for helpful discussions and a stimulating environment. The manuscript was improved by the thoughtful comments of Claude Desplan, Corinne Zaffran, Heather Broihier, and Ulrike Heberlein. This work was supported in part by NSF grant no. IBN-9728140 to J. E. T., and in part by Skirball Institute start-up funds. M. B. was supported by the Centre National de la Recherche Scientifique, l’Institut National de la Santé et de la Recherche Médicale and l’Hôpital Universitaire de Strasbourg. U.W. was supported by a grant from the Deutsche Forschungsgemeinschaft (WA 556/4-1).