wingless signaling in the Drosophila eye and embryonic epidermis

The wingless (wg) gene is the best characterized member of the Wnt family, which contains over fifty genes in organisms ranging from nematodes to humans (Nusse and Varmus, 1992). Wnt genes encode cysteine-rich proteins containing signal sequences and several members, including wg, have rigorously been shown to be secreted (Bradley and Brown, 1990; Fradkin et al., 1995; González et al., 1991; Papkoff and Schryver, 1990; Van den Heuvel et al., 1989; Van Leeuwen et al., 1994).

In Drosophila melanogaster, wg is required throughout embryogenesis and larval development for a wide range of patterning events (Klingensmith and Nusse, 1994; Siegfried and Perrimon, 1994). Some of these include specifying cell fate in the embryonic epidermis (Baker, 1988; Bejsovec and Martinez-Arias, 1991; Dougan and Dinardo, 1992), CNS (Chu-Lagraff and Doe, 1993), mesoderm (Baylies et al., 1995; Wu et al., 1995) and endoderm (Hoppler and Bienz, 1995). In larval development, wg is required for patterning in leg (Couso et al., 1993; Diaz-Benjumea and Cohen, 1994; Struhl and Basler, 1993; Wilder and Perrimon, 1995) and wing (Couso et al., 1994; Diaz-Benjumea and Cohen, 1995; Phillips and Whittle, 1993) imaginal discs. In the eye, wg has recently been shown to be necessary for proper spacing of morphogenetic furrow initiation (Ma and Moses, 1995; Treisman and Rubin, 1995). How one signal can produce so many responses remains an important unanswered question in developmental biology.

Consistent with being a secreted molecule, wg is thought to execute most of its functions in a paracrine manner. In the best documented cases, the range of wg action can vary from one (Vincent and Lawrence, 1994) to several (Hoppler and Bienz, 1995) cell diameters, though the exact limits of wg diffusion remain unclear (Axelrod et al., 1996; Peifer et al., 1991; Theisen et al., 1994). In a few cases, wg regulates gene expression in the same cells in which it is expressed, e.g. the activation of cut expression at the wing margin (Couso et al., 1994) and the regulation of its own expression in the embryo (Bejsovec and Wieschaus, 1993; Hooper, 1994; Yoffe et al., 1995). This embryonic autoregulation has been referred to as ‘autocrine wg signaling’ but it is not clear whether wg works in a truly autocrine manner. However, recent evidence indicates that wg autoregulation may have different genetic requirements than the paracrine signaling pathway of wg (Hooper, 1994; Manoukian et al., 1995; see discussion).

Three genes with embryonic phenotypes very similar to that of wg have been described (Klingensmith et al., 1989; Peifer and Wieschaus, 1990; Perrimon et al., 1989; Perrimon and Mahowald, 1987), porcupine (porc), dishevelled (dsh) and armadillo (arm). Another gene, zeste white-3 (zw3; also known as shaggy) has a mutant phenotype (Perrimon and Smouse, 1989; Siegfried et al., 1992) very similar to that of embryos where wg has been expressed ubiquitously (Noordermeer et al., 1992). Genetic epistasis (Noordermeer et al., 1994; Peifer et al., 1994b; Siegfried et al., 1994) have ordered these genes in the following genetic pathway:

porc has been shown to be involved in either secretion or subsequent diffusion of the wg protein (Siegfried et al., 1994; van den Heuvel et al., 1993a) and the other three genes are thought to be required for receiving the wg signal (Klingensmith and Nusse, 1994; Siegfried and Perrimon, 1994).

Recent work has revealed that many aspects of this embryonic wg signaling pathway are conserved in larval Drosophila tissues as well as in other organisms. Analysis of dsh, zw3 and arm mutations in leg and wing imaginal discs indicates that these genes are required for wg signaling (Couso et al., 1994; Diaz-Benjumea and Cohen, 1994; Klingensmith et al., 1994; Peifer et al., 1991; Theisen et al., 1994). This has been best shown in the developing wing margin, where these genes mediate wg regulation of the acheate (ac) gene (Couso et al., 1994; Blair, 1994). The vertebrate homologs of these three genes have been shown to play a role in inducing dorsal mesoderm in Xenopus in a manner consistent with functioning in a Wnt signaling pathway (Dominguez et al., 1995; He et al., 1995; Heasman et al., 1994; Pierce and Kimelman, 1995; Rothbacher et al., 1995; Sokol et al., 1995).

The wg signaling pathway described above was first postulated based on extensive genetic analysis, but recent work indicates that some of the gene products may function directly with wg in a biochemical pathway. The arm gene encodes the Drosophila homolog of β-catenin (Peifer and Wieschaus, 1990), a component of vertebrate adherens junctions (Kemler, 1993). A similar junctional complex is found in flies (Peifer, 1993) but a substantial pool of cytoplasmic arm protein also exists (Peifer et al., 1994b; Van Leeuwen et al., 1994). wg signaling causes an accumulation of cytoplasmic arm protein (Peifer et al., 1994b; Van Leeuwen et al., 1994) caused by a dramatic decrease in arm protein turnover (Van Leeuwen et al., 1994). This accumulation is correlated with a reduction in phosphorylation of arm (Peifer et al., 1994a). This increase in arm protein is thought to somehow transduce the wg signal to the nucleus (Klingensmith and Nusse, 1994; Siegfried and Perrimon, 1994).

Consistent with the proposed genetic pathway, mutations in the other components of the wg pathway affect arm protein levels. The normal segmentally repeated accumulation of arm protein is absent in wg, porc and dsh mutants (Peifer et al., 1994b; Riggleman et al., 1990), while zw3 mutants have uniformly high levels of arm protein (Peifer et al., 1994b; Siegfried et al., 1994). The dsh gene encodes a novel protein (Klingensmith et al., 1994; Theisen et al., 1994) containing a PDZ domain (Kennedy, 1995) that is phosphorylated in response to wg in embryos and cultured cells, and this phosphorylation is correlated with the ability of dsh to stabilize the arm protein (Yanagawa et al., 1995). zw3 encodes a serinethreonine protein kinase that is homologous with mammalian glycogen synthase kinase-3 (Ruel et al., 1993a; Siegfried et al., 1992). At the present time, it is not clear whether any of the regulatory steps in the pathway are direct or how many missing components remain to be identified.

One new candidate for functioning in the wg pathway is the product of the Notch (N) gene, which encodes a transmembrane protein found on the surface of cells. N protein is thought to act as the receptor for the Delta (Dl) gene product in a signaling pathway involved in many aspects of development (Muskavitch, 1994; Artavanis-Tsakonas et al. 1995). Its potential role in the wg pathway is based on strong genetic interactions between N and wg mutations in several tissues, but primarily in the wing (Couso and Martinez Arias, 1994; Hing et al., 1994). It is possible that the role of N in the separate but oft-used pathway with Dl could mask a requirement for N in wg signaling when N mutant embryos or clones are examined. Because N is expressed at the cell surface and appears to act as a receptor, it has been postulated that wg encodes a ligand for the N protein (Couso and Martinez Arias, 1994).

This report describes a phenotype created by ectopic expression of wg during eye development. These transgenic animals lack the mechanosensory bristles normally surrounding each facet of the compound eye. This is the exact opposite effect seen in the wing, where wg is required for bristle formation (Couso et al., 1994; Phillips and Whittle, 1993). Despite this difference in regulation, the wg signal transduction machinery found in the embryo and wing also functions in the eye. Finally, the role of N in wg signaling was examined in the eye and in the embryonic epidermis, where, in the complete absence of N protein, wg signaling appears to occur normally. These data argue against a direct role for N in wg signaling.

The mutant alleles in components of the wg signaling pathway used in this study were: wg^IL, wg^IN, wg^CX4, porc^I8, porc^2E, dsh^V26, dsh⁴⁷⁷, arm^XM19, arm^25B, sgg^D127and zw3^M11. wg^CX4 (van den Heuvel et al., 1993a,b), dsh^V26 (Yanagawa et al., 1995) and sgg^D127 (Ruel et al., 1993b) are null alleles, wg^IN encodes a non-secreted wg protein (van den Heuvel et al., 1993a,b), wg^ILis a temperature-sensitive allele (Baker, 1988) and the rest are characterized phenotypically as strong alleles (Klingensmith, 1993; Siegfried et al., 1992), except for the arm alleles, which are hypomorphs but are the strongest alleles that are cell viable when homozygous (Peifer et al., 1991). Two null alleles of N, N^264.40 and N⁵⁴¹⁹ (S. Artavanis-Tsakonas, personal comm.) and the temperature-sensitive alleles N^ts1 (Cagan and Ready, 1989b), Dl^6E (Dietrich and Campos-Ortega, 1984) and Dl^RF (Parody and Muskavitch, 1993) were also used. For further information, see Lindsley and Zimm (1992).

A P-element construct placing the wg ORF under the control of the sevenless (sev) promoter (P[sev-wg]) was made by inserting the XbaI/ClaI (blunt ended) fragment of the wg cDNA, pCV (Rijsewijk et al., 1987) into the XbaI and BglII (blunt ended) sites of pSEWa (Fortini et al., 1992), between the sev proximal promoter and 3′ processing elements. pSEWa also contains three tandem repeats of the sev enhancer 5′ of the promoter. yw⁶⁷ embryos were coinjected with P[sev-wg] and pπ25.7 as described previously (Rubin and Spradling, 1982) and several independent lines were established using standard balancer stocks. A stock containing the lacZ coding sequences under the control of the sev enhancer (three tandem repeats) and hsp70 proximal promoter (P[sev-lacZ]; R. Carthew, personal communication) was obtained from Todd Laverty (UC Berkeley, CA).

The following heat-shock strains were used: P[hs-wg] (Noordermeer et al., 1992), P[hs-zw3] (Siegfried et al., 1992) and P[hs-dsh] (Axelrod et al., 1996). P[hs-wg] is on the third chromosome, the other two on the second. The following chromosomes were created by recombination. P[sev-wg; w⁻], P[hs-zw3; w⁺] (the white (w) gene in the P[sev-wg; w⁺] transgene was inactivated by EMS mutagenesis). A P[sev-wg] insert on chromosome 3L was recombined with a Dl^RF mutation to make P[sev-wg], Dl^RF. Two different P[hs-dsh; w⁺], wg^IL recombinants were created, one using a wg^ILcn bw sp chromosome and the other a wg^ILbr pr, since both chromosomes contain a different lethal mutation unrelated to wg (Couso et al., 1994). Both P[hs-dsh], wg^IL recombinants were placed over a SM5a-TM6B compound chromosome, so that homozygotes could be identified by the absence of the Tubby pupal marker.

Pupal eyes were dissected and then immunostained as described (Blochlinger et al., 1993). Embryo stainings were performed essentially as previously described [Frasch et al., 1987; Grossniklaus et al., 1992). Affinity-purified rat α-cut antisera was generously provided by K. Blochinger (Fred Hutchinson Institute, WA), mouse α-ac monoclonal antibody was a gift of Sean Carroll (University of Wisconsin at Madison). Rabbit α-lacZ antisera was from Cappel and affinity purified rabbit α-wg antisera was kindly provided by C. Harryman-Samos (Stanford University, CA). Mouse α-N monoclonal antibody was provided by S. Artavanis-Tsakonas (Yale Univ. CT) and mouse α-en antisera by T. Kornberg (UCSF, CA). The primary antibodies were used at the following dilutions: ac, 1:3 to 1:5, wg, 1:20, N, 1:100, cut and en, 1:300, lacZ, 1:500. For histochemistry, secondary antibodies were either biotinylated (goat α-mouse, horse α-rabbit and rabbit α-rat; all from the Elite ABC kit, Vectastain, used at a 1:500 dilution) or goat αrabbit conjugated to alkaline phosphatase (from Vector, used at 1:300). For fluorescence microscopy either donkey FITC α-mouse (1:100) or donkey Cy3 α-rabbit (1:200) were used (Jackson Immunochemicals). Confocal images were collected with a Bio-Rad MRC 1000 confocal laser setup attached to a Zeiss Axioscope microscope. Images were imported into Adobe Photoshop for presentation.

In situ hybridization to whole-mount embryos using digoxigenin-labeled probes (Tautz and Pfeiffle, 1989) and antibody/in situ double stainings (Manoukian and Krause, 1992) were performed as described (detailed protocol available upon request).

All whole-mount stainings were photographed with a Nikon Microphot-FXA microscope and slides were scanned into Adobe Photoshop for presentation.

Mutant alleles of dsh, zw3 and arm were recombined unto a P[hs-neo; FRT]18A chromosome, porc onto P[hs-neo; FRT]19A, wg onto P[hsneo; FRT]40A and a P[sev-wg; w⁺] mapping to 3L onto P[hs-neo; FRT]80A, all in a w background. w clones were induced in animals heterozygous with the appropriate P[mini-w⁺], P[FRT] chromosome: P[mini-w⁺; hs-πM]5A, 10D, P[hs-neo; FRT]18A; P[mini-w⁺]18A, P[hs-neo; FRT]19A; P[mini-w⁺; hs-πM]21C, 36F, P[hs-neo; FRT]. All FRT derivatives are as described (Xu and Rubin, 1993) except for P[mini-w⁺]18A, which is from the Jan lab enhancer detection collection (Bier et al. 1989). FLP recombinase was provided from the FLP99 chromosome (Chou and Perrimon, 1992). Clones were induced by a one hour heat shock (37°C) 24-48 hours (at 25°C) after egg laying and scored for the absence of pigmentation in the adult eye.

For production of N germ-line clones, the N null alleles were recombined onto a P[mini-w⁺; FRT]¹⁰¹ chromosome (Chou and Perrimon, 1992). N, P[mini-w⁺; FRT]¹⁰¹/FM7 females were crossed to a w ovo^D1, P[mini-w⁺; FRT]¹⁰¹/Y ; P[hs-FLP]³⁸ stock (Chou and Perrimon, 1992) and progeny were heat shocked late 3rd instar/early pupation for 2 hours at 37°C (earlier heat shocks resulted in high lethality due to somatic clones). Mosaic mothers were crossed to P[ftzlacZ]C males (Hiromi and Gehring, 1987) or P[ftz-lacZ]C; P[hswg]/TM3 males. Embryos with no β-gal staining lacked both maternal and zygotic expression of N.

The P[hs-wg] phenotype was induced by multiple heat shocks as previously described (Noordermeer et al., 1992). Late larval/early pupal temperature shifts were performed by submerging glass vials in a water bath of the appropriate temperature (37°C for heat shocks). At all other times, larvae and pupae were kept at 25°C. Formation of white pupae was used as the reference point (0 hours APF).

Flies were prepared for scanning electron microscopy by serial dehydration in ethanol and Freon 113 (EM Sciences) as described (Kimmel et al., 1990). Dried samples were mounted with colloidal graphite, and a 10 nm gold-platinum coat was applied with a Hummer sputter coater. The samples were viewed with an AMR1000 SEM and photographed using Polapan 400 film (Kodak). Pupal eyes were surface stained with Co(NO₃)_2.6H20 and (NH₄)₂S as described (Kimmel et al., 1990).

During the course of our attempts to create a dominant adult wg mutant through limited misexpression of wg during larval development, we found a highly penetrant phenotype when wg was placed under the control of the eye-specific promoter sev. As shown in Fig. 1, the eyes of P[sev-wg] flies appear normal, except that the interommatidial bristles, normally found at alternating vertices in the compound eye’s hexagonal array, are almost completely missing. Sections through adult eyes (data not shown) and surface staining of pupal eyes with cobalt sulfide (Fig. 1E,F) revealed no other detectable abnormality in adult eyes. The bristles are replaced in the repeated structure of the eye with tertiary pigment cells. Thus, at the level of ectopic wg expressed from the P[sev-wg] transgene, the effect of wg on eye development is very specific.

Interommatidial bristles are mechanosensory organs composed of four cells that are derived from a single sensory organ precursor (SOP; (Cagan and Ready, 1989a). Larval SOP determination has been best described in the wing imaginal disc (Campuzano and Modolell, 1992; Jan and Jan, 1993b). The process begins with small groups of cells expressing basic helix-loop-helix proteins such as acheate (ac) and scute (Cubas et al., 1991; Skeath and Carroll, 1991). All the cells in these proneural clusters have the ability to become the SOP, however, in a wild-type background, only one does. This cell is thought to become the SOP by reaching a threshold level of ac and/or scute after which it inhibits these genes’ expression in its neighbors (Ghysen et al., 1993; Simpson, 1990). This lateral inhibition is mediated by the neurogenic pathway, in which the products of the Dl and N genes are thought to act as ligand and receptor, respectively (Artavanis-Tsakonas et al., 1995; Muskavitch, 1994). The initiation of SOP development is correlated with the expression of a new set of genes, such as neuralized (Huang and Dambly-Chaudière, 1991) and for some SOPs, cut (Blochlinger et al., 1993). The SOP undergoes to two divisions to generate the four cells that will give rise to the mature bristle organ (Bodmer et al., 1989; Hartenstein and Posakony, 1989).

The events leading to SOP formation in the eye have many similarities to those occuring in other tissues. ac protein becomes detectable shortly after white prepupa formation (data not shown). At 3 hours after the white prepupa stage (3 hours APF), the ac gene is expressed in small clusters of cells throughout the eye (Fig. 2C). Unlike the photoreceptors and cone cells, the appearance of the ac-positive cells is not related to the distance from the morphogenetic furrow, although the cells anterior of the furrow do not express ac (see arrows in Fig. 2C). By 6 hours APF, only one cell per cluster still expresses ac, again with the anterior-most portion of the eye showing a less mature pattern (data not shown). At 15 hours APF, after the eye disc everts, ac protein is gone, but the daughters of the SOPs can be observed by staining with α-cut antisera (Fig. 2A). Because of the complicated morphogenetic movements associated with the eye/head disc eversion, we have been unable to stain tissue between 6 and 15 hours APF.

In the P[SEV-wg] eyes, ac expression is greatly reduced compared to controls though not completely absent (Fig. 2D,F). After disc eversion, no SOPs are found, as judged by cut staining (Fig. 2B) and an enhancer detector line for the neuralized gene (data not shown). Thus, wg appears to act at the level of the proneural genes, i.e., ac, to inhibit SOP formation.

The activity of the sev promoter has been well studied in third instar larva, by monitoring endogenous sev expression (Tomlinson et al., 1987) and with chimeric constructs (Bowtell et al., 1989) using sev enhancer and promoter elements similar to the ones in P[sev-wg]. The enhancer is active in the cone cells and in a subset of the underlying photoreceptor precursors. No description of sev expression has been reported after pupation, so the possibility existed that wg was expressed in the proneural cells of P[sev-wg] eyes, suggesting a possible autocrine effect.

This question was addressed by examining the distribution of wg protein in P[sev-wg] eyes. Though wg is a secreted protein, it is found at the highest levels on the surface of the same cells that synthesize it (Bejsovec and Wieschaus, 1995; Couso et al., 1994; van den Heuvel et al., 1993). In P[sev-wg] eyes, the highest levels of wg protein were found around the four cone cells (Fig. 3A) and accumulated on their apical surface (Fig. 3B). In more basal sections of the eye, wg protein was associated with the photoreceptors, which extend basally to the same plane as the ac-positive cells (Fig. 3C). There was no significant overlap between wg protein and the remaining cells expressing ac.

To confirm that the sev enhancer was not active in the proneural clusters, we stained eyes of flies that contained a P[sev-lacZ] transgene (see Materials and Methods) for products of lacZ and ac. As found for wg in P[sev-wg] eyes, most of the β-gal was found in the cone cells (data not shown). In the same focal plane as the ac-expressing cells, there is no overlap (Fig. 3D). Thus, the inhibitory effect of wg on ac expression is paracrine in nature.

Extensive genetic analysis, confirmed by recent biochemical experiments, has identified four genes that encode probable components of the wg signaling pathway, porc, dsh, zw3 and arm (Klingensmith and Nusse, 1994; Siegfried and Perrimon, 1994; see introduction). Mosaic analysis (using the w gene as a marker) was performed to determine if these genes were required for the P[sev-wg]dependent bristle inhibition. Control clones still lack bristles (Fig. 4A), as do clones mutant for the endogenous wg gene (Fig. 4C). In clones that lack the P[sev-wg] transgene, bristles are found almost to the clonal boundary (Fig. 4B). Likewise, 89% of the mutant clones for porc, dsh and arm had the full array of bristles within the clone (Fig. 4D-F and Table 1) and an additional 9% had a partial rescue of the bristleless phenotype. The remaining 2% that still lacked bristles were small in size and probably not completely mutant since the absence of the w gene cannot be detected on the surface of the eye at the cellular level. These experiments indicate that porc, dsh and arm are required for wg-dependent bristle inhibition.

zw3 is unique among the known genes required for wg signaling because it must be inhibited for the wg signal to be transduced (Klingensmith and Nusse, 1994; Siegfried and Perrimon, 1994). Thus, loss of zw3 should be equivalent to activation of wg signaling. Therefore, a zw3 mutant clone in the eye might be expected to lack bristles. This straightforward analysis cannot be employed because the cells in zw3 clones in the eye imaginal disc do not differentiate into eye tissue (Treisman and Rubin, 1995; data not shown). This is probably due to the fact that high levels of wg signaling activity prevent the morphogenetic furrow from progressing, blocking any subsequent differentiation (Treisman and Rubin, 1995).

If zw3 must be inhibited for the wg signal to be transduced, then flooding cells with zw3 protein might titrate out the signal. This has been shown to be the case in Xenopus where overexpression of the homologue of zw3, glycogen-synthase kinase 3, blocks Wnt gene induction of dorsal mesoderm (Dominguez et al., 1995; He et al., 1995). We attempted a similar experiment by creating flies with one copy of P[sev-wg] (we chose one of the weaker P[sev-wg] lines, which at one copy has approximately 20 bristles/eye) and one or two copies of a heatshock construct expressing the zw3 gene, P[hs-zw3] (Siegfried et al., 1992). zw3 was induced by heat shock shortly before and twice after the onset of pupation (see Fig. 5 legend for details). Though the results were not entirely conclusive (Fig. 5), many pupal eyes showed a significant response especially when the ratio of P[hs-zw3]/P[sev-wg] is two (Fig. 5C). Other heat-shock regimes were not as effective at suppressing the P[sev-wg] phenotype. These results are consistent with the current model for zw3 function in wg signaling.

Overexpression of dsh has previously been found to mimic wg signaling in cultured cells (Yanagawa et al., 1995), frog embryos (Sokol et al. 1995; Rothbacher et al. 1995) and in the wing imaginal disc (Axelrod et al., 1996). The same P[hs-dsh] transgenic stock used in the wing can also duplicate the effect of wg in the eye. Induction of dsh at 3 hours (data not shown) or 6 hours APF (Fig. 6B) could block bristle formation, but heat shock at 9 hours APF (Fig. 6C) failed to inhibit bristles in the interior of the eye, though inhibition still occurred toward the periphery. This can be explained by previous work(Cagan and Ready, 1989a,b), which showed that SOP determination occurs first in the center of the eye and radiates outward concentrically. The same time requirements were seen when the bristles were inhibited using P[hs-wg] (data not shown).

Genetic and biochemical evidence places dsh downstream of wg in the signal transduction pathway (Klingensmith et al., 1994; Noordermeer et al., 1994; Theisen et al., 1994; Yanagawa et al., 1995), suggesting that the overexpression of dsh can bypass wg function. However, in the wing, where dsh overexpression causes an expansion of the wing margin, it appears that wg gene activity is needed to see the dsh effect (Axelrod et al., 1996). In the eye, the opposite appears to be true. In pupa homozygous for a wg temperature-sensitive mutation, induction of dsh after 6 hours at the restrictive temperature still inhibited SOP formation (Fig. 6E). Thus it appears that dsh in the eye can act independently of wg, though caveats remain (see discussion).

A strong interaction between mutations in the N and wg genes has been described (Couso and Martinez Arias, 1994; Hing et al., 1994), which suggests that the two genes have common developmental targets in some tissues. One report suggested that wg encodes a ligand for N, based on these genetic interactions and the fact that N encodes a transmembrane receptorlike protein (Couso and Martinez Arias, 1994). In the eye, N activity is required for almost every differentiated cell type (Cagan and Ready, 1989b), so examining N clones in a P[sevwg] background was not possible. Therefore, we utilized N^ts1, a temperature-sensitive allele (Cagan and Ready, 1989b). When these flies were reared at the restrictive temperature for 3-11 hours APF in a P[sev-wg] background, a strong suppression of the wg bristleless phenotype was seen (Fig. 7B). This is consistent with a proposed role for N in transducing the wg signal. However, removal of Dl activity for the same time period also suppresses the P[sev-wg] phenotype (Fig. 7C).

N and Dl are key components in the lateral inhibition pathway (functioning as receptor and ligand, respectively) that insures the proper number of bristles in the eye (Cagan and Ready, 1989b; Parody and Muskavitch, 1993; note the abnormally high bristle density in Fig. 7B and C). This pathway is independent of wg, since mutant clones of wg, porc, dsh and arm in an otherwise wild-type background have the normal number of bristles (data not shown; see also Fig. 4D-F). Thus, the observation that loss of Dl activity can suppress the P[sevwg] phenotype as well if not better than loss of N raises the possibility that the interaction between N and wg in the eye is due to the role of N in the lateral inhibition pathway.

If a higher level of wg expression is used (via a heat-shock promoter) all the bristles in the N^ts1 background can be inhibited (data not shown; pupa were placed at the restrictive temperature for 6 hours before a 30 minute heat-shock pulse was given at 6 hours APF). However, it is known that the N^ts1 allele does not completely remove N activity (Couso and Arias, 1994; Hartenstein et al., 1992) so this result is inconclusive. In the eye, it is not possible to determine whether wg works through N or in a parallel pathway converging at proneural gene expression.

In order to more rigorously test the requirement of N for wg signaling, a tissue is needed where a putative N-wg connection can be separated from the wg-independent functions of N. One suitable place is the embryonic epidermis. Embryos mutant for N undergo a dramatic neural hyperplasia; almost all of the cells of the epidermis delaminate and become neuroblasts (CamposOrtega, 1993). However, the epidermis remains relatively intact until full germ-band extension, after significant wg signaling has already occurred. Null N embryos were generated by making germ-line clones (Chou and Perrimon, 1992; see Materials and Methods). Antibody staining revealed no detectable N protein in N germline clones that have received a paternal Y chromosome (Fig. 8F). Thus we can examine wg signaling in a tissue that has never contained N protein.

Two well-characterized targets of wg signaling in the embryo are the engrailed (en) gene (DiNardo et al., 1988; MartinezArias et al., 1988) and the wg gene itself (Bejsovec and Wieschaus, 1993; Hooper, 1994; Yoffe et al., 1995). Careful analysis of expression of both genes has revealed that, in wg mutants, wg transcripts begin to fade before the embryo reaches full germband extension (stage 9; all stages according to (Campos-Ortega and Harten-stein, 1985), and is gone by the beginning of stage 10 (Manoukian et al., 1995). en protein in the adjacent posterior cells fades shortly thereafter. By mid-stage 10, both en protein and wg transcripts are completely gone from wg^IN homozygous embryos (Fig. 8B). In N null embryos at early stage 10, wg and en patterns are indistinguishable from wild type (data not shown). At mid-stage 10, both sets of stripes are still clearly present (Fig. 8C,D). The stripes do appear a little ragged, and we believe this is a con sequence of the beginning of the disintegration of the epidermis, which is well underway by late stage 10 (about 1520 minutes later than the embryos shown in Fig. 8).

Despite the results in Fig. 8, it might be argued that in N mutants, perhaps wg and en expression no longer depended on wg activity. To address this, we examined the affect of global wg expression on en transcript distribution in a N mutant background. As previously reported (Noordermeer et al., 1992, 1994), overexpression of wg via a heat-shock promoter in an otherwise wild-type background causes a dramatic posterior expansion of the en stripes so that they are about twice as wide as normal (compare Fig. 9A and B). This expansion in still seen in embryos lacking N protein (Fig. 9D) and is dependent on the presence of the P[hs-wg] transgene (Fig. 9C). In the complete absence of N protein, wg signaling appears normal as late as we can reliably assay for it.

The interommatidial bristle is a 4-cell sensory organ that arises from a single SOP which is selected from a group of cells expressing proneural basic helix-loop-helix proteins (Campuzano and Modolell, 1992; Jan and Jan, 1993a). Our data strongly suggests that P[sev-wg]-derived wg protein blocks SOP formation in the eye by inhibiting proneural gene expression. Levels of ac protein are much lower in P[sev-wg] eyes (at 3 hours APF) compared to controls (Fig. 2C-F). 12 hours later, after the eye disc has everted, no SOP daughter cells are seen in the transgenic eyes (Fig. 2A,B). Though disc eversion prevents us from directly showing that no SOPs ever form in P[sev-wg] eyes, the time window when P[hs-wg] or P[hs-dsh] can inhibit bristle formation (no later than 6 hours APF for the central portion of the eye; Fig. 6 and results) is consistent with the model that, once an SOP is determined, wg signaling activity can no longer influence its fate.

The ac protein is the only proneural gene product monitored in this study and we are by no means suggesting that the wg signaling pathway acts directly on the ac promoter. In fact, lost of the ac gene alone does not result in complete elimination of interommatidial bristles; a related gene, scute (sc) must also be removed (Brown et al., 1991). The expression patterns of ac and sc are nearly identical (Cubas et al., 1991; Skeath and Carroll, 1991). This is most likely achieved by a combination of shared enhancer elements (Gómez-Skarmeta et al., 1995) and autoand transactivation between the two genes (Martinez and Modolell, 1991; Skeath and Carroll, 1991; Van Doren et al., 1992). In addition, there are important negative inputs from other bHLH proteins such as extramacrocheate (Cubas and Modolell, 1992; Van Doren et al., 1992) and hairy (Brown et al., 1991; Van Doren et al., 1994). wg could be acting to inhibit ac (and presumably sc) expression at any of these regulatory levels. Further studies are needed to address this issue.

The P[sev-wg] bristleless phenotype was unexpected, because in the wing imaginal disc, wg has been shown to have the opposite effect, i.e., it promotes bristle development. In the absence of wg activity, the proneural ac-positive clusters fail to form (Couso et al., 1994; Phillips and Whittle, 1993). It is not clear why wg activates ac in one tissue and inhibits it in another, but this is a simple example of how one signal can generate different responses in various tissues.

wg is not normally expressed in the interior of the eye, but it is present at the periphery, forming a ring around the pupal eye (Cadigan and Nusse, unpublished data). Interestingly, the edge of the eye lacks bristles (Cagan and Ready, 1989b; Fig. 1A). Clones of arm at the periphery contain ectopic bristles (Cadigan and Nusse, unpublished data), suggesting that wg normally inhibits bristles there. However, large wg clones do not show this effect. We are currently examining this in more detail.

A genetic pathway for wg signal transduction has been elucidated in which the gene products work in the following order: (Klingensmith and Nusse, 1994; Siegfried and Perrimon, 1994). Studies in the wing and leg imaginal disc have indicated that dsh, zw3 and arm are also required there for wg signaling (Couso et al., 1994; DiazBenjumea and Cohen, 1994; Klingensmith et al., 1994; Peifer et al., 1991; Theisen et al., 1994; Wilder and Perrimon, 1995). This study extends these findings; porc, dsh and arm are clearly required for the ability of wg to inhibit eye bristles (Fig. 4; Table 1). The overexpression experiments with zw3, while not as conclusive (Fig. 5), are entirely consistent with the favored model, where wg acts by antagonizing zw3 gene activity. While there may be exceptions (see below), it seems that most tissues use the same wg signaling components to achieve a variety of effects.

The mammalian counterpart of zw3, glycogen synthase kinase-3, has been shown to function in ras-dependent signaling (Stambolic and Woodgett, 1994). This raises the possibility that members of the ras and wg pathways share components in flies. In the eye, differentiation of photoreceptor cells is absolutely dependent on ras-dependent signaling (Simon et al., 1991). However, in clones of dsh and arm, all photoreceptors are present (S. Kaech, K.M. Cadigan and R. Nusse, unpublished observations). In the wing, clonal analysis with members of the ras pathway demonstrated that, unlike wg, they were not required for wing margin development (DiazBenjumea and Hafen, 1994). Thus, no interaction between these two pathways has yet been observed in Drosophila.

wg expression is subject to positive autoregulation in the embryo (Bejsovec and Wieschaus, 1993; Hooper, 1994; Yoffe et al., 1995) and recent evidence suggests that this occurs through a distinct signaling mechanism (Hooper, 1994; Manoukian et al., 1995). Some discrepancies exist between the two reports, but Manoukian et al. (1995) provide strong evidence that wg autoregulation requires porc but not dsh, zw3 and arm. They suggest a model where porc functions only in wg autoregulation and the other three genes in wg paracrine functions.

Our results in the eye indicate that, at least in the eye, porc is required for wg paracrine signaling. While we could clearly see sev enhancer-driven wg expression in cone cells and photoreceptors, we found no expression in the proneural clusters, the targets of wg action (Fig. 2). The endogenous wg gene was not required for the P[sev-wg]-dependent bristle inhibition (Fig. 4C), ruling out a paracrine-autocrine circuit. Our results indicating a role for porc in paracrine wg signaling are consistent with the observation that secretion or diffusion of wg protein is blocked in porc mutant embryos (Siegfried et al., 1994; van den Heuvel et al., 1993a).

Overexpression of dsh can mimic the action of wg in the eye (Fig. 6) as has been shown previously in the wing (Axelrod et al., 1996) and in cultured cells (Yanagawa et al., 1995). In the wing, this effect of dsh required wg. This does not appear to be the case in the eye (Fig. 6E). This is an important point because it speaks as to whether dsh can completely bypass the requirement for wg or whether overexpression of dsh simply potentiates wg signaling. It may be that there is residual wg activity left in our experiments (we could only rear the animals for 6 hours at the restrictive temperature before induction of dsh; longer times killed the organism before disc eversion). Another possibility is that a much higher threshold of wg activity is needed to transform wing blade to wing margin than is needed to inhibit eye bristles. The data of Axelrod et al. (1996) show that the transformation of identity is more penetrant closest to the normal wing margin, where wg is expressed. Thus, overexpression of dsh in the wing blade may not easily reach the necessary level of signaling to trigger the change in cell fate. In the eye, dsh is able (at 3 hours APF) to inhibit bristles in the middle of the eye (far from endogenous wg expression) just as efficiently as bristles closer to the periphery. That dsh can bypass the need for wg is also supported by the cell culture experiments (Yanagawa et al., 1995) where no detectable wg protein was observed under conditions where dsh could stabilize arm protein. In addition, Park et al. (1996) have recently shown that overexpression of dsh in the embryo can induce wg targets in a wg null background.

On the basis of genetic interactions between mutations in the two genes, the N protein was proposed to be a receptor (or part of a receptor complex) for wg (Couso and Martinez Arias, 1994). In the eye, we also observed strong genetic interactions between wg and N (Fig. 7). However, the interpretation of these experiments are complicated, since N is known to affect bristle development independently of wg, and because, for technical reasons, we could not completely remove N activity to determine whether wg signaling could still occur. Likewise, the previously published genetic interactions involve animals where wg and N activities are only partially removed (many of the experiments were done with double heterozygotes of various wg and N alleles), and are therefore subject to the same limits of interpretation.

Unlike the eye, wg signaling in the complete absence of N activity can be assayed in the embryonic epidermis until just after germ-band extension is complete (mid stage 10), right before the absence of N causes most of the epidermis to delaminate and become neuroblasts. We found no significant change in the expression of wg and en in N null mutants at this time (Fig. 8), even though their expression fades at early stage 10 in wg mutants and mutants in dsh or arm (Manoukian et al., 1995; Van den Heuvel et al., 1993b). In addition, the effect of overexpression of wg on the en stripes is still seen in a N mutant background (Fig. 9). Couso and Martinez Arias (1994) reported that the en stripes were affected in about half the N mutants they examined, but they used hyperplasia of the nervous system as their method for determining which embryos were N mutants. This happens after mid-stage 10, thus any effect on the stripes may be a secondary consequence of the epidermis falling apart. Therefore, we conclude that in N mutant embryos, wg signaling occurs normally, at least with regard to the two markers we assayed.

A similar conclusion with regards to N-wg interactions has been reached in the wing (Rulifson and Blair, 1995). They showed that wg could still regulate ac expression in homozygous clones for a N null allele. These mutant clones should completely lack N, barring prolonged perdurance of the N protein. Of equal importance is their finding that N activity is required for wg expression at the wing margin (see also DiazBenjumea and Cohen, 1995; Doherty et al., 1996). This means that all of the genetic interactions between wg and N in the wing can potentially be explained by a reduction in N activity causing a reduction in the amount of wg signal, not the ability of wg to signal.

Another link between wg and N has been proposed by Axelrod et al. (1996), who have presented evidence that dsh protein can bind to and inhibit N activity in the wing imaginal disc. They suggest that part of the ability of wg to induce bristles in the wing is achieved by inhibition of N through dsh. Such an antagonistic relationship does not appear to be occuring in the eye since wg, dsh and N all inhibit bristle formation, although we can not rule out a mechanism where wg and dsh activate N to inhibit ac expression.

A subtle role for N in transducing the wg signal cannot be entirely ruled out. However, our results and those of Rulifson and Blair (1995) argue that in tissues where the direct test can be done, i.e., can wg signaling occur in cells that lack N protein, N is not required. A better candidate for a wg receptor is the product of the Drosophila frizzled2 gene, which can bind to wg and tranduce the wg signal in cultured cells (Bhanot et al. 1996). N showed no activity in this wg-binding assay. In the absence of any biochemical data suggesting that the proteins interact, the simplest models for wg signal transduction should exclude a direct role for N.

Special acknowledgment and thanks to Monty Laskosky, for superb operation of the SEM. We would also like to thank Dr Mike Simon for the pSEWa construct, Drs Shu-wen Wang, Mark Muskavitch, Esther Siegfried, Todd Laverty, Juan Pablo Couso, Alfonso Martinez Arias, Kathy Matthews and the Bloomington Stock Center for various fly stocks, and especially Drs. Jeff Axelrod and Norbert Perrimon for providing the P[hs-dsh] flies prior to publication. Thanks also to all the researchers who provided antibodies (see Materials and Methods) and to Drs Sofia Lopes da Silva, Harsh Thaker, Andreas Wodarz, Diane Spillane and Derek Lessing for critical reading of the manuscript. Sectioning of the P[sev-wg] eyes was performed by Sue Kaech, to whom we are grateful. We would like to thank Dr Matt Scott for use of his confocal microscope, and Dr Andreas Wodarz for instruction on its proper use. Mike Ollman and Brent Wilson helped with Fig. 5B. These studies were supported by the Howard Hughes Medical Institute, of which R. N. is an investigator and K. M. C. is an associate, and by a grant from the USAMRAMC, Grant number DAMD17-94-J-4351.