The cell adhesion molecule Echinoid defines a new pathway that antagonizes the Drosophila EGF receptor signaling pathway

The Drosophila compound eye is composed of approximately 800 ommatidia, each of which contains eight photoreceptor cells (R1-R8), four non-neuronal cone cells, and eight accessory cells arranged in a highly ordered pattern. In third instar larvae, patterning of the ommatidial field begins at the posterior margin of the eye imaginal disc, with the morphogenetic furrow sweeping across the disc epithelium in a posterior to anterior direction. Behind the furrow, different cell types are recruited sequentially. The R8 photoreceptor is the first cell to differentiate, followed by R2/R5, R3/R4, R1/R6 and R7. Addition of cone and accessory cells to the photoreceptor cluster produces the final ommatidial unit (reviewed in Wolff and Ready, 1993). The Drosophila epidermal growth factor receptor (EGFR) is required for differentiation of all the cell types, with the exception of R8 (Freeman, 1996; Dominguez et al., 1998). A model describing the reiterative activation of the EGFR by the opposing action of the EGFR ligand, Spitz (SPI) and an antagonist, Argos (AOS), has been proposed to explain the successive recruitment of each cell type in the developing eye (Freeman, 1997).

The Drosophila EGFR signaling pathway is subject to modulation at multiple levels by various positive and negative mechanisms (reviewed in Perrimon and Perkins, 1997; Wasserman and Freeman, 1997; Schweitzer and Shilo, 1997). There are multiple EGFR ligands, SPI, Vein and Gurken that activate the receptor tyrosine kinase (RTK) at different stages of development. In addition, Rhomboid (RHO), a multiple transmembrane domain protein, can potentiate EGFR signaling by regulating SPI processing and presentation (reviewed in Wasserman and Freeman, 1997; Bang and Kintner, 2000). The activated EGFR triggers a conserved signal transduction cascade that includes DRK, SOS, RAS1, KSR, RAF, MAPK (reviewed by Perrimon and Perkins, 1997). While SOS activates RAS1 by promoting the exchange of GDP for GTP, the GTP-activating protein, GAP1, inactivates RAS1 by stimulating its intrinsic GTPase activity (Gaul et al., 1992). Furthermore, activated MAPK is thought to propagate the RAS cascade signal into the nucleus by phosphorylating two members of the ETS family of transcription factors, YAN (AOP – FlyBase) and Pointed (PNT) (Brunner et al., 1994a; Brunner et al., 1994b; O’Neill et al., 1994). PTP-ER, a cytoplasmic tyrosine phosphatase, binds to and downregulates activated MAPK (Karim and Rubin, 1999). Therefore, both GAP1 and PTP-ER act in the cytoplasm to negatively regulate EGFR signaling. YAN is a transcriptional repressor that inhibits the production of both photoreceptor and cone cells (Lai and Rubin, 1992). Upon phosphorylation, YAN moves to the cytoplasm for degradation, leading to the differentiation of both cell types (Rebay and Rubin, 1995). In addition, tramtrack (ttk), which encodes two alternatively spliced, zinc finger proteins, TTK69 and TTK88 (Read and Manley, 1992; Xiong and Montell, 1993), plays a central role in photoreceptor cell differentiation. TTK88 functions as a transcriptional repressor to inhibit photoreceptor but not cone cell differentiation (Lai et al., 1996; Yamamoto et al., 1996). Upon EGFR activation, two other nuclear proteins, Phyllopod (PHYL), which is transcriptionally regulated by the RAS/MAPK pathway, and Seven In Abstentia (SINA), form a complex that binds to TTK88. This association leads to the targeting of TTK88 for degradation which results in the differentiation of photoreceptor cells (Tang et al., 1997; Li et al., 1997).

In addition to the cytoplasmic (GAP1, PTP-ER) and nuclear (YAN and TTK88) repressors, EGFR signaling can also be down-regulated by the production of AOS, Kekkon 1 (KEK1), and Sprouty (STY), via negative feedback loops following EGFR activation. AOS, a secreted molecule functions nonautonomously to repress EGFR signaling by blocking both the receptor dimerization and the ability of SPI to bind the EGFR (Freeman et al., 1992; Schweitzer et al., 1995; Jin et al., 2000). KEK1, an adhesion molecule protein, might execute its inhibitory effect through direct association of its extracellular and transmembrane domain with the EGFR (Ghiglione et al., 1999). Finally, the intracellular but inner membrane-associated STY has been proposed to act through its direct binding to DRK and GAP1, thereby blocking the activation of RAS1 (Casci et al., 1999).

Studies of the EGFR have also revealed that heterologous pathways can also modulate the activity of this RTK (reviewed in Moghal and Sternberg, 1999; Tan and Kim, 1999). For example, when growth hormone binds to its receptor, it activates JAK2, which promotes tyrosine phosphorylation of a GRB2 binding site on the EGFR, thereby leading to the activation of RAS and MAPK (Yamauchi et al., 1997). In addition, occupied and aggregated integrins can collaborate with growth factors by triggering tyrosine phosphorylation of EGFR (Miyamoto et al., 1996; Moro et al., 1998). In contrast, extracellular matrix proteins such as collagen, can promote retinoid-induced differentiation of normal human bronchial epithelial cells, by reducing the level of EGFR-dependent MAPK activation (Moghal and Neel, 1998). Thus, molecules involved in cell adhesion can act either positively or negatively to regulate EGFR signaling.

In this report, we describe the identification of the gene echinoid (ed) as a novel negative regulator of EGFR signaling. Ommatidia of ed mutant flies contain extra photoreceptor and cone cells. In contrast, ectopic expression of ed in the eye leads to a reduction in the number of photoreceptors. We show that ed is not transcriptionally regulated by EGFR signaling and that it encodes a putative cell adhesion protein which is primarily localized to the plasma membrane of every cells throughout the eye disc. Our genetic analyses demonstrate that ed acts nonautonomously to generate extra R7 cells by a mechanism that is sina-independent but upstream of TTK (TTK88). Together, our results support a model whereby ED defines an independent pathway that antagonizes EGFR signaling by regulating the activity of the TTK88 transcriptional repressor.

The stocks used were ed^slH8, ed^slA12, ed^lF20 (de Belle et al., 1993); l(2)k01102 (Torok et al., 1993); Elp^B1 (Baker and Rubin, 1989); Gap1^B2 (Chou et al., 1993); sos^e4G (Simon et al., 1991); hs-aos (Freeman, 1994); rl^SEM (Brunner et al., 1994); sev^d2 (Simon et al., 1991); sev-Ras^V12 (Fortini et al., 1992); sev-Ras^N17 (Karim et al., 1996); sev-tor⁴⁰²¹Egfr (Reichman-Fried et al., 1994); sev-tor⁴⁰²¹Raf (Dickson et al., 1992); pnt¹²⁷⁷, pnt^Δ88 (Brunner et al., 1994); sevyan^ACT (Rebay and Rubin, 1995); sina², sina³ (Carthew and Rubin, 1990); ttk⁰²¹⁹ (Lai et al., 1996); GMR-ttk88 (Tang et al., 1997); sprouty^S88 (Casci et al., 1999); UAS-kek1 (Ghiglione et al., 1999); hsrho (Dominguez et al., 1998); GMR-GAL4 (Freeman, 1996); and MS1096-GAL4 (Capdevila and Guerrero, 1994).

Mitotic clones in the eye were induced by using ey-FLP (kindly provided by B. Dickson). The genotype of these flies was w,sev^d2/Y; ed^slA12,FRT^40A/ P[w+]^30C, FRT^40A; ey-FLP/+.

A 8 kb EcoRI genomic DNA flanking the l(2)k01102 P-element insertion was isolated by plasmid rescue and used to screen an eye disc cDNA library (provided by Dr A. Cowman). Two classes of cDNAs were isolated and sequenced. One is 5.1 kb and encodes an open reading frame (ORF) of 1332 amino acids. We refer to it as ed cDNA. The other cDNA is 4.5 kb and encodes a noncoding RNA: the longest open reading frame (ORF) would encode a polypeptide of 102 amino acids but its AUG codon is in a poor position for translation initiation (Cavener, 1987). The ORF region of the ed cDNA was generated by PCR and inserted into the pCaSpeR-hs and pUAST transformation vectors (Brand and Perrimon, 1993) to create hs-ed and UAS-ed. UAS-ed^Δintra was made by inserting by PCR a stop codon 31 amino acids after the transmembrane domain of ED. All the constructs were sequenced. Transgenic lines were generated by P-element-mediated transformation (Spradling and Rubin 1982). hs-ed was tested for its ability to rescue the lethality of ed^slH8 by heat shock in a 37° water bath for 20 minutes every 12 hours throughout development.

ed-specific RT-PCR was performed as follows. Five late third instar larval eye discs of different genotypes were dissected and subjected to cDNA synthesis with a Cells-to-cDNA system (Ambion). 20 cycles of PCR amplifications were carried out with primer pairs from the seventh (CGATGCCCGGAAATGAATGG) and ninth exon (GCGTATGACGCGACGGTTT) of ed genomic DNA. 18S rRNA primers (Ambion) were used as internal controls.

Fixation, embedding and sectioning of adult retina were performed as described by Wolff and Ready (Wolff and Ready, 1991). Scanning electron micrographs were prepared as described by Kimmel et al. (Kimmel et al., 1990). Cobalt sulphide staining of pupal retinas was performed as described by Wolff and Ready (Wolff and Ready, 1991). Immunohistochemical staining of imaginal discs was performed as described in Xu and Rubin (Xu and Rubin, 1993). Polyclonal rabbit α-ED antibodies were generated against a synthetic peptide, corresponding to the N-terminal region of ED (MRRKTVTKGTAIVNSRSARRAATTI) and were used at a dilution of 1: 200. α-ELAV (rat, 1: 250, Developmental Studies Hybridoma Bank); αi-Cut (mouse, 1: 5, Developmental Studies Hybridoma Bank); α-β-galactosidase (rabbit, 1: 1000, Cappel); α-TTK88 (mouse, 1:100); and α-Boss (mouse, 1:1000); Cy3-, Cy5-, FITC-conjugated secondary IgGs are from Jackson Immunological Laboratories. Confocal microscopy was performed using a Zeiss Model 310.

The accession number for the ed sequence reported in this paper is AF275903.

Elp^B1 is a gain-of-function allele of the Egfr (Baker and Rubin, 1989). We carried out a genetic modifier screen for components of the EGFR pathway that dominantly enhance or suppress the rough eye phenotype caused by Elp^B1 (Fig. 1B). 1×5 was isolated as an EMS induced mutation which strongly enhances the rough eye phenotype associated with Elp^B1 (Fig. 1C). The dominant enhancer activity of 1×5 is similar to the effect of Gap1 (Fig. 1D) or yan (data not shown) mutations, two known negative regulators of the EGFR signaling pathway.

Consistent with the genetic interaction with Elp^B1, 1×5 also enhances the eye phenotype caused by sev-tor⁴⁰²¹Egfr (Fig. 1E,F), another constitutively active form of the EGFR (Reichman-Fried et al., 1994). To define further the role of 1×5 in the EGFR signaling pathway, we examined the genetic interactions between 1×5 and rho, a specific activator of EGFR pathway, and aos, a specific EGFR inhibitor. Interestingly, we found that 1×5 enhances the rough eye phenotype caused by ectopic expression of rho (Fig. 1G,H), and suppresses the rough eye phenotype caused by misexpression of aos (Fig. 1I,J).

Further genetic interactions between the EGFR pathway and 1×5 were also detected in the wing. 1×5 enhances the extra wing-vein phenotype caused by the overactive Elp^B1 mutation, as well as rl^SEM, a constitutively active MAPK (Brunner et al., 1994; data not shown). In addition, flies heterozygous for both 1×5 and Gap1, or both 1×5 and sty^S88 (data not shown), exhibit extra vein materials, although heterozygosity for either mutation alone causes no phenotype. Therefore, the genetic interactions observed between 1×5 and several components of the EGFR pathway suggest that 1×5 is a negative regulator of the EGFR signaling pathway during eye and wing vein development.

1×5 was mapped to 24D3-4 using three overlapping deficiencies: Df(2L)ed1 (24A3-4;24D3-4), Df(2L)ed-dp (24C3-5;25A2-3) and Df(2L)M24F11 (24D3-4;25A2-3). This region contains the ed gene and we found that ed^lF20 (de Belle et al., 1993) fails to complement 1×5 and enhances the Elp^B1 rough eye phenotype, as well as the extra wing vein phenotype of rl^SEM (data not shown). Thus 1×5 is allelic to ed, and we refer to it as such below. All ed mutations are pupal lethal in homozygotes with the exception of ed^slH8, which is a weaker allele. Homozygous ed^slH8, as well as ed^slH8 in combination with all other ed alleles, including Df(2L)ed-dp, are semi-lethal. Emerging adults have rough eyes (Fig. 2B) and extra wing veins (see Fig. 5J). When sectioned, 33% of ommatidia contain extra R7-like cells with small and centrally positioned rhabdomeres (Fig. 2D). To exclude that these extra cells with small rhabdomeres are R8, third instar larval imaginal discs of ed^1×5/ed^slH8 transheterozygotes were stained with anti-Boss, an R8-specific antibody. Single R8 cell was seen in each mature ommatidium (Fig. 2E), confirming that the extra photoreceptor cells are indeed R7. In addition, 26% of ommatidia exhibit extra outer-photoreceptor cells while 6% of the ommatidia show reduced outer-photoreceptor cells. Further, ed^slH8 hemizygotes animals have more R7 cells than ed^1×5/ed^slH8 transheterozygote animals indicating that the ed alleles are loss of function. ed^slH8 hemizygotes have 1.68 R7 cells in average (n=64), compared with 1.34 (n=164) in ed^1×5/ed^slH8. To determine the origins of the extra photoreceptor cells, ed^1×5/ed^slH8 transheterozygote discs were stained with the anti-Elav neural marker (Fig. 2F). Extra Elav-positive cells were first detected in rows 2 and 3, where R8/R2/R5 are located. However, these four-cell clusters contain only single R8. In addition, one or two extra Elav-positive mystery cells were detected adjacent to R3 and R4 cells four row of cells behind the furrow. Mystery cells will normally leave the five-cell precluster and disappear; however, as in sty or yan mutants (Casci et al., 1999; Lai et al., 1992), they are transformed into neuronal photoreceptor cells in the ed mutant discs.

We also examined the ed mutant phenotype during pupariation. At this stage there are four cone cells and two primary pigment cells in wild-type discs (Fig. 2G). However, 69% of ommatidia in ed^1×5/ed^slH8 transheterozygotes exhibit five or six cone cells (Fig. 2H) and 10% contain three primary pigment cells (data not shown). Together, the over-recruitment of photoreceptor, cone and pigment cells in ed mutants is consistent with ED acting as a negative regulator of EGFR because previous analyses have shown that EGFR is required for differentiation of these three cell types (Freeman, 1996).

We identified a P-element insertion, l(2)k01102 (Torok et al., 1993) located at 24D3-4, that failed to complement either ed^1×5 or ed^lF20. To characterize ed molecularly, we recovered the DNA region flanking l(2)k01102 by plasmid rescue (see Materials and Methods). A 8 kb genomic DNA was isolated and used to screen an eye disc cDNA library. A 5.1 kb cDNA was isolated and sequence analysis revealed that it encodes an open reading frame of 3996 bp, which predicts a protein of 1332 amino acids (Fig. 3B). The translated protein contains six immunoglobulin (Ig) C2 type domains (Williams and Barclay, 1988) (Fig. 3C), a fibronectin type III domain (Hynes, 1986) (Fig. 3D) and a transmembrane domain, followed by a 315 amino acid C-terminal tail with no identifiable functional motif. A comparison of the genomic and cDNA sequence indicates that l(2)k1102 is inserted in the first intron, which is upstream of the coding region (Fig. 3A). To establish that the 5.1 kb cDNA identifies ed, we expressed the cDNA under the control of a heat shock promoter. Following heat shock treatments, we found that the hs-ed transgene rescues the lethality associated with the weak ed^slH8 allele (data not shown).

To detect the expression pattern of ED, we stained embryos with an antibody generated against the N-terminal ED peptide. The ED protein is widely expressed in the epidermis and is localized to the plasma membrane (Fig. 4A). Further, we find that ED is uniformly detected in all cells throughout the third instar larval eye and wing disc (data not shown).

The expression of aos and kek1, two other negative regulators of the EGFR pathway, is regulated by the EGFR pathway. To determine whether ed is regulated by the EGFR pathway, we examined the expression of ed in GMR-aos (Fig. 4B-D) and sev-Ras^V12 (data not shown) eye discs. In each case, the level of ed mRNA is not affected, as revealed by either the X-gal staining of the P insertion l(2)k1102. (Fig. 4B,C) or the ed-specific relative RT-PCR (Fig. 4D). These results indicate that ed, unlike aos and kek1, is not transcriptionally regulated by the activation of EGFR pathway.

As shown above, loss of ed function is required for the formation of photoreceptor, cone and primary pigment cells. To determine the effect of overexpression of ed in the eye, we expressed UAS-ed using the GMR-Gal4 driver. GMR-Gal4; UAS-ed flies exhibit a small rough eye (Fig. 5A) and a reduced number of photoreceptors (Fig. 5B); this effect correlates with the reduced number of Elav-positive cells (Fig. 5C) in the eye disc. There are only four or five Elav-positive cells per cluster. In contrast, no obvious defects in the formation of cone cells were observed in response to ed overexpression, as most ommatidia still contain four Cut-positive cells (data not shown). Flies carrying two copies of GMR-GAL4-driven UAS-ed exhibit complete absence of the eye (data not shown).

To further document the interaction between ED and the EGFR pathway, we examined the effect of ectopic expression of ed in flies where other regulators were overexpressed. Overexpression of UAS-sty alone by GMR-GAL4 produces small rough eye (Fig. 5D). This phenotype can be partially suppressed by halving the dose of ed (Fig. 5E), and enhanced by GMR-GAL4-driven UAS-ed (Fig. 5F). Similar genetic interactions can also be observed between ed and kek1. The rough eye phenotype caused by GMR-GAL4-driven UAS-kek1 (Fig. 5G) is enhanced by GMR-GAL4-driven UAS-ed (Fig. 5H). Therefore ed, like sty and kek1, is a repressor of EGFR signaling during eye development.

Similarly, during wing vein development, ed genetically interacts with several components in the EGFR pathway. Flies of ed^1×5/ed^slH8 have increased size of wing and extra wing vein (Fig. 5J). However, ectopic expression of ed using MS1096 GAL4 results in severe reduction in the size of wing, ranging from one quarter to one fifth of normal wing size. In addition, there is no vein material present (Fig. 5K).

ED contains six Ig domains and a 315 amino acid intracellular domain. To determine whether the intracellular domain of ED is required for the repression of the EGFR signaling, we generated UAS-ed^Δintra flies. Overexpression ofUAS-ed^Δintra usingGMR-GAL4 had no phenotypes in the eye, indicating that the cytoplasmic domain of ED is required for the repression of the EGFR signaling pathway.

To determine where in the RAS/RAF/MAPK signaling pathway ed acts, we conducted a number of genetic epistasis experiments. sev^d2 is a loss-of-function sevenless (sev) allele (Simon et al., 1991) and sev^d2 mutant flies lack R7 cells (Fig. 6A). Although ommatidia within a ed^1×5/ed^slH8 mutants contain an average of 1.34 R7 cell (Fig. 2F), we found that ommatidia within a sev^d2; ed^1×5/ed^slH8 double mutant contain an average of 1.37 R7 cells (n=61) (Fig. 6B). This demonstrates that in ed mutants, the formation of supernumary R7 cells is independent of sev function. In addition, ed^1×5 enhances the rough eye phenotype caused by overexpressing constitutive active forms of either the EGFR (Fig. 1C,F), RAS1 (Fig. 6C,D), or RAF (Fig. 6G,H). Conversely, ed^1×5 suppresses the rough eye phenotype caused by overexpressing dominant negative RAS1 (Fig. 6E,F). While 61% of ommatidia in a sev-Ras^N17/+ mutant lack R7 cells (n=164), only 10% of ommatidia in ed^1×5/ed^slH8; sev-Ras^N17/+ double mutants lack R7 photoreceptors. In addition, at 25°, ed^1×5 also rescues the lethality of Raf^HM7, a temperature-sensitive Raf allele. Therefore, ed acts either downstream of the Ras/Raf pathway or in parallel.

To determine whether ED acts in the nucleus, we generated flies double mutant for ed;pnt, ed;yan or ed;sina. We found that pnt^Δ88/pnt¹²⁷⁷ (Fig. 6I) and sev-yan^ACT/+ (Fig. 6K) ommatidia contain an average of 0.69 (n=264) and 0.05 (n=250) R7 cells, respectively. However, ed^1×5/ed^slH8; pnt^Δ88/pnt¹²⁷⁷ (Fig. 6J) and ed^1×5/ed^slH8; sev-yan^ACT/+ ommatidia (Fig. 6L) contain an average of 1.44 (n=102) and 1.01 (n=125) R7 cells, respectively. Strikingly, ed^1×5/ed^slH8; sina²/sina³ ommatidia (Fig. 6N) contain an average of 1.29 (n=180) R7 cells, as compared with 0.01 (n=173) R7 cells in sina²/sina³ mutant (Fig. 6M). Therefore, in ed mutants, the formation of supernumary R7 cells is independent of sina function. Finally, loss of ttk activity has been shown to produce ectopic R7 cells in a sina-independent manner (Lai et al., 1996). To determine whether ed acts downstream of ttk, we overexpressed ttk in ed mutants. Overexpression of TTK88 under the control of either the GMR enhancer that completely inhibits photoreceptor cell development (Fig. 6L) or the sev enhancer that only deletes R3, R4 and R7 photoreceptors (data not shown). However, this TTK88-mediated neuronal repression cannot be suppressed by removing ed activity (Fig. 6P), indicating that ed acts upstream of ttk to specify R7 development. Together, our genetic epistatic analysis suggests that ed acts either parallel or downstream of Ras, Raf, pnt, yan and sina, but upstream of ttk to specify R7 cell fates.

Our genetic epistatic analyses suggest that ed acts upstream of ttk88 to specify R7. ed might regulate ttk88 mRNA expression or TTK88 protein levels. Alternatively, ed might regulate the activity of TTK88 through protein modification, like phosphorylation. To determine whether ed regulates ttk expression, we examined the expression of ttk in ed mutant disc using the X-gal staining of the P-element insertion ttk⁰²¹⁹ (Li et al., 1997) and detected no obvious changes (Fig. 7B). Furthermore, TTK88 is expressed at high levels in the cone cells but is not expressed in developing photoreceptor cells (Li et al., 1997; Dong et al., 1999). To determine whether ed regulates TTK88 protein levels, we examined TTK88 levels in ed (Fig. 7D) and GMR-Gal4; UAS-ed eye discs (data not shown). In each case, the level of TTK88 was unaffected. Together, our results suggest that ED does not regulate ttk88 mRNA expression or TTK88 protein stability.

To determine in which cells ed is required, we used ey-FLP (Newsome et al., 2000) to generate clones of homozygous ed^slA12 mutant cells in a sev^d2 background. As shown in Fig. 6A, no R7 cells develop in the sev^d2 background. We scored 54 mosaic ommatidia that contain R7-like cells. Among them, 57% of the R7-like cells were ed⁻ (Fig. 8A), while 43% were ed⁺ (Fig. 8B). Similar results were obtained when we generated ed mutant clones in sina and sev-yan^ACT mutant backgrounds (data not shown). The observation that R7 cells can be derived from either wild-type or ed mutant cells, leads us to propose that the ed mutation acts cell non-autonomously in the generation of supernumerary R7 cells.

ED is uniformly expressed in the follicle cells during stage 1-10 oogenesis (data not shown). To determine whether ed acts during oogenesis in the establishment of EGFR-dependent dorsal/ventral polarity, we examined the eggs derived from ed^slH8/Df(2L)ed-dp females. These females are fertile and do not exhibit any overt morphological defects (data not shown). As loss-of-function mutations in many cell adhesion molecule have subtle mutant phenotypes (Ghiglione et al., 1999), we overexpressed UAS-ed in the follicle cells using the GAL4 drivers T155 or CY2 (Queenan et al., 1997; Ghiglione et al., 1999). The eggs derived from such females have completely normal dorsal appendages (data not shown) suggesting that ED does not interfere with EGFR signaling in follicle cells.

The EGFR plays important roles at various stages of Drosophila development and is subject to modulation by multiple positive and negative regulators. We have identified ED as a novel adhesion molecule-like protein that negatively regulates the EGFR signaling pathway. ed genetically interacts with several components in the EGFR pathway. Flies of ed mutant produce extra photoreceptor and cone cells. Conversely, ectopic overexpression of ed in the eye leads to reduction of photoreceptor number. We demonstrate that ED acts by converging on TTK88, the most downstream component known in EGF receptor signaling. Our results not only demonstrate the active role of an adhesion molecule in the EGFR signal transduction pathway but also identify a previously unknown regulatory mechanism.

ED contains six Ig domain with extensive homology to vertebrate neural adhesion molecule L1. The L1-family of cell adhesion molecules exert their functions through homophilic or heterophilic interactions with other Ig domaincontaining adhesion molecule (Hortsch, 1996). We found that ED is expressed in every cell of the eye disc. In addition, our genetic analysis demonstrate that ed acts in a cell nonautonomous manner to generate extra R7 cells. If ED transmits the negative signal from the sending cell via homophilic interaction to the receiving cell, loss of ed in either sending or receiving cells would result in the same phenotype, owing to the failure to receive the inhibitory signal. Therefore the extra R7 cells found in the receiving cells could be either wild type or mutant for ed. However, if ED transmits the negative signaling via heterophilic interaction, ed is only required in the sending cells but not the receiving cells. Therefore, the extra R7 cells found in the receiving cells could be either wild type or mutant for ed. Alternatively, ED might act as a ligand that activates an unidentified receptor on receiving cells. All three models are consistent with our results showing that ed functions cell nonautonomously. However, only the homophilic interaction model would require the cytoplasmic domain of ED to be required in both the sending and receiving cells. Since we found that the cytoplasmic domain of ED is required for the repression of the EGFR pathway, we favor the homophilic interaction model between ED molecules to specify photoreceptor cell formation (Fig. 9).

Studies on RTK signaling in both vertebrates and invertebrates have converged on an evolutionarily conserved DRK/RAS/ RAF/MAPK signaling cassette that is required to transmit the signal from the receptor to the nucleus. Previous studies on TORSO (TOR) signaling, however, indicated that TOR RTK transduces its signals through both a RAS-dependent and an unidentified RAS-independent pathways that converge on RAF (Hou et al., 1995). In addition, genetic analysis of daughter of sevenless (Dos) has revealed that it functions upstream of RAS1 and defines a signaling pathway that is independent of direct binding of DRK/GRB2 to the SEV RTK (Raabe et al., 1996; Herbst et al., 1996). Nevertheless, either a RAS-or KSR-independent pathway still acts underneath a RTK.

In contrast to the previous examples, our genetic data indicates that ED functions either downstream of RAS1/MAPK/PNT/YAN or in a parallel pathway. However, based on the following data we argue that ED is unlikely to act downstream of PNT/YAN/SINA in the nucleus, but instead defines an independent pathway that antagonizes EGFR signaling (Fig. 9). First, antiserum against ED N-terminal peptide localizes ED to the plasma membrane, but not the nucleus, of every cell in the eye disc. Second, ED functions non-autonomously in the signal-sending cells. Third, ed is not transcriptionally regulated by the activation of the EGFR pathway, a situation that is different from other negative regulators of EGFR such as aos, sty and kek1. Fourth, SINA has been shown to form a complex with PHYL to target TTK for degradation. The production of ectopic R7 cells in mutations of most negative regulators, like Gap1 and yan, all require sina (Gaul et al., 1992; Lai and Rubin, 1992). However, the formation of extra R7 cells in ttk mutant is only partially sina dependent. This observation led Lai et al., to suggest that the production of extra R7 cells in ttk mutant is partly influenced by both the normal R7 developmental signals, which are sina dependent, and another sin-independent signaling (Lai et al., 1996). Our genetic analysis demonstrates that the production of ectopic R7 cells in ed mutant is completely sina independent. The observation that ed functions upstream of ttk88, implies that the independent inhibitory pathway, although sina independent, converges on TTK.

ED, a putative cell adhesion protein, is constitutively expressed on every cells on the eye disc. There are several ways in which ed expression can influence signaling. For example, it could induce polarization and adherens junction formation of undifferentiated cells. The EGFR is localized to the apical microvillar border where it binds its inductive ligand (Zak and Shilo, 1992). The apical restriction of EGFR may concentrate these receptors at a high density and allows efficient capture of the SPI ligand, thus restricting SPI diffusion over a long distance. In the absence of ED, the EGFR might diffuse to the basolateral membrane and the density of EGFR may be too low to capture SPI efficiently. According to this model, SPI may diffuse to distant cells. These cells which normally do not encounter the ligand would then differentiate extra photoreceptor or cone cells. In this case, ED would function as a mechanical force to affect the binding efficiency of EGFR and the diffusion distance of SPI. However, we do not favor this mechanism because we observe that halving the dose of ed can enhance the rough eye phenotype caused by constitutively active EGFR (sev-tor⁴⁰²¹Egfr), which does not require SPI.

The other possibility is that ED, via homophilic interactions, may directly transmit a negative signal. This signal would counteract the basal ligand-independent activity of RTK, caused by nonspecific RTK oligomerization, and establish an inherent inhibitory network to prevent cells from differentiating as photoreceptor or cone cells. According to this model, only when a cell receives its ligand can it activate its RTK and antagonize this negative effect and differentiates. Thus, loss of ed activity behind furrow produces ectopic photoreceptor and cone cells. In this case, a photoreceptor differentiation response would be elicited in region where only the RTK signaling pathway was activated. Thus, RTK activation functions in a permissive manner.

Loss of ed gene activity results in ectopic photoreceptor and cone cells formation. Thus ED, like GAP1 and YAN, functions as a general repressor of differentiation in the developing eye. The negative signal that ED transmits might be mediated through TTK to repress photoreceptor cell formation and another negative regulator to repress cone cells formation. How does ed transmit an inhibitory signal into the receiving cells? The neurite outgrowth and axonal fasciculation mediated by L1-family neural adhesion molecule require both the homotypic or heterotypic interactions of extracellular Ig domains and the conserved ankyrin binding site (FIGQY) in the cytoplasmic domain (Dubreuil et al., 1996; Hortsch et al., 1998). Unlike members of L1-family, ED contains a 315 amino acid cytoplasmic domain with no apparent sequence homology to the conserved FIGQY ankyrin-binding site. However, deletion analysis indicates that the intracellular domain is required for the repression of the EGFR pathway. Finally, the repressing effect of ED signaling in photoreceptors is mediated through TTK. There are several ways that ED signaling might affect TTK. It might elevate the repressing activity of TTK through posttranslational modification, like phosphorylation, to increase its DNA binding affinity to its target genes. Alternatively, it might directly upregulate ttk mRNA or its protein levels. The activation of EGFR signaling has been shown to downregulate TTK88 protein stability (Tang et al., 1997; Li et al., 1997). We favor the former possibility, as we found that the levels of both ttk mRNA and ED protein are unaffected in ed mutant background.

ED is widely expressed at various stages of Drosophila development (J.-C. H., unpublished observations). Our results demonstrate that ED is a negative regulator of EGFR and Sev signaling pathways during eye and wing development. However, ED does not appear to be involved in EGFR signaling during oogenesis. Therefore, ED differentially functions as an inhibitor of RTK in a tissue-specific manner.

We thank M. B. Sokolowski, I. Kiss, G. M. Rubin, Z.-C. Lai, M. Freeman, M. Simon, C. Klambt, B. Dickson, E. Hafen, H. Y. Sun and the Bloomington Stock Center who sent us fly stocks, and Z.-C. Lai, S. L. Zipursky and R. W. Carthew, who sent us anti-TTK88 antibody. We also thank D. F. Ready for help with cobalt sulphide staining, C.-T. Chien for help with confocal microscopy, C.-K. Chen for help with antibody staining and C.-Y. Tang for help with microinjection. We are grateful to M. P. Zeidler, C.-T. Chien and Henry Y. Sun for helpful discussion and critical reading of the manuscript. This study was supported by grants to J. C. H. from the National Science Council (NSC), Taiwan, Republic of China, and a grant from the US Army Breast Cancer Research (N. P).