Ebi modulates wing growth by ubiquitin-dependent downregulation of Crumbs in Drosophila

Notch signaling is a conserved mechanism that regulates diverse developmental events, including tissue growth, cell fate specification and cell polarity (Artavanis-Tsakonas et al., 1999; Lai, 2004). Thus, how Notch signaling is regulated is an important developmental issue. Notch signaling is activated by the interaction between the extracellular domains (ECDs) of the Notch receptor and its transmembrane protein ligands, Delta and Serrate (de Celis et al., 1996). Ligand binding to Notch results in the cleavage of the intracellular domain (ICD) by γ-secretase (Mumm et al., 2000). The ICD fragment of Notch enters the nucleus, leading to transcriptional activation of specific target genes (Lieber et al., 1993; Lai, 2004).

Crumbs (Crb) is a transmembrane protein with EGF-like repeats in the extracellular domain, a structural feature shared with Notch (Tepass et al., 1990). Crb has multiple functions, including roles in apicobasal cell polarity (Tepass et al., 1990; Wodarz et al., 1995; Tanentzapf et al., 2000), morphogenesis (Izaddoost et al., 2002; Pellikka et al., 2002), Hippo signaling (Parsons et al., 2010; Laprise, 2011; Ribeiro et al., 2014) and mitosis (Yeom et al., 2014). Interestingly, recent studies have provided evidence that Notch activity is affected by Crb. For example, loss of Crb in the eye imaginal disc results in head overgrowth by increased cell proliferation due to ectopic Notch activity. This activation of Notch signaling is correlated with an increased endocytosis of Notch and its ligand Delta. Hence, independent of its function in cell polarity, Crb acts as an inhibitory factor to Notch activation by limiting endocytosis (Richardson and Pichaud, 2010). Crb is also involved in the inhibition of Notch signaling in vertebrates like zebrafish, in which Notch activity is necessary for the apical mitosis of neuroepithelial cells during embryogenesis (Ohata et al., 2011). The inhibition of Notch activity by Crb is due to a direct interaction between the extracellular domains of Crb and Notch. Further, Mosaic eyes (Moe, a homolog of Drosophila Yurt) antagonizes the Crb function in a positive feedback loop to maintain the apical basal gradient of Notch activity in neuroepithelial cells, thus restricting their mitosis to the apical area (Ohata et al., 2011).

These interactions between Crb and Notch described above suggest that Crb plays distinct roles in the regulation of Notch signaling in different developmental contexts. In Drosophila, activation of Notch signaling at the DV boundary of wing discs is pivotal for wing growth (de Celis et al., 1996). It has been shown that while crb gene expression is induced by Notch signaling, Crb protein antagonizes Notch signaling by interfering with γ-secretase (Herranz et al., 2006), a protease necessary for Notch signaling (Schweisguth, 2004). Thus, the level of Crb at the DV boundary needs to be downregulated to maintain Notch signaling in wing disc. However, the mechanism underlying the regulation of Crb remains to be understood.

Genetic studies have shown that the ebi gene is required for wing growth by activating or de-repressing transcription of Notch target genes in the DV boundary (Marygold et al., 2011). Although both Ebi and Crb are involved in the modulation of Notch signaling at the DV boundary of wing disc, the relationship between these two proteins has not been studied. Ebi protein contains a LisH domain and an F box-like motif in the N-terminal region and eight WD40 repeats in the C-terminal region (Marygold et al., 2011). Interestingly, Ebi is known to function in the degradation of specific target proteins by interacting with an E3 ubiquitin ligase Seven-in-absentia (Sina), the founding member of the SIAH family proteins (Matsuzawa and Reed, 2001; Tsuda et al., 2006). This raises a possibility that Ebi might be involved in downregulation of Crb by directly interacting with a specific region of Crb. Crb protein consists of a large extracellular domain (Crb^ext) and a short intracellular domain (Crb^intra) (Tepass et al., 1990). While the Crb^intra domain and its interaction partners have been extensively studied (Bachmann et al., 2001; Hong et al., 2001; Izaddoost et al., 2002; Nam and Choi, 2003; Tepass, 2012), few binding partners of the Crb^ext domain have been identified thus far (Hafezi et al., 2012; Roper, 2012; Zou et al., 2012; Letizia et al., 2013; Pocha and Knust, 2013).

In this work, we identified Ebi as an interacting partner of the Crb^ext domain. Genetic evidence indicates that Ebi is antagonistic to Crb function in wing development. We show that Ebi is required for ubiquitin-dependent downregulation of Crb. This study provides a novel mechanism for the regulation of Notch signaling through the antagonistic interaction between Ebi and Crb.

To gain insights into the function of Crb^ext, we searched for binding partners of a conserved domain of Crb^ext using a yeast two-hybrid screen. The bait was the fourth laminin G domain (Crb^lam4). This domain is most closely related to the third laminin G domain of human CRB1 in which many mutations associated with retinal diseases have been identified (den Hollander et al., 2004). One of the Crb^lam4-interacting clones contained a 230 bp cDNA insert that encodes WD40 repeats 7 and 8 located in the C-terminal region of Ebi (Fig. 1A). To further confirm the binding between Ebi and Crb, we examined their interaction by co-immunoprecipitation (co-IP). In the assay using S2 cells transfected with Crb^lam4TM-V5 and Flag-Ebi^full, these two proteins were coimmunoprecipitated (Fig. 1B), indicating that the extracellular domain of Crb can form a protein complex with Ebi.

To identify the specific region of Ebi involved in Crb^lam4 binding, we performed pull-down experiments using GST-fusion proteins. The N-terminal region of Ebi has a LisH domain, while the C-terminal part consists of eight WD40 repeats. Consistent with the two-hybrid interaction, pull-down assays confirmed the specific binding between Crb^lam4 and the WD40 7-8 repeats of Ebi (hereafter Ebi^WD7/8). Additional binding assays with truncated mutant Ebi proteins showed that Crb^lam4 did not bind the WD40 1-6 repeats or the N-terminal region of Ebi (Fig. 1C). Furthermore, other domains of Crb such as the laminin G domains 1-3, the EGF repeat domain or the intracellular domain also did not bind to the Ebi^WD7/8 (Fig. 1D). These data indicate that the Crb^lam4 domain specifically interacts with the Ebi^WD7/8 repeats.

To check for physiological relevance of the binding between Crb and Ebi, we tested whether these two genes show any genetic interaction in wing development. Consistent with a previous study (Marygold et al., 2011), RNAi knockdown of Ebi in the DV boundary region driven by C96-Gal4 (C96>ebi RNAi) caused notching along the wing margin resulting in a 33±3% (n=10) reduction of the wing size (Fig. 2B,G) compared with the control (Fig. 2A). Three independent ebi RNAi lines showed similar wing defects (Fig. S1B-D). Because Notch signaling in the wing disc induces expression of its target gene cut along the DV boundary (de Celis et al., 1996; Micchelli et al., 1997), we examined the effects of ebi knockdown on the level of Cut protein. As expected from the notching phenotype in the adult wing, ebi RNAi by C96-Gal4 resulted in reduced Cut expression along the DV boundary (50±2% reduction, n=5) (Fig. S1F). This ebi RNAi wing phenotype was significantly suppressed in 70% (n=10) of examined wings by reducing the crb gene dosage in the crb^11A22/+ heterozygote. The size of ebi RNAi wings was increased by 85±3% (n=10) in the crb^11A22/+ background (Fig. 2C,G) compared with the control (Fig. 2A). Similarly, reduced Cut expression by ebi RNAi was significantly restored by crb^11A22/+ (Fig. S1G). These results suggest that ebi and crb function antagonistically.

Since Ebi binds to the extracellular domain of Crb, we checked whether there is any specific genetic interaction between Ebi and Crb^ext. ebi RNAi in the wing pouch by nub-Gal4 led to a 47±3% (n=10) reduction of the wing size at 25°C (Fig. S2B). Similarly, overexpression of Crb^ext (Crb^extTMGFP) by nub-Gal4 resulted in 50±3% (n=10) reduction of the wing size (Fig. S2C). Both nub>crb^extTMGFP and nub>ebi RNAi wings also showed loss of crossveins in the wing (Fig. S2B,C). ebi RNAi and Crb^ext overexpression caused a 80±3% reduction of wing size at 29°C (Fig. 2D,D′,H). Under these conditions, overexpression of wild-type Ebi (Ebi^full) resulted in partial suppression of the reduced wing phenotype caused by Crb^extTMGFP (Fig. 2E,E′,H), although overexpression of Ebi alone as a control did not affect wing development (Fig. S2D). These results indicate that Ebi can antagonize the effects of Crb^ext overexpression.

To test whether the WD40^7/8 repeats that bind to Crb are important for the function of Ebi, we generated UAS-ebi^ΔWD7/8 to express a mutated Ebi protein deleted in the WD40^7/8 repeats. In contrast to the wild-type Ebi (Ebi^full) (Fig. 2E,E′), Ebi^ΔWD7/8 failed to suppress the nub>crb^extTMGFP phenotype (Fig. 2F,F′,H). Hence, the WD40^7/8 domains of Ebi are not only important for binding Crb^lam4 but also required for Ebi to antagonize Crb.

To understand the mechanism for the antagonistic genetic interaction between Ebi and Crb, we checked whether Ebi regulates the level of Crb protein. As shown previously (Herranz et al., 2006), the crb-lacZ reporter is strongly expressed in the DV boundary region of all wing discs examined (Fig. 3D-D′′). In contrast to crb-lacZ expression, Crb protein is expressed uniformly in most wing discs with no significant increase in the DV boundary region (Fig. 3A,A′), although increased Crb expression in the DV boundary was found in about 10±2% (n=50) of wing discs examined. Thus, we reasoned that excessive Crb protein induced in the DV boundary might be unstable and subject to degradation. If Ebi is involved in the downregulation of Crb, reduction of Ebi may increase the Crb protein level in the DV boundary region. Indeed, Ebi knockdown by C96>ebi RNAi resulted in upregulation of Crb in the DV boundary region compared with the surrounding areas (42±2% upregulation in ∼50% of wing discs examined, n>50) (Fig. 3B). We also tested whether Ebi knockdown in the anterior compartment by Ci-Gal4 is sufficient to increase the Crb level in the targeted region. Consistent with the result from C96>ebi RNAi, Ebi knockdown by Ci-Gal4 led to upregulation of Crb in the anterior region but not in the posterior control region in about 50% of wing discs examined (n>30) (Fig. 3C-C′′).

To determine whether Ebi regulates crb expression at the transcriptional level, we examined the effects of Ebi knockdown on the expression of a crb-lacZ reporter (Herranz et al., 2006). Crb-lacZ in the wild-type background was strongly induced along the DV boundary compared with surrounding regions of the wing pouch (Fig. 3D′). In contrast to the increase of Crb protein level by ebi RNAi, Ci>ebi RNAi did not significantly alter the level of crb-lacZ expression in the anterior wing compartment (Fig. 3D-D″). Thus, Ebi seems to downregulate the level of Crb protein post-translationally rather than by transcriptional repression.

In addition, we tried to examine the effects of an ebi null mutation on Crb expression, but mutant clones could not survive or were too small to determine the effect, as also previously reported (Marygold et al., 2011). As an alternative, we generated ebi RNAi clones by flp-out recombination (Theodosiou and Xu, 1998) to compare the Crb expression level in ebi RNAi clones and the adjacent wild-type cells. An early induction of flippase resulted in large areas of ebi RNAi clones marked with GFP. Similar to the results from Ci>ebi RNAi, the level of Crb was increased in the DV boundary region within ebi RNAi clones compared with the wild-type control area (Fig. 3E-E″). Taken together, these results show that Ebi is required for downregulation of Crb protein in the DV boundary.

To confirm the role of Ebi in Crb downregulation, we examined the level of Crb protein from fly extracts by western blot analysis. Crb protein levels in crb^11A22/+ heterozygotes were reduced by 48±2% (n=3) compared with wild-type level (Canton-S or w¹¹¹⁸) (Fig. 4A,A′). In contrast, the level of Crb protein was about 60±2% (n=3) higher in ebi^E4/+ heterozygotes than the wild-type level (Fig. 4A,A′). These results suggest that Ebi is required for the downregulation of Crb protein level.

It has been shown that Ebi is involved in proteasome-dependent degradation (Matsuzawa and Reed, 2001; Tsuda et al., 2002). Thus, we postulated that Crb might be a target for ubiquitin-mediated degradation by Ebi. To test this idea, first we checked whether reducing proteasome activity could increase the Crb level. In control wing discs without ebi RNAi, anti-ubiquitin staining was almost uniformly distributed within the wing pouch area (Fig. 4B,B″). Interestingly, when a 26S proteasome subunit was knocked down by Prosβ6 RNAi with C96-Gal4, the level of ubiquitin was consistently reduced in the DV boundary region (Fig. 4C). In contrast, the level of Crb was enhanced in the area of Prosβ6 knockdown (Fig. 4C′,C′′). These results suggest that Ebi might be involved in ubiquitin/proteasome-dependent degradation of Crb in the DV boundary region.

Next, we tested whether Crb is ubiquitylated to be targeted by the proteasome. Western blots of protein extracts from wild-type and ebi/+ heterozygous adult tissues showed similar patterns of many ubiquitin-labeled protein bands stained by an anti-Ubiquitin antibody (Fig. S3). For identification of ubiquitylated Crb, we carried out immunoprecipitation of Crb from protein extracts of wild-type and ebi heterozygous mutant adult flies. From immunoprecipitation with anti-Crb, the majority of Crb was detected as a 60 kDa band whereas the 220 kDa full length Crb was a minor form (Fig. 4D), suggesting that the 60 kDa band is a cleaved form of Crb. Interestingly, only the 60 kDa form of Crb was ubiquitylated whereas the full length Crb was not, as if ubiquitylated full-length Crb proteins were degraded. The amount of the full length Crb was increased in two different alleles of ebi/+ heterozygotes (Fig. 4D). In contrast, ebi^E4/+ heterozygous flies showed strong reduction of ubiquitylation in the 60 kDa Crb (Fig. 4E,E′), whereas the levels of α-tubulin were similar (Fig. 4D′). Similar results were found in another allele, ebi^k16213/+. Taken together with increased Crb levels in ebi mutants and Ebi-depleted wing discs, these data suggest that Ebi is required for ubiquitin-dependent downregulation of Crb.

It has been reported that the extracellular domains of Crb and Notch physically interact in mammalian cells (Ohata et al., 2011). We tested whether Crb puncta of wing disc cells overlap with Notch detected by anti-NICD (Notch intracellular domain) antibody. At the apical level of wing disc epithelium, Crb and Notch showed nearly identical patterns of membrane staining (Fig. S4A). Many NICD puncta were detected at more basal sections (Fig. S4B). Interestingly, approximately 40% of NICD puncta overlapped with Crb puncta based on ImageJ quantification. Ebi overexpression by C96-Gal4 did not significantly alter the pattern of NICD and Crb staining (Fig. S4C,D). In contrast, Ebi knockdown by C96>ebi RNAi increased the Crb level in the apical region of the DV boundary (Fig. S4E′). Numbers of both Notch and Crb puncta were also increased by Ebi knockdown (Fig. S4F,F′,G). These results suggest that loss of Ebi leads to increased internalization of NICD and Crb and accumulation of Crb in the apical cell membrane for Notch inhibition.

In this study, we have shown that Ebi promotes wing development by downregulating the Crb protein level. Crb is known to antagonize Notch signaling by interfering with γ-secretase, which is necessary for Notch cleavage (Herranz et al., 2006). Our data indicate that the level of Crb protein is regulated post-translationally by Ebi. Thus, Notch signaling at the DV boundary in the wing disc seems to be regulated by two opposing activities, that is, the activation of crb transcription by Notch and the downregulation of Crb protein by Ebi. In a model proposed in Fig. 4F, Notch is activated in the DV boundary for active growth of wing discs, which will also induce upregulation of crb mRNA and protein. When Notch signaling reaches a level that is higher than optimal, accumulated Crb protein may antagonize Notch signaling (Herranz et al., 2006; Ohata et al., 2011). However, excessive Notch inhibition by Crb can be prevented by Ebi-dependent downregulation of Crb, thus maintaining the homeostasis of growth signaling. This feedback control of the Crb protein level might help keep Notch signaling at a proper level during wing development.

Ebi and mammalian Ebi orthologs can act as corepressors/coactivator exchange factors (Tsuda et al., 2002; Perissi et al., 2008). Thus, it has been proposed that Ebi might participate in the dissociation or degradation of the corepressor(s) via the proteasome and facilitates the recruitment of coactivators for transcriptional activation of Notch target genes in wing development (Marygold et al., 2011). However, knockdown of Ebi did not significantly alter the level of crb-lacZ (Fig. 3D′), a reporter gene induced by Notch signaling (Herranz et al., 2006). Instead, our data suggest that Ebi activates Notch signaling in wing development by proteasome-dependent downregulation of Crb. This function of Ebi is consistent with the role of Ebi in degradation of Tramtrack in Drosophila (Dong et al., 1999; Boulton et al., 2000) and the mammalian Ebi function in β-catenin downregulation by ubiquitylation (Matsuzawa and Reed, 2001).

While the intracellular domain of Crb has been extensively studied, little is known about protein partners that physically interact with Crb^ext. In zebrafish, the extracellular domains of the Crb family proteins can bind to the Notch extracellular domain to inhibit Notch activation (Ohata et al., 2011). In addition, Crb^ext is known to dimerize with itself for cell adhesion between photoreceptors in zebrafish (Zou et al., 2012; Pocha and Knust, 2013). In this study, we have shown that the lam4 domain of Crb^ext directly interacts with repeats 7-8 of the WD40 domain of Ebi. The mechanism underlying this physical interaction between Ebi and the extracellular domain of Crb is currently unknown. One possibility is that Crb protein may be internalized into the cytoplasm by endocytosis (Roeth et al., 2009). In this case, Crb^ext will be located inside the endosomal vesicles, hence cytosolic Ebi still cannot access Crb^ext. Secondly, Ebi might interact with the extracellular domain of newly synthesized Crb protein prior to its targeting to the cell membrane. Thirdly, Ebi protein may be secreted from the cell and bind to Crb^ext. It remains to be determined whether Ebi uses one of these methods or another unknown mechanism to interact with the lam4 domain of Crb. It would also be interesting to see whether the Ebi-dependent downregulation of Crb plays a conserved role for Notch signaling in other organisms.

All Drosophila strains were maintained at room temperature, unless stated otherwise. Ebi knockdown was induced by crossing UAS-ebi RNAi lines with nub-Gal4 and C96-Gal4. Gal4 and RNAi lines were obtained from the Bloomington Stock Center, Vienna Drosophila Resource Center (Austria) and National Institute of Genetics (Japan). UAS-Crb^extTMGFP fly strain was provided by Elizabeth Knust (Max Planck Institute, Dresden, Germany). ebi^E4 and ebi^k16213 mutants were provided by Leo Tsuda (National Center for Geriatrics and Gerontology, Japan). Ebi overexpression was induced by crossing UAS-ebi^Exell (ebi^full) with C96-Gal4. Flp-out ebi RNAi clones were made by crossing: yw; hs-flp22; ebi RNAi with Ay≫Gal4,UAS-GFP (Kyoto Stock Center #107-724). F1 progeny was heatshocked at 38°C for 60 min at first instar larval stage. Then, wing imaginal discs from 3rd instar larvae were analyzed by immunostaining.

Matchmaker Gold yeast two-hybrid system (Clontech) was used to screen for Crb^lam4 binding proteins. Firstly, the Crb^lam4 coding sequence was cloned in pGBKT7 vector to make a fusion construct of Gal4 DNA-binding domain and Crb^lam4 in the yeast strain Y2HGold. The culture of Y2HGold containing the fusion bait construct was mixed with the adult Drosophila Mate and Plate cDNA library (Clontech), which expresses Drosophila proteins fused with the Gal4 activation domain in the yeast strain Y187. After 24 h, mated diploid cells contain four reporter genes, HIS3, ADE2, MEL1 and AUR1-C that can be activated in response to two-hybrid interaction. Approximately 10⁷ transformants were screened, and positive clones obtained were selected by the ability to activate β-galactosidase and grow in restricted medium by expression of four reporter genes under the control of Gal4-responsive promoters. The strength of interactions was determined by the growth rate and color of clones on selective plates. Finally, inserts from positive clones were sequenced and identified by BLAST search in Flybase.

Wing imaginal discs were fixed with 1% paraformaldehyde for 15 min and 100% methanol for 5 min. Fixed discs were immunostained, as described (Yeom et al., 2014). Briefly, after washing twice with PBS (pH 7.3), disc samples were blocked in 0.5% BSA. Samples were incubated with primary antibodies at the following concentrations: rabbit anti-Ebi (from Leo Tsuda, National Center for Geriatrics and Gerontology, Japan) at 1:200, rat anti-Crb^intra at 1:100, rabbit anti-Dlg (from Kyungok Cho, Korea Advanced Institute of Science and Technology, Daejeon, Korea) at 1:100, sheep anti-GFP at 1:100 (Bio-Rad, 4745-1051), mouse anti-Cut at 1:100 (DSHB, 2B10) and rabbit anti-ubiquitin at 1:50 (Abcam, ab8134). Then, secondary antibodies conjugated with Cy3 (1:600), Cy5 (1:400) or FITC (1:100) (Alexa Fluor, Molecular Probes) were used. Specificity for anti-Crb^intra antibody was tested in crb null mutant clones (Fig. S5). Fluorescent images were acquired using a Carl Zeiss LSM710 confocal microscope. The puncta colocalization was quantified by ImageJ-Fiji plugin colocalization analysis (http://fiji.sc/Cookbook). Manders’ coefficients were used for percentage puncta colocalization measurements.

A DNA fragment for the fourth laminin domain of Crb was amplified and cloned into the pMal vector. MBP-Crb^lam4 fusion protein was expressed in E.coli R2 cells by isopropyl β-D-1-thiogalactopyranoside (IPTG) induction. Purified fusion protein was used to inject into rats to raise antibody (Abfrontier, Seoul, Korea). Anti-Crb^ext was validated by immunostaining in western blots (Fig. S6).

S2 cells were grown in M3 medium (Sigma) supplemented with 10% heat-inactivated fetal bovine serum (Sigma) and 1% penicillin-streptomycin solution. For transfection, 3×10⁶ to 4×10⁶ cells were seeded in 4.5 ml medium per 50 ml flask and allowed to adhere. Then, transfection was performed using Cellfectin II reagent, according to the manufacturer's protocol (Invitrogen). A total of 1-2 μg DNA was used for each transfection. Cells were collected for analysis 36-48 h after transfection. For immunoprecipitation, cells were lysed in 0.1% CHAPS buffer, and the lysates (1 mg total protein) were precleared by incubating with Protein G-Sepharose beads (Amersham Bioscience) for 1 h at 4°C. The beads were removed after centrifugation and the clear lysates in the supernatant were kept. The beads were coupled with Flag-tag (Abcam) polyclonal antibodies at 4°C for 2 h and were incubated with the clear lysates overnight at 4°C. Samples were washed four times with IP buffer (20 mM HEPES pH 7.5, 100 mM KCl, 0.05% Triton X-100, 25 mM EDTA, 5 mM DTT, 5% glycerol and protease inhibitor cocktail). Immunoprecipitates were resuspended in 5× SDS sample buffer, heated for 5 min at 95°C, separated by SDS-PAGE and transferred to nitrocellulose membrane. Proteins were detected by a standard immunostaining protocol and enhanced chemiluminescence (ECL) (Gendepot, USA).

Crb^lam1 [amino acids (aa) 1-465], Crb^EGF (aa 466-981), Crb^lam23 (aa 982-1575), Crb^lam4 (aa 1576-1792) and Crb^intra (aa 2108-2146) were cloned into pMal (New England) to generate N-terminal MBP fusion proteins. Ebi^Full (aa 1-696), Ebi^ΔWD7/8 (aa 1-613), Ebi^Nter (aa 1-343) and Ebi^WD7/8 (aa 614-696) were cloned into pGex4T1 (Pharmacia) to generate N-terminal GST fusion proteins. Crb^lam4TM (aa 1576-2106) was cloned into pAC5.1V5 (Invitrogen) and Ebi^Full was cloned into pAC5.1Flag (Invitrogen) for expressing protein in S2 cells.

For GST pull down, IPTG-inducible E. coli R2 cells (BL21 derivative) were transformed with plasmid constructs for fusion proteins MBP-Crb^lam4, GST Ebi^Full, GST Ebi^ΔWD7/8, GST Ebi^Nter and GST Ebi^WD7/8. Bacterial cell lysates were prepared using a standard method. Equal amounts of blocked glutathione Sepharose 4B beads (Bioprogen) with GST, GST fusion protein or beads alone were incubated in pull down buffer (20 mM Tris-HCl, pH 7.5, 150 mM NaCl, 0.5 mM EDTA, 10% glycerol, 0.1% Triton X-100, 1 mM DTT and protease inhibitor cocktail), 1× sample buffer was added, beads were boiled, and proteins were resolved on SDS-PAGE. Proteins were electrophoretically transferred onto nitrocellulose membranes, blocked in 5% skimmed milk (Bio-Rad) for 1 h and incubated with primary goat anti-GST (Santa Cruz) or rabbit anti-MBP antibody (Santa Cruz). Goat anti-rabbit or anti-goat antibodies (Molecular Probes) were used as secondary antibodies. Pulled down proteins were visualized using an ECL kit (Thermo Fisher Scientific).

The following constructs and primers were used: crb^lam1, Fwd-ATGAGGATCCGCCAATGCGTCACTGTCGCAA and Rev-ATGAGTCGACTGGCAGCAGTCATATTGTGCTG; crb^EGF, Fwd-ATGAGAATTCTGCTTCCAGTCAGACTGCAAA and Rev-ATGAGTCGACCAGTAAATGTGGTGTTCGTAACAC; crb^lam23, Fwd-ATGAGTCGACAGGTAACTCATCTGTACAACTGC and Rev-ATGAGTCGACCAGTAAATGTGGTGTTCGTAACAC; crb^lam4, Fwd-ATGAGTCGACGTGTTACGAACACCACATTTACTG and Rev-ATGAGTCGACTCAATCGCCCTCGAATCCAGGCTG; crb^intra, Fwd-ATGAGAATTCATGGCCAGGAACAAGCGAGC and Rev-ATGAGTCGACCTAAATTAGTCGCTCTTCCGG; crb^lam4TM, Fwd-ATGAGTCGACGTGTTACGAACACCACATTTACTG and Rev-ATGAGAATTCGCTCGCTTGTTCCTGGCCAT; Ebi^full, Fwd-ATGAGAATTCATGAGTTTTTCCAGCGACGAGG and Rev-ATGACTCGAGTCAGAACTTTCGCAGGTCCAAC; Ebi^ΔWD7/8, Fwd-ATGAGAATTCATGAGTTTTTCCAGCGACGAGG and Rev-ATGACTCGAGTCACCACAGTCTTACCGTGGAAT; Ebi^Nter, Fwd-ATGAGAATTCATGAGTTTTTCCAGCGACGAGG and Rev-ATGACTCGAGTATTTCGATGTTCTCGTCAATC; Ebi^WD7/8, Fwd-ATGAGAATTCGACGTGGAGAGGGGCAGCTG and Rev-ATGACTCGAGTCAGAACTTTCGCAGGTCCAAC.

crb and ebi cDNA (Flybase) were used as templates for PCR amplification.

We are grateful to Elizabeth Knust for UAS-Crb^extTMGFP fly strain, Kyung-Ok Cho for Dlg antibody, Leo Tsuda for ebi mutant flies and Ebi antibody, and Macro Milan for crb-lacZ stock. We are indebted to the Bloomington Drosophila Stock Center, the Vienna Drosophila Resource Center, the National Institute of Genetics (Japan) and the Developmental Studies Hybridoma Bank for fly stocks and antibodies. We thank Kyung-Ok Cho and Sang-Chul Nam for helpful comments on the manuscript.