Loss of the extraproteasomal ubiquitin receptor Rings lost impairs ring canal growth in Drosophila oogenesis

Germline cells of many species ranging from insects to mammals develop as syncytia (Matova and Cooley, 2001). This is also the case in Drosophila melanogaster oogenesis, in which germline cells stay connected by intercellular bridges, so-called ring canals. In the germarium of the Drosophila ovary, a germline stem cell divides asymmetrically to produce a daughter stem cell and a daughter cystoblast. The cystoblast undergoes four rounds of mitosis with incomplete cytokinesis, generating a 16-cell cyst with 15 nurse cells and one oocyte (Ong and Tan, 2010). The polyploid nurse cells contribute mRNAs, proteins, and organelles to the oocyte. Actin filaments are recruited to the ring canals together with several actin-binding proteins and proteins phosphorylated at tyrosine residues (phosphotyrosine [PY]). Hu-li tai shao–ring canal (Hts-RC) and Filamin also localize to the ring canals in a strictly controlled manner. Later, ring canals are further stabilized by the actin-bundling protein Kelch (Kel; Xue and Cooley, 1993; Kelso et al., 2002). Kel interacts with the ubiquitin E3 ligase Cullin-3, and both are essential for the growth of ring canals (Hudson and Cooley, 2010).

During the growth and expansion of the ring canals, proteins involved in ring canal assembly and maturation are subject to rapid turnover. One way to increase the turnover of proteins is by their degradation by the ubiquitin–proteasome system. Ubiquitinated proteins can be directly recognized by subunits of the 26S proteasome or by so-called ubiquitin receptors, which shuttle the ubiquitinated proteins to the proteasome. Ubiquitin receptors bind to ubiquitin via their ubiquitin-associated (UBA) domain and to the proteasome via their ubiquitin-like (UBQ) domain (Schauber et al., 1998; Bertolaet et al., 2001; Grabbe and Dikic, 2009).

In this paper, we present a molecular and phenotypic analysis of the Drosophila ubiquitin receptor CG4420, which we have named Rings lost (Rngo) according to its mutant phenotype in oogenesis. Like Ddi1/Vsm1, its homologue in yeast (Liu and Xiao, 1997; Lustgarten and Gerst, 1999), Rngo possesses an N-terminal UBQ and a C-terminal UBA domain, which flank a central retroviral-like aspartate protease (RVP) domain (Krylov and Koonin, 2001). Ddi1/Vsm1 regulates the proteasomal degradation of a variety of proteins, including cell cycle regulators, homothallism endonuclease, and a member of the Skp–Cullin–F-box complex (Kaplun et al., 2005; Díaz-Martínez et al., 2006; Ivantsiv et al., 2006; Gabriely et al., 2008). In addition, Ddi1/Vsm1 controls SNARE-mediated membrane fusion events (Lustgarten and Gerst, 1999; Marash and Gerst, 2003). Ddi1/Vsm1 and its homologues are the only members of the ubiquitin receptor family that possess an RVP domain (Elsasser and Finley, 2005; Gabriely et al., 2008). The catalytic center of aspartate proteases is formed by two aspartate residues and an activated water molecule, which is needed for hydrolysis (Sirkis et al., 2006). Retroviral aspartate proteases only possess one catalytic aspartate, so the proteins have to form homodimers to build a catalytic center.

We show that Rngo forms homodimers and binds to ubiquitin and to the proteasome. rngo is an essential gene, and the loss of rngo function in female germline cells affects the growth of ring canals and leads to cell fusion. Both the lethality and the oogenesis defects of rngo mutant animals can be fully rescued by transgenes encoding full-length (FL) Rngo-GFP and versions of Rngo lacking either the UBQ or the UBA domain. In contrast, the catalytic function of the RVP domain is essential for the function of Rngo. To our knowledge, this study is the first to demonstrate a specific function for a ubiquitin receptor in Drosophila development.

The Drosophila genome contains a single homologue of Ddi1/Vsm1 encoded by the CG4420 transcription unit on the X chromosome (Fig. 1 a). Because of its mutant phenotype (see following paragraphs), we named this gene rings lost. rngo encodes an acidic protein (isoelectric point = 4.78) of 458 amino acids and a calculated molecular mass of 50,500 D. Rngo possesses an N-terminal UBQ, a central RVP, and a C-terminal UBA domain (Fig. 1 b). An alignment of the RVP domains of several eukaryotic Ddi1/Vsm1 homologues with the sequence of Rngo suggested that aspartate 257 most likely is the catalytic center of the Drosophila protein (Fig. 1, b and c).

To study the expression pattern and subcellular localization of Rngo, we raised an antibody against an FL GST-tagged Rngo protein. Rngo was ubiquitously expressed during embryogenesis and in larval and adult tissues (Fig. S2 f and not depicted). Because germline clones of the rngo^GE3956-null allele (see following paragraphs) showed severe defects during oogenesis, we focused our analysis on the subcellular localization of Rngo in ovaries. Rngo was ubiquitously expressed both in the somatic follicle cells and in the germline at all stages of development (Fig. 2, a–d; and Fig. S1). Rngo was uniformly localized in the cytoplasm, whereas in the nucleus, it was enriched in subnuclear structures. This was particularly obvious in the polyploid nurse cell nuclei and in the oocyte nucleus (Fig. 2, a and b). Because Rngo is a predicted proteasomal shuttle factor, we tested whether Rngo colocalized with ubiquitin and with the proteasome by performing double stainings against Rngo and ubiquitin (Fig. 2 c) and the proteasome (Fig. 2 d). These stainings showed that Rngo colocalized with ubiquitin and with the proteasome in female germline cells.

The specificity of the anti-Rngo antibody was tested in stainings of ovaries containing germline clones of the rngo^GE3956 allele (see following paragraphs; Fig. 2 e). The specificity of the anti-Rngo antibody was also tested in Western blots using protein extracts of rngo^GE3956 mutant L3 larvae in comparison with L3 larvae possessing the wild-type rngo allele (Fig. 2 f). In both stainings and Western blots, the signal obtained with the anti-Rngo antibody was completely abolished in tissues or protein extracts of rngo^GE3956 mutant animals, demonstrating that this antibody is highly specific for Rngo.

Next, we asked whether Rngo physically associates with ubiquitin and with the proteasome. To test the binding to ubiquitin, we generated a series of FL and partially deleted Rngo fusion proteins N-terminally tagged with the maltose-binding protein (MBP) for expression in Escherichia coli (Fig. 3 a). The Rngo fusion proteins were incubated with agarose beads covalently attached to ubiquitin. Only variants of Rngo that possessed the UBA domain bound to ubiquitin (Fig. 3 b) as expected.

Because Ddi1/Vsm1 binds to Rpn10, a regulatory subunit of the 26S proteasome (Saeki et al., 2002), we tested whether Rngo can also bind to Rpn10. To that aim, we incubated GST–FL Rngo together with MBP-Rpn10 and performed a pull-down with glutathione agarose beads. We detected MBP-Rpn10 in the precipitate with GST–FL Rngo but not in the control (Fig. 3 c), demonstrating that Rngo forms a complex with Rpn10 in vitro. The binding of Rngo to ubiquitin and to Rpn10 was also confirmed in vivo by coimmunoprecipitation (IP; co-IP; Fig. 3 d).

It was previously found that proteins containing an RVP domain can form homodimers to establish a catalytic center around two conserved aspartate residues (Miller et al., 1989). This is also true for Ddi1/Vsm1 (Sirkis et al., 2006; Gabriely et al., 2008). Thus, we used a series of Rngo constructs (Fig. 3 a) to determine the structural requirements for homodimerization by co-IP. We coexpressed FL Rngo tagged at the C terminus with GFP and either FL Rngo or one of the mutated versions, RngoΔUBA, RngoΔUBQ, RngoΔRVP, or RngoD257A, tagged at the C terminus with the HA epitope in S2 cells and performed co-IP with an anti-GFP antibody. We found that only RngoΔRVP, which lacks the RVP domain, is unable to bind FL Rngo-GFP (Fig. 3 e). The D257A substitution, which is predicted to abolish the catalytic activity of the RVP domain, did not interfere with dimerization (Fig. 3 e).

To analyze the function of rngo in development, we used the transposon insertion line rngo^GE3956, in which the EP element (Rørth, 1996) is inserted 203 bp downstream of the ATG start codon in the first exon of rngo (Fig. 1 a). The transposon insertion in rngo^GE3956 is predicted to result in the premature termination of translation, and according to our immunohistochemical and Western blot data, it represents a null allele of rngo (Fig. 2, e and f). Animals homo- or hemizygous for rngo^GE3956 were lethal at pupal stages. This late lethality is probably caused by maternal rngo expression that rescues an earlier requirement for Rngo because we detected Rngo in rngo^GE3956 mutant embryos (not depicted) but not in L3 larvae of the same genotype (Fig. 2 f). The lethality of rngo^GE3956 mutant animals and the ovarian phenotype of rngo^GE3956 germline clones (see following paragraphs) were fully rescued by ubiquitous expression of FL Rngo-GFP (Fig. S2, a and d).

To investigate the requirement for rngo during female germline development, we generated germline clones of the mutant rngo^GE3956 allele using the Flipase–Flipase recombination target technique (Harrison and Perrimon, 1993). Although control clones showed the wild-type staining pattern of Rngo, clones mutant for rngo^GE3956 completely lacked Rngo staining in the germline (Fig. 2 e and Fig. S1). We noted that egg chambers with rngo^GE3956 mutant germline cells did not develop beyond stages 8 and 9 of oogenesis because of degeneration and apoptosis of germline cells (Fig. S1 i). Closer analysis of the development of mutant egg chambers revealed severe abnormalities with respect to the localization of ring canal–associated proteins and the shape of the germline cells. In wild-type egg chambers, Hts-RC, Kel, Cheerio (Cher; Drosophila Filamin), and PY epitopes exclusively localized to the ring canals in the germline (Fig. 4, a, d, and f; Video 1; and Video 3). In contrast, Hts-RC, Cher, and PY epitopes spread over the whole plasma membrane of the germline cells in rngo^GE3956 mutant germline clones (Fig. 4, b, c, e, and g; Video 2; and Video 3). Kel still localized to remnants of ring canals and to the plasma membrane of germline cells but was also found in the cytoplasm (Fig. 4 e). The defects in ring canal morphogenesis ranged from an almost complete loss of ring canals (Fig. 4 c) to the formation of small, rudimentary ring canals connecting two nurse cells (Fig. 4, e and g) to the formation of unusually large ring canals connecting a nurse cell with the oocyte (Fig. 4 b). The quantification of these defects revealed that the growth of ring canals connecting two nurse cells was arrested at stage 5 in rngo^GE3956 germline clones (Fig. 4 h), whereas ring canals connecting a nurse cell with the oocyte were always larger than their wild-type counterparts and continued to grow until stage 8 (Fig. 4 i). To our knowledge, rngo is the first mutation that differentially affects the growth of ring canals between nurse cells and between nurse cells and the oocyte. Ring canal defects and the mislocalization of ring canal–associated proteins were already detectable at early stages of oogenesis (Fig. S1), demonstrating that Rngo function is required throughout oogenesis.

We wondered whether the mislocalization of ring canal–associated proteins in rngo^GE3956 germline clones was accompanied by a change in their expression levels. To address this question, we performed Western blotting of ovaries dissected from wild-type females and from females with egg chambers carrying rngo^GE3956 germline clones. Concomitant with a reduction in Rngo protein level, the levels of Hts-RC, Kel, Cher, and PY epitopes were significantly reduced, in contrast to the tubulin control (Fig. 5 a). This finding was confirmed by confocal analysis of control and rngo^GE3956 germline clone egg chambers that were positioned next to each other in the same ovariole. Although the levels of Hts-RC, Kel, Cher, and PY epitopes were similar in adjacent wild-type and control germline clone egg chambers (Fig. 5, b, d, and f), they were strongly reduced in rngo^GE3956 germline clone egg chambers (Fig. 5, c, e, and g).

In addition to defects in ring canal morphogenesis, we found that the oocyte nucleus frequently was misshapen in rngo^GE3956 mutant germline clones. Instead of being round and compact as in wild type (Fig. 4 a, inset), the nucleus in rngo^GE3956 mutant oocytes was enlarged and elongated (Fig. 4 c, inset), pointing to abnormal packing of the chromatin. This phenotype was fully penetrant after stage 7 of oogenesis.

Compared with wild-type controls, rngo^GE3956 mutant egg chambers showed an abnormal organization of cortical F-actin. Although wild-type egg chambers always contained 15 nurse cells and one oocyte that were separated from each other by plasma membranes with an associated cortical actin filament network, egg chambers with rngo^GE3956 germline clones frequently had <16 germline cells based on the counting of cell outlines (Fig. 4 c), whereas the number of nuclei was unaffected. Consequently, germline cells with more than one nucleus were frequently observed in rngo^GE3956 germline clones. This was particularly obvious for the oocyte, which in addition to the oocyte nucleus, always contained one or more nurse cell nuclei at later stages of oogenesis (Fig. 4, c and j). This phenotype most likely results from the fusion of germline cells, which may be caused by decreased stability of the cortical actin cytoskeleton. Loss of rngo function in the follicle epithelium did not cause any obvious abnormalities in actin organization (Fig. S2 e), pointing to a germline-specific function of rngo in oogenesis.

To test which domains of Rngo are required for its function in vivo, we performed rescue experiments using the upstream activating sequence (UAS)–GAL4 system (Brand and Perrimon, 1993). To that aim, we ubiquitously expressed FL Rngo-GFP and different GFP-tagged mutant versions of Rngo in animals mutant for rngo^GE3956. Although RngoΔUBQ-GFP and RngoΔUBA-GFP (Fig. 3 a) rescued both the lethality and the oogenesis defects of rngo^GE3956 mutant animals to the same extent as FL Rngo-GFP (Fig. S2, a–d), neither RngoΔRVP-GFP (Fig. 3 a) nor RngoD257A-GFP, a point mutation that is predicted to destroy the catalytic activity of the RVP domain, showed any rescue. These results demonstrate that binding to ubiquitin and binding to the Rpn10 subunit of the proteasome are dispensable for the function of Rngo in our rescue assay, whereas the catalytic activity of the RVP domain is essential.

In this study, we characterize Rngo, the Drosophila homologue of the ubiquitin receptor Ddi1/Vsm1 from yeast. Rngo shares many features with Ddi1/Vsm1, including the binding to ubiquitin and to the proteasome and its ability to form homodimers (Bertolaet et al., 2001; Sirkis et al., 2006; Gabriely et al., 2008). In contrast to Ddi1/Vsm1, which is dispensable for the growth and survival of yeast (Lustgarten and Gerst, 1999; Clarke et al., 2001), rngo is an essential gene and is required for proper development of female germline cells. The defects observed in rngo mutants were less severe than those of hts mutations but were stronger than mutations in kel and cher (Yue and Spradling, 1992; Robinson et al., 1994; Li et al., 1999; Sokol and Cooley, 1999; Hudson and Cooley, 2010). Egg chambers with rngo germline clones showed mislocalization of Hts-RC, Cher, Kel, and PY epitopes and, unlike the aforementioned mutations, never developed beyond stage 9 of oogenesis and showed fusion of germline cells. These phenotypic differences between rngo and other genes involved in ring canal morphogenesis point to additional functions of Rngo, possibly in membrane trafficking.

How could these phenotypes relate to the biochemical functions assigned to Rngo? The predicted function of Rngo as a ubiquitin receptor implies that it binds to ubiquitinated proteins and either shuttles them to the proteasome or prevents their proteasomal degradation by interfering with the recognition of the ubiquitin modification by a proteasomal ubiquitin receptor. This may be important for the proper assembly of ring canals because their growth requires the dynamic remodeling of the actin cytoskeleton and a high turnover of actin-binding proteins. Evidence supporting this hypothesis comes from a recent study that demonstrated a function for Kel as a component of an E3 ubiquitin ligase complex together with Cullin-3 (Hudson and Cooley, 2010). This study proposed that Kel is required as an adapter protein to ubiquitinate various ring canal–associated proteins, which need to be degraded to disassemble the actin cytoskeleton of the growing ring canal. Indeed, the staining intensity for F-actin, Cher, and Hts-RC was strongly increased in kel mutant egg chambers (Hudson and Cooley, 2010). We have now shown that the levels of Kel, Cher, and Hts-RC are strongly reduced in rngo mutant egg chambers. Thus, Rngo may antagonize the activity of the Kel–Cullin-3 E3 ligase complex. Our genetic rescue experiments demonstrated that the catalytic activity of the RVP domain of Rngo is essential for its function, in contrast to the UBA and UBQ domains. It is therefore tempting to speculate that the protease activity of the RVP domain may be directly involved in the deubiquitination of specific ring canal proteins, thus protecting them from degradation in the proteasome. To test this hypothesis, it will be important to identify the protease targets of Rngo and to elucidate whether these are indeed ubiquitinated by the Kel–Cullin-3 E3 ligase complex.

The following stocks were used in this study: P{EP}CG4420^GE3956/FM7, renamed rngo^GE3956 in this study (GenExel), y¹arm⁴ w*/FM7c, ovo[D2] P{FRT(w^hs)}9-2/Y; P{hsFLP} (stock no. 1,843), p{ftz/lacZ}YH1 (stock no. 616), P{FRT(w^hs)}9-2 (stock no. 5,749), P{Ubi-GFP.nls}P{FRT(w^hs)}9-2 (stock no. 5,832), P{FRT(w^hs)}9-2; P{hsFLP}38 (stock no. 1,843), daughterless-GAL4 (stock no. 5,460), and y¹ w* N¹/FM7c, P{GAL4-twi.G}108.4, P{UAS-2xEGFP}AX (stock no. 6,873; Bloomington Drosophila stock center). Expression of UASP-Rngo-GFP, UASP-RngoΔUBA-GFP, UASP-RngoΔUBQ-GFP, UASP-RngoΔRVP-GFP, and UASP-RngoD257A-GFP in transgenic flies was performed with the UAS-GAL4 system (Brand and Perrimon, 1993). Germline clones and follicle cell clones for rngo^GE3956 were generated as previously described using a heat shock promoter-driven flipase on the second chromosome (Chou and Perrimon, 1996). To obtain hemizygous rngo^GE3956 mutant L3 larvae, living embryos lacking GFP fluorescence derived from the FM7[twi::GFP] balancer chromosome were separated from their GFP-positive siblings under a fluorescent stereo microscope and raised in separate vials. To rescue the lethality and phenotype of hemi- or homozygous rngo^GE3956 mutant flies, rngo^GE3956/FM7 females were crossed to males that were homozygous for one of the UASP-Rngo-GFP constructs recombined with the daughterless-GAL4 driver on the third chromosome and carried the FM7 balancer. Rescued hemizygous males carrying the rngo^GE3956 allele and the GFP construct were crossed to rngo^GE3956/FM7 females also carrying the GFP construct to establish a viable and fertile stock homozygous for the mutant rngo^GE3956 allele.

An antibody against Rngo was generated by immunizing a rabbit with a GST fusion protein containing the FL Rngo protein (Eurogentec). Final bleed serum was affinity purified and used for immunohistochemistry. For immunohistochemical stainings of ovaries the following antibodies were used: rabbit anti-Rngo, affinity purified at 1:1,000, mouse anti-26S proteasome (Rpn10) at 1:400 (Santa Cruz Biotechnology, Inc.), mouse anti-PY at 1:1,000 (Sigma-Aldrich), rabbit anti-GFP at 1:1,000 (Invitrogen), mouse antiubiquitin at 1:6,000 (FK2; Enzo Life Sciences), rat anti–N-filamin at 1:3,000 (Sokol and Cooley, 1999), provided by L. Cooley, Yale University, New Haven, CT, mouse anti–Hts-RC at 1:20 (Hts-RC; Developmental Studies Hybridoma Bank), mouse anti-Kel at 1:1 (kel1B; Developmental Studies Hybridoma Bank), and mouse anti-Orb at 1:20 (4H8; Developmental Studies Hybridoma Bank). Secondary antibodies conjugated to Cy2, Cy3 (Jackson Immunoresearch Laboratories, Inc.), and Alexa Fluor 647 (Invitrogen) were used at 1:200. To visualize F-actin, ovaries were incubated with 2 U rhodamine-conjugated phalloidin in PBS with 0.1% Triton X-100. DNA was stained with DAPI. Ovaries were fixed in 4% formaldehyde/PBS and stained according to standard procedures. Images were taken on a confocal microscope (LSM 510 Meta; Carl Zeiss) using 25× Plan-Neofluar, NA 0.8, and 63× Plan-Apochromat, NA 1.40, objectives and processed using Photoshop (Adobe).

Lysates of L3 larvae, ovaries, and S2 cells were prepared in TNT buffer (150 mM NaCl, 50 mM Tris-HCl, pH 8.0, and 1% Triton X-100) supplemented with protease inhibitors (Roche). Western blotting and co-IPs were performed as previously described (Wodarz, 2008). In brief, for Western blotting, cell lysates were subjected to SDS-PAGE, blotted onto nitrocellulose membranes, and blocked in 3% dry milk and 1% BSA in TBS with Tween (TBS-T). The following primary antibodies were incubated with the nitrocellulose filters in blocking buffer: rabbit anti-Rngo, affinity purified at 1:2,000, mouse anti–α-tubulin at 1:25 (12G10; Developmental Studies Hybridoma Bank), mouse anti-HA at 1:1,000 (12C25; Roche), mouse anti-GFP at 1:1,000 (Roche), mouse anti-MBP at 1:10,000 (New England Biolabs, Inc.), rabbit anti-GST at 1:10,000 (Sigma-Aldrich), mouse anti-PY at 1:1,000, mouse antiubiquitin at 1:1,000, mouse anti-26S proteasome at 1:500, rat anti–N-filamin at 1:1,000 (Sokol and Cooley, 1999), mouse anti–Hts-RC at 1:5, and mouse anti-Kel at 1:2. After repeated washing in TBS-T, blots were incubated with HRP-conjugated secondary antibodies in blocking buffer and washed again with TBS-T. Protein bands were visualized with BM chemiluminescence blotting substrate peroxidase (Roche). For co-IP, rabbit anti-GFP, mouse anti-26S proteasome, and mouse antiubiquitin antibodies were incubated with cell lysates followed by pull-down of the antibody–protein complexes with protein A/G–agarose beads. Rabbit anti-GST and mouse anti-MBP antibodies were used as negative controls. Washed beads were boiled in sample buffer and subjected to SDS-PAGE and Western blotting.

For GST pull-down the FL rngo sequence was subcloned into pGEX-4T-1 (GE Healthcare), and the FL sequence of Rpn10 was subcloned into pMGWA (Drosophila Genomics Resource Center). Each GST or MBP fusion protein expressed in E. coli was purified with glutathione Sepharose (GE Healthcare) or amylose resin (New England Biolabs, Inc.), respectively. GST and GST-Rngo were bound to 15 µl glutathione Sepharose resin in binding buffer (100 mM NaCl, 50 mM Tris-HCl, pH 7.5, and 0.5% Triton X-100 supplemented with protease inhibitors). The beads were subsequently incubated with 200 µl binding buffer plus MBP-Rpn10 for 2 h at 4°C and were then washed with binding buffer and eluted by adding 15 µl of 2× SDS sample buffer and boiling for 10 min at 96°C. The supernatants were separated by SDS-PAGE and analyzed by Western blotting.

For ubiquitin pull-down, the FL rngo sequence and the sequences of RngoΔUBA, RngoΔUBQ, RngoΔRVP, RngoD257A, RngoΔUBAΔUBQ, RngoΔUBAΔRVP, and RngoΔUBQΔRVP were subcloned into pMGWA. MBP fusion proteins expressed in E. coli were purified with amylose resin. MBP fusion proteins were diluted in 500 µl binding buffer (100 mM NaCl, 50 mM Tris-HCl, pH 7.5, and 0.5% Triton X-100), added to 15 µl ubiquitin–agarose (Enzo Life Sciences) and incubated for 2 h at 4°C. The beads were then washed with binding buffer and eluted by adding 15 µl of 2× SDS sample buffer and boiling for 10 min at 96°C. The supernatants were separated by SDS-PAGE and analyzed by Western blotting.

Fig. S1 is related to Fig. 2 and Fig. 4 and shows the requirement for Rngo already at the earliest stages of oogenesis and also the degeneration of rngo mutant egg chambers after stages 8 and 9 of oogenesis. Fig. S2 is related to Fig. 4 and shows stainings and Western blots of rngo mutant animals rescued by different Rngo-GFP transgenes. Video 1 is related to Fig. 4 and shows the 3D reconstruction of the wild-type staining pattern of Hts-RC in an egg chamber at stage 8. Video 2 is related to Fig. 4 and shows the 3D reconstruction of the staining pattern of Hts-RC in an rngo mutant egg chamber at stage 8. Video 3 is related to Fig. 4 and shows the 3D reconstruction of the staining pattern of Hts-RC in two adjacent egg chambers, one at stage 6 that is wild type for rngo and the other one at stage 8 that is mutant for rngo.

We thank Lynn Cooley, the Bloomington Drosophila stock center at the University of Indiana, and the Developmental Studies Hybridoma Bank at the University of Iowa for sending fly stocks and antibodies. We also thank Katja Brechtel-Curth for technical assistance and Ernst A. Wimmer and members of the Wodarz laboratory for discussions.

This work was supported by grants from the Deutsche Forschungsgemeinschaft to A. Wodarz (Sonderforschungsbereich 523, Teilprojekt B15, and Forschungszentrum für Molekularphysiologie des Gehirns).