Dystroglycan is required for polarizing the epithelial cells and the oocyte in Drosophila

Muscular dystrophies are genetic disorders that are characterized by progressive muscle degeneration. These diseases are caused by mutations in different members of the Dystrophin-associated glycoprotein complex (DGC),which is composed of multiple cytocortical, transmembrane and extracellular proteins (Burton and Davies,2002; Henry and Campbell,1999; Winder,2001). How these mutations cause the observed muscle defects is not fully understood and no cures for the diseases exist. In addition to muscle cells, the DGC is required in other cell types such as epithelial cells and neural cells (Durbeej et al.,1995; Durbeej and Campbell,1999; Williamson et al.,1997; Michele et al.,2002; Moore et al.,2002). Reduced expression of the DGC components is observed in breast and prostate cancers (Henry et al.,2001a). Dystroglycan (DG), a transmembrane protein, is a central player of the DGC. It acts as a receptor for the extracellular matrix (ECM)component Laminin (Ibraghimov-Beskrovnaya et al., 1992), and connects to the actin cytoskeleton through an actin-binding protein, Dystrophin. The cellular function and regulation of these interactions remain elusive, however. Drosophila melanogasteris an excellent model organism with which to study basic cellular functions of evolutionarily conserved genes, particularly human disease genes(Bernards and Hariharan, 2001). We have analyzed the DG homolog in Drosophila and shown that it is required for the establishment of cellular polarity.

How polarity is established at a cellular level is one of the most fundamental questions in biology. Many cell types undergo certain degrees of polarization to fulfill their specific functions. For example, neurons polarize to form axons and dendrites in order to convey signals; polarization of T cells is needed for their migration. Epithelial cells, however, have a pronounced apicobasal polarity, which is needed for them to cope with different extracellular environments. Studies using genetic model systems such as Drosophila and mammalian culture cells have revealed three groups of protein complexes that are involved in the specification and regionalization of the plasma membrane and cortex of the polarized epithelium:Crumbs, Par and Lgl complexes (Tepass et al., 2001). The transmembrane protein, Crumbs (Crb) and its cytoplasmic-binding partners, the PDZ domain proteins Discs Lost (Dlt) and Stardust as well as the Par-complex [Bazooka(Par3)/DmPar6(Par6)/atypical protein kinase C (aPKC)] are located on the apical membrane and are required for the establishment of this domain(Bachmann et al., 2001;Bhat et al., 1999;Hong et al., 2001;Petronczki and Knoblich, 2001;Tepass et al., 1990;Wodarz et al., 2000). The Lgl-complex [Lethal Giant Larvae (Lgl)/Discs Large (Dlg)/Scribble (Scrib)] is located at the lateral region of the epithelium and is required to restrict Crb to the apical side (Bilder and Perrimon, 2000; Bilder et al.,2000; Woods and Bryant,1991).

Cell polarity can also be the basis for a body axis. InDrosophila, the body polarity is built upon the polarity of the oocyte, and in C. elegans, polarization of the single-cell embryo determines the anteroposterior (AP) body axis (reviewed byWodarz, 2002). In fact, the process of polarity formation in the developing Drosophila oocyte provides an excellent model with which to study how the polarity of the cytoskeleton is dynamically regulated. The AP asymmetry of the oocyte cytoskeleton, which is the basis for morphogen localization within different compartments of the egg, is established by a series of dynamic steps (reviewed by Riechmann and Ephrussi,2001). First, centrioles and the microtubule organizing center(MTOC) are located at the anterior end of the oocyte at stage 1. By stage 3 the MTOC has moved to the posterior of the oocyte. This posterior movement of the MTOC requires function of the Par proteins (Par-complex and Par1) and the action of maelstrom gene product(Cox et al., 2001a;Cox et al., 2001b;Huynh et al., 2001a;Huynh et al., 2001b;Clegg et al., 2001). At stage 6, posterior follicle cells send an unidentified signal back to the oocyte to re-orient the oocyte microtubule (MT) polarity, which requires the function of an ECM protein, Laminin (Deng and Ruohola-Baker, 2000). At each step, the proper MT polarity is required for localization of key molecules in the oocyte. In addition to MTs,the oocyte has an enriched cortical array of actin cytoskeleton that plays an important role in localizing posterior morphogens in the oocyte(Baum et al., 2000;Erdelyi et al., 1995).

Although the oocyte and the epithelial cells differ profoundly in their morphology and function, polarization of these two cell types uses some of the same genes. For example, the Par genes are required to establish the polarity of both the oocyte and epithelial cells(Cox et al., 2001a;Cox et al., 2001b;Huynh et al., 2001a;Huynh et al., 2001b;Petronczki and Knoblich, 2001;Wodarz et al., 2000). This similarly raises the possibility that some common strategies may exist for cellular polarization. We show that DG, a receptor for multiple ECM proteins,is required cell-autonomously to polarize both the epithelial cells and the oocyte in Drosophila. We also show a separate, non-cell-autonomous function for DG: disruption of DG affects the organization of the basal actin cytoskeleton in neighboring cells, which suggests the involvement of DG in cell-cell communication.

Drosophila Dystroglycan (Dg) gene was identified by EP(Rørth, 1996) screen and independently by homology to mouse DG using the Protein Blast program. LD04782 that maps to the genomic region identified by this search contains the 3′ half of the gene. We therefore isolated several cDNAs from the 5′ region by PCR using an embryonic cDNA library. The 5′ end ofDg was mapped near the insertion site of EP(2)2241 and ca.4 kb downstream of Rho 1 (Fig. 1A). Several EST 5′ sequences overlap the 5′ UTR ofDg. We analyzed the following ESTs: LD11619, GH09323 and SD06707. We fully sequenced LD11619 and partially sequenced three of the cDNAs isolated and found the intron-exon structure to be different from the Gadfly prediction in two incidents (Gadfly annotation has been corrected). We further noticed that at least two exons [exon 8 (265 amino acids) and 9 (83 amino acids)] are subjected to alternative splicing (Fig. 1A). We tested five different cDNAs for presence of exon 8 and 9. Two lacked exon 9, among them LD11619 (1179 amino acids). Two lacked both exon 8 and exon 9, among them SD06707 (914 amino acids). One lacked exon 8(GH09323, 997 amino acids). None of the cDNAs tested contained both exon 8 and exon 9 (Fig. 1A shows the exon/intron structure and the alternatively spliced exons).

Drosophila melanogaster stocks were raised on standard cornmeal-yeast-agar medium at 25°C. Dg allelesDg²⁴⁸, Dg⁶² andDg³²³ were obtained by imprecise excision of EP(2)2241(Fig. 1A). EP(2)2241 was mobilized by crossing the line to a Δ2-3 transposase containing line. 330 White-eyed flies of the F1 generation were established as balanced stocks,and the homozygous lethal (12%) or semi-lethal lines (5%) were tested for complementation with a CG8414-allele (EP(2)0525), a deficiency lineDf(2)JP6 and a Rho1 allele (Rho1^E3.10)(Halsell et al., 2000). All three alleles contain deletions that remove the putative transcription start site and the 5′UTR of the Dg gene(Fig. 1A). The breakpoints of the deletion mutants were mapped by PCR using combinations of the following genomic primers: forward, GATCAGGGCCAAGGTGTGTCCAGC and AAGCCGCTTTGGCGTTGC;reverse, GCTCACTCCCACACAAGCGC and GAGCCCAATGATCCGTGGAAAGCG.

PCR fragments including the breakpoints of Dg²⁴⁸ andDg³²³ were sequenced. Dg²⁴⁸ contains a 785 bp deletion between bp 32,514 and bp 33,299 of DS03910. Fifteen basepairs of the inverted repeat of the P element are still present (lowercase inFig. 1). An A to Tmutation and a 2 bp deletion is found at the distal breakpoint:GGAGCATTCCTTGCT--ATGTTatgttatttcatcatgGGCAGGAGAGTCCCGAAT.

Dg³²³ contains a 3155 bp deletion between bp 32,345 and bp 35,669 of DS03910. A C to T change was found near the proximal breakpoint: AAAATGGCAGCGTACTTTCG/TTTTGCTTTGCGCTTCTCTG.

The cDNA corresponding to the cytoplasmic domain and 670 bp of the 3′UTR of Drosophila Dystroglycan (CTGTTGCCTGCA to TTGCTTGCATGTTTTTTTTTTTT) was directionally cloned into theKpnI/HindIII sites in pBluescript II (Stratagene) to form an intermediate vector pKS-dg. The 1 kb KpnI/BamHI fragment of pKS-dg was excised and subcloned into pEGFP-N1 (Clontech) and then digested with NheI and BamHI. The NheI/BamHI fragment was inserted together with the 148 bp Sau3A fragment of pEGFP-N1 back into the BamHI/SpeI-digested pKS-dg (triple ligation). TheKpnI 2.2 kb fragment was subcloned into pUAST. Thus, we constructed aDg hairpin-loop plasmid (pUAST-dg-L-gd). The construct was verified by sequencing and then injected to embryos to obtain stable transformant lines, UAS-dsDG-RNAi (dsDG). To drive expression ofdsDG in animals, we crossed the transgenic flies with thetubP-Gal4 line (Lee and Luo,1999), which shows ubiquitous expression in follicle cells.tubP-Gal4/dsDG causes reduction of DG expression, leading to semilethality, consistent with the fact that dg deletion alleles are homozygous lethal. However, some tubP-Gal4/dsDG escapers were observed and analyzed for their oogenesis phenotypes.

In order to generate mutant cell clones, Dg alleles were recombined to the FRT chromosome (Xu and Rubin, 1993). To obtain follicle cell clones, 1- to 5-day-old flies were heat-shocked as adults for 60 minutes at 37°C and put in freshly yeasted vials with males for 2 or 3 days. To obtain germline clones,2- to 3-day-old larvae were heat-shocked at 37°C for 2 hours each time during 2 consecutive days. Ovaries from adult female flies at 3-5 days of age were harvested. To generate mutant clones in imaginal discs, the flies were allowed to lay eggs for 24 hours at 25°C and the eggs were allowed to develop 48 hours at 25°C. Thereafter, the larvae were heat shocked for 30-40 minutes at 38°C and returned to 25°C. After 2 days, wandering third instar larvae were collected and dissected for antibody staining.

Three antibodies against DG protein were raised in rabbits: one against the extracellular domain corresponding to exon 8 (amino acids 243-507) and two against the intracellular regions (one against 18 C-terminal amino acids and one against 102 C-terminal amino acids). Most of the staining shown in this paper use the antibody raised against 102 C-terminal amino acids. DG cytoplasmic domain (102 amino acids) was synthesized by PCR using the primer pair: CGGGATCCAAAGGAGCGGCAAAATGGAG and GCTCTAGAAAGCGGCCGCCGTACGTCCCAGTAAGT(Gibco BRL) and the template GH09323. The PCR product was digested usingBamHI and NotI and the fragment was cloned in vector pGEX5X. The DGcyto-GST fusion protein was produced in JM109 after induction with IPTG. Then bacteria were harvested and lyzed by French Press. To purify the fusion protein, GST resin was used to bind the protein, and 10 mM gluthione 50 mM Tris (pH 7.5) was used to elude the protein. Polyclonal antisera were produced by R & R rabbitury, and affinity purified by the fusion protein.

UAS-DG was constructed by cloning the KpnI/NotI insert of an EST clone LD11619 into the pUAST transformation vector. UAS-DGcyto contains a tandem Flag tag sequence inserted at amino acid 37 of DG (amino acids 1-27 constitute the putative signal peptide) fused to the transmembrane (starting WPIVI...) and cytoplasmic domains. The constructs were injected to embryos to obtain stable transformant lines. The UAS-DG and UAS-DGcyto fly-lines were then crossed to different Gal4 driver lines.

To generate DG overexpression follicle cell clones, hsFLP, UAS-Dgmales were crossed to virgin female act<FRT-CD2-FRT<Gal4;UASGFP flies (Pignoni and Zipursky,1997). The F₁ progeny were heat shocked at 37°C for 1 hour and raised at 25°C for 3 days, dissected and analyzed.

Ovarian antibody staining and confocal microscopy was performed as described previously (Deng et al.,2001). Basal actin staining was performed according to the protocol provided by Frydman (Frydman and Spradling, 2001). Imaginal disc staining was as described previously (Woods et al.,1997). A two-photon laser scanning microscope (Leica TCS SP/MP)was used to detect DAPI staining.

The following antibodies were used: rabbit anti-Dlg (1:500)(Woods and Bryant, 1991);mouse anti-Orb (1:20) (Lantz et al.,1994); rabbit anti Baz (1:500)(Wodarz et al., 2000); mouse anti-Crb (CQ4,1:20) (Tepass et al.,1990); rabbit anti DG (1:3000; this study); mouse or rabbit anti-βGal (1:5000, Sigma); mouse anti-Neurotactin (BP102; 1:20)(Hortsch et al., 1990); rabbit anti-Laminin (1:3000) (Fessler et al.,1987); ant-Dlt (1:1000) (Bhat et al., 1999); anti-β-HSpectrin (1:1000)(Thomas and Kiehart, 1994);Alexa 488, 568 or 633 goat anti-mouse (Molecular Probes); Alexa 568 or 633 Goat anti-Rabbit or Rat (Molecular Probes); and Alexa 568 or 633 Phalloidin(Molecular Probes). DAB staining was done with the Vectastain Kit (Vector Laboratories, CA).

To identify genes that affect the polarity of the Drosophilaoocyte, we used the EP/Gal4 system(Rørth, 1996) to screen for genes that, when overexpressed in follicle cells, cause a polarity defect in the underlying oocyte. From the over 2000 EP insertions screened, two components of the mammalian DGC were identified: Laminin A (Deng and Ruohola-Baker, 2001) and Drosophila DG (EP(2)2241;Fig. 1A). An independent homology search with mouse DG protein verified the identification of CG18250 as the Drosophila homolog of the mammalian Dystroglycangene.

Conceptual translation of the longest cDNA (LD11619) reveals an open reading frame of 1179 amino acids (Fig. 1A). This deduced Drosophila DG protein contains all the hallmarks of vertebrate DG: a mucin-like domain, a transmembrane domain and a C-terminal region with WW-, SH2- and SH3-binding domains(Fig. 1B,D). The best conserved region between human and Drosophila is the C-terminal half of the protein showing 31% identity (Fig. 1B). The last 12 amino acids of the C terminus include the WW domain-binding motif (PPxY), which is the Dystrophin binding site. Of 12 amino acids within the C terminus, 10 are perfectly conserved in Drosophila(Fig. 1C). Vertebrate DG contains a second PPxY motif in its cytoplasmic domain, which is also conserved in Drosophila. In addition, two of the six putative SH3 binding sites and all three SH2-binding sites in the cytoplasmic domain of vertebrate DG can be found in Drosophila(Fig. 1D). The putative C. elegans homolog DGN-1 (T21B6.1) shows 20% identity to Drosophilain the C-terminal half. However, T21B6.1 contains no mucin-like domain,Dystrophin-binding site or second PPxY motif(Fig. 1D).

To analyze the expression pattern of DG protein, we raised antibodies against the cytoplasmic domain. Five major bands can be detected on a western blot of wild-type embryonic extracts: 75 kDa, 105 kDa, 120 kDa, 180 kDa and 200 kDa (Fig. 1E). None of these major bands could be seen in the extracts from the deficiency[Df(2R)JP4, Df(2R)JP6) embryos that completely delete the Dglocus, suggesting that all five forms are specific for DG(Fig. 1E)]. Strong Dgmutants were isolated by imprecise excisions of EP(2)2241 element and by generating a transgenic line expressing a double-stranded DG-RNA construct that destroys DG RNA by RNAi-mechanism(Kennerdell and Carthew, 2000)(Fig. 1A). InDg²⁴⁸ or Dg³²³ mutant embryos, of the five major bands derived from the Dg locus only the 105 kDa band can be detected weakly (Fig. 1E),indicating that the level of DG expression is highly reduced in these mutants. Furthermore, to test the specificity of the antibodies in tissue samples we analyzed the expression in the follicle cell epithelium. A high level of DG is observed on the basal side of the epithelium, while a lower level is detected on the apical side. This signal is absent in follicle cell clones homozygous for Dg²⁴⁸ or Dg³²³, suggesting that the signal observed with the antibody in the tissue is specific for DG (arrow in Fig. 1F). Similarly,Dystroglycan protein level was highly reduced or patchy because of the expression of DG-RNAi construct (tubP-Gal4/dsDG) in follicle cells(Fig. 1G and data not shown).

Since DG is highly expressed in the follicle cells, we first asked whetherDrosophila DG plays a role in establishing or maintaining epithelial morphology in this tissue. The follicle cell epithelium (FE) has a typical apical-basal polarity, with its apical side facing the germline cells(Fig. 2A). As all follicle cells are derived from two to three somatic stem cells, mosaic analysis provides an excellent tool with which to study gene functions in epithelial development (Margolis and Spradling,1995).

We employed the FLP/FRT system (Xu and Rubin, 1993) to generate follicle cell clones of all threeDg alleles and applied the Gal4/UAS-mediated RNAi technique(Kennerdell and Carthew, 2000)to silence DG expression in all follicle cells (tubP-Gal4/dsDG). Similar phenotypes are observed in these different Dg mutant backgrounds. Some mutant cells lost their epithelial appearance and formed multiple layers (Fig. 2B,E), a typical terminal phenotype for polarity defects in epithelial cells. Within the multi-layer groups, the mutant cells from the mosaic egg chambers were frequently excluded from the layer that contacts the germline cells. Discontinuity of the epithelium was also visible in egg chambers containingDg follicle cell clones but not in control egg chambers(Fig. 2B) (yellow arrow). These phenotypes are similar to loss-of-function phenotypes of crb, dlt,dlg or lgl in follicle cells(Tanentzapf et al., 2000;Bilder et al., 2000) and suggest that DG is required for proper epithelial polarity. The mutant follicle cells eventually died off, as we rarely saw mutant clones 9-10 days after heat shock, while sister clones (twin spots) were readily observed.

To characterize the apicobasal polarity defect in more detail, we examined the expression and distribution of molecular markers in mutant cells that still maintained their columnar shape(Table 1). In Dgfollicle cell clones and tubP-Gal4/dsDG follicle cells,mislocalization of apical markers, Dlt and β-Heavy-Spectrin(βH-Spec) (Table 1) was observed. Instead of a strict apical localization, Dlt and βH-Spec were present at both the apical and basal sides of the mutant epithelia(Figs 2C,C′,F′,F″). Dlg, a basolateral marker,exhibited a significant reduction of staining in the basolateral domain inDg RNAi follicle cells (Fig. 2G′). The function of DG in apicobasal polarity formation was not restricted to the FE, as mislocalization of Dlt to the lateral and basal sides was also observed in the mutant epithelial cells in an antennal disc (Fig. 2D, arrows). Taken together, these results suggest that DG is required in different epithelial cells for proper formation or maintenance of apicobasal polarity.

To ask whether DG, when overexpressed, is sufficient to interfere with the epithelial cell polarity we used two UAS constructs, the full-length DG-construct (UAS-DG; Fig. 3A) and the short construct with cytoplasmic and transmembrane domains (UAS-DGcyto; Fig. 3B) and expressed them in the FE and in the embryonic salivary glands (Fig. 3). Both constructs expressed proteins of the expected sizes(Fig. 3A,B) and were induced by the following Gal4 driver lines: daughterlessGal4 (daGal4),for maternal expression: elavGal4, for the salivary gland expression;and the flip-out Gal4 system (Pignoni and Zipursky, 1997) for the FE expression. Similar defects in epithelial polarity were observed with all three drivers.

In wild-type salivary glands, Crb was localized to the apical membrane of the epithelium, facing the lumen of the gland(Fig. 3D,G), while DG expression was undetectable (Fig. 3G). Embryos that overexpress DG showed strong ectopic DG staining on both the apical and basolateral membranes of the salivary gland(Fig. 3H). In about 75% of these salivary glands, the expression of Crb was strongly reduced(n=50; Fig. 3E,H,H′). Whereas Crb localization was disrupted by overexpression of full-length DG (UAS-DG), it was unaffected by overexpression of the form of DG lacking the extracellular domain(UAS-DGcyto; data not shown). These results suggest that the mislocalization of Crb was not due to nonspecific interference with the secretory apparatus but due to a defect on cell polarity. The lateral membrane domain was unaffected as assayed by the localization of Neurotactin, a lateral marker (Table 1; data not shown). As seen in the salivary glands, we found that follicle cells that overexpressed DG (Fig. 3I,arrow) lost the apical markers: βH-Spec and Bazooka (Baz), while normal apical localization of these proteins was observed in neighboring wild-type cells (Fig. 3J,J′,K,K′; Table 1). Again, overexpression of the DGcyto-form did not cause any obvious defects in the follicle epithelial polarity.

As Laminin A is required in the posterior follicle cells for proper oocyte polarity at stages 7-10 (Deng and Ruohola-Baker, 2000), we attempted to ask whether DG functions in the germline cells to receive the polarity signal from the Laminin ECM by clonal analysis. Unfortunately, egg chambers bearing germline clones of all deletion alleles are arrested at previtellogenic stages(Fig. 4A, arrow), prior to the stage we could detect signaling between the posterior follicle cells and the oocyte. Therefore, we concentrated on analyzing the establishment of oocyte polarity in earlier stages, a process that is marked by a posterior movement of the MTOC (Fig. 4B,C). During these stages, a low-level expression of DG is detected at the oocyte membrane(data not shown).

To detect whether the early oocyte polarity is properly established inDg germline clones, we examined the localization of two MTOC markers,Nod-β-Galactosidase (Nod-β-Gal) and ORB(Table 1), which (in the wild type) are localized at the anterior of the oocyte at stage 1(Fig. 4B) and move to the posterior in later stages (Fig. 4C,F). Mislocalization of both markers was observed in the mutant germline clones [Nod-β-Gal mislocalization: 60%, n=32; ORB mislocalization: 76% in Dg³²³ (n=25); 60%Dg²⁴⁸ (n=38)]. In half of the mislocalization cases, the markers either remained in the anterior of the oocyte or surrounded the nuclei after stage 3 (Fig. 4D,E,G). The remaining egg chambers exhibited diffuse staining(data not shown). Compared with wild type, the staining was significantly reduced. Furthermore, no accumulation of α-tubulin was observed in the mutant oocytes, while normal posterior accumulation was detected in the control oocytes between stages 2 and 6 (data not shown)(Clegg et al., 2001;Cox et al., 2001a). In conclusion, these data suggest that DG is required in the early oocyte for the maintenance or translocation of the MTOC from the anterior to the posterior of the oocyte (Fig. 4J-K). This step is crucial in establishing AP polarity in the oocyte and the future embryo (Riechmann and Ephrussi,2001).

Although links between DG and MT cytoskeleton have been suggested(Lumeng et al., 1999), the linkage between DG and the actin cytoskeleton via dystrophin/utrophin is far more evident. We therefore examined the actin distribution in the developing oocyte in the wild-type and Dg germline clones. Previous studies demonstrated that actin is enriched at the cortex of early wild-type oocytes(Fig. 4H)(González-Reyes and St Johnston,1998). Interestingly, this actin enrichment is disrupted in theDg germline clones (Fig. 4I). In addition, `spreading' of the ring canals normally observed in stage 1-2 oocyte is not detected in egg chambers that lack germline DG(Fig. 4I, arrowhead).

At the basal side of the FE, actin filaments have a planar polarity that is perpendicular to the long axis, the AP axis, of the egg chamber(Fig. 5A,C). Integrins and receptor tyrosine phosphatase Lar are involved non-cell-autonomously in organizing this basal actin orientation(Bateman et al., 2001). In our analysis of the βH-Spec staining in follicle cells that expressdsDG, we noticed that βH-Spec is mislocalized to the basal side of the FE to bind the basal actin fibers. Noticeably, the fibers decorated with βH-Spec in different follicle cells appeared to be oriented in a random fashion. To test whether this defect reflects problems in basal actin orientation, we analyzed planar polarity of the actin arrays in control egg chambers and in the mutant Dg follicle cell clones. Instead of normal perpendicular orientation to the AP axis, random misorientation was observed in the Dg mutant egg chambers(Fig. 5B,D). Moreover, the basal actin fibers in follicle cells adjacent to the mutant clones were also misoriented, revealing a non-cell autonomous requirement for DG function(Fig. 5E,F). Although the actin filaments were not organized perpendicular to the AP axis in the mutant cells,they aligned with the neighboring cells, suggesting that some communication of the orientation from one cell to the other still existed. These results suggest that DG has a non-cell-autonomous role in organizing the actin cytoskeleton in the follicle cells, similar to other receptors such as Integrin and Lar. Losing any one of these receptors still allows some orientation transfer but the global direction is defective(Frydman and Spradling, 2001;Bateman et al., 2001) (this study) suggesting that multiple receptor-ECM interactions are required for precise orientation.

Previous data have shown that Laminin stripes in the basement membrane of the FE are organized in the same orientation as the basal actin fibers(Gutzeit et al., 1991;Bateman et al., 2001)(Fig. 5C,G), suggesting an instructive interaction between the actin cytoskeleton and the ECM through a receptor(s). One explanation for the non-cell-autonomous role of DG in basal actin organization is that DG functions through organizing the Laminin ECM to affect the basal actin in the neighboring cell. To test this idea further, we analyzed the orientation of Laminin stripes in the wild-type and theDg mutant follicle cells. Instead of the orientation perpendicular to the AP axis in the wild type (Fig. 5G), overall reduction and misorganization of Laminin ECM occurred in the mutant clone and neighboring regions(Fig. 5H).

To test whether DG is sufficient to organize the Laminin ECM, we asked whether overexpression of DG had any effect on Laminin localization. In stage 10 follicle cells, the majority of the Laminin staining is observed at the basal side (Fig. 5I,arrowhead). Noticeably, Laminin is accumulated at the lateral and apical sides of the follicle cells that overexpressed DG(Fig. 5I, arrow), which is consistent with the fact that high-level DG expression is visible at the apical and basal surfaces of these cells(Fig. 3I). This result suggests that DG can effectively organize the Laminin ECM in Drosophila. The dotted instead of stripe/line appearance of ectopic Laminin because of DG overexpression is consistent with a previous report that DG is required for Laminin binding, while Integrin is required for further formation of the Laminin stripe/line-like structures (Henry et al., 2001b).

Disruption of DGC function is tightly linked with the pathogenesis of various forms of muscular dystrophy. Previous work has revealed a central role for the transmembrane protein DG in this protein complex (reviewed byWinder, 2001). The cellular function and regulation of DG interactions remain elusive, however. In this paper, we describe the isolation and analysis of Drosophila mutants of Dg. Using genetic tools in this model organism, we show that DG is required cell-autonomously for polarizing two different cell types: epithelial cells and the oocyte. In Dg mutant epithelial cells, apicobasal polarity is disrupted, while in oocytes, anteroposterior polarity is abnormal. Loss of DG function in follicle and disc epithelia results in expansion of apical markers to the basal side of the cells and overexpression results in a seemingly opposite phenotype, reduced localization of apical markers.Dg mutations in the germline, however, disturb the enrichment of the oocyte cortical actin and the movement of the MTOC to the posterior oocyte: a process that is the prerequisite for the establishment of all polarity within the egg chamber and embryo. In addition, DG has a non-cell-autonomous effect on the planar polarity of basal F-actin in follicle cells. The non-cell-autonomous phenotype probably results from a lack of instructive interaction between the actin cytoskeleton and the ECM, as DG is required for and sufficient in organizing Laminin in follicle cells(Fig. 5H-J).

Drosophila DG contains most of the hallmarks of vertebrate DG, but is significantly longer than its vertebrate orthologs, due to a ∼250 amino acid duplication in the extracellular domain. The N-terminal half of fly DG harbors a mucin-like domain, similar to vertebrate DG, but is otherwise only weakly conserved. As the mucin-like sugars have been implicated in Laminin binding it is interesting to note that splicing variants of DG that lack exon 8, also lack most of this domain. In addition, altered glycosylation of DG is related to two forms of congenital muscular dystrophy(Brockington et al., 2001;Hayashi et al., 2001;Michele et al., 2002;Moore et al., 2002), and reduced expression of DG is observed in a mouse model of Duchene's muscular dystrophy (Ervasti and Campbell,1993).

The C-terminal half of Drosophila DG is conserved with 31%identity (46% similarity, Fig. 1A,B). Especially well conserved are the protein-protein interaction sites in the cytoplasmic domain of DG, including the binding site for Dystrophin. Seven of the eight amino residues, which are crucial for Dystrophin binding (Huang et al.,2000) are conserved in Drosophila. Recent studies demonstrate that phosphorylation of the tyrosine residue within the dystrophin/utrophin binding motif can interfere with binding to utrophin,leading to recruitment of SH2 domain proteins(Sotgia et al., 2001;Ilsley et al., 2002). The putative SH2-binding motif involved in this interaction is conserved inDrosophila. The third protein-protein interaction described for vertebrate DG is the binding of the SH2-SH3 adaptor GRB2. GRB2 helps initiate the Ras-MAP kinase signal transduction cascade and is involved in controlling cytoskeletal organization (Yang et al.,1995). However, the SH3-binding motif, thought to mediate GRB2 binding, is not fully conserved in Drosophila.

Reduced expression of DG is often associated with tumor formation,suggesting that DG can act as a tumor suppressor(Henry et al., 2001a). It is likely that loss of DG function in some cancers leads to abnormal cell-ECM interactions and thus contributes to progression to a metastatic state. Defects in epithelial interactions normally result in cell death; when associated with abnormal cell growth and division; however, such defects could induce metastasis. Our analysis supports this hypothesis: lack of DG function in Drosophila results in tumor-like structures(Fig. 2B) and abnormal cell movement because of the lack of epithelial integrity and cellular polarity.

Reduction of DG function expands the apical domain and overexpression of DG reduces this domain in epithelial cells. In Dg loss-of-function follicle cell clones, a component of the Lgl-complex, Dlg, is mislocalized. This mislocalization could explain the expansion of apical markers in the clones, as Dlg and Scrib are each required for the lateral localization of each other and their function is essential to restrict the apical markers Crb and Dlt to the apical surface (Bilder et al., 2000; Bilder and Perrimon,2000). Further experiments are directed to distinguish whether mislocalization of Dlg is caused directly by lack of physical interaction with DG or indirectly by lack of proper cytoskeletal arrangements.

Drosophila oocyte polarity is essential for morphogen localization and therefore for the formation of the major body axes. The establishment of oocyte polarity is a gradual process that involves multiple steps (reviewed byRiechmann and Ephrussi, 2001). Key events in the process are cytoskeletal rearrangements. First, the MTOC is present in the anterior region of an early oocyte. By stage 3, the first rearrangement has occurred and the MTOC is positioned in the posterior portion of the developing oocyte. By the end of stage 6, a signal from the posterior follicle cells has initiated a new MT rearrangement, the posterior MTOC disappears and a new anterior MTOC forms. Although this signaling pathway remains a mystery, several molecules including Laminin A have been shown to be involved (Riechmann and Ephrussi,2001; Deng and Ruohola-Baker,2000). As for the first rearrangement, genes encoding theDrosophila Par3/Par6/aPKC-complex, Par-1, and Maelstrom are required,(Cox et al., 2001a;Cox et al., 2001b;Huynh et al., 2001a;Huynh et al., 2001b;Clegg et al., 2001;Vaccari and Ephrussi, 2002). However, the mechanism for the MTOC movement or anchoring is not clear. We show that DG, similar to the Par proteins, is required in the germline for this first rearrangement step. As Dg germline clones also exhibited a defect in cortical actin enrichment in the oocyte, it is possible that the cortical actin plays an important role in MTOC movement and/or anchoring. Alternatively, as DGC contains proteins that can interact with either actin or microtubular cytoskeletons, it could play a role in coordinating actin and microtubule functions in this process.

The fact that DG is required for both epithelial and oocyte polarity re-iterates the idea that common strategies may exist for polarizing these two very different cell types. In addition to DG, Par proteins also act in polarity formation in both cell types, suggesting that the Par proteins and DG complex have functional similarities. Interestingly, DG can affect localization of the Par complex as one of the members, Baz (Par3), is mislocalized when DG is overexpressed. In addition, both Par-proteins and the DG-complex interact with molecules that can associate with either actin or microtubular cytoskeletons. Par-1 associates with Myosin II heavy chain and also phosphorylates a MT-associated protein(Drewes et al., 1997;Guo and Kemphues, 1996). DG can interact with actin through Dystrophin-like proteins. Furthermore, the Dystrophin-associated protein, Syntrophin, interacts with MT-associated proteins via a two-hybrid assay (Lumeng et al., 1999). It is possible that both Par proteins and the DG complex facilitate interactions between actin and microtubules and that these interactions between the two cytoskeletal systems are key regulators for establishment of polarity in both cell types.

To our surprise, Dg mutant follicle cell generated actin defects in neighboring cells; the basal actin was misoriented in adjacent follicle cells (Fig. 5F). How would a defective DG in one cell alter the dynamics of actin organization in the neighboring cell? We propose that the interaction between ECM and DG is bi-directional (Fig. 5J): on one hand, DG organizes the Laminin ECM architecture(Henry et al., 2001b; this study), suggesting that a defect in DG will be transmitted to a defect in ECM organization; on the other hand, a defective Laminin lattice will extend to the surface of the neighboring cell and there this architectural information could be transmitted to the cellular actin cytoskeleton by DG in the neighboring cell (Colognato et al.,1999). Three pieces of evidence support this hypothesis. First,Drosophila DG is capable of organizing the Laminin lattice(Fig. 5H,I). Second, the Laminin lattice in the basal side of follicle cells is oriented in the same orientation as the underlying basal actin lattice(Fig. 5C,G)(Gutzeit et al., 1991;Bateman et al., 2001). Third,Laminin, similar to DG, could also be involved in basal actin organization(Bateman et al., 2001;Frydman and Spradling, 2001). Interestingly, two other Laminin receptors, Integrin and Lar, are also required for basal actin planar polarity in follicle cells(Bateman et al., 2001;Frydman and Spradling, 2001). It is possible that one connector alone would not give enough rigidity or allow enough flexibility in relaying information between the ECM and the basal actin.

In summary, we have shown that DG has two separate functions in cell polarity: cell autonomous in apical-basal and anteroposterior polarity, and non-cell-autonomous in planar polarity. Future research aims to take advantage of Drosophila as a model organism to genetically dissect the partners of DG in these two functions.

We thank Drs Chamberlain and Hauschka for helpful discussions during this study. We are grateful for technical help from C. Althauser, M. Mogren and T. Ho. We also thank Drs Hauschka, Angelo, Campbell, Berg, Cheng, Schaeffer and Baker for helpful comments on the manuscript; and Drs Bhat, Bilder, Bryant,Edgar, Fessler, Giniger, Goodman, Knoblich, Knust, Laverty, Oda, Peifer,Parkhurst, Schedl, Tepass, Thomas, Volk, Warrior and Wodarz; the Bloomington stock center; and Developmental Studies Hybridoma Bank for fly-lines and antibodies. This work was supported by grants from Swedish natural science research council (NFR) to S. B.; Crafoodska stiftelsen and Kungl Fysiografiska Sällskapet i Lund to M. S.; and by grants from the National Institutes of Health, American Heart Association, American Cancer Society and Pew Memorial Trust and Pew Scholarship in the Biomedical Sciences to H.R.-B.