Apical, Lateral, and Basal Polarization Cues Contribute to the Development of the Follicular Epithelium during Drosophila Oogenesis

The polarization of epithelial cells is a fundamental process in animal development. The study of mammalian culture cells and the genetic analysis of epithelial differentiation in Drosophila have made significant contributions to the understanding of the mechanisms involved in epithelial polarization (Tepass 1997; Yeaman et al. 1999; Müller 2000). Transmembrane proteins that specifically localize to one of three surface domains found in epithelial cells mediate asymmetric cues that control cell polarization. Work on mammalian culture cells, which allow the reversible induction of epithelial morphology, has revealed important roles for cadherin- and integrin-mediated adhesion in epithelial polarization. Cadherin and integrin activity causes the assembly of a domain-specific cytocortex at the lateral and basal membrane, respectively. Cell surface polarization is followed by a reorganization of the cytoskeleton that in turn facilitates asymmetric distribution of organelles and the polarized targeting of transport vesicles to the apical or basolateral membranes. Polarized delivery of proteins and lipids is critical for solidifying and maintaining the polarized membrane domains of epithelial cells (Drubin and Nelson 1996; Yeaman et al. 1999). Genetic studies in Drosophila have also revealed evidence for lateral and basal polarization cues (for review, see Tepass 1997). However, the best understood factor that controls epithelial polarization in Drosophila is Crumbs (Crb), a single pass transmembrane protein that is part of the apical membrane (Tepass et al. 1990; Tepass 1996). Crb is a powerful regulator of epithelial polarization as lack of Crb causes the apical membrane to disappear, and overexpression of Crb leads to an “apicalization” of much of the cell surface. Both conditions cause the breakdown of epithelial cell and tissue structure (Tepass et al. 1990; Tepass and Knust 1990; Wodarz et al. 1993, Wodarz et al. 1995).

Analysis of epithelial development in Drosophila has led to a distinction between primary and secondary epithelia (Tepass and Hartenstein 1994a; Tepass 1997). Primary epithelia derive directly from the blastoderm epithelium without mesenchymal intermediates, and differentiate a zonula adherens (ZA) as part of their junctional complex. In contrast, secondary epithelia form by a mesenchymal–epithelial transition and do not contain a ZA. Both types of epithelia require lateral adhesion mediated by DE-cadherin to maintain integrity (Tepass and Hartenstein 1994b; Tepass et al. 1996; Uemura et al. 1996; Haag et al. 1999). Differences in the mechanisms that orient the apical–basal axis in primary and secondary epithelia exist as primary epithelia depend on apical Crb for maintaining polarity, whereas secondary epithelia require basal cues that involve Laminin (reviewed in Tepass 1997). Evidence for a role of basal cues in the polarization of primary epithelia is lacking. On the other hand, Crb is not expressed in secondary epithelia, and an alternative apical polarization factor in secondary epithelia is not known. Mammalian culture cells used in studies on epithelial polarization are derived from epithelia that form late in development in the presence of extracellular matrix. For example, MDCK and CACO-2 cells, which are related to kidney epithelia and intestinal epithelial cells, respectively (Pinto et al. 1983; Yeaman et al. 1999), depend on integrins for normal polarization and could therefore be regarded as models for secondary epithelia. A well-conserved homologue of Crb has recently been described in humans, but its role in epithelial development remains unclear (den Hollander et al. 1999). Other apical polarization factors are currently not known in vertebrates or other invertebrates, although circumstantial evidence suggests that apical polarity might be specified independent of lateral and basal cues in early sea urchin, Xenopus, and mouse embryos (Reeve and Ziomek 1981; Nelson and McClay 1988; Müller and Hausen 1995). Taken together, these findings suggest that epithelial polarity is governed by two polarizing cues in a given cell type. Lateral cadherin-mediated adhesion is critical for the assembly of the lateral membrane domain, and organizes epithelial cells into two-dimensional sheets. One additional polarization cue, either apical or basal, is required to establish the apical–basal axis.

This study characterizes the mechanisms that are involved in the formation of the follicular epithelium (FE) during Drosophila oogenesis. Drosophila follicle cells are widely used as a model to study pattern formation and morphogenesis in an epithelial context. The FE combines features of primary and secondary epithelia as it has a ZA and expresses Crb (Mahowald 1972; Tepass and Knust 1990). On the other hand, the FE forms through a mesenchymal–epithelial transition, during which follicle cells are in contact with a basement membrane (King 1970). Our observations suggest that the FE uses mechanisms operating on the apical, lateral, and basal membranes for epithelial differentiation. Contact of follicle cells to the basement membrane leads to an initial polarization of follicle cells, generating a basal membrane that is distinct from the rest of the plasma membrane.

Full epithelial polarization is initiated by the contact of follicle cells to germline cells. This contact leads to a demarcation of apical and lateral membranes and the accumulation of Crb and the Crb-associated PDZ domain protein Discs Lost (Dlt; Bhat et al. 1999; Klebes and Knust 2000) at the apical membrane. Crb and Dlt are involved in the formation of the FE and at least Crb is also required for its maintenance. Although Crb and Dlt are likely to form a complex at the apical membrane, our data also suggest that Dlt interacts with a second apical factor different from Crb, which may explain the variable defects observed in crb mutant clones. Finally, we explored the requirement of the cadherin–catenin complex in the differentiation of the FE by analyzing mutations in armadillo (arm), which encodes the Drosophila homologue of β-catenin (Peifer and Wieschaus 1990). An essential role for arm in the formation of embryonic epithelia has been documented (Cox et al. 1996; Müller and Wieschaus 1996). We were therefore surprised to find that disruption of the cadherin–catenin complex does not prevent follicle cells from forming a FE. However, arm mutant follicle cells show defects that indicate a disruption of the lateral and apical membrane domains.

We used Oregon R as wild-type stock. Agametic ovaries were recovered from female flies homozygous mutant for oskar³⁰¹ that were raised at 18°C (Lehmann and Nusslein-Volhard 1986). Other mutations and transgenic lines are described below.

Homozygous mutant clones for crb, dlt, and arm were generated with the FRT/FLP system (Golic 1991; Xu and Harrison 1994). Two different strategies were used to induce clones either before the formation of the FE (“early clones”), or after the FE has been established (“late clones”). For induction of early clones, FLP expression was driven in follicle stem cells using the system developed by Duffy et al. 1998. Here, the FLP gene, which is under the control of upstream activator sequences (UAS>FLP), is driven by the Gal4 line e22c-Gal4 that is expressed in follicle stem cells. Late clones were induced by driving FLP under the control of a heat shock promoter (hsFLP; Xu and Harrison 1994).

Homozygous crb mutant clones were induced for the null alleles crb^11A22 and crb^B82 (Tepass and Knust 1990; Brook et al. 1993). The genotypes of flies, in which clones were induced are as follows. Early clones: w; e22c-Gal4, UAS>FLP/+; FRT^82B crb⁻/FRT^82B+. Late clones: w hsFLP1/+; FRT^82B crb⁻/FRT^82B+.

Homozygous dlt mutant clones were induced for the null allele dlt^MY10 and the hypomorphic, protein-positive allele dlt^dre1 (Bhat et al. 1999). The dlt^MY10 allele is a deletion that uncovers the genes dlt, cdc37, and α-spectrin. To compensate for the loss of cdc37, clones were induced in the background of a cdc37 rescue construct, P[w⁺, cdc37^res]. However, dlt^MY10 clones are mutant for both dlt and α-spectrin. FRT, dlt recombinant lines were provided by Manzoor Bhat (Mount Sinai School of Medicine, New York, NY). The genotypes of flies, in which clones were induced are as follows. Early clones: w; e22c-Gal4 UAS>FLP/P[w⁺, cdc37^res]; FRT^80B dlt^MY10/FRT^80B+. w; e22c-Gal4 UAS>FLP/+; FRT^80B dlt^dre1/FRT^80B+. Late clones: w hsFLP1/+; P[w⁺, cdc37^res]/+; FRT^80B dlt^MY10/FRT^80B+. w hsFLP1/+; FRT^80B dlt^dre1/FRT^80B+.

Homozygous arm mutant clones were induced for the null allele arm^YD35, the strong hypomorph arm^XK22, and the intermediate hypomorph arm^XP33 (Peifer and Wieschaus 1990). The genotypes of flies in which clones were induced are as follows. Early clones: w arm⁻ FRT¹⁰¹/w, FRT¹⁰¹; e22c-Gal4 UAS>FLP/+. Late clones: w, arm⁻ FRT¹⁰¹/w FRT¹⁰¹; hsFLP2/+.

For early clones, flies where grown at 25° or 29°C; adults where maintained in yeasted vials and dissected 2–4 d after eclosure. For late clones, freshly eclosed females were collected, heat shocked for 1–2 h at 37°C, and transferred to yeasted vials for 2–5 d before dissection.

Clonal overexpression of full-length Crb (UAS>crb^wt2e; Wodarz et al. 1995) and of a truncated form of Crb that lacks most of the extracellular part, including all EGF-like domains and the LG domains (UAS>crb^intra2b; Wodarz et al. 1995) was induced following the technique of Ito et al. 1997. Clones were induced in flies of the genotypes w hsFLP1; AyGal4 UAS>lacZ/UAS>crb^wt2e or w hsFLP1; AyGal4 UAS>lacZ/+; UAS> crb^intra2b/+. Newly eclosed females were heat shocked for 15 min at 37°C and kept in yeasted vials for 2–4 d before dissection.

A full-length cDNA encoding Dlt (accession No. AF 274350) as isolated by screening a Drosophila embryo (4–6 h) cDNA expression library (Novagen) with a peptide corresponding to the carboxy terminus of the Inscuteable protein. Polyclonal antibodies were generated by immunizing rabbits with a peptide (NH₃-SMGAEPDLIPDWRN-COOH) corresponding to amino acids 858–872 of Dlt. An amino-terminal cysteine was added to facilitate keyhole limpet hemocyanin conjugation and to couple the peptide to a Sulfolink column (Pierce Chemical Co.) used to affinity purify the anti–Dlt serum.

For immunostainings, the following primary antibodies were used: rat monoclonal antibody anti–DE-cadherin (DCAD2, 1:50; Oda et al. 1994); rat mAb anti–DN-cadherin (Ex8, 1:50; Iwai et al. 1997); mouse mAb anti–Crb (Cq4, 1:25; Tepass and Knust 1990); mouse mAb anti–Fasciclin III (7G10, 1:50; Patel et al. 1987); mouse mAb anti–Armadillo (N2-7A1, 1:100; Peifer et al. 1993); rabbit polyclonal antibody anti–Armadillo (N2, 1:200; Riggleman et al. 1990); mouse mAb anti–α-spectrin (3A9, 1:100; Byers et al. 1987); rabbit pAb anti–β_Heavy-spectrin (243, 1:500; Thomas and Kiehart 1994); rabbit pAb anti–Dlt (1:250); rabbit pAb anti–β-galactosidase (1:1,000; Cappel); mouse mAb anti–β_PS-integrin (G11, 1:50; a gift from L. Fessler, University of California at Los Angeles, Los Angeles, CA).

Anti–DE-cadherin and anti–DN-cadherin stainings were carried out as previously described (Niewiadomska et al. 1999). For other antibody stainings, ovaries from 2–5-d-old well-fed female flies were dissected in PBS and fixed in 5% formaldehyde PBS, pH 7.4, for 10 min, and then treated with methanol for 5 min. Methanol treatment was not used for all phalloidin stainings, and the β_PS-integrin, DE-cadherin double-labeling experiment. Tissues were washed in PBT (PBS, 0.2% Triton X-100) for 4 × 15 min, followed by a 1-h incubation in blocking solution PBTB (PBT, 0.2% BSA, 5% goat serum). Incubation with primary antibodies, diluted in blocking solution, was done at 4°C overnight. Ovaries were washed in PBT for 4 × 15 min and incubated in blocking solution for 1 h. Secondary antibodies conjugated with Cy3, Cy5, FITC (Jackson ImmunoResearch Laboratories), or Oregon green 488 (Molecular Probes) were used at a dilution of 1:400 in PBTB at 4°C overnight. Ovaries were washed in PBT for 4 × 15 min and mounted in Antifade [70% glycerol + 2.5% DABCO (Sigma-Aldrich) in PBS].

F-actin was detected with phalloidin. After antibody staining, ovaries were washed with PBS, incubated in Oregon green 488-phalloidin (Molecular Probes) at a dilution of 1:20 in PBS at 4°C overnight, washed in PBS, and mounted in Antifade. Cell nuclei were visualized with Picogreen (Molecular Probes). After antibody staining, ovaries were treated with 0.4 mg RNaseA/ml in PBT for 1 h, rinsed with PBT, incubated with Picogreen at a dilution of 1:1,000 in PBT at 4°C overnight, washed in PBT, and mounted in Antifade.

Confocal images were obtained with a scanning laser confocal microscope (LSM510; Carl Zeiss, Inc.) using Plan-Neofluar 40×/1.30 oil and Plan-Apochromat 100×/1.40 oil lenses. Images were processed in Adobe Photoshop or Adobe Illustrator.

Immunoelectron microscopy (IEM) on embryos was carried out as described previously (Tepass 1996). For IEM on follicles, ovaries were dissected in PSS buffer (100 mM Na-glutamate, 25 mM KCl, 15 mM MgCl₂, 5 mM CaSO₄, 2 mM sodium phosphate buffer, pH 6.9) (Woodruff and Tilney 1998), and fixed for 30 min in 8% formaldehyde, 0.02% glutaraldehyde in 0.05 M phosphate buffer, pH 7.2. Further treatment was as described previously (Tepass 1996). The dilution of the anti–Dlt antibody in these experiments was 1:150.

Drosophila ovaries are composed of ovarioles, each representing an anterior–posterior series of follicles of increasing age (King 1970; Spradling 1993). The assembly of follicles takes place in the germarium that is located at the anterior tip of each ovariole (Fig. 1). During follicle formation, ∼30 follicle cells form an epithelial monolayer that surrounds a cluster of 16 germline cells (Margolis and Spradling 1995). Whereas the germline cells increase in size but not in number, the cells in the FE proliferate so that the FE in a mature follicle contains ∼650 cells. A second population of follicle cells that does not establish direct contact with germline cells forms short stacks by cell intercalation, giving rise to the interfollicular stalk (Godt and Laski 1995).

Crb is initially expressed in all follicle cells at stage 1 of oogenesis, but its expression becomes rapidly restricted to the cells of the FE, and is not maintained in the interfollicular stalk (Fig. 2 A; Tepass and Knust 1990). Crb is found in the apical membrane of the cells of the FE (Fig. 2 A), similar to many other Drosophila epithelia (Tepass et al. 1990; Tepass and Knust 1990). Dlt shows a similar temporal and spatial distribution as Crb in follicle cells (Fig. 2, B–D), suggesting that both proteins might interact physically in follicle cells as they do in embryonic epithelia (Bhat et al. 1999; Klebes and Knust 2000). Dlt also associates in vitro with Neurexin IV, a transmembrane component of the septate junction (Baumgartner et al. 1996; Bhat et al. 1999). The septate junction, which blocks paracellular diffusion similar to the chordate tight junction, is located basally to the ZA (Mahowald 1972; Tepass and Hartenstein 1994a,Tepass and Hartenstein 1994b). To reconcile these data, it was proposed that Dlt interacts with Crb apically to the ZA, with the ZA, and with Neurexin IV basally to the ZA (Bhat et al. 1999).

To clarify the subcellular localization of Dlt, we determined the distribution of Dlt in the FE and in the embryonic ectoderm and epidermis by IEM (Fig. 3). IEM reveals that Dlt, similar to Crb (Tepass 1996), is confined to the apical membrane. Within the apical membrane, Dlt accumulates at the marginal zone, an area of cell–cell contact apical to the ZA, and shows a lower concentration throughout the apical surface (Fig. 3). The signal for Dlt at the ZA or basally to it does not exceed background levels, and no signal is observed at septate junctions (Fig. 3 F). These findings suggest that Dlt does not physically interact with Neurexin IV at the septate junction, and further corroborates the notion that Dlt and Crb form a complex at the apical membrane of epithelial cells.

Crb and Dlt are also expressed in the germline, although the distribution of both proteins overlaps only partially. Dlt and Crb colocalize at the membrane of the nurse cells during early and mid oogenesis. High levels of Crb are seen in the plasma membrane of the oocyte, whereas Dlt is not detectable (Fig. 2C and Fig. D; Niewiadomska et al. 1999). Dlt accumulates transiently at the ring canals that connect nurse cells (data not shown). During late stages of oogenesis, when the content of the nurse cells is rapidly transferred to the oocyte (Cooley and Theurkauf 1994), Dlt is found at the nuclear membrane in close association with actin filaments that connect the nuclei of the nurse cells with the plasma membrane (data not shown). Previous germline clone experiments did not reveal a function for crb in the germline (Tepass and Knust 1990). In contrast, maternal expression of dlt is essential for egg production, suggesting that dlt plays an important role in germline development (Bhat et al. 1999). The function of dlt in germline development remains to be studied in more detail.

The follicle cells are generated by two stem cells that are located in the middle of the germarium at the boundary of region 2a and 2b (Margolis and Spradling 1995). Offspring of these stem cells establish contact with the basement membrane that surrounds the germarium and the follicles (King 1970; Fig. 1). Analysis of agametic ovaries has shown that the contact between follicle cells and germline cells is required for the formation of the FE. Follicle cells continue to proliferate in agametic ovaries and form a column that is two to three cells wide (Margolis and Spradling 1995; Goode et al. 1996b). To determine whether, and to what extent, contact between follicle cells and the basement membrane contributes to the polarization of the FE, we examined follicle cells in agametic ovaries (Fig. 4). In agametic ovaries, β_PS-integrin localizes to the basal membrane of follicle cells as in wild type (Fig. 4A and Fig. B). Markers that normally localize to the lateral membrane [Fasciclin III (Fig. 4 C), and DN-cadherin (data not shown)], to the lateral and apical membranes [DE-cadherin (Fig. 4A and Fig. B), Armadillo (data not shown)], or to the apical membrane (β_Heavy-spectrin, Fig. 4 D) are excluded from the basal cell pole. Apical and lateral markers show an overlapping distribution at the non-basal cell surface, but are concentrated at the cell pole that opposes the basal membrane. These findings suggest that contact to the basement membrane causes a partial polarization of follicle cells. The basal membrane is established and an asymmetric distribution of apical and lateral markers is observed, but the apical and lateral membrane domains are not clearly demarcated. As the follicle cells express the cadherin–catenin complex constitutively (Peifer et al. 1993; Oda et al. 1997; Godt and Tepass 1998; Niewiadomska et al. 1999; this work), it appears that basal polarization cues together with the activity of the cadherin–catenin complex are insufficient to fully polarize follicle cells.

Interestingly, the presence of germline cells is required for the accumulation of Crb in follicle cells, as Crb is not detected in follicle cells in agametic ovaries (Fig. 4 E).

To study the role of crb and dlt in the development of the FE, we generated homozygous mutant follicle cell clones. Clones were induced for the two protein-negative crb null alleles crb^11A22 and crb^B82, for the dlt protein negative null allele dlt^MY10, and the hypomorphic protein-positive allele dlt^dre1. We examined two types of mutant follicle cell clones. Clones were generated either before the formation of the FE by inducing mitotic recombination in the follicle stem cells or their immediate offspring. Alternatively, clones were induced after the formation of the FE by inducing mitotic recombination in region 2b or stage 1 follicles (Fig. 1; see Materials and Methods). Early induced clones can be distinguished from late induced clones by their larger size as all follicle cells derive from only two stem cells, and follicle cells continue to proliferate during early stages of follicle development (Margolis and Spradling 1995). We considered clones as induced early (early clones) if they comprise 15% or more of the cells of the FE of an individual follicle. Early clones can make up the entire FE of a single follicle. On the other hand, late-induced clones (late clones) typically comprise less than 10 cells at mid to late stages of oogenesis.

A variable phenotype is observed when crb mutant follicle cells are generated before the FE forms. Many experimental follicles show an incomplete FE with areas in which the germline cells are not covered by follicle cells (Fig. 5A and Fig. D), a defect that is never seen in wild-type follicles. The area of a follicle not covered by a FE varied greatly in size and in some cases comprised most of the follicle. These gaps in the FE are likely caused by the failure of crb mutant follicle cells to integrate into the FE. We were unable to track the missing follicle cells and presume that they degenerate within the germarium. On the other hand, some crb mutant follicle cells can form a FE with apparently normal morphology (Fig. 5D and Fig. E). The distribution of various markers was analyzed in crb mutant follicle cells that were part of the FE. Such crb mutant follicle cells retained apical Dlt, although the level of Dlt associated with the apical membrane was reduced in these cells (Fig. 5D and Fig. E). The level of β_Heavy-spectrin is also slightly reduced compared with wild-type follicle cells (Fig. 5 F). These findings indicating that the apical localization of Dlt and β_Heavy-spectrin depends on Crb, but also on other mechanisms. No significant alteration was noticed in the level or distribution of Arm (Fig. 5 G).

dlt^MY10and dlt^dre1 mutant follicle cells clones, if generated before the formation of the FE, displayed gaps of variable size in the FE similar to those caused by crb mutations (Fig. 6, B–D). As dlt^dre1 is a protein-positive allele, we were not able to determine whether some dlt^dre1 mutant follicle cells become part of the FE. In contrast to crb mutant follicle cell clones, we did not find any dlt^MY10 mutant follicle cells that participate in the formation of the FE in early clones. dlt^MY10 is a deletion that removes the genes encoding cdc37 and α-spectrin in addition to dlt (Bhat et al. 1999; our unpublished observations). The loss of cdc37 is compensated for by a transgene (see Materials and Methods). Thus, dlt^MY10 mutant clones lack dlt and α-spectrin, raising the possibility that the observed mutant phenotype is the consequence of a synergistic effect between dlt and α-spectrin mutations, although it was shown that α-spectrin is not required for the formation of the FE (Lee et al. 1997). These findings suggests that Dlt is required, and may be essential for the formation of the FE.

Large crb mutant clones within the FE that survive until mid-oogenesis often form a multilayered epithelium indicating that Crb is important for the maintenance of the FE. The multilayering of follicle cells in crb mutant clones is limited to posterior follicle cells that cover the oocyte (Fig. 5 B). To determine whether removal of Crb and Dlt from cells of the FE disrupts the integrity of the FE, we also examined small crb or dlt mutant follicle cell clones that were induced after the FE had been established. Small crb mutant cell clones developed normally until late in oogenesis (Fig. 5 C), although they showed a substantial reduction in the level of Dlt (data not shown). Cells in small dlt mutant clones develop with apparently normal morphology, and show a normal distribution of F-actin and Arm (Fig. 6G and Fig. H). Crb was lost from most dlt mutant clones, indicating that Dlt is important for maintaining apical Crb (Fig. 6E and Fig. F). Some dlt mutant clones retained Crb, suggesting that a mechanism other than binding to Dlt can contribute to maintaining Crb at the apical membrane. Crb was also undetectable in many cell clones mutant for dlt^dre1. Dlt protein in dlt^dre1 mutant cells forms a “cap” in the center of the apical membrane rather than being distributed throughout (Fig. 6 C). Together, these findings suggest that the dlt^dre1 allele may carry a mutation that specifically disrupts the interaction of Dlt and Crb, but not the interaction of Dlt with other apical factors.

We examined the consequences of Crb overexpression in the FE to analyze the response of the FE to mislocalization of Crb, and to further study the interactions between Crb and Dlt. Overexpression of Crb in embryonic epithelia causes a severe disruption of epithelial integrity that includes an extension of the apical cell surface and multilayering of epithelial tissues (Wodarz et al. 1995; Grawe et al. 1996; Klebes and Knust 2000). We first expressed a truncated form of Crb that lacks all EGF and laminin G domains (UAS>crb^intra; Wodarz et al. 1995). In embryos, this construct causes a similar phenotype as overexpression of full-length Crb, but, as it is not recognized by our anti–Crb antibody, the effect of the expression of crb^intra on the distribution of endogenous Crb can be examined. Expression of UAS>crb^intra causes a strong reduction in the level of endogenous Crb, suggesting that crb^intra acts competitively with Crb for interaction to binding partners that are important for maintaining Crb in the plasma membrane (Fig. 7 A). Overexpression of full-length Crb leads to misdistribution of Crb into the lateral membrane, at levels that are similar to the apical membrane. In most of these cell clones, we did not see a significant mislocalization of Dlt (Fig. 7 B), suggesting again that the apical localization of Dlt is at least in part Crb independent. A fraction of the Crb overexpressing follicle cells show mislocalization of Dlt, a thinning of the epithelium, and a strong reduction of lateral markers such as Arm and Fasciclin III (Fig. 7C and Fig. D). These findings indicate that overexpression of Crb disrupts the lateral membrane in follicle cells.

The cadherin–catenin complex plays a major role in epithelial polarization as cadherin-mediated adhesive contacts cause the assembly of the lateral surface domain. Consequently, lack of cadherin activity compromises epithelial integrity in many tissues or epithelial cell culture lines (Drubin and Nelson 1996; Tepass 1997). Removal of DE-cadherin from follicle cells causes only mild defects in the development of the FE (Godt and Tepass 1998). In follicle cells that lack DE-cadherin, some Arm is retained at adherens junctions, raising the possibility that the FE coexpresses a second cadherin that interacts with Arm. Indeed, we find that DN-cadherin (Iwai et al. 1997) is expressed in follicle cells in a pattern that overlaps with DE-cadherin in early to mid oogenesis (Fig. 8; data not shown). DN-cadherin disappears from the FE at stage 10 of oogenesis, whereas DE-cadherin is expressed throughout oogenesis. In contrast to DE-cadherin, DN-cadherin is not expressed in the cells of the germline (data not shown). On the other hand, Arm appears to be the only β-catenin homologue in Drosophila, in contrast to vertebrates, where β-catenin can be functionally replaced by plakoglobin in the cadherin–catenin complex (e.g., Huelsken et al. 2000). Thus, to effectively remove the cadherin–catenin complex from follicle cells, we generated clones that lack Arm, which was previously shown to interact with both DE- and DN-cadherin (Peifer 1993; Oda et al. 1993, Oda et al. 1994; Iwai et al. 1997).

Clones were induced for three different mutant arm alleles that carry premature stop codons in the 6th (arm^YD35), 7th (arm^XK22), and 10th (arm^XP33) arm repeat. arm^XP33 is an intermediate hypomorph, arm^XK22 is a strong hypomorph, and arm^YD35 is believed to be a null allele (Peifer and Wieschaus 1990). Embryos derived from germline clones mutant for arm^XP33 show a dramatic disruption of epithelial morphology that occurs at the onset of gastrulation and is substantially more severe than the defects in epithelial structure seen in crb null mutant embryos (Cox et al. 1996; Müller and Wieschaus 1996; Tepass 1996, Tepass 1997). Moreover, it has been shown previously that Crb is needed for the formation of the ZA in embryonic epithelia (Tepass 1996; Grawe et al. 1996). If the failure to form the ZA is the major consequence of compromising Crb or Dlt activity in the FE, we would expect the lack of the cadherin–catenin complex to cause similar defects as seen in crb and dlt mutant follicle cells; that is, a failure to form a FE. Surprisingly, we find that follicle cells mutant for any of the three arm alleles form a FE (Fig. 8A, Fig. D, and Fig. G). No follicles were observed in these experiments that show epithelial discontinuities, as seen in crb and dlt mutant follicles. arm mutant follicle cells often show an irregular morphology at early stages of follicle development. The irregularities in epithelial structure increase in severity until the FE is compromised and the follicle degenerates at mid to late oogenesis. To determine whether adherens junctions were effectively disrupted in arm mutant follicle cells, we examined the expression of DE- and DN-cadherin in those cells. Neither DE- nor DN-cadherin are detected in follicle cells mutant for any of the three arm alleles studied (Fig. 8B, Fig. C, Fig. E, Fig. F, Fig. H, and Fig. I). Taken together, these findings suggest that cadherin-based adherens junctions are not essential for the formation of the FE, but are important for maintaining its epithelial structure.

We took advantage of the fact that arm mutant cells in the FE are maintained for several days and analyzed their molecular architecture. arm^XP33 mutant follicle cells (see also Müller 2000), which in most cases have a normal cuboidal to columnar shape, show a decrease of F-actin and α-spectrin at the lateral membrane, and an accumulation of these molecules at the apical cell pole (Fig. 9A and Fig. B). In contrast, the apical marker β_Heavy-spectrin shows a normal distribution in arm^XP33 mutant cells (Fig. 9 C), suggesting that the apical spectrin cytoskeleton is intact. Follicle cells mutant for arm^XK22 or arm^YD35 often develop a squamous cell morphology (Fig. 9, D–F) or show a multilayered structure. α-Spectrin remains associated with the narrow lateral membranes in squamous arm mutant cells (Fig. 9 D). β_Heavy-Spectrin, on the other hand, is lost from these follicle cells, suggesting that the apical spectrin cytoskeleton is disrupted (Fig. 9 E). To further examine the apical surface domain of arm mutant follicle cells, we studied the distribution of Crb and Dlt in these cells. arm^XP33 mutant follicle cells show typically a normal apical localization of Crb and Dlt (Fig. 10A and Fig. B). In contrast, Crb is lost from the apical membrane of follicle cells mutant for strong arm alleles, whereas apical Dlt is retained in these cells (Fig. 10, C–E). Similar to dlt^dre1 mutant cell clones, Dlt forms a cap in the center of the apical membrane of arm mutant follicle cells. These observations suggest that the disruption of adherens junctions leads to a breakdown of the lateral membrane domain, as expected, but also compromises the apical surface domain. The differential behavior of Crb and Dlt in strong arm mutant cell clones again emphasizes that Dlt can rely on a Crb-independent apical targeting mechanism, and shows that apical Dlt can be retained in the absence of an apical spectrin cytoskeleton.

The polarization cues that contribute to the differentiation of the FE described here participate in the development of a number of other epithelia (Eaton and Simons 1995; Drubin and Nelson 1996; Tepass 1997; Yeaman et al. 1999). They include contact with the basement membrane, lateral adhesion mediated by the cadherin–catenin complex, and apical polarization that, in Drosophila, involves Crb. The FE is unique among epithelia characterized to date as it is the first epithelium for which polarization cues operating in parallel at all three cell surfaces have been documented. Further, our observations concerning the interactions between Crb and Dlt suggest that both proteins interact functionally in epithelial polarization, but also that the activity of Dlt is in part independent of Crb.

Contact of follicle cells with basement membrane appears sufficient to establish a basal membrane from which apical and lateral markers are excluded. Contact to extracellular matrix (ECM) material is important for the polarization of a number of epithelia including the Drosophila midgut epithelium and the dorsal vessel (Yarnitzky and Volk 1995; Haag et al. 1999), as well as vertebrate kidney epithelia and cell culture lines such as MDCK (Eaton and Simons 1995; Yeaman et al. 1999). Contact of MDCK cells to ECM not only establishes a basal membrane, but also leads to the formation of microvilli throughout the nonbasal plasma membrane (Vega-Salas et al. 1987; Ojakian and Schwimmer 1988). Such long-range effects of basal contacts in the differentiation of epithelial membrane domains may also contribute to the differentiation of the FE as apical and lateral markers, such as β_Heavy-Spectrin and DE-cadherin, are not uniformly distributed over the nonbasal plasma membrane of follicle cells in agametic follicles, but are concentrated at the cell pole that opposes the basement membrane. These findings suggest that follicle cells express extracellular matrix receptors that contribute to epithelial polarization. However, these extracellular matrix receptors together with the activity of the cadherin–catenin complex are not sufficient to cause full epithelial polarization of follicle cells.

One unusual feature that sets the FE apart from most other epithelia is that its apical surface is not a free surface but contacts the cells of the germline. Contact of follicle cells to the germline is essential for the formation of the FE (Margolis and Spradling 1995; Goode et al. 1996b). Interaction between germline and follicle cells is also required for the accumulation of Crb in follicle cells, as indicated by the absence of Crb from follicle cells in agametic ovaries. In wild-type ovaries, Crb and Dlt are initially detected in all follicle cells, but both proteins rapidly disappear from the cells that form the interfollicular stalk. Both proteins are retained only in those cells that give rise to the FE in which they are found at the apical membrane. The subcellular distribution of Crb and Dlt in the FE and the fact that both proteins are required for the differentiation of the FE suggest that the mechanism of apical polarization of the FE is similar to epithelia that have a free apical surface (Tepass et al. 1990; Tepass and Knust 1990; Bhat et al. 1999).

Crb and Dlt are the first identified structural components of the FE that are essential for the formation of this epithelium. The activity of two other genes, brainiac (brn), which encodes a secreted protein, and egghead (egg), which encodes a transmembrane protein, are also important for the formation of the FE as mutations in both genes cause epithelial discontinuities in the FE similar to mutations in Crb and Dlt (Goode et al. 1992, Goode et al. 1996a,Goode et al. 1996b). Both brn and egg are expressed exclusively by the germline, suggesting that these genes contribute to germline-follicle cell interactions critical for FE formation. Interestingly, the EGF receptor (Egfr) pathway mediates interactions between germline and follicle cells that contribute to the formation of the FE. Compromising the activity of Gurken, a transforming growth factor-α–like ligand of Egfr expressed by the germline, or of Egfr causes gaps in the FE similar to those seen in crb or dlt mutant follicle cell clones (Goode et al. 1996a). These findings raise the possibility that the germline-dependent expression of Crb in follicle cells may be controlled by Egfr signaling.

Crb is also required for the maintenance of the FE as Crb mutant clones develop a multilayered structure by mid oogenesis. This defect, which is seen in posterior follicle cells that cover the oocyte, was only observed in large crb mutant clones induced before the formation of the FE, but not in small crb mutant clones. Absence of structural defects in small crb mutant cell clones might be due to perdurance; that is, the cytoplasmic inheritance of crb gene product, after a crb mutant clone was induced. We believe that this possibility is unlikely as small crb mutant clones do not contain detectable levels of Crb 2 d after they were induced, and are then maintained in the FE for several days until late oogenesis. An alternative possibility that would explain the differences in the behavior of small and large crb mutant clones is to assume that the structural defect caused by the lack of Crb is rather weak, so that effects on epithelial tissue structure are manifested only as a “community effect;” that is, when small structural defects in a large number of cells enhance each other, leading eventually to a collapse of normal tissue architecture.

Mutations in a number of other genes, including α-spectrin, brn, egg, and Notch, also cause multilayering of the FE that is limited to the follicle cells that cover the oocyte and develop by mid oogenesis (stage 7 and beyond; Goode et al. 1996b; Lee et al. 1997). Defects in epithelial architecture that are more severe than those seen in large crb mutant cell clones are detected in follicles mutant for discs large, lethal giant larva, and scribble (Manfruelli et al. 1996; Goode and Perrimon 1997; De Lorenzo et al. 1999; Bilder et al. 2000), as well as arm (this work). These four genes encode cytoplasmic factors that are predominantly associated with the lateral cytocortex of epithelial cells. Taken together, the multitude of factors now known to contribute to maintaining the FE, and the fact that most factors cause only moderate defects in epithelial structure if removed, suggest that the stability of the FE is controlled by several molecular pathways that act in parallel and may overlap in function.

The cadherin–catenin complex is required for maintaining the integrity of the FE, but, surprisingly, not for its initial formation. Embryos mutant for the intermediate allele arm^XP33 show a dramatic and rapid collapse of the ectodermal epithelium at gastrulation after the blastoderm epithelium has been established (Cox et al. 1996; Müller and Wieschaus 1996). In contrast, follicle cell clones mutant for this allele of arm are maintained with normal gross morphology for several days. The lateral cytocortex is disrupted in these cells as F-actin and α-spectrin are strongly decreased in concentration at the lateral plasma membrane, and have apparently relocalized to the apical cell pole, where they may associate with the apical membrane or the ZA. Also, follicle cells mutant for an arm null allele form a FE, but display an irregular morphology already at early stages of oogenesis. These mutant cells often develop a squamous cell shape, which suggests a decrease in the size of the lateral membrane. A correlation between the expression level of the cadherin–catenin complex and epithelial cell shape has also been found in Xenopus ectodermal cells that develop a cuboidal or columnar shape in response to the overexpression of N-cadherin (Detrick et al. 1990; Fujimori et al. 1990).

The finding that the cadherin–catenin complex is dispensable for the formation of the FE, and that in its absence integrity is lost not rapidly, but slowly, over a period of several days, implies that the FE has mechanisms that can compensate for the loss of this complex, although this complex plays a key role in epithelial development in other Drosophila and vertebrate epithelia (Tepass 1997; Yeaman et al. 1999). One possibility is that the basal and apical polarization cues are sufficient to establish epithelial polarity in follicle cells, and align these cells in an epithelial layer. Moreover, apical adhesion to germline cells and basal adhesion to a basement membrane might constrain the follicle cells in a monolayered sheet. Finally, the FE might express an alternative lateral adhesion system that can assemble a lateral cytocortex and even a ZA, as exemplified by Caenorhabditis elegans epithelia that form and are maintained in the absence of a classic cadherin and associated catenins (Costa et al. 1998; Raich et al. 1999).

One interesting observation in arm mutant follicle cells is the disruption of the apical cell pole. Whereas several apical markers are maintained in arm^XP33 mutant follicle cells, Crb and β_Heavy-Spectrin are lost from the apical surface in cells mutant for strong or null alleles of arm. In contrast, Dlt still associates with the apical membrane of arm mutant follicle cells, suggesting that the apical membrane has not disappeared entirely, but shows specific molecular defects. Previously, it was shown that Crb is needed for the formation of the ZA in early Drosophila embryos (Grawe et al. 1996; Tepass 1996) and that β_Heavy-spectrin is required for maintaining a continuous ZA in imaginal disc epithelia and in the FE (Thomas et al. 1998; Zarnescu and Thomas 1999). Taken together, these data reveal mutual dependencies in the stability of molecular complexes that contribute to epithelial polarity. These complexes may interact directly at the transition zone between the apical and lateral membranes, where both complexes are enriched, Crb and the apical spectrin cytoskeleton at the marginal zone, and the cadherin–catenin complex at the ZA.

Dlt was shown to interact with the cytoplasmic tail of Crb in vitro and in vivo, an interaction that involves the COOH-terminal five amino acids in Crb and presumably the first of four PDZ domains in Dlt (Bhat et al. 1999; Klebes and Knust 2000). In epithelia that express Crb and Dlt, both proteins colocalize at the apical membrane, where they concentrate just apical to the ZA, in the marginal zone of the apical membrane, and are found at a lower concentration in the remaining apical membrane (Fig. 4; Tepass 1996). IEM shows that Dlt, like Crb, straddles the apical edge of the ZA, but is not a component of the ZA or the septate junction. This finding suggests that Dlt does not interact with Neurexin IV at the septate junction, as was proposed previously (Bhat et al. 1999).

Comparative analysis of the crb and dlt mutant phenotypes thus far did not directly support a functional interaction between these genes. Removal of Dlt from embryos causes severe defects at pregastrula stages that mask potential defects at later embryonic stages when Crb and Dlt are co-expressed (Bhat et al. 1999). Dlt is also required for the development of imaginal disc epithelia, whereas Crb is not, although Crb and Dlt are coexpressed in those epithelia (Tepass and Knust 1990; Bhat et al. 1999). Thus, our finding that crb and dlt mutant follicle cell clones show similar defects is the first direct phenotypic evidence that these genes interact functionally to support epithelial development. Dlt is required to maintain Crb at the apical surface. On the other hand, four observations support the conclusion that apical Dlt localization is at least in part independent of Crb. (a) Dlt is found at the apical membrane, although at reduced levels, in Crb mutant follicle cell clones. Not only is the retention of Dlt at the apical membrane independent of Crb, but also its initial localization as follicle cells that have been rendered crb mutant before they would normally express Crb still show apically localized Dlt. (b) Dlt is also retained apically in dlt^dre1 mutant follicle cells that have lost Crb, and (c) in arm^XK22 and arm^YD35 mutant follicle cells that have lost Crb as well. (d) Dlt retains its apical localization in many follicle cells that overexpress Crb and in which Crb accumulates at the lateral membrane. Moreover, reduced levels of Dlt are retained at the apical membrane in embryonic and imaginal epithelia that have not lost integrity in crb mutants (our unpublished observations). These findings suggest that Dlt interacts with another apical factor or factors different from Crb, which remain to be identified.

Assuming that Crb activity is mediated through apical Dlt, the Crb-independent retention of Dlt might explain the variability in the crb mutant phenotype observed in follicle cells and other epithelia (Tepass and Knust 1990). A finding that corroborates this explanation of the variable nature of the crb mutant phenotype is that dlt^MY10 mutant follicle cells induced before the formation of the FE show no variability and, consistently, do not form a FE. A caveat of this conclusion is that dlt^MY10, the only available null allele of dlt, is a deletion that uncovers α-spectrin and cdc37, in addition to dlt (Bhat et al. 1999). The function of cdc37 has been restored by a cdc37 transgene in our experiments so that dlt^MY10 mutant clones lack only Dlt and α-spectrin. Although it was shown that the lack of α-Spectrin does not interfere with the formation of the FE (Lee et al. 1997), we cannot rule out the possibility that the consistent defects cause by dlt^MY10 are a result of synergistic interactions between α-spectrin and dlt. Analysis of a second, hypomorphic dlt allele, dlt^dre1, which also causes epithelial discontinuities in the FE, demonstrates that dlt similar to crb is required for the formation of the FE. Together, these results suggest that Crb and Dlt form a complex at the apical membrane of follicle cells that plays an important role in the formation and maintenance of the FE.

We thank Manzoor Bhat, Lisa Fessler, Mark Peifer, Graham Thomas, Tadashi Uemura, Eric Vieschaus, the Developmental Studies Hybridoma Bank, and the Bloomington Drosophila Stock Center for providing reagents. We thank Bart Kuss for his help with the initial characterization of Discs Lost. We are grateful to Dorothea Godt for stimulating discussion, help with the dissection of oskar ovaries, and for providing 4 C. We thank Dorothea Godt for critical reading of the manuscript.

This work was supported by a grant of the Medical Research Council of Canada (U. Tepass), and by the National Cancer Institute of Canada with funds from the Canadian Cancer Society (J. McGlade).