Multiple regions of Crumbs3 are required for tight junction formation in MCF10A cells

The defining characteristic of cell polarity is the asymmetric distribution of proteins and macromolecules into distinct domains within a cell. This property is crucial for many cellular processes, including cell migration, asymmetric division of stem cells, and the specialized functions of neurons and epithelial cells. Genetic studies in Caenorhabditis elegans and Drosophila have identified a set of evolutionarily conserved proteins involved in the determination of epithelial cell polarity. Among these conserved proteins are two complexes known in mammals as the Par6/Par3/atypical protein kinase C (aPKC) complex (or Par6 complex) and the Crumbs/protein associated with Lin Seven-1 (PALS1)/PALS1 associated tight junction protein (PATJ) complex (or Crumbs complex) (for reviews, see Macara, 2004; Roh and Margolis, 2003). The Par6 complex is made up of the PDZ domain-containing proteins Par6 and Par3, along with atypical protein kinase C. The Crumbs complex is made up of the transmembrane protein Crumbs, the MAGUK protein PALS1 and the multiple PDZ domain-containing protein PATJ. In this complex, PALS1 acts as an adaptor linking Crumbs and PATJ; the PDZ domain of PALS1 binds directly to the C-terminal tail of Crumbs, whereas the L27 domain of PALS1 binds the L27 domain of PATJ (Roh et al., 2002). It has been shown that in Drosophila both the Par6 and Crumbs protein complexes are required for the development of apical-basal polarity in embryonic epithelia, and genetic evidence suggests that both complexes function together in a common pathway (Bilder et al., 2003; Tanentzapf and Tepass, 2003). More recently it was shown that the Crumbs and Par6 complexes also act cooperatively to regulate morphogenesis of the photoreceptor apical membrane domains in Drosophila (Hong et al., 2003; Nam and Choi, 2003).

Three isoforms of mammalian Crumbs have been identified, and two have been characterized. Crumbs1 (CRB1) is expressed primarily in the central nervous system (den Hollander et al., 2002; den Hollander et al., 1999), whereas Crumbs3 (CRB3) is an epithelia-specific isoform (Lemmers et al., 2004; Makarova et al., 2003). Both the CRB3 and Par6 complexes have been localized to the tight junctions of mammalian epithelial cells (Izumi et al., 1998; Joberty et al., 2000; Roh et al., 2003). Tight junctions are specialized apical junctional complexes that serve to form a tight seal between adjoining epithelial cells and function as a selective barrier to paracellular diffusion through the intercellular space. Tight junctions also serve to restrict intramembrane diffusion of proteins and lipids between the apical and basolateral domains, and thus are essential to the maintenance of epithelial cell polarity (Schneeberger and Lynch, 2004; Tsukita et al., 2001).

Numerous studies now support a role for the proteins of the Par6 complex in the development of epithelial tight junctions (Gao et al., 2002; Hirose et al., 2002; Joberty et al., 2000; Mizuno et al., 2003; Suzuki et al., 2002; Suzuki et al., 2001). However, comparatively little is known of the role of the Crumbs complex in epithelial tight junction formation. We have recently shown that the Crumbs/PALS1/PATJ complex is physically linked to the Par6 complex through a direct interaction between PALS1 and Par6 (Hurd et al., 2003). This suggests that the Crumbs and Par6 complexes may act together to regulate tight junction formation. Indeed, we and others have found that overexpression of CRB3 in MDCKII cells delays the formation of tight junctions in a calcium-switch assay (Lemmers et al., 2004; Roh et al., 2003). This effect was found to be dependent upon the C-terminal PDZ binding motif of CRB3; a CRB3 mutant that lacked the C-terminal PDZ binding motif, and thus the ability to bind the PDZ protein PALS1, was unable to affect tight junction formation. This finding implicated PALS1, and potentially the Par6 complex, as an important downstream mediator of CRB3 function in tight junction regulation. Additional evidence that the CRB3 complex is important in tight junction regulation came from a recent study in which RNAi-mediated suppression of PALS1 expression in MDCKII cells also led to defects in tight junction formation (Straight et al., 2004).

In this paper we present new evidence that the CRB3 protein plays a key role in the development of epithelial tight junctions. Mammary epithelial MCF10A cells express little endogenous CRB3, and are unable to form tight junctions when grown under standard tissue culture conditions. We find that exogenous expression of CRB3 is sufficient for the development of functional tight junctions in these cells. Mutations in both the CRB3 C-terminal PDZ binding motif, and the putative FERM binding motif compromise the ability of CRB3 to induce tight junctions. We propose that CRB3 plays an essential role in tight junction formation, and that interaction with both PALS1 and an unknown FERM protein are required for this process.

The rabbit polyclonal antibodies against CRB3, PALS1 and PATJ have been previously described and were used for immunostaining and western blotting (Makarova et al., 2003; Roh et al., 2002). The mouse monoclonal 9E10 anti-myc antibody was used for all myc immunoprecipitations. The mouse monoclonal 4A6 anti-myc antibody (Upstate Biotechnology, Lake Placid, NY) was used for all myc immunostaining and western blotting. Immunostaining and western blotting were performed with the following antibodies: mouse or rabbit anti-ZO-1 (Zymed Laboratories, San Francisco, CA), mouse anti-β-catenin (BD Transduction Laboratories), rabbit anti-claudin-1 (Zymed Laboratories) and rabbit anti-occludin (Zymed Laboratories). Secondary anti-mouse and anti-rabbit antibodies conjugated to Alexa 488 and 594 fluorochromes (Molecular Probes, Eugene, OR) were used in immunostaining experiments. Nuclei were visualized with 4′,6-diamidino-2-phenylindole (DAPI).

MCF-7, T47D, SKBr3, MDA-MB-231 and MDCKII cells were cultured in DMEM plus 10% fetal bovine serum supplemented with penicillin, streptomycin and L-glutamine. MCF10A cells were cultured essentially as described (Debnath et al., 2003) and grown in DMEM/F12 plus 5% horse serum supplemented with penicillin, streptomycin, L-glutamine, 20 ng/ml epidermal growth factor, 0.5 μg/ml hydrocortisone, 10 μg/ml insulin and 100 ng/ml cholera toxin. All cells were maintained on standard tissue culture plastic in an 8% CO₂ humidified incubator at 37°C. For immunostaining experiments, cells were seeded onto permeable 24-mm Transwell clear polyester filters, 0.4 μm pore size (Corning, NY), and allowed to grow until confluency. Thereafter, media was changed every day until desired time points were reached.

Cell monolayers grown on permeable Transwell filters were fixed in 4% paraformaldehyde and permeabilized with 1% SDS in PBS for 10 minutes at room temperature, except for samples immunostained for mycCRB3 with the myc 4A6 antibody. These cells were instead permeabilized with 0.1% Triton X-100 in PBS to visualize CRB3 expression. Samples were then washed with PBS and immunostained as previously described (Straight et al., 2004). Regular fluorescence microscopy was performed with an inverted Leica DM IRB fluorescent microscope and data analyzed with SPOT image software (Diagnostic Instruments). Confocal images were taken with a Zeiss lSM 510 Axiovert 100 M inverted confocal microscope or with an Olympus FV500 inverted confocal microscope. Confocal images were analyzed with the Zeiss LSM 5 Image Browser software. All fluorescent images were prepared for publication with Adobe Photoshop and Adobe Illustrator software (San Jose, CA).

The png retroviral vector was generously provided by Steven Ethier (Barbara Ann Karmanos Cancer Institute, Detroit, MI). The mycCRB3-png, mycCRB3ΔERLI-png, mycCRB3mutFERM-png and mycCRB3mutFERMΔERLI-png retroviral constructs were produced by respective PCR amplification from the previously described mycCRB3-pSecTag, mycCRB3ΔERLI-pSecTag, mycCRB3FERMmut-pSecTag and mycCRB3FERMmutΔERLI-pSecTag constructs (Roh et al., 2003). The amplified mycCRB3 PCR products were cloned into the NotI/XhoI sites of the png retroviral vector. The mycCRB3mutRP-png construct was produced by PCR-directed site-specific mutagenesis using the mycCRB3-pSecTag vector as template. The mycCRB3mutRP sequence was then cloned into the NotI/XhoI sites of the png vector. The sequences of all constructs were verified by automated sequencing at the University of Michigan DNA Sequencing Core.

Amphotropic retroviruses were produced by transfection of DNA constructs into the retroviral producer Phoenix-Ampho cell line. Viral supernatants were collected 48 hours post-transfection, filtered through a 0.45 μm membrane, and either used immediately or stored at –80°C. MCF10A cells were infected with viral supernatants according to standard procedures as described (Debnath et al., 2003). 48 hours post-infection cells were passaged into selection media containing 1 μg/ml puromycin to select for cells stably expressing the retroviral vector. Cells were selected for 5-6 days before use, or until all mock-infected cells had died.

Cell lysates were prepared from confluent plates of cells as previously described (Straight et al., 2004). Protein concentrations of cell lysates were determined using the Bio-Rad Protein Assay (Bio-Rad Laboratories). For immunoprecipitations, 1 mg protein from each cell lysate sample was incubated overnight with 5 μl purified antibody at 4°C. The next day samples were incubated with 60 μl of 50% slurry protein A-sepharose (Zymed) for an additional 1-2 hours, then washed three times with wash buffer (50 mM HEPES, pH 7.5, 150 mM NaCl, 0.1% Triton X-100 and 10% glycerol) and finally resuspended in LDS loading buffer (Invitrogen). Resolution of samples by gel electrophoresis was performed using the NuPage Novex gel system (Invitrogen) as previously reported (Straight et al., 2004). Gels were transferred to nitrocellulose membranes and immunoblotted as previously described (Straight et al., 2004). Equal loading was confirmed by Ponceau S staining.

Cells on Transwell filters were fixed with 2.5% glutaraldehyde in 0.1 M Sorensen's buffer pH 7.4 containing 0.1% Ruthenium Red for 30 minutes at 4°C. After washing twice with Sorensen's buffer, the filters were post-fixed in 1% osmium tetroxide in Sorensen's buffer for 15 minutes at 4°C. After two further washes with Sorensen's buffer, cells were stained with 3% uranyl acetate in water for 15 minutes at room temperature and dehydrated in ethanol/water mixtures. The filters were embedded in LX112 resin (Ladd Research). Ultrathin sections (50-70 nm) were placed on carbon/Formvar-coated copper grids and stained post sectioning with uranyl acetate and Reynold's lead citrate. The grids were imaged under a Philip CM100 electron microscope.

Cells were grown on permeable Transwell filters, and transepithelial electrical resistance (TER) measurements were determined with a Millicell-ERS volt-ohm meter (Millipore, Billerica, MA) according to manufacturer's instructions. Background resistance was determined using cell-free filters. Samples for each time point were measured in triplicate.

For quantification purposes, a tight junction structure was defined as a completely enclosed ring of smooth, contiguous, apical ZO-1 staining. Cell samples were immunostained for ZO-1 and examined with a Leica DM IRB fluorescent microscope. For each cell sample, the number of tight junction structures were counted in three different fields under 20 × magnification (>100 cells per field) and the mean number of tight junction structures per field was determined. Experiments were performed three times and the mean number of tight junction structures per field from three independent experiments was calculated. Data were plotted using the GraphPad Prism data analysis software (GraphPad Software). Error bars represent s.e.m.

The mammalian CRB3 protein is expressed in a variety of epithelial tissues, and has been studied primarily in the Madin-Darby canine kidney (MDCK) epithelial cell line (Lemmers et al., 2004; Makarova et al., 2003; Roh et al., 2003). We used immunoprecipitation and western blotting to examine the expression of endogenous CRB3 in several human mammary epithelial cell lines (Fig. 1A). We found that CRB3 is expressed at very low levels in the non-tumorigenic MCF10A cells, but at high levels in several other cell lines, including the MCF7 cell line. In agreement with previous reports, we found that MCF7 cells form well-defined tight junctions in tissue culture (van Deurs et al., 1987). However, we found that MCF10A cells were unable to form tight junctions in monolayer culture. Staining in MCF10A cells for the tight junction marker ZO-1 revealed fragmented staining at apical junctions, as opposed to the smooth, continuous staining seen in both MCF7 and MDCKII cells (Fig. 1B and data not shown). For these cell lines at least, we find a correlation between expression levels of CRB3 and the formation of well-defined tight junctions.

We wished to determine the effect of exogenous CRB3 expression in MCF10A cells. Thus, we constructed a retrovirus expressing myc-tagged human CRB3 and used this to infect MCF10A cells. Pools of stably expressing cells were selected. These pools overexpressed CRB3 relative to the wild-type cells (Fig. 2). Immunofluorescent analysis showed heterogeneous CRB3 expression, with more than 80% of cells showing detectable mycCRB3 expression (see Fig. 6 and data not shown).

Cells expressing mycCRB3 as well as cells infected with the control png retroviral vector (png cells) were seeded onto permeable membranes and grown for several days past confluence. By 2 days post confluent, a dramatic difference in tight junction formation could be seen. In the mycCRB3-expressing pools, patches of cells formed in which ZO-1 staining exhibited the smooth, contiguous `honeycomb' pattern of apical staining, which is a hallmark of tight junctions. As cells were grown longer in culture, these patches of smooth ZO-1 staining expanded (Fig. 3A). We noted that unlike the tight junctions seen in MDCK cells, the tight junction structures induced in MCF10A cells were of irregular shape, and often appeared to encircle multiple cells at the apical epithelial layer (Fig. 3A-C). Staining of MCF10A cells for the adherens junction protein β-catenin showed that β-catenin delimits the basal lateral membranes of both individual control and mycCRB3-expressing cells, showing that a syncytium has not been formed (Fig. 3B). Similar results were seen with staining for the adherens junction marker protein, E-cadherin (data not shown). These results suggest that adherens junctions form properly in wild-type MCF10A cells and are not disturbed by CRB3 overexpression. Further investigations revealed that the appearance of irregularly shaped tight junctions encircling multiple nuclei appears to arise from two factors. These insights came from examining additional Z sections of the cells as well as transmission electron microscopy. First, the MCF-10A cells are irregularly shaped with the apical surface often larger or smaller than the basal surface. This leads to a misalignment of nuclei and tight junctions when viewed on X-Y sections. Second, MCF-10A often grow in multilayers on filters leading to the appearance of multiple nuclei per cell on X-Y sections when actually one cell is sitting on top another.

We co-stained mycCRB3-expressing MCF10A cells for ZO-1 and a variety of known tight junction marker proteins. PALS1 and PATJ are cytoplasmic CRB3-associated proteins, which are specifically localized to tight junctions in epithelial cells (Lemmers et al., 2002; Roh et al., 2002). Claudin-1 and occludin are integral membrane proteins that constitute the intermembrane strands of the tight junction complex; the claudin proteins are believed to form pores that directly mediate the selective permeability properties of tight junctions (Schneeberger and Lynch, 2004). PALS1 and PATJ display a diffuse, cytoplasmic staining pattern in control MCF10A cells, whereas claudin-1 and occludin display a fragmented apical junctional staining (Fig. 4A). Upon expression of exogenous CRB3, all four proteins were recruited to tight junction structures containing ZO-1, and all displayed a smooth, continuous staining pattern (Fig. 4A). Western blotting experiments showed that the introduction of exogenous CRB3 does not alter the total expression levels of these tight junction proteins, or of ZO-1 (Fig. 4B).

To confirm the formation of tight junctions, we further analyzed our cell lines by transmission electron microscopy (TEM). Fig. 5A shows a mycCRB3-expressing cell displaying a tight junction (arrow) and desmosome (asterisk). Tight junctions were not detected in TEM analysis of control MCF10A cells. We also performed measurements of transepithelial electrical resistance (TER) to determine the functionality of the tight junctions. Immunofluorescence analysis had shown that the number of tight junctions formed by mycCRB3-expressing MCF10A cells increases with time in culture. Our TER measurements are in agreement with this, showing that the TER of cells expressing exogenous CRB3 increases with time, whereas the TER of control cells is significantly lower and does not increase (Fig. 5B).

We attempted to correlate CRB3 expression in individual cells with the formation of tight junctions. In MDCK II cells, endogenous CRB3 is localized to both tight junctions and throughout the apical plasma membrane. Overexpression of CRB3 in MDCKII cells also leads to expression at lateral membranes (Roh et al., 2003). We found that endogenous CRB3 in MCF10A cells is not detectable by immunofluorescence with our CRB3 antibody, probably due to its low expression level (data not shown). However, both our CRB3 antibody and an antibody against the myc epitope recognize the exogenously expressed mycCRB3 construct in MCF10A cells. Expression of mycCRB3 is heterogeneous in our cell pools; in highly expressing cells, mycCRB3 is localized to the entire cell membrane and throughout the cytoplasm (Fig. 6 and data not shown). Interestingly, there did not appear to be a strict correlation between CRB3 expression levels and tight junction formation in individual cells; tight junctions could be seen surrounding even those expressing very low or undetectable amounts of CRB3 (Fig. 6, arrows). We speculate that low levels of CRB3 expression are sufficient to induce the tight junction structures or that expression of CRB3 in one cell can induce tight junctions in neighboring cells. Indeed, higher levels may even be detrimental to tight junction formation, as overexpression of CRB3 in MDCK cells has been shown to interfere with tight junction development (Lemmers et al., 2004; Roh et al., 2003).

We undertook a mutagenesis study to determine the sequences of CRB3 that are required for the induction of tight junctions. The CRB3 intracellular sequence contains two motifs that are conserved with Drosophila Crumbs and all identified Crumbs isoforms: a juxtamembrane FERM binding motif with similarity to that found in glycophorin C, and a C-terminal PDZ binding motif, `ERLI' (Izaddoost et al., 2002; Klebes and Knust, 2000; Makarova et al., 2003). A comparison of the CRB3 sequence across several mammalian species (mouse, human, dog, pig and cow) also revealed an intracellular conserved `RxPPxP' sequence, which is specific to the CRB3 isoform and has similarity to an SH3 domain binding site (see Fig. S1 in supplementary material). We generated specific mutations in these three motifs (Fig. 7A) and expressed these CRB3 mutants as myc-epitope tagged proteins from retroviral vectors. These retroviral constructs were used to infect MCF10A cells and stably expressing pools were selected. Western blotting of these pools showed that all mycCRB3 mutants were expressed at the same level as wild-type mycCRB3 (Fig. 7B). Immunofluorescence analysis showed comparative levels of protein expression in individual cells, and also similar cellular localization (data not shown). As expected, deletion of the ERLI PDZ binding motif abolished interaction between CRB3 and PALS1 in MCF10A cells. Mutations in the FERM binding motif and RxPPxP motif, however, had no effect on CRB3 binding to PALS1 in these cells (Fig. 7C).

MCF10A cells expressing the different mycCRB3 constructs were seeded onto permeable membranes and grown for 4 days past confluence as before. Staining for the ZO-1 protein revealed a marked difference in tight junction formation between the different cell lines. Point mutations in the RxPPxP motif did not affect the ability of CRB3 to induce tight junction structures in MCF10A cells; however, loss of the ERLI sequence, or point mutations in the FERM binding motif, severely compromised the ability of CRB3 to induce tight junctions (Fig. 8A-C). Mutations in the FERM binding motif appeared to result in a more severe defect than loss of the PDZ binding motif, as tight junctions were still sometimes seen in the mycCRB3delERLI cells (Fig. 8A, arrows). In order to verify this, we quantified the number of tight junction structures induced by the different CRB3 constructs. A tight junction structure was defined as an enclosed ring of contiguous apical ZO-1 staining; Fig. 8B shows the average number of such tight junction structures formed per field in 4 days post-confluent cells. These quantitative results confirm that the mycCRB3ΔERLI protein retains a partial, if weak, ability to induce tight junctions above levels seen in the vector control MCF10A cells. In contrast, the mycCRB3mutFERM protein is almost completely unable to induce tight junctions at this time point.

We have noted that the number of tight junction formed in MCF10A cells increases with time in culture. Therefore, we cultured our cells for 9 days post confluence and compared the mycCRB3ΔERLI and myCRB3mutFERM cell lines at this time point (Fig. 8C). After 9 days, patches of ZO-1-positive tight junction-like structures have expanded to cover the entire sample of wild-type mycCRB3-expressing cells. However, these structures continue to be very rare in the control vector MCF10A cells, even after 21 days in culture (Fig. 8C and data not shown). By 9 days post confluence a small number of tight junction structures could be seen forming in the mycCRB3mutFERM cells. However, more tight junction structures as defined by ZO-1 staining had formed in the mycCRB3ΔERLI cells. These results are consistent with those seen at the earlier 4-day time point, and show that both the CRB3ΔERLI and CRB3mutFERM proteins have partial, albeit weak ability to induce ZO-1-containing junctions. In contrast, a mycCRB3 construct which lacked both the ERLI sequence and contained mutations in the FERM binding motif was completely unable to induce these junctions in MCF10A cells, even after 9 days post-confluent growth (Fig. 8C).

In this study we report that expression of exogenous CRB3 protein is sufficient to induce the formation of tight junctions in mammary epithelial MCF10A cells. These tight junction structures are marked by the smooth, contiguous staining of a variety of tight junction marker proteins, including claudin-1 and occludin, in a typical honeycomb-like apical staining pattern. This is in strong contrast to control MCF10A cells, in which tight junction marker proteins display either a fragmented junctional or diffuse cytoplasmic staining pattern. Analysis by transmission electron microscopy and measurements of transepithelial electrical resistance confirm the formation of functional tight junctions. Mutations in either the ERLI PDZ binding motif or the FERM binding motif of CRB3 compromise the ability of CRB3 to induce tight junction formation. These results indicate that CRB3 probably plays an important role in the organization of epithelial tight junctions.

How may CRB3 act to promote the formation of tight junctions? Numerous studies now support a critical role for the proteins of the Par6 complex in the development of epithelial tight junctions. One possibility is that CRB3 acts to recruit and/or stabilize the Par6 complex at the developing tight junction. Genetic evidence in Drosophila suggests that the proteins of the Crumbs and Par6 complexes are mutually dependent upon one another for proper targeting and stabilization at the subapical domain (SAC), or marginal zone, a domain in Drosophila epithelia that is spatially analogous to the mammalian tight junction (Bilder et al., 2003; Tanentzapf and Tepass, 2003). We have recently shown that the mammalian Crumbs complex is physically linked to the Par6 complex through a direct interaction between PALS1 and Par6 (Hurd et al., 2003). In MDCK cells, mislocalization of PALS1 or inhibition of its expression both result in the mislocalization of aPKCζ, a member of the Par6 complex, away from tight junctions (Hurd et al., 2003; Straight et al., 2004). Defects in tight junction development also result. It may be that CRB3 is required to recruit/stabilize PALS1 at tight junctions, which in turn leads to the recruitment and stabilization of the Par6 complex there. This hypothesis is consistent with genetic evidence in Drosophila that Crumbs, Stardust (the Drosophila PALS1 homologue), and PATJ are all mutually dependent upon one another for proper localization to the subapical complex (Bachmann et al., 2001; Hong et al., 2001; Tepass et al., 2001).

Our mutagenesis experiments support the idea that CRB3 binding to PALS1 is crucial to its ability to promote tight junction formation. Loss of the conserved PALS1 binding ERLI sequence severely compromises the ability of CRB3 to induce tight junctions. In addition to binding Par6, PALS1 also mediates association between CRB3 and PATJ. PATJ contains multiple PDZ domains and binding partners for many of these domains remain unidentified. It is possible that PATJ, too, may act downstream of PALS1 to regulate tight junction formation through as yet unknown effectors. A recent study reported that the Drosophila homologue of PATJ can interact directly with Drosophila Par6, potentially providing another mechanism by which PALS1 may indirectly promote association of CRB3 with the Par6 complex (Nam and Choi, 2003). Moreover, another recent study has reported that the ERLI sequence of CRB3 may itself bind directly to Par6 (Lemmers et al., 2004). Thus, the conserved ERLI motif may link CRB3 to the Par6 complex through multiple mechanisms.

Although loss of the ERLI sequence resulted in defects in the ability of CRB3 to induce tight junctions, we observed that the CRB3delERLI protein retained a partial ability to induce tight junction formation. This suggests that other domains in CRB3 are also involved in tight junction formation. We found that mutations in a conserved `RxPPxP' motif did not affect the ability of CRB3 to induce tight junctions. However, mutations in the FERM binding motif severely disturbed the ability of CRB3 to promote tight junction formation and led to a more severe loss of function than did loss of the ERLI sequence. The CRB3mutFERM protein retained the ability to interact with PALS1, suggesting that a PALS1-independent mechanism is required for tight junction formation. This is supported by experiments in which we have found that overexpression of PALS1 alone in MCF10A cells was not sufficient to induce tight junction formation (data not shown).

The FERM binding motif of CRB3 is expected to interact with a protein containing a FERM domain. The FERM domain is a protein-protein interaction domain found in a diverse set of proteins, many of which function as adaptors that link transmembrane proteins to the cortical actin cytoskeleton (Chishti et al., 1998). Such a FERM protein may be involved in linking CRB3 to the actin cytoskeleton, and such linkage may stabilize the CRB3 complex at the plasma membrane. Additionally, recruitment and reorganization of cytoskeletal elements through the FERM protein may be involved in tight junction formation. At this moment, the FERM protein that binds to the Crumbs FERM binding motif is still unknown. In Drosophila embryos and S2 cells, a physical interaction has been reported between Crumbs and the FERM-containing protein dMoesin, a protein that has homology to the mammalian ezrin/radixin/moesin family of ERM proteins (Medina et al., 2002). However, definitive evidence that dMoesin functions in the Crumbs polarity pathway is still lacking. An interesting new study identified Mosaic Eyes, or Moe (not to be confused with moesin) as a novel FERM protein required for tight junction formation in the zebrafish retinal pigmented epithelium (Jensen and Westerfield, 2004). Genetic experiments in zebrafish suggest that Moe may act in the same pathway as Crumbs. However, a physical interaction between these proteins has not been confirmed. In our own studies, we have found that the FERM binding domain of CRB3 is able to interact with several different FERM proteins in vitro (our unpublished data). Thus the identification of the functional FERM protein partner of Crumbs remains an important challenge.

Previous experiments in MDCKII cells have suggested that CRB3 acts to regulate epithelial tight junction formation. Overexpression of CRB3 in MDCKII cells leads to a delay in tight junction formation in the calcium switch assay. This phenotype was dependent on the presence of the ERLI sequence: a CRB3 construct that lacked the ERLI sequence was unable to delay tight junction formation when overexpressed in these cells (Lemmers et al., 2004; Roh et al., 2003). However, mutations in the FERM binding motif did not affect the ability of overexpressed CRB3 to disturb tight junction formation in MDCKII cells (Roh et al., 2003). Our current study thus provides the first evidence that the CRB3 FERM binding motif is involved in tight junction formation. Experiments in Drosophila have shown the importance of the FERM binding motif in epithelial cell polarization, and have also shown that the Crumbs FERM and PDZ binding motifs can act separately to regulate distinct processes. Both the FERM binding and PDZ binding motifs are required for the rescue of epithelial cell polarity in Drosophila embryos mutant for the crumbs gene. Crumbs overexpression as well as underexpression also leads to defects in epithelial cell polarity in Drosophila; however, the FERM binding motif, but not the ERLI sequence, is dispensable for the overexpression phenotype (Klebes and Knust, 2000). Recently, the function of Crumbs has also been studied in the development of Drosophila photoreceptors. Overexpression in Drosophila photoreceptors of a Crumbs construct lacking the ERLI sequence but retaining the FERM binding domain led to ectopic localization of adherens junctions, but did not alter the localization of the Drosophila PATJ homologue. On the other hand, overexpression of a Crumbs construct containing the ERLI sequence but lacking the FERM binding motif led to mislocalization of PATJ, but not the mislocalization of adherens junctions (Izaddoost et al., 2002).

It is not known whether the FERM and PDZ binding motifs act through completely parallel signaling pathways that are able to partially compensate for one another, or if some level of cross-talk exists between pathways that may partially compensate for mutations in either motif. In this respect, it is interesting to note that PALS1 also contains a FERM binding motif (Kamberov et al., 2000), although PALS1 interaction with a FERM protein has not been demonstrated. FERM domains may simultaneously interact with multiple partners (Han et al., 2000); it is therefore possible that a FERM protein may simultaneously interact with both CRB3 and PALS1 and act to help stabilize the complex in vivo. Such a FERM protein may act as a bridge to recruit a low level of PALS1 into the complex even in the absence of the CRB3 ERLI sequence. Similarly, PALS1 may act to recruit a small amount of FERM protein into the complex even if CRB3 itself contains mutations in its FERM binding motif. Such interactions may explain the partial activities we see in our CRB3 mutants. We found that loss of both the PDZ binding and FERM binding motifs led to a complete loss of the ability of CRB3 to induce tight junction formation in our MCF10A cell system.

The formation of tight junctions is intimately involved with the development and maintenance of epithelial cell polarity. When cultured in a three-dimensional matrix of collagen or reconstituted basement membrane, MDCKII cells are able to form multicellular cysts with defined lumens and apico-basal polarity (O'Brien et al., 2001). We and others have shown that overexpression of CRB3 or PALS1 in MDCKII cells disrupts the polarization of such cysts, and leads to defects in lumen formation (Lemmers et al., 2004; Straight et al., 2004). Mammary MCF10A cells also form hollow, polarized cysts when cultured in a three-dimensional matrix of reconstituted basement membrane (Debnath et al., 2003; Muthuswamy et al., 2001). However, we have found that overexpression of CRB3 appears to have little effect on the development and polarization of MCF10A cysts (our unpublished data). Immunostaining for ZO-1 and other tight junction markers in MCF10A cysts has not been reported, and our own attempts have been complicated by technical difficulties; therefore, the effect of exogenous CRB3 expression on tight junction formation in three-dimensional MCF10A cyst structures cannot be assessed.

In summary, we have found that expression of exogenous CRB3 is sufficient to induce tight junction formation in a cell line that normally lacks tight junctions. When grown in monolayer culture, wild-type MCF10A cells express some tight junction marker proteins such as ZO-1, claudin-1 and occludin at the apical/basolateral boundary; however, these proteins appear only in a fragmented staining pattern. This fragmented staining may represent nascent tight junction structures in MCF10A cells. Expression of exogenous CRB3 leads to a coalescence of such nascent structures into the smooth, contiguous, apical structures typical of tight junctions. This leads us to conclude that Crumbs3 expression appears crucial for proper tight junction formation.

We thank the members of our laboratory for helpful discussions. We give special thanks to Sam Straight for the alignment of mammalian CRB3 homologues and assistance with figure preparation. We thank Steven Ethier for the kind gift of the retroviral png vector, and we thank Kathy Ignatoski for advice on MCF10A culture and retroviral infections. We also thank the University of Michigan Microscopy and Image Analysis Laboratory and the University of Michigan Diabetes Morphology and Imaging Core for the use of their confocal microscopes. V.F. was supported by the National Institutes of Health Postdoctoral Training Grant in Organogenesis 5-T32-HD07505-06. This work was also supported by NIH grant DK69605. B.M. is an investigator of the Howard Hughes Medical Institute.