A Conserved Functional Domain of Drosophila Coracle Is Required for Localization at the Septate Junction and Has Membrane-organizing Activity

All fly stocks were maintained on standard corn meal, yeast, molasses, agar medium. Canton S was used as the wild-type stock. w¹¹¹⁸ flies were used for the transformation of coracle transgenes using the mini-w⁺ vector pCaSpeR-hs (Thummel et al., 1988). The coracle alleles used in this study are described elsewhere. Nrx mutant fly stocks (Nrx⁴⁶, Nrx⁴⁰²⁵, Nrx⁴³⁰⁴, and Nrx¹⁴) were described previously (Baumgartner et al., 1996). Df(3L) vin⁷ was obtained from the Bloomington Drosophila stock center at Indiana University (Bloomington, IN).

Genomic DNA was obtained using standard methods from flies homozygous for either cor⁴ or cor⁶, which had been rescued to adulthood by ubiquitous expression of a coracle cDNA transgene (Fehon et al., 1994). coracle genomic DNA was amplified using intron-specific coracle primers. The resulting PCR products were sequenced using the AmpliCycle sequence kit (Perkin Elmer Corp., Branchburg, NJ). The sequence was verified on both strands.

The transformation vector pCaSpeR-hs (Thummel et al., 1988) was modified to include an amino-terminal myc epitope by annealing the two primers consmyc S (sense) 5′-AAT TCA CCA TGG AGC AAA AGC TCA TTT CTG AAG AGG ACT TGA GGC CTA A and consmyc A (antisense) 5′-GAT CTT AGG CCT CAA GTC CTC TTC AGA AAT GAG CTT TTG CTC CAT GGT G, which produce a duplex with overhanging EcoRI and BglII ends. This fragment was subsequently cloned into an EcoRI/BamHI cut pCaSpeR-hs vector (all restriction enzymes were obtained from New England Biolabs Inc., Beverly, MA). Digestion of the modified vector with StuI allowed PCR generated coracle transgenes to be cloned in-frame immediately downstream of the myc epitope.

The following coracle transgenes were generated by PCR amplification from a full length coracle cDNA clone (Fehon et al., 1994): cor 1–383 (bp 370–1,516), cor 1–323 (bp 370–1,337), cor 1–167 (bp 370–868), cor 106–167 (bp 685–868), cor 106–383 (bp 685–1,516), cor 207–383 (bp 988–1,516), cor 378–1698 (bp 1,501–5,544), cor 378–1591 (bp 1,501–5,141), cor 1585–1698 (bp 5,122–5,544). The sequence of all PCR amplified regions was confirmed by sequencing using standard methods (Fehon et al., 1991).

To specifically immunoprecipitate cor⁴ mutant protein from a large collection of embryos, a transformation construct providing genetic rescue of coracle mutations that lacked the epitope recognized by the guinea pig anti-Coracle antibody was constructed by fusing the Drosophila ubiquitin promoter (Lee et al., 1988) to the cDNA2 isoform of coracle (Fehon et al., 1994) and placing this expression cassette within pCaSperR-4 (Thummel et al., 1988). For this construct, a 1.3-kb EcoRI/PvuII fragment representing a portion of the 3′ end of cDNA2 was ligated into an EcoRI/PvuII (partial) cut pRmHa-3 construct containing the 3′ end of coracle. The resulting plasmid was digested with EcoRI and XbaI to excise a 2.8-kb fragment which contained the 3′ half of cDNA2 and the ADH polyadenylation signal. This fragment was ligated along with an 885-bp SacI/EcoRI fragment containing the 5′ half of cDNA2 into a SacI/XbaI digested RHXpHSS7-pUp2 vector. A 6.1-kb NotI fragment containing the ubiquitin promoter, the coracle cDNA2 coding sequence and the ADH polyadenylation signal was then excised and subcloned into a NotI digested pCaSpeR-4 vector. Transformation of all these constructs was performed as described (Rebay et al., 1993). For each construct from 4–10 independent transformant lines were established.

The maintenance, transfection, induction, and immunofluorescent analysis of Schneider 2 (S2) cells were performed as described (Fehon et al., 1990). The heat shock regimen consisted of a 30-min heat shock at 38°C followed by 2 h recovery at 25°C.

A DLG fusion protein was generated by cloning a PCR-generated fragment of nucleotides 2,653–3,360 comprising amino acids 757–993 (Woods and Bryant, 1991) into SmaI cut pGEX 2 (Smith and Johnson, 1988; Pharmacia Diagnostics AB, Uppsala, Sweden). The fusion protein was purified as described (Frorath et al., 1991). Polyclonal antiserum was raised against the DLG fusion protein in mice. Mouse monoclonal antiserum against the myc epitope was obtained as monoclonal supernatant from cell line 1-9E10.2 (Evan et al., 1985) using standard procedures (Harlow and Lane, 1988). Guinea pig polyclonal antiserum against Coracle has been previously described (Fehon et al., 1994). Rabbit polyclonal antiserum against NRX was provided by H. Bellen (Baylor College of Medicine, Baylor, TX) and has been previously described (Baumgartner et al., 1996). All secondary antibodies were obtained from Jackson ImmunoResearch Laboratories (West Grove, PA).

Mutant embryos were unambiguously identified either using marked balancers that express β-gal driven by the ftz promoter or by counterstaining the embryos with antibodies against Coracle or NRX. Immunostaining of embryos was performed as described (Fehon et al., 1991). For the genetic rescue of NRX localization, wild-type embryos or embryos of the genotypes cor⁵; hscor 1–383 or cor⁵; hscor 378–1698 were collected for 2 h at 25°C, aged for 4 h at 25°C, heat shocked for 1 h at 38°C, and then allowed to recover for 7 h at 25°C. The embryos were subsequently fixed and stained as described above. A Zeiss LSM410 laser scanning confocal microscope with a krypton/argon laser (Carl Zeiss Inc., Thornwood, NY) was used for the confocal microscopy. Antibodies were used at the following dilutions: anti-Coracle 1:10,000; anti-DLG 1:250; anti-myc 1:100; and anti-NRX 1:5,000. Cuticle preparations were mounted in Hoyer's medium (Ashburner, 1989).

Wild-type and cor⁴ mutant embryos ubiquitously expressing the cDNA2 isoform of Coracle (Fehon et al., 1994) were collected for 3 h at 25°C, and then aged for ∼12–14 h at room temperature. Transgenic embryos expressing Cor 1–383 or Cor 378–1698 were collected for 3 h at 25°C, aged for ∼10–12 h at room temperature, and were then heat shocked for 1 h at 38°C followed by a 2 h recovery at 25°C. Embryos were then dechorionated in 50% commercial bleach (12% sodium hypoclorite) for 3 min and rinsed several times in ice-cold lysis buffer (20 mM Hepes, pH 7.0, 50 mM NaCl, 1 mM EDTA, 0.5 mM EGTA, 10 mM DTT, 1% Triton X-100). Embryos were homogenized in a 10× vol of ice-cold lysis buffer that included PMSF and leupeptin at a 1:1,000 dilution in a tight-fitting dounce homogenizer. The homogenate was drawn through a 27 gauge tuberculin needle 5–10 times, and then centrifuged at 16,000 g for 15 min at 4°C. After removing the upper lipid layer, the soluble lysate was collected and incubated for 30–60 min with primary antibody. Primary antibodies used were guinea pig anti-Coracle at 1:200 and anti-myc 1-9E10.2 at 1:1. Immunoreactive complexes were precipitated by centrifugation (8,000 g for 1 min at 4°C) following incubation with zysorbin (Zymed Labs Inc., South San Francisco, CA) for 30–60 min at 4°C. The complexes were washed 8–10 times in lysis buffer, and then boiled in 1× SDS sample buffer (Laemmli, 1970). Protein samples separated on 3–15% SDS-PAGE were transferred to nitrocellulose for 1 h at 100 V at 4°C. Blots were blocked in 5% nonfat milk in TBS plus 0.1% Tween-20 for 1 h at room temperature, and then incubated overnight at 4°C in primary antibody (anti-NRX at 1:10,000, anti-DLG at 1:5,000). After washing in TBS Tween-20 and incubation in blocking reagent with horseradish peroxidase–coupled secondary antibodies for 3–4 h at room temperature, the immunoreactive proteins were visualized using chemiluminescent autoradiography (Pierce, Rockford, IL).

A Coracle fusion protein was constructed by cloning the entire coracle cDNA3 sequence (Fehon et al., 1994) into the SmaI site of pGEX 2TK (Pharmacia Diagnostics AB). Purified Coracle fusion protein was prepared as described (Frangioni and Neel, 1993) with the exception that 1% Triton X-100 was used in place of sarkosyl. The glutathione-S-transferase (GST) moeity was removed by thrombin cleavage followed by adsorption to glutathione agarose (Sigma Chemical Co., St. Louis, MO). Purified Coracle was labeled with ³²P using γ-³²P dATP and the catalytic subunit of protein kinase (from bovine heart, Sigma Chemical Co.) according to manufacturers instructions. Labeled protein was purified from unincorporated nucleotides through a G25 column (Pharmacia Diagnostics AB) and fractions containing the highest concentration of TCA precipitable counts were used for the binding reactions. A Coracle fusion protein of amino acids 608–1,285 has been previously described (Fehon et al., 1994), and was purified as just described. A NRX fusion protein was constructed by cloning the cytoplasmic domain of NRX (bp 3,993–4,127) into the SmaI site of pGEX 3. NRX fusion protein was purified as just described. GST-NRX was covalently linked to CNBr-activated Sepharose (Pharmacia Diagnostics AB) at a concentration of ∼1 mg protein per ml beads and blocked in binding buffer (20 mM Hepes, pH 7.0, 50 mM NaCl, 1 mM EDTA, 0.5 mM EGTA, 5 mM DTT) plus 1% BSA for at least 2 h at 4°C. As a control, purified GST was covalently linked to CNBr-activated Sepharose (Pharmacia Diagnostics AB) at a concentration of ∼1 mg protein per ml beads. 100 μl of a 50% slurry of NRX or GST-linked beads (∼50 μg protein) was incubated with 5–10 μg of ³²P-labeled Coracle in 250–400 μl of binding buffer ± a 10-fold molar excess of unlabeled proteins for 30 min at room temperature with gentle shaking. After washing three times with binding buffer, the beads were added to 3 ml of scintillation fluid (Safety-Solve, Research Products International Corp., Mount Prospect, IL) and counted in a scintillation counter (LS6000SC; Beckman, Fullerton, CA). Three replicates were performed for each treatment and means plus standard errors were determined.

Coracle protein is normally localized to the septate junction in ectodermally derived epithelial cells (Fehon et al., 1994). Using optical microscopy, this localization is most obvious in the polarized epithelia of the salivary glands and hindgut (Fig. 1, B and C, respectively). To identify regions of Coracle responsible for this localization, we examined its subcellular localization in a collection of coracle point mutations with moderate to severe phenotypes. Two severe alleles, cor⁴ and cor⁶, produce Coracle protein which is mislocalized to the cytoplasm (Fig. 1, D–G). The Coracle protein in cor⁴ mutant embryos is expressed at nearly wild-type levels (Fig. 1), but with a molecular mass roughly 10 kD smaller than the wild-type protein (data not shown). Sequence analysis revealed that the molecular lesion responsible for this allele is an in-frame deletion of 78 amino acids (amino acids 68–146) in the conserved amino-terminal region (CNTR; Fig. 1,A). Analysis of the cor⁶ mutation showed that it is a missense mutation of Leu³⁹ to Gln in the CNTR (Fig. 1,A). Coracle protein in embryos homozygous for cor⁶ is expressed at much lower levels based upon immunofluorescence (Fig. 1) and by immunoblot (data not shown). These results, together with the observation that both cor⁴ and cor⁶ behave genetically as strong hypomorphs (Lamb, R.S., R.E. Ward, and R.G. Fehon, manuscript in preparation), suggest that correct subcellular localization is essential for Coracle function.

To address more definitively the role of the CNTR in subcellular localization, we initiated a structure/function analysis of Coracle by generating the coracle deletion mutations shown in Fig. 2. These constructs were transfected into S2 Drosophila cultured cells, heat shocked to induce expression, and analyzed by indirect immunofluorescence to determine the subcellular localization of the expressed proteins. Although S2 cells do not form an epithelium or display junctions, transiently transfected full-length Coracle is found predominately associated with the plasma membrane (Fig. 2,B), consistent with its membrane association in embryonic epithelia and in imaginal discs. Similarly, Cor 1–383, encoding the entire CNTR (Fehon et al., 1994), is also predominately associated with the plasma membrane, indicating that this region is sufficient for membrane localization in S2 cells (Fig. 2,C). To delimit the region responsible for this subcellular localization, we generated several smaller fragments of the amino-terminal region and expressed them in S2 cells. In every case, the resulting recombinant protein was found in the cytoplasm and showed no appreciable association with the plasma membrane (Fig. 2, D–G). Even Cor 1–323, which only removes 60 amino acids from the carboxyl terminus of the CNTR, was cytoplasmically localized (Fig. 2,D), indicating that an intact CNTR is necessary for correct subcellular localization. Consistent with this notion, Cor 378–1698, which contains all of Coracle except the CNTR, is also found throughout the cytoplasm as are all smaller fragments derived from this region (Fig. 2, H–J).

To assess the in vivo function of these recombinant proteins, we generated transgenic flies using P-element– mediated germline transformation and determined the subcellular localization of the induced proteins during embryogenesis by indirect immunofluorescence. In all cases, the subcellular localization of these recombinant proteins was consistent with our observations in the S2 cell assay. Specifically, the CNTR fragment, Cor 1–383, is associated with the membrane in the region of the apical junctional complex in a pattern that is very similar to the localization of Coracle expressed in wild-type embryos (compare Figs. 2,K to 1 B). Cor 1–383 displays the same pattern of localization in embryos homozygous for cor⁵ (data not shown), a null genetic background that has no detectable Coracle protein (Lamb., R.S., manuscript in preparation), indicating that sequences within the CNTR alone are responsible for this localization. Also consistent with the S2 cell assay, Cor 378–1698 in transgenic embryos is found throughout the cytoplasm and is excluded from the nucleus (Fig. 2 L). The expression levels of the smaller fragments derived from the CNTR were usually substantially reduced, although in all cases where the recombinant proteins were detected, they were found throughout the cytoplasm and not associated with the membrane (data not shown). Taken together, these experiments demonstrate that the amino-terminal 383 amino acids are both necessary and sufficient for correct subcellular localization in cultured S2 and embryonic epithelial cells.

Given the evidence that protein 4.1 localizes to the erythrocyte membrane through an interaction with the cytoplasmic tail of glycophorin C (Marchesi, 1985), it seems likely that Coracle localizes to the plasma membrane via an interaction with a septate junction resident transmembrane protein. Two recent results implicate NRX in this possible interaction with Coracle at the septate junction. First, Coracle localization is altered in Nrx mutant embryos (Baumgartner et al., 1996), and second, 8 of the 12 amino acid residues shown to be required for the binding of protein 4.1 to glycophorin C are conserved in NRX (Littleton et al., 1997). We therefore sought to examine further the interaction between Coracle and NRX through comparison of their genetic phenotypes, by more extensive investigation of the localization requirements for each protein, and through biochemical assays for direct physical interactions.

A previous study (Baumgartner et al., 1996) has reported defects in dorsal closure in severe Nrx alleles similar to those reported for coracle alleles (Fehon et al., 1994). To understand better the relationship between these genes, we compared the embryonic phenotypes of coracle and Nrx mutant embryos. Cuticle preparations of coracle and Nrx mutant embryos display similar dorsal closure defects (Fig. 3, A and B), although the penetrance of this phenotype is greater in coracle than in Nrx mutant embryos. The most penetrant phenotype shared between coracle and Nrx is a “weakening” of the cuticle in which the ventral denticle belts are faint, with finer, less opaque, and less abundant denticles. Additionally, both coracle and Nrx mutants display a very penetrant necrosis of the salivary glands (Fig. 3, C and D). Two other phenotypes reported for Nrx mutants (Baumgartner et al., 1996) are apparent in strong coracle alleles: the tracheae fail to inflate, and muscular contractions are severely diminished late in embryogenesis (data not shown).

Baumgartner and colleagues (1996) have recently reported that in embryos mutant for a strong Nrx allele, Coracle protein was no longer tightly localized to the septate junction. To confirm and extend this observation, we have examined the subcellular localization of Coracle in embryos homozygous for null (Df[3L]vin⁷) and strongly hypomorphic (Nrx⁴⁶, Nrx⁴⁰²⁵, and Nrx⁴³⁰⁴) Nrx alleles by indirect immunofluorescence. In all these cases, Coracle protein is mislocalized in the mutant embryos (Fig. 4). Rather than being tightly associated with the septate junction, the majority of the Coracle protein, although still membrane associated, has a more basolateral distribution (Fig. 4, compare D with G). In addition, some of the Coracle protein is distributed throughout the cytoplasm. Furthermore, the overall expression level of Coracle in the internal epithelia of the salivary gland and hindgut is considerably reduced relative to the expression in the epidermis. This mislocalization of Coracle is specific and does not reflect a gross alteration in the integrity of the septate junction, as judged by the nearly wild-type appearance of DLG in these embryos (compare Fig. 4 E with 4 H). This result suggests that the targeting of Coracle to the septate junction requires either a direct interaction with NRX or a NRX- dependent complex.

A previous study (Baumgartner et al., 1996) has indicated that in contrast to the dependence of Coracle localization on NRX expression, NRX does not depend on Coracle expression for its localization. However, the coracle allele used for this study, cor², expresses a protein with an intact CNTR that localizes to the cell membrane (Fehon et al., 1994) and retains partial function (Lamb, R.S., manuscript in preparation). It is, therefore, still possible that NRX is dependent on Coracle for proper subcellular localization. To address this question, we took advantage of recently isolated coracle alleles (Lamb, R.S., manuscript in preparation). In weaker coracle alleles, NRX protein targets correctly to the septate junction in embryonic salivary glands (Fig. 5,D). However, in null and strongly hypomorphic coracle alleles (cor⁵ and cor⁶, respectively), NRX extends basolaterally so that it is expressed along the entire lateral membrane (Figs. 5,F and 6,B). In addition, we have observed punctate cytoplasmic NRX staining in the salivary glands and in the epidermis of both weakly and strongly hypomorphic coracle embryos (Fig. 5, C and E). This effect on the localization of NRX is specific, as we find no detectable difference in the localization of DLG in coracle mutant embryos (data not shown).

Given the observations that Coracle function is required for maintenance of NRX localization at the septate junction and that the CNTR of Coracle is sufficient to correctly target Coracle to the septate junction, it is possible that sequences within the CNTR are sufficient for the maintenance of NRX localization. To test this idea, we induced expression of the CNTR in the cor⁵ null genetic background and examined the subcellular localization of NRX by indirect immunofluorescence. NRX protein is mislocalized along the entire basolateral membrane in cor⁵ mutant embryos (Fig. 6,B). In contrast, cor⁵ embryos that have been induced to express Cor 1–383 display NRX localization identical to that of wild-type embryos (Fig. 6, compare C with A). However, cor⁵ mutant embryos that express the remainder of Coracle (Cor 378–1698) fail to “rescue” the NRX localization defect and display a NRX distribution identical to cor⁵ mutant embryos (Fig. 6, compare D with B). Taken together, these results indicate that there is an interdependence of Coracle and NRX for proper localization to the septate junction, and that sequences within the CNTR of Coracle are essential for this function.

The observed codependence of Coracle and NRX for normal localization to the septate junction suggests that these proteins interact at the cytoplasmic face of the septate junction. We have addressed this idea in two ways, by asking if Coracle and NRX coprecipitate in extracts of embryonic tissues, and by examining the binding of Coracle and NRX in vitro. Consistent with the notion that they do interact, immunoprecipitation of Coracle from wild-type embryonic lysates results in a coprecipitation of NRX (Fig. 7,A). This coprecipitation is specific to Coracle and NRX, because DLG, another septate junction associated protein, is not detected in anti-Coracle immunoprecipitates (Fig. 7 B).

Based upon the observation that the CNTR of Coracle is sufficient for the maintenance of NRX localization, it seemed likely that the coprecipitation of NRX with Coracle was mediated through the CNTR. To examine this possibility, we prepared lysates from embryos homozygous for cor⁴ (a deletion of 78 amino acids within the CNTR) carrying a cor⁺ transgene that lacks the epitope recognized by the guinea pig antibody, but provides full genetic function (see Materials and Methods; Schweizer, L.Z., and R.G. Fehon, unpublished observations). Immunoprecipitation of Coracle from this lysate precipitates only the cor⁴-derived Coracle protein. NRX protein does not coprecipitate with the cor⁴ protein (Fig. 7,A), indicating that the CNTR of Coracle is necessary for the observed interaction with NRX. Furthermore, immunoprecipitates of just the CNTR (Cor 1–383) from transgenic embryos did coprecipitate NRX, whereas immunoprecipitates from embryos expressing the rest of Coracle (Cor 378–1698) did not (Fig. 7 C), indicating that the CNTR is not only necessary but also sufficient for interaction with NRX. Thus the essential sequences for binding NRX or a NRX-containing complex are contained within the amino-terminal 383 amino acids of Coracle.

To determine whether the coprecipitation of Coracle and NRX in embryonic lysates was mediated through a direct interaction between these two proteins, we examined binding in vitro between bacterially expressed recombinant proteins. A Coracle fusion protein representing the cDNA3 isoform of coracle (Fehon et al., 1994) specifically bound to the cytoplasmic domain of NRX immobilized onto a column. This interaction was efficiently competed with a 10-fold molar excess of unlabeled Coracle (Table I). However, this interaction was not significantly affected with a 10-fold molar excess of unlabeled GST nor with a Coracle fusion protein containing the central divergent region of Coracle (amino acids 608–1,285). Similar binding was observed between the NRX cytoplasmic tail and just the CNTR of Coracle (data not shown). These results demonstrate that Coracle and NRX interact directly via the CNTR of Coracle and the cytoplasmic tail of NRX.

We have dissected the Coracle protein structurally to identify functionally important regions of this member of the protein 4.1 superfamily. Using in vitro mutagenesis, we have shown that the amino-terminal 383 amino acids of Coracle are sufficient and necessary for membrane association in S2 Drosophila cultured cells, and for correct subcellular localization to the septate junction in transgenic embryos. Genetic mutations that affect this domain disrupt the subcellular localization of Coracle and severely affect its genetic functions. Furthermore, the localization of Coracle and the transmembrane protein Neurexin to the septate junction display an interdependent relationship, suggesting that they interact with one another. Consistent with this notion, immunoprecipitation and in vitro binding assays demonstrate that the amino-terminal 383 amino acids of Coracle and the cytoplasmic domain of Neurexin interact directly.

The finding that the amino-terminal 383 amino acids (the CNTR) of Coracle is responsible for septate junction localization correlates well with similar studies of other protein 4.1 family members. The corresponding region of erythrocyte protein 4.1, the 30-kD domain, has been shown to be required for membrane association through binding to the cytoplasmic domain of glycophorin C and the anion exchanger (Anderson and Lovrien, 1984; Pasternack et al., 1985). Likewise, in ERM proteins, the amino-terminal conserved region has been shown to be responsible for localization to the cytoplasmic face of the cell membrane in cultured cells (Algrain et al., 1993), possibly through an interaction with CD44 (Tsukita et al., 1994). Taken together, these results suggest that sequences conserved in all protein 4.1 superfamily members serve as a subcellular targeting domain.

Although the CNTR is well conserved between different members of this superfamily, it is likely that the differences in sequence between them may affect interactions with different transmembrane partners and thereby provide subcellular specificity. For example, in Drosophila, Coracle is localized to the septate junction whereas the other family members, Moesin, Merlin, and Expanded, are found primarily in the region of the adherens junction (Boedigheimer and Laughon, 1993; McCartney and Fehon, 1996). Once correctly targeted to a specific subdomain of the membrane, each protein may then serve unique functions. Erythrocyte protein 4.1 and the ERM proteins appear to function as membrane-cytoskeletal crosslinkers, whereas other family members, such as the protein tyrosine phosphatases, have divergent functions that are not directly related to cytoskeletal binding (McCartney and Fehon, 1997).

The identification and characterization of the cor⁴ mutation, which results from an in-frame deletion of 78 amino acids within the CNTR, provides insight into the functions of Coracle and its localization to the septate junction. Although cor⁴ protein is expressed at essentially normal levels, it does not localize correctly to the septate junction (Fig. 1, D and E). Similarly, cor⁶, which results from an amino acid substitution at Leu³⁹ within the CNTR (but outside of the region deleted in cor⁴), produces a protein that is also mislocalized to the cytoplasm. Attempts to subdivide this region into smaller pieces using in vitro mutagenesis result in proteins that cannot localize to either the cell membrane or to the septate junction (Fig. 2). Taken together, these mutagenesis studies indicate that the CNTR operates as a single functional unit. Furthermore, the observation that both cor⁴ and cor⁶ mutant embryos display severe embryonic phenotypes (Lamb, R.S., R.E. Ward, and R.G. Fehon, manuscript in preparation) suggests that an intact CNTR is essential not only for subcellular localization, but also for function in the context of the developing embryo. However, the CNTR does not appear to provide full genetic function as determined by the inability of hscor 1–383 to rescue a null coracle allele by repeated daily heat shocks (data not shown). Further studies are underway to more fully examine the properties of this functional domain.

Previous work has indicated that whereas Coracle is dependent on NRX for its localization, NRX is not dependent on Coracle (Baumgartner et al., 1996). We have confirmed that Coracle localization is dependent on NRX, but find in addition that in strong coracle alleles, NRX protein is no longer tightly confined to the septate junction and instead spreads basolaterally (Fig. 5). However, the mislocalization of NRX in coracle mutants is qualitatively different from that of Coracle in Nrx mutants. Whereas Coracle shows no apparent localization to the septate junction in Nrx mutants, in coracle mutants there is an enrichment of NRX at the septate junction. One interpretation of this result is that targeting of NRX to the septate junction occurs properly in cells lacking Coracle, but that subsequent maintenance of NRX localization is disrupted. Alternatively, it is possible that there is residual Coracle function in even the strongest alleles, although phenotypic and biochemical evidence indicates that this is not the case (Lamb, R.S., R.E. Ward, and R.G. Fehon, manuscript in preparation).

The evidence presented here indicates that Coracle functions to organize the transmembrane protein NRX to the septate junction. Similarly, previous studies have shown that transfection of ezrin into mouse thymoma cells leads to the redistribution of ICAM-2 in cellular protrusions known as uropods (Helander et al., 1996). In addition, the observed correlation between the rapid redistribution of ERM proteins and dynamic cell surface projections upon growth factor stimulation (Bretscher, 1989) suggests that this membrane-organizing function is a general property of the protein 4.1 superfamily. Our observation that this activity is contained entirely within the CNTR indicates that this conserved domain has similar functions in all family members.

The interdependence between Coracle and NRX for proper localization suggests that at least one other protein in the presumptive septate junction serves as the initial target for both proteins to be properly localized. Based upon the protein 4.1 paradigm of a ternary complex consisting of protein 4.1, glycophorin C, and p55 (Marfatia et al., 1995), we would predict that a PDZ repeat-containing protein is a part of the complex containing Coracle and NRX. The most likely existing candidate for this additional protein is DLG based upon its extensive sequence similarity to p55. DLG is expressed maternally, and initially is uniformly distributed along the lateral membrane and to a lesser extent throughout the cytoplasm. Coincident with the expression of Coracle and NRX, this subcellular localization is refined to the presumptive septate junction (Woods and Bryant, 1991). This expression pattern might be expected of a protein that serves to “prepattern” the septate junction. However, we have failed to detect any interaction between Coracle and DLG by immunoprecipitation (Fig. 7), and we have observed no genetic interaction between coracle and dlg mutant alleles (data not shown). In addition, the embryonic defects associated with dlg mutants (Perrimon, 1988) are different from those of coracle and Nrx. These results suggest that DLG is not involved in a ternary complex together with Coracle and NRX, despite its structural similarity with p55. The question of whether there is another PDZ repeat–containing protein that functions to stabilize Coracle-NRX interactions remains to be answered, although the structural similarities between the respective Drosophila and human proteins strongly suggest that one exists. Additionally, the recent identification of EBP50 as a PDZ repeat–containing protein that associates with ERM proteins (Reczek et al., 1997) suggests that an interaction with a PDZ repeat-containing protein may be a ubiquitous feature of protein 4.1 members. Regardless, the results described here strongly suggest that at least one other component is involved in Coracle/NRX localization and function. Further studies are currently underway to identify this other possible component using a combination of genetic and biochemical approaches.

The structural components and signaling molecules that function together at the septate junction implicate this region in essential signaling mechanisms required for proper cellular communication and epithelial integrity. The phenotype resulting from zygotic loss of dlg function suggests a requirement for septate junctions in the control of cell proliferation and apicobasal polarity (Stewart et al., 1972; Woods and Bryant, 1989; Woods et al., 1997). Nrx mutant embryos display neuromuscular defects apparently due to a breakdown in the blood-brain barrier, and subtle morphological defects in the ultrastructure of the septate junction, suggesting that the septate junction acts as a barrier to paracellular flow (Baumgartner et al., 1996). In addition, the extracellular domain of NRX has EGF and laminin G repeats (Baumgartner et al., 1996), raising the possibility that it serves as a cell surface receptor. In support of this hypothesis, Neurexins were first identified as synaptic receptors (Petrenko et al., 1991), and have subsequently been proposed to function as receptors in synaptic targeting of neurons (Ullrich et al., 1995). The interdependence between Coracle and NRX for proper localization and their similar embryonic phenotype suggests that they function together in signaling events. The molecular details of these signaling processes are currently unclear, as are their functions. However, the range of observed coracle phenotypes, including embryonic dorsal closure and proliferation defects (Lamb, R.S., R.E. Ward, and R.G. Fehon, manuscript in preparation), suggests that either more than one signaling mechanism is involved, or that this signal has multiple functions during development. Given the membrane-organizing activity of the CNTR of Coracle that we describe here, it is likely that Coracle and the other protein 4.1 family members function to target receptors and their intracellular effectors to specific membrane domains. Precedence for this type of signal-mediating molecular scaffold has been found in the multiple PDZ repeat–containing protein InaD, which functions to bring components of the phototransduction pathway together into a “transducisome” (Tsunoda et al., 1997). Given the increasing evidence that protein 4.1 family members function in concert with PDZ repeat–containing proteins (Marfatia et al., 1994, 1995; Reczek et al., 1997), the hypothesis that Coracle and other protein 4.1 family members link together the molecular components of a signal cascade is particularly intriguing. The identification and characterization of other components of this protein complex at the cytoplasmic face of the septate junction will add to our understanding of the functions of Coracle and the other members of this highly conserved protein superfamily in the context of signal integration during development.

We would like to thank H. Bellen (Baylor College of Medicine, Baylor, TX) for providing NRX antibodies and fly stocks, D. Woods (University of California, Irvine, CA) for providing dlg clones, and the Bloomington stock center for fly stocks (Bloomington, IN). We thank L. Schweizer (University of Zurich, Zurich, Switzerland) and D. Mootz (Harvard University, Cambridge, MA) for technical help; and I. Rebay (Whitehead Institute, Massachusetts Institute of Technology, Cambridge, MA), the anonymous reviewers, and our colleagues in the Fehon lab for valuable suggestions and discussions.

This work was supported by grants from the National Science Foundation, No. IBN-92066555, and the American Cancer Society, No. DB-84846, to R.G. Fehon. R.E. Ward, and R.S. Lamb were supported by National Institutes of Health Training Grants, No. 5T32 GM07754 and No. GM07184, respectively.

CNTR

conserved amino-terminal region

DLG

discs large

ERM

ezrin/radixin/moesin

GST

glutathione- S-transferase

MAGUK

membrane-associated guanylate kinase

NRX

neurexin

PDZ

PSD-95, DLG, ZO-1