Septate junctions (SJs), similar to tight junctions, function as transepithelial permeability barriers. Gliotactin (Gli) is a cholinesterase-like molecule that is necessary for blood–nerve barrier integrity, and may, therefore, contribute to SJ development or function. To address this hypothesis, we analyzed Gli expression and the Gli mutant phenotype in Drosophila epithelia. In Gli mutants, localization of SJ markers neurexin-IV, discs large, and coracle are disrupted. Furthermore, SJ barrier function is lost as determined by dye permeability assays. These data suggest that Gli is necessary for SJ formation. Surprisingly, Gli distribution only colocalizes with other SJ markers at tricellular junctions, suggesting that Gli has a unique function in SJ development. Ultrastructural analysis of Gli mutants supports this notion. In contrast to other SJ mutants in which septa are missing, septa are present in Gli mutants, but the junction has an immature morphology. We propose a model, whereby Gli acts at tricellular junctions to bind, anchor, or compact SJ strands apically during SJ development.
Permeability barriers have important roles in many tissues in both vertebrates and invertebrates. The blood–brain barrier, for example, is essential in vertebrates to keep the brain isolated from blood-borne growth factors, neural active compounds, and fluctuating blood ion levels that can severely impact neuronal physiology and brain function (Rubin and Staddon, 1999). Similarly, in insects, the high potassium concentration of the hemolymph can block action potentials in neurons, and, thus, cause paralysis, if the blood–brain barrier is disrupted (Auld et al., 1995; Baumgartner et al., 1996). Epithelial permeability barriers are formed by tight junctions (TJs)* in chordates and by septate junctions (SJs) in most invertebrates. TJs and SJs differ in their ultrastructure, position in epithelial cells, and molecular composition, yet they share certain organizational similarities that enables them to form effective permeability barriers (Lane et al., 1994; Tepass et al., 2001; Tsukita et al., 2001).
SJs are located in the apical portion of the lateral membrane of invertebrate epithelial cells, immediately below adherens junctions (AJs). TJs in contrast, lie apical to AJs in vertebrate epithelial cells. SJs are characterized by a ladderlike array of cross-bridges or septa that span the 15–20-nm intermembrane space of cell–cell contacts. TJs, on the other hand, appear as multiple “kissing-points,” in transmission electron micrographs where adjacent plasma membranes are in direct contact (Lane et al., 1994; Tsukita et al., 2001). Both SJs and TJs are composed of multiple strands with some variation in strand number depending on cell type. For example, 10 or more strands typically compose an SJ in a locust epithelial cell, whereas 4–7 strands are found in a TJ of kidney distal tubule epithelial cell. SJ strands are tightly arrayed parallel to each other, whereas TJ strands are less compact and are organized into overlapping or anastomizing networks (Claude and Goodenough, 1973; Lane and Swales, 1982). The multi-stranded composition of both SJs and TJs appears to be necessary to effectively block the paracellular flow of substances. In insects, permeability studies have shown that heavy metal tracer dyes are often able to penetrate deep into the stacked arrays of SJ strands before they are blocked from paracellular passage (Swales and Lane, 1985). Similarly, studies of “tight” and “leaky” TJs in vertebrate epithelia have shown a positive correlation between strand number and TJ permeability (Claude and Goodenough, 1973).
In Drosophila, two types of SJs, smooth (sSJs) and pleated SJs (pSJs), have been observed (Tepass and Hartenstein, 1994). Smooth SJs and pSJs vary morphologically and have different tissue distributions, but they are functionally equivalent (Lane et al., 1994). In freeze-fracture electron micrographs, pSJ strands appear to lie in membrane depressions or grooves that are absent in sSJs (Lane and Swales, 1982). pSJs are found in ectodermally derived tissue, such as the foregut, hindgut, tracheae, and glia, whereas sSJs are found in endodermally derived tissue, such as the midgut (Tepass and Hartenstein, 1994).
TJs and SJs have been described to encircle epithelial cells as a continuous belt, though this view has been challenged. EM studies of epithelial cells, in both vertebrates and insects, have shown that the continuity of TJ and SJ belts is interrupted at sites of tricellular contact by “pores” or “channels” that span the depth of the epithelium (Fristrom, 1982; Graf et al., 1982; Noirot-Timothée et al., 1982; Walker et al., 1985). It is unclear what the function of these specialized structures may be. In insects, it has been proposed that diaphragms associated with these channels may serve as anchors for SJ strands (Fristrom, 1982; Graf et al., 1982; Noirot-Timothée et al., 1982). Similarly, developmental studies on rat olfactory epithelium suggest that de novo synthesis of TJs strands may occur at sites of tricellular contact (Menco, 1988). Recent studies on human umbilical vein cultures have suggested that the localized disruption of TJs at endothelial tricellular corners is important during acute immune responses as it enables neutrophils to migrate across capillaries and reach sites of inflammation or infection (Burns et al., 2000).
To more thoroughly appreciate the similarities and differences between vertebrate TJs and Drosophila SJs, a molecular characterization of these junctions has been performed over the last decade. Of the Drosophila SJ–associated proteins, four have clear roles in SJ formation. Mutations in neurexin-IV (Nrx), discs large (Dlg), scribble (Scrib), or coracle (Cora) prevent the formation of septa (Baumgartner et al., 1996; Woods et al., 1996; Lamb et al., 1998; Bilder et al., 2003). Nrx is a transmembrane protein and a member of the neurexin family of synapse-associated proteins (Ushkaryov et al., 1992; Baumgartner et al., 1996). Dlg and Scrib are cytosolic, PSD-95/Dlg/ZO-1 (PDZ) domain-containing proteins that also have roles in establishing epithelial cell polarity before SJ development (Perrimon, 1988; Woods et al., 1996; Bilder and Perrimon, 2000; Bilder et al., 2003). Cora is a band 4.1–related protein that possesses a four point one/ezrin/vadixin/moesin domain and physically associates with Nrx (Fehon et al., 1994; Ward et al., 1998). In the case of vertebrate TJs, at least 25 claudins have been identified that are believed to play critical roles in TJ development (Furuse et al., 1998; Gow et al., 1999; Morita et al., 1999; Tsukita and Furuse, 2000). Claudins can interact in a homophilic or heterophilic fashion, and their mixing ratio is believed to moderate the permeability of TJ transepithelial barriers (Furuse et al., 1999). In addition, numerous other transmembrane or cytoplasmic factors have been found to be associated with TJs (Tsukita et al., 2001).
No significant similarities between the molecular composition of vertebrate TJs and Drosophila SJs have been noted so far (Tepass et al., 2001). Moreover, vertebrate homologues of Drosophila Nrx and Cora have been identified that localize to mammalian paranodal junctions, at the interface of axons and glia (Menegoz et al., 1997; Poliak et al., 1999; Bhat et al., 2001; Boyle et al., 2001). These junctions are morphologically very similar to Drosophila SJs. It is important to more thoroughly characterize SJs, TJs, and paranodal junctions, in order to establish a better understanding of the relationship between these junction types and to gain insight into the mechanism of permeability barrier formation.
Gliotactin (Gli) is a noncatalytically active cholinesterase-like molecule that is a member of a class of adhesion proteins termed the electrotactins (Auld et al., 1995, Botti et al., 1998). In the peripheral nervous system, Gli is necessary for glial ensheathment of axons, and for the formation of the glial-based blood–nerve barrier (BNB; Auld et al., 1995). Although pSJs between glial wraps constitute the molecular seal of the BNB, it has not been determined if the BNB defects seen in Gli mutants arises from defective pSJ development or from inadequate axonal ensheathment (Baumgartner et al., 1996; for review see Carlson et al., 2000). Here, we investigate the role of Gli in the formation of SJs through genetic and cell biological approaches.
Expression profile of Gli in the epidermis
To test the hypothesis that Gli is involved in the formation of SJs, we first examined if the tissue and subcellular distribution of Gli matches that reported for SJs (Fehon et al., 1994; Tepass and Hartenstein, 1994; Baumgartner et al., 1996; Lamb et al., 1998). Embryos were doubled labeled for Gli and the pSJ markers Nrx and Cora to investigate their temporal and spatial overlap. Gli first appears in the ectoderm at stage 11 of embryogenesis, shortly after Nrx, and persists throughout embryonic development (Fig. 1). Gli expression at this stage appears to be due to zygotic gene activity as no maternal Gli mRNA is detected in Northern blots of 0–6-h embryos (stage 1–10; unpublished data). The distribution of Gli and Nrx protein is similar in the epidermis at stage 11, and both molecules are distributed evenly over the lateral membrane of epithelial cells (Fig. 1 A). In the homozygous Gli mutants (GliAE2Δ45, GliAE2Δ4b, GliDV3, and GliCQ1), no Gli staining was observed at any stage of embryogenesis (Fig. 1 E) demonstrating that the Gli 1F61D4 mAb is specific to Gli.
At stage 13, the localization of Nrx and Gli changes. In en face views of the epidermis, Gli becomes concentrated at the tricellular corners of abutting epithelial cells, whereas Nrx remains distributed around the cell circumferences (Fig. 1 B, arrowhead). Occasionally, patchy Gli staining is observed in some of the elongated epithelial cells in the dorsal epidermis (Fig. 1 B, solid arrow). Interestingly, at the leading edge of the epidermis there is little overlap between Gli and Nrx staining (Fig. 1 B, concave arrow). In the amnioserosa, Gli and Nrx staining is absent, consistent with the fact that this tissue lacks SJs (Tepass and Hartenstein, 1994). In addition to the epidermis, Gli is also found at the tricellular junctions of all epithelial tissues that express Nrx and Cora, and in which pSJs have been observed at the EM level (Fehon et al., 1994; Tepass and Hartenstein, 1994; Baumgartner et al., 1996; Lamb et al., 1998). These tissues include the trachea (Fig. 1 F, asterisk), salivary glands (see Fig. 5), PNS glia (Auld et al., 1995), the chordotonal organs (unpublished data), and the foregut and the hindgut (Fig. 1, G and H). Gli was not observed in tissues that contain sSJs.
In addition to becoming redistributed in the planar axis of epithelial cells, Gli also undergoes a redistribution along the apical–basal axis. At stage 13 of embryogenesis, Gli is localized along the entire length of the lateral membrane. However, unlike Nrx, Gli is typically concentrated in multiple discrete patches within the lateral membrane (Fig. 1 C). By stage 15 of embryogenesis, Gli expression is restricted to the apical half of the lateral membrane where it colocalizes with Nrx at the presumptive pSJ domain (Fig. 1 F). The tissue distribution and subcellular localization profile of Gli suggest that it is a component of pSJs, but its unique localization to tricellular junctions suggests that it plays a distinct role from Nrx and Cora.
SJ markers are mislocalized in Gli mutants
To investigate the effect of loss of Gli on epithelial cell development, we stained homozygous Gli mutants for a variety of epithelial markers. We focused on stage 15 embryos that in wild-type display the mature distribution pattern of SJ markers. Four different strong Gli loss of function alleles (GliAE2Δ45, GliAE2Δ4b, GliDV3, and GliCQ1) were tested and gave indistinguishable results.
Epidermal cells within the abdominal segments of Gli mutants are slightly taller and have a more uniformly columnar appearance than wild-type cells as revealed through α-spectrin labeling, which outlines cell profiles (Fig. 2, A and B). Next, we double stained Gli mutant embryos for SJ and AJ markers. In stage 15 Gli mutants, the SJ markers Dlg, Nrx, and Cora are all mislocalized (Fig. 2, C–H). Rather than being confined to the apical half of the lateral membrane, these SJ markers extend to the extreme basal side of epithelial cells, although an apical emphasis in their distribution is retained (Fig. 2, D, F, and H). The mislocalization of Dlg is less severely affected than that of Cora or Nrx in Gli mutants. Failure of SJ markers to redistribute to the apico-lateral membrane is already seen at stage 13, at a time when the first septa appear in wild-type embryos (Tepass and Hartenstein, 1994). In contrast to SJ markers, AJ markers (Armadillo [arm] and DE-cadherin) are normally localized in Gli mutants (Fig. 2, C–F and G and H, respectively). Together, the generally normal columnar morphology of epithelial cells seen in Gli mutants, and the correct apical localization of AJ markers suggests that Gli does not have a significant role in specifying apical–basal epithelial polarity. However, the mislocalization of SJ markers in Gli mutants indicates that Gli has a specific role in the maturation of pSJs.
Nrx and Cora mutant embryos have defects in dorsal closure, which manifest as dorsal holes or scabs in cuticle preparations (Baumgartner et al., 1996; Lamb et al., 1998). In Gli mutants, cuticles appear normal, however, small dorsal “holes” are observed at a low frequency in the epidermis of stage 16 embryos (unpublished data). These observations suggest that dorsal closure is delayed in some Gli mutant embryos; however, it is completed successfully before cuticle deposition.
Gli's localization at epidermal tricellular corners is dependent on Nrx
To determine if pSJ formation is necessary for Gli to become localized to the tricellular corners of epithelial cells, Gli distribution in Nrx46 homozygous mutants was analyzed. Nrx46 is a severe loss of function mutation and septae are absent from pSJs in these mutants (Baumgartner et al., 1996). We found that Gli does not localize normally to the tricellular corners of epithelial cells in late Nrx46 mutant embryos (Fig. 3). In en face views of the epidermis of stage 15, Nrx46 mutants with a severe dorsal closure phenotype, Gli is abnormally distributed around the circumference of cells (Fig. 3 C). Similarly, in cross section views of these mutants, Gli is not restricted to the apical half of lateral membrane, but instead extends basally (Fig. 3 D). In stage 15, Nrx46 mutants with mild dorsal closure phenotypes, Gli is less severely mislocalized (Fig. 3, E and F). Thus, the severity of the dorsal closure phenotype in the Nrx46 mutants correlates with the severity of Gli mislocalization in the epithelium. The localization profile of Gli in stage 15 Nrx46 mutants resembles that of Gli in wild-type embryos at stage 13 of development (compare Fig. 1 B with Fig. 3, C and E; compare Fig. 1 C with Fig. 3, D and F). These results indicate a reciprocal dependence between Gli and Nrx for localization, but also suggest that pSJs must develop to enable Gli to localize to tricellular junctions.
pSJ septae are present in Gli mutants, but not compacted into clusters
Ultrastructural analysis of Gli mutants was performed to further characterize the nature of the pSJ defect. We examined pSJ structure at bicellular contacts, as well as at tricellular junctions, in the epidermis (Fig. 4). In wild-type animals pSJ septa are present in the apical 1/3 to 2/3 of the lateral membrane and are typically organized into clusters (Fig. 4 A). We counted the number of pSJ septa, and the number of clusters of septa per cell–cell contact in stage 17, wild-type, and Gli mutant embryos. It was discovered that total number of septa at a cell–cell contact is statistically equivalent in wild-type and Gli mutants; however, there are more clusters containing fewer septa/cluster in Gli mutants (Fig. 4, A, C, and E). 16.4 septa/cell–cell contact (SD 7.2, n = 47) occur in wild-type animals, as compared with 15.5 septa/cell–cell contact in Gli mutants (SD 5.3, n = 98). 3.3 (SD 1.7) septal clusters/cell–cell contact occur in wild-type animals as compared with 8.3 (SD 3.2) in Gli mutants. The distribution of septa in Gli mutants, resembles that reported for stage 14 embryos, suggesting that Gli is necessary for the maturation of pSJs (Tepass and Hartenstein, 1994).
pSJ defects are also observed at tricellular corners (Fig. 4). In wild-type pSJ, septa are typically present at all bicellular contacts of a tricellular junction (Fig. 4 B); however, in Gli mutants, pSJ septa are often absent at one or more bicellular contacts, in the region immediately flanking the tricellular channel (Fig. 4 D).
In addition to the pSJ defects observed in Gli mutants, gaps between the lateral membranes of adjacent epithelial cells were observed more frequently in wild type (Fig. 4, F and G). 30.4% of cell–cell contacts in Gli mutants show large intercellular spaces in the basal portion of the lateral membrane in contrast to (5.1%) in wild type (Fig. 4 G, asterisk). In addition, the gaps are of a smaller size in wild type. These results raise the possibility that Gli may contribute to cell adhesion.
Transepithelial barriers are compromised in Gli mutants
The mislocalization of Dlg, Nrx, and Cora, as well as the abnormal distribution of pSJ septa, in the epidermis of Gli mutants, suggests that pSJs are unable to form effective transepithelial barriers in these animals. To test this hypothesis, a dye permeability assay was performed in the salivary glands (Lamb et al., 1998). Gli is localized to the tricellular corners of salivary gland epithelial cells as in the epidermis (Fig. 5, A and B), and overlaps with Cora at pSJs when the gland is viewed in cross section (Fig. 5 C). As in the epidermis, the SJ markers Dlg, Cora, and Nrx are all mislocalized in the salivary glands of Gli mutants (unpublished data).
To test the integrity of the pSJ transepithelial barrier in salivary glands, a 10-kD rhodamine–dextran conjugate was injected into the hemocoel of wild-type (n = 23) and homozygous GliAE2Δ45 mutant (n = 37) embryos. In these studies, we were careful to choose stage 17 embryos that are expected to have functional pSJs (Lamb et al., 1998). In all wild-type animals tested, the rhodamine–dextran dye was effectively blocked from passing into the lumen of the salivary gland for up to 40 min after injection (Fig. 5, D–F), after which the animals would hatch and crawl away. In these animals, the dye was able to pass between salivary gland epithelial cells, in a basal to apical direction, up to the pSJ where it stopped without penetrating into the lumen (Fig. 5 E, arrowhead). In contrast, in every Gli mutant tested, the tracer dye was found to pass into the salivary gland lumen within 10 min of injection (Fig. 5, G–I). In these animals, the dye was observed to pass between epithelial cells at multiple consecutive cell–cell contacts (Fig. 5, H and I, arrowheads). In a given longitudinal section (2 μm confocal slice) of a salivary gland, tricellular corners are not transected in many consecutive cells due to their small size and position at the corners of these hexagonally shaped cells. Therefore, this result suggests that the dye penetrates pSJ domains, not solely at tricellular corners of epithelial cells in Gli mutants, but rather around their circumferences. These findings indicate that the Gli is required for the formation of functional transepithelial barriers.
Rescue of Gli mutants by Gli, but not Neuroligin transgenes
To show that pSJ transepithelial barrier defects observed in GliAE2Δ45 mutants are specifically due to loss of Gli and not due to potential second-site mutations in the genetic background, rescue experiments were performed. We used the GAL4/UAS system (Brand and Perrimon, 1993) to express a wild-type Gli transgene (UAS-Gliwt) in homozygous GliAE2Δ45 mutant embryos, and then scored for the ability of these rescued embryos to hatch into first instar larvae. A variety of GAL4 drivers were tested. They included, repoGAL4, hsp-G303–7, and daughterless GAL4 (da.G32; Wodarz et al., 1995; Leiserson et al., 2000). Rescue of lethality was obtained with the ubiquitously expressed da.G32 driver suggesting that there are no second site lethal mutations in the GliAE2Δ45 mutant strain. In da.G32:UAS-Gliwt –rescued embryos, Gli localizes correctly to the tricellular corners of epithelial cells in the epidermis (unpublished data). 78% (n = 39) of rescued embryos hatch and survive to adulthood. Rescued adults are fertile and can be maintained as a stable stock, however, 58% (n = 51) have severe leg defects. The metatarsus and tibia are typically bent at 45° toward the midline, and necrotic tissue is often present on the medial aspect of the limbs and limb joints. Similar phenotypes are also seen in adult escapers of a GliDV5 hypomorphic strain (unpublished data), as well as in Nrx, and Cora hypomorphic mutants (Baumgartner et al., 1996; Lamb et al., 1998). In carrying out the rescue experiments, we were not able to rescue the lethality of homozygous Gli mutants with the glial-specific driver repoGAL4. This result suggests that the lethality of Gli mutants is not solely due to a disrupted glial-based BNB (Auld et al., 1995).
Gli is a member of a large family of cholinesterase-like molecules called the electrotactins, which are structurally very similar to one another (Botti et al., 1998). We wondered if the closest Drosophila family member, Drosophila neuroligin (Dnl), which has the highest sequence similarity to Gli, of the electrotactins in Drosophila (Gilbert et al., 2001), might be functionally similar enough to Gli be interchangeable. Despite that Gli and Dnl share 40% amino acid identity across their extracellular serinesterase-like domains, a Dnl transgene could not rescue Gli lethality. These results suggest that Gli and Dnl are not functionally interchangeable electrotactin molecules.
Gli shows a tissue distribution pattern similar to that of the pSJ proteins Cora and Nrx. However, the subcellular distribution of Gli varies from other known pSJ proteins in that Gli is restricted to tricellular junctions. Three lines of evidence suggest that Gli is necessary for pSJ formation. First, the pSJ markers Dlg, Cora, and Nrx are mislocalized in Gli mutant epithelial cells. Second, Gli mutants do not form effective transepithelial barriers as determined through a salivary gland dye permeability assay. Third, ultrastructural analysis indicates that pSJs are malformed in Gli mutants.
The mutant phenotype of Gli, in many respects, is similar to that of Nrx and Cora, however, there are also clear differences between these mutants. In all mutants, SJ markers are mislocalized, but apical–basal polarity appears unaffected. All mutants have dorsal closure defects: whereas in Nrx and Cora mutants dorsal closure fails, and is accompanied with prominent cuticular defects, in Gli mutants, dorsal closure is slightly delayed at a low frequency but eventually completed (Fehon et al., 1994; Baumgartner et al., 1996; Lamb et al., 1998). The most striking difference between Gli and Nrx/Cora mutants is that at the ultrastructural level they have distinct pSJ morphologies. In Nrx/Cora mutants, septa fail to form (Baumgartner et al., 1996; Lamb et al., 1998), whereas in Gli mutants, a normal number of pSJ strands forms, but they are not tightly arrayed (Fig. 4). This suggests that Nrx/Cora are necessary for pSJ strand synthesis, whereas Gli appears to be required for the maturation of the pSJ, which involves the compaction of SJ strands. pSJs in wild-type embryos, at stage 14 of development, have a morphology that resembles those of late stage Gli mutants (Tepass and Hartenstein, 1994). Evidently, compaction of pSJ septa is essential to form an effective transepithelial barrier because Gli mutants have leaky salivary glands. This result is in agreement with the findings of Swales and Lane (1985), who have shown in locust that tracer dyes are often able to cross individual septa, as well as small groups, but that the wild-type distribution of multiple large groups of septa do form effective permeability barriers. It is also possible that the detachment of lateral membranes observed in Gli mutants may contribute to a compromised permeability barrier.
Gli, SJ development, and the “tricellular plug” (TCP) model
Several ultrastructural studies of various insects, including Drosophila, have demonstrated specialized structures at the tricellular corners of epithelial cells that are linked to SJs (Fristrom, 1982; Graf et al., 1982; Lane and Swales, 1982). Graf et al. (1982) performed a detailed freeze-fracture EM analysis of epithelial tissue in crustaceans and cockroaches and identified channels at the tricellular corners of abutting epithelial cells (Fig. 6). These channels span the length of the cells. In the region of the SJ domain, the channels are filled with what appears to be a series of diaphragms that are stacked on top of each other. The diaphragms make contact with the lateral membranes of all three epithelial cells comprising a channel. In the vicinity of the tricellular corners, SJ strands run parallel to the axis of the channel. This organization is different from that of SJ strands elsewhere in the cell. SJ strands typically run parallel to the axis of the apical membrane domain. Graf et al. (1982) suggested that SJ strands anchor on the stacked arrays of diaphragms at tricellular corners. Other researchers referred to these structures as TCPs, and suggested that they serve as occlusive devices during transepithelial barrier formation in addition to acting as anchors for SJ strands (Fristrom, 1982; Lane and Swales, 1982).
The development of pSJs occurs in a step wise process (Lane and Swales, 1982). Early in pSJ development, intermembrane particles (IMPs), the building blocks of SJ strands, are homogeneously distributed throughout the lateral membrane. They polymerize at random sites in the lateral membrane (in small depressions) to form short pSJ strands. These, in turn, lengthen and “stack” to form pSJ placodes, which eventually anchor on TCP diaphragms. TCPs are not observed early in pSJ development and, thus, they are believed to be mature features of pSJs. Our observations that Gli is localized to the tricellular corners of epithelial cells, and that it is necessary for pSJ maturation is consistent with this TCP model. The EM images of Graf et al. (1982), which show that TCP diaphragms are associated with pSJ strands in the apical half of tricellular channels, agree with our observations that Gli is restricted to the apical half of tricellular corners. These data suggest that Gli is an integral component of TCPs, a notion that could be confirmed by future immuno-EM experiments.
Combining the TCP model with the EM observations of Lane et al. (1994), and our analysis of Gli, we propose a model to suggest how Gli is involved in the formation of pSJ (Fig. 6). During SJ development, Gli may serve as a linker between SJ strands and tricellular channel diaphragms (TCDs). Gli has the potential to associate with TCDs, through its extracellular domain and with SJ strands through its intracellular tail. Nrx and Cora (and possibly Dlg and Scrib) could be integral components of pSJ strands and may represent the pSJ IMPs reported by Lane et al. (1994). Nrx and Cora behave like IMPs in that they are homogeneously distributed in the lateral membranes during stage 12 of embryogenesis, and are concentrated in the apical half of the lateral membrane at stage 15 (Fehon et al., 1994; Baumgartner et al., 1996; unpublished data). It is possible that Gli could physically associate with Nrx and, thus, be linked to SJ strands, because Gli-like vertebrate neuroligins have been shown to bind to neurexin-1β via their cholinesterase-like domains (Ichtchenko et al., 1995). However, Nrx is more similar to the α-neurexins than the β-neurexins, and the latter is not known to bind neuroligins (Ichtchenko et al., 1995; Tabuchi and Südhof, 2002). Gli and Nrx also fail to interact in S2 cell aggregation assays (Auld et al., 1995; Baumgartner et al., 1996). Alternatively, Gli may associate through its intracellular domain with SJ strands as it contains a COOH-terminal PDZ recognition sequence, and Dlg and Scrib are PDZ domain proteins. Also in vertebrates, it has been shown that PSD-95 (a Dlg-related protein) binds the intracellular tail of various neuroligins (Irie et al., 1997). Irrespective of the mechanism, our results support the notion that Gli is physically associated with SJ strands, as Dlg, Nrx, and Cora fail to be confined to the pSJ domain within the lateral membrane, and pSJ strands are unorganized in Gli mutant epithelial cells.
Many interesting questions remain to be addressed regarding the specific role of Gli in TCP and pSJ development. Is Gli only necessary to anchor pSJ strands to TCP diaphragms, or is it also directly involved in the formation of the TCP diaphragms? The molecular nature of the diaphragms is not known. One possibility is that the diaphragms are composed of secreted molecules that are linked via Gli to the pSJ strands (Fig. 6). This linkage may be critical for the compaction of SJ strands.
Gli, tricellular channels, tight and paranodal junctions
Tricellular channels are not features unique to the insect epithelium. Walker et al. (1985) performed a freeze-fracture EM analysis of vertebrate epithelial tissue, and found the organization of TJs at tricellular corners to be strikingly similar to that of SJs in insects. It will be interesting to determine if any of the vertebrate neuroligins are localized to the tricellular corners of epithelial cells, and if they have a role in TJ maturation similar to Gli's role in SJ development. Of particular interest are human Neuroligins 3 and 4, and rat neuroligins 2 and 3, which are not nervous system–specific and which have a broader tissue distribution than other vertebrate neuroligins (Philibert et al., 2000; Bolliger et al., 2001; Gilbert et al., 2001). Given the structural and molecular similarities between Drosophila SJs and mammalian paranodal junctions (Tepass et al., 2001), it will also be interesting to determine if neuroligins, or tricellular junctions are present at axon–glial contacts. However, at least two neuroligins, 1 and 3, are not found at paranodal junctions in mice or rat, rather neuroligin 3 appears to be associated with nonmyelinating Schwann cells (unpublished data).
Materials And Methods
The wild-type stock, Oregon R, was used as a control in all experiments, except for the transepithelial barrier assay (see below). GliAE2Δ45 and GliAE2Δ4b are P element excision deletion alleles and protein nulls (Auld et al., 1995). GliDV3 and GliCQ1 are ethyl methane sulfonate alleles; both contain stop codons in the extracellular domain of Gli and are null alleles (Venema, 2003). The SJ mutant Nrx46 is a strong ethyl methane sulfonate allele and putative protein null (Baumgartner et al., 1996). The da.G32-GAL4 driver is ubiquitously expressed (Wodarz et al., 1995), hsp-GAL4 303–7 is also ubiquitously expressed, but nonhomogeneously (Leiserson et al., 2000), and repo-GAL4 is expressed in all glia with exception of the midline glia (Sepp et al., 2001). The Gli-GAL4 driver J29GAL4#2 and the UAS-Gliwt reporter line were generated as outlined below (next section). The J29GAL4#2 driver is expressed in a Gli-like distribution. UAS-gapGFP, (Bloomington Stock Center, Indiana University, Bloomington, IN) was used to label cell membranes. UAS-Neuroligin was obtained from G. Boulianne (Hospital for Sick Children, Toronto, Canada). Marked balancer chromosomes used for mutant analysis were CyO, P(ry+, enwglacZ) and CyO, P(w+, actinGFP), and TM6B,P(w+, iab-2 [1.7]lacZ),Tb.
Generation of transgenic lines
For the J29GAL4#2 driver, 3.7 kb of Gli's 5′ regulatory sequence (BamH1-Xba1 fragment of J29LamdaG5 genomic clone; Auld et al., 1995) was subcloned upstream of the GAL4 cDNA in a pCasper2A+ P element transformation vector (Brand and Perrimon, 1993). For the UAS-Gliwt line, the 3.9-kb Gli cDNA (AE2 7.41; Auld et al., 1995) was subcloned into the EcoR1 site of the pP(UAST) transformation vector (Brand and Perrimon, 1993). Engineered constructs (200 ng/μl) were injected into w1118 embryos together with the pπ25.7wcΔ2–3 (400 ng/μl) using standard techniques (Rubin and Spradling, 1982), and insertion lines were isolated and balanced.
Antibody staining of embryos was performed as described by Halter et al. (1995). Homozygous Gli and Nrx mutants were identified with marked balancers (blue or GFP) by staining for a lack of β-galactosidase (β-Gal) or GFP expression. Embryos were staged according to Hartenstein (1993). Stained embryos were mounted in Vectashield (Vector Laboratories), and imaged with a Radiance Plus confocal microscope (40× oil and 63× oil objective lenses, Bio-Rad Laboratories). Single, 2-μm optical slices were recorded in all experiments. Confocal files were processed with Image-J 1.24 and Adobe Photoshop 5.5. Primary antibodies and the dilutions used for embryo staining were: guinea pig anti-Dlg at 1/300 (Woods and Bryant, 1991), mouse anti-Gli (1F61D4) at 1:1 (Auld et al., 1995), mouse anti–Drosophila α-spectrin (3A9) at 1:5 (Developmental Studies Hybridoma Bank, University of Iowa, Iowa City, IA), mouse anti-arm (N2 7A1) at 1:5 (Developmental Studies Hybridoma Bank), mouse anti–β-Gal at 1:500 (Sigma BioSciences), mouse anti-Cora (9C and C615–16B cocktail) at 1:100 (Fehon et al., 1994), rabbit anti-Nrx at 1:200 (Baumgartner et al., 1996), rabbit anti–β-Gal at 1:400 (Cappel, ICN Pharmaceuticals Inc.), rabbit anti-GFP at 1:200 (Abcam Ltd.), and rat anti–DE-cadherin at 1:50 (Oda et al., 1994). All the following secondary antibodies (Molecular Probes) were used at 1:300 dilution: goat anti–guinea pig A488, goat anti–mouse A488 and A568, goat anti–rabbit A488 and A568, and goat anti–rat A488. All the secondary antibodies were highly cross adsorbed, except for the rat secondary. The 1F61D4 mAb was preadsorbed before use, by adding 100 μl of 4-h-old embryos (fixed and blocked) to 900 μl of 1F61D4 containing 10% normal goat serum (Sigma BioSciences) and incubating at 4°C overnight.
Transepithelial barrier assay
Dye injections were performed as described by Lamb et al. (1998) with the following changes. The salivary glands of wild-type and Gli mutant embryos were labeled with gapGFP to facilitate detection using the GAL4/UAS expression system (Brand and Perrimon, 1993). For wild-type embryos, da.G32 females were mated to w;UAS-gapGFP/CyO males. Labeled Gli mutant embryos were obtained by crossing w; GliAE2Δ45,UAS-gapGFP/ CyO females to w;GliAE2Δ45/CyOactinGFP;da.G32/TM6,Tb males. The da.G32 GAL4 driver is strongly expressed in salivary gland tissue. Crosses were performed at 21°C, and eggs were collected at 1-h intervals, and aged for 24 h to obtain stage 17 embryos. Embryo staging was confirmed by scoring for mouth hooks and prominent trachea. Embryos were then dechorionated and spread on glass microscope slides for sorting using a GFP dissecting scope. Mutant Gli embryos of the genotype w; GliAE2Δ45,UAS-gapGFP/GliAE2Δ45; da.G32/+ were identified as having brightly fluorescing salivary glands, but lacking CyOactinGFP expression. Rhodamine-labeled dextran (10,000 mol wt; Molecular Probes) was reconstituted in ddH2O to 3 mM.
GAL4 drivers capable of rescuing Gli lethality were identified by crossing GliAE2Δ45,UAS-Gliwt/CyOactinGFP females with GliAE2Δ45/CyOactinGFP; GAL4 driver/TM6,Tb males at 23°C. Progeny were screened with a GFP dissecting scope for non-GFP first instar larval escapers. For each rescue experiment ∼2,000 (mixed genotype) embryos were screened. To quantify the frequency at which w; GliAE2Δ45,UAS-Gliwt/GliAE2Δ45; da.G32/+ embryos were able to hatch, all progeny of the parental rescue cross were dechorionated and screened with a GFP scope to identify homozygous Gli mutants. Embryos of the correct genotype were arrayed on apple juice plates, overlaid with halocarbon oil, and incubated at 23°C until hatching occurred. Percent survival was quantified as the number of first instar larval escapers/total number mutant eggs arrayed.
For standard transmission EM (TEM), late stage 17 wild-type and Gli mutant embryos were injected with fixative according to the protocol of Prokop et al. (1998). Embryos were embedded in Epon-Araldite. For high-pressure freezing and freeze substitution, stage 17 embryos were placed in brass specimen holders, using hexedecene as a support medium, and frozen with a high pressure freezer (model HPM010; BAL-TEC, Liechtenstein). Frozen embryos were transferred to acetone substitution medium (2% OsO4, 8% dimethoxypropane, 0.1% uranyl acetate, and kept at −80°C for 3 d, −20°C for 24 h, 4°C for 2 h, and RT for 2 h. Embryos were washed twice with fresh acetone, and embedded in Spurrs resin (25% resin for 3 h, 50% resin for overnight, 75% resin for 3 h, and 100% resin for overnight). Ultrathin sections were stained after with uranyl acetate and lead citrate, and imaged with a Hitachi 7100 STEM.
We thank M. Bhat, P. Bryant, V. Budnik, R. Fehon, M. Gorczyca, H. Oda, E. Wieschaus, and D. Woods for providing antibodies; M. Bhat, G. Boulianne, W. Leiserson, and H. Keshishian for fly stocks; B. Argiropoulos for embryo injection training; H. Hong for TEM sectioning; M. Pellikka for statistical analysis; and K. Sepp, K. Norman, and C. Roskelley for many helpful discussions and comments on the manuscript.
This work was supported by grants to V.J. Auld from the Canadian Institute of Health Research (MOP42439), and the Howard Hughes Medical Institute (75197-526603). J. Schulte was supported by a studentship from the Rick Hansen Neurotrauma Initiative.
Abbreviations used in this paper: AJ, adherens junction; arm, Armadillo; β-Gal, β-galactosidase; BNB, blood–nerve barrier; Cora, coracle; Dlg, discs large; Dnl, Drosophila neuroligin; Gli, gliotactin; IMP, intermembrane particle; Nrx, neurexin-IV; PDZ, PSD-95/Dlg/ZO-1; pSJ, pleated SJ; Scrib, scribble; SJ, septate junction; sSJ, smooth SJ; TCD, tricellular channel diaphragm; TCP, tricellular plug; TEM, transmission EM; TJ, tight junction.