Snakeskin, a membrane protein associated with smooth septate junctions, is required for intestinal barrier function in Drosophila

The epithelium isolates the body from the outer environment and separates distinct fluid compartments within the body. To accomplish these functions as barriers, epithelial cells have specialized intercellular junctions, termed occluding junctions, that restrict the free diffusion of solutes across the cellular sheets through the paracellular pathway (Furuse and Tsukita, 2006). In vertebrates and tunicates, tight junctions (TJs) create a diffusion barrier within the intercellular space of epithelial and endothelial cells (Lane et al., 1994). Membrane proteins of the claudin family, which have four membrane-spanning domains, are the major players in the core structure of TJs and determine the barrier and channel property of TJs (Angelow et al., 2008; Anderson and Van Itallie, 2009; Furuse, 2010).

In contrast to vertebrate species, epithelial cells in invertebrates generally lack TJs, although a few exceptions have been reported (Lane and Chandler, 1980). Instead, they possess another membrane specialization, called septate junctions (SJs), as their occluding junctions (Lane et al., 1994; Tepass et al., 2001). In ultra-thin section electron microscopy, SJs are visualized as parallel plasma membranes of adjacent cells with an obvious intercellular gap containing ladder-like septa. Morphological variations of SJs exist across invertebrate phyla. Some animals, including coelenterates, arthropods, echinoderms and hemichordates, possess multiple types of SJs depending on the ectodermal or endodermal origin of epithelial cells (Green and Bergquist, 1982; Lane et al., 1994). However, it is largely unknown whether the different SJs are evolutionary variations from a common origin because of a lack of information on their molecular composition. In light of the concept that the paracellular diffusion barrier is a fundamental cellular mechanism in metazoans, the variation in SJs is of great interest in terms of cell biology, comparative physiology and the evolution of animal species.

In arthropods, SJs are divided into two types, pleated SJs (pSJs) and smooth SJs (sSJs) (Lane et al., 1994; Tepass and Hartenstein, 1994). Although both SJs have ladder-like septa, those along the plasma membranes in tracer-infiltrated specimens show zigzagging and smooth lines in pSJs and sSJs, respectively (Lane et al., 1994). In freeze-fracture replicas, rows of intramembranous particles, separated from one another, are observed in pSJs, whereas rows of particles fused into ridges are prominent in sSJs (Lane and Swales, 1982). In Drosophila, pSJs are found in ectodermal epithelia, such as the epidermis, foregut, hindgut and tracheae, and in glia, whereas sSJs are observed in endodermal epithelia, such as the midgut (Tepass and Hartenstein, 1994). The outer epithelial layer of the proventriculus and the Malpighian tubules also have sSJs, although these cells are originally derived from ectoderm during development.

Molecular genetic analyses of Drosophila have identified and characterized many pSJ-associated proteins (Tepass et al., 2001; Banerjee et al., 2006; Furuse and Tsukita, 2006; Nelson and Beitel, 2009). Among these proteins, Nrx (Baumgartner et al., 1996), Nrg (Genova and Fehon, 2003), Cont (Faivre-Sarrailh et al., 2004) and Cora (Tepass et al., 2001) are orthologs of the vertebrate proteins CASPR, neurofascin, contactin and protein 4.1, respectively, which are molecular components of paranodal junctions between neurons and myelinated glial cells, i.e. oligodendrocytes and Schwann cells (Bhat, 2003). Interestingly, Sinu (Wu et al., 2004), Mega (Behr et al., 2003) and Kune (Nelson et al., 2010) are Drosophila homologs of vertebrate claudins, suggesting that pSJs also share some features with TJs. Loss-of-function studies of many pSJ-associated proteins in Drosophila have revealed that pSJs are involved in epithelial barrier function and tracheal tube morphogenesis (Baumgartner et al., 1996; Lamb et al., 1998; Behr et al., 2003; Genova and Fehon, 2003; Llimargas et al., 2004; Wu et al., 2004; Paul et al., 2007; Wu et al., 2007; Nelson et al., 2010). The size and length of tracheal tubes are determined by a highly organized chitinous extracellular matrix. pSJ-related genes have been shown to be required for normal secretion of chitin-modifying enzymes into the lumen of these epithelial tubes (Luschnig et al., 2006; Wang et al., 2006). Furthermore, the pSJ-associated proteins Dlg (Woods et al., 1996), Lgl (Manfruelli et al., 1996) and Scrib (Bilder and Perrimon, 2000), which are not considered to be core components of pSJs (Oshima and Fehon, 2011), are involved in the formation of epithelial polarity and in the regulation of cell growth as Drosophila tumor suppressors. This implies that pSJs act as signaling platforms during epithelial morphogenesis in addition to forming paracellular barriers. Among the pSJ-associated proteins, Cora (Fehon et al., 1994), Nrx (Baumgartner et al., 1996), Lac (Llimargas et al., 2004) and Drosophila claudins, including Sinu (Wu et al., 2004), Mega (Behr et al., 2003) and Kune (Nelson et al., 2010), have all been reported to be highly, or exclusively, expressed in ectodermally derived epithelial tissues and in the glial cells that ensheath the central nervous system (CNS), possibly implying that they are pSJ-specific proteins.

Morphological and physiological studies suggest that sSJs also function to restrict or regulate the diffusion of solutes through the paracellular pathway (Skaer et al., 1987; Beyenbach, 2003). However, in contrast to pSJs, few analyses have been carried out on sSJs in arthropods using molecular biological and genetic approaches. Several studies have described the isolation of sSJ-enriched membrane fractions and have demonstrated protein bands within these fractions by SDS-PAGE, but these proteins have not been characterized further (Green et al., 1983; Lane and Dilworth, 1989; Baldwin and Hakim, 1999). Although the localization of ankyrin, αβ-spectrin, fasciclin III (Fas III) (Baumann, 2001) and Dlg (Maynard et al., 2010) to the sSJ region has been shown in a part of the midgut, so far, no molecular components specific for sSJs have been identified.

Here, the molecular structure of sSJs is clarified as a first step towards a better understanding of the functional and evolutionary aspects of sSJs, and of the biological significance of the two different SJs in arthropods. We identify a new membrane protein that we call ‘Snakeskin’ (Ssk), which is the first demonstration of an sSJ-specific protein conserved within arthropod species. We show that Ssk is essential for Drosophila development and is required for sSJ formation and for the paracellular barrier function in the Drosophila midgut.

To study the molecular architecture of sSJs we identify new molecular components of sSJs in silkworms, from which we can obtain abundant sSJ-containing materials, and then characterize their orthologs in Drosophila by using a combination of genetic and cell biological approaches. We initially isolated the plasma membrane fraction containing sSJs from midgut epithelial cells of the silkworm Bombyx mori (Fig. 1A). This fraction was subsequently injected into rats to generate monoclonal antibodies (mAbs). During screening by immunofluorescence staining of frozen sections of the Bombyx midgut, we identified a mAb clone, designated clone 45, which recognized the upper region of the lateral membrane of epithelial cells in the midgut, where sSJs occur (Fig. 1B). mAb clone 45 precipitated a ~17 kDa protein, tentatively designated as Ag45, from the NP-40-solublized midgut-derived membrane fraction (Fig. 1C). SDS-PAGE-separated Ag45 protein was collected and the N-terminal amino acid sequence was determined. By conducting a database search of the Bombyx mori genome sequence, and hybridization screening of a Bombyx midgut cDNA library, we cloned the cDNA encoding full-length Ag45 (Fig. 1D). Further database searching identified orthologs of Ag45 in other arthropods, including mosquitoes (GenBank accession no. XM_311473), flies (NM_140927) and honeybees (XM_003249611), but not in vertebrates. The Drosophila ortholog for Ag45 is the gene product of CG6981, which has two alternatively spliced transcripts sharing a common protein coding region (Fig. 3A). We cloned CG6981 cDNA and further analyzed it in this study (Fig. 1D).

Bombyx Ag45 and Drosophila CG6981 encode 157- and 162-amino-acid proteins, respectively, and contain four hydrophobic regions that are predicted to be transmembrane domains by the SOSUI algorithm (Hirokawa et al., 1998) (Fig. 1D). Their C-termini do not have typical PDZ-domain-binding motifs (Lee and Zheng, 2010). To study the membrane-spanning topology of CG6981, we overexpressed N-terminally GFP-tagged CG6981 in Drosophila S2 cells and examined the accessibility of an anti-GFP mAb or an anti-CG6981 polyclonal antibody (pAb) raised against the C-terminal 12 amino acids of CG6981. Both antibodies labeled the plasma membranes around the cell borders of the S2 cells expressing the N-terminally GFP-tagged CG6981 only when the cells were permeabilized with Triton X-100 (Fig. 1E). This indicates that the N- and C-termini of CG6981 are exposed to the cytoplasm. Therefore it is reasonable to speculate that CG6981 consists of four transmembrane domains, two short extracellular loops, the N- and C-terminal cytoplasmic domains, and a cytoplasmic turn similar to other cell–cell-junction-associated membrane proteins with four transmembrane domains, such as claudins (Fig. 1F). As a notable feature, two lysine residues are included in the predicted first transmembrane domain, and two (in CG6981) or one (in Ag45) negatively charged amino acids are contained in the predicted fourth transmembrane domain (Fig. 1D). We named CG6981 Snakeskin (Ssk) because of its immunofluorescence staining pattern in the Drosophila midgut (see below).

When we analyzed various stages of Drosophila development by western blotting, Ssk expression was first detected as a protein band of ~15 kDa after AEL 12 hours (Fig. 2A). To investigate further the expression profile of Ssk, Drosophila embryos, larvae and adults were analyzed by immunofluorescence microscopy with anti-Ssk pAb. Ssk protein initially appears at stage 12 in midgut rudiments and its expression is sustained until the adult stage throughout the midgut and Malpighian tubules (Fig. 2B–H). During morphogenesis of the embryonic midgut, Ssk is initially localized along the plasma membrane (Fig. 2C). By late stage 17, Ssk is concentrated into a clear belt surrounding each epithelial cell of the midgut and Malpighian tubules, where sSJs occur (Fig. 2D–H). At the midgut–foregut and midgut–hindgut junctions, Ssk expression is detected clearly only in the midgut (Fig. 2D,G).

The subcellular localization of Ssk in the midgut of the third-instar larvae was examined in more detail by immunoelectron microscopy. sSJs in the Drosophila midgut are distinguished by parallel plasma membranes of adjacent cells with obvious intercellular gaps containing ladder-like septa (Fig. 2I). As shown in Fig. 2J, a substantial part of the lateral membrane from the apical tip, corresponding to the location of sSJs, was strongly labeled with anti-Ssk pAb. When observed under higher magnification, Ssk could be seen colocalizing with parallel plasma membranes, often containing septa of sSJs (Fig. 2K). From these observations, we conclude that Ssk is a membrane protein specifically located at sSJs.

To investigate the physiological function of Ssk, we generated animals whose expression of Ssk was suppressed by ssk-targeting RNA interference (ssk-RNAi) using the ubiquitous da-GAL4 driver (da>ssk RNAi) (Fig. 3A). We also generated a chromosome-deficient strain lacking the ssk (CG6981) gene locus, by flippase-mediated recombination between pairs of flippase recognition target (FRT) sites in the genomic region 77A1 on the left hand of the third chromosome [Df(3L)ssk] (Fig. 3A). An unavoidable problem in the generation of da>ssk RNAi and Df(3L)ssk is that CG42674 is encoded on the complementary DNA strand of ssk. Flybase describes that the CG42674 gene encodes five transcripts, CG42674-RA to -RE. The ssk DNA sequence used for the construction of ssk-RNAi is included within the second intron of CG42674-RE. Furthermore, the first and second exons for CG42674-RE are deleted in Df(3L)ssk, implying that the expression of CG42674-RE is affected in Df(3L)ssk.

Following immunofluorescence staining, only trace levels of Ssk expression were observed in the midgut in da>ssk RNAi embryos compared with that in wild-type (WT) embryos (Fig. 3B–C′). As expected, we could not detect Ssk signals in embryos homozygous for Df(3L)ssk (Fig. 3D,D′). Immunofluorescence staining of stage 16 embryos with the antibody to FasIII, which is highly expressed in mesoderm surrounding the midgut (Patel et al., 1987), revealed that the midgut in da>ssk RNAi and Df(3L)ssk embryos showed constrictions similar to WT embryos (Fig. 3B′–D′). Immunofluorescence staining of da>ssk RNAi embryos at late stage 17 with anti-FasIII antibody showed that the midgut exhibited the normal elongated and folded appearance of that in WT embryos (Fig. 3F,G). These observations suggest that the loss of Ssk does not impair the gross midgut morphogenesis during embryonic development. Notably, however, most of these embryos failed to hatch. Only a small number of embryos hatched to first-instar larvae, and those that did were sluggish in their movements and did not survive to the second-instar stage (data not shown), indicating that Ssk is essential for development.

Furthermore, phase-contrast microscopy revealed that the midgut in WT embryos of late stage 17 appeared to be clear (Fig. 3H) whereas, in da>ssk RNAi embryos, it contained dark material (Fig. 3I), indicating a difference in the contents of the midgut between these embryos. SDS-PAGE analyses detected a triplet of protein bands around ~45 kDa in da>ssk RNAi embryos but not in WT embryos in late stage 17 embryos (Fig. 3J). The sizes of these protein bands corresponded well to those of yolk proteins (Warren and Mahowald, 1979). This observation suggests that the digestion of yolk proteins in the midgut is impaired by the loss of Ssk.

To examine further the effect of the lack of Ssk, we analyzed the morphology of the midgut epithelial cells of da>ssk RNAi and Df(3L)ssk embryos. When the isolated midguts from late stage 17 embryos were labeled with fluorescently tagged phalloidin to delineate the plasma membrane, the anterior midgut epithelial cells in da>ssk RNAi and Df(3L)ssk embryos were more elongated in the apical-basal direction than those in WT embryos (Fig. 4A–C″). Ultra-thin section electron microscopy confirmed this observation. The epithelial cells of the anterior midgut in WT embryos at late stage 17 were cuboidal in appearance, with a slightly rounded apical surface composed of microvilli (Fig. 5A). By contrast, in both da>ssk RNAi (Fig. 5B) and Df(3L)ssk (Fig. 5C) embryos, apical membranes with microvilli often protruded into the lumen although the cells appeared to retain their apicobasal polarity. These observations suggest that Ssk is required for the morphogenesis of midgut epithelial cells.

Next, we compared the detailed structure of sSJs in WT and Ssk-deficient midguts by using electron microscopy. It has been reported that sSJ formation starts at late stage 17 and is completed in the larva (Tepass and Hartenstein, 1994). We confirmed the existence of sSJ structures, distinguished by parallel plasma membranes of adjacent cells with septa, at this stage in WT embryos (Fig. 5E). By contrast, in the midgut of da>ssk RNAi embryos at late stage 17, parallel plasma membranes between cells were markedly reduced and intercellular spaces of irregular width were often observed. The intercellular space in the midgut of da>ssk RNAi embryos appeared to lack electron-dense material and the septa were observed at a much lower frequency in these embryos than in WT embryos (Fig. 5F). In the midgut of Df(3L)ssk embryos, the phenotype was similar to that of da>ssk RNAi embryos, and the characteristic septa of sSJs in the midgut were barely detectable (Fig. 5G).

Next, we attempted to rescue the phenotype of Df(3L)ssk strain by re-expression of Ssk using the da-GAL4 driver, which induces UAS-mediated transgene expression in various tissues, including the midgut and salivary gland. The expression of Ssk in the midgut in these embryos [Df(3L)ssk and da>ssk] at late stage 17 was much less than that in WT embryos, as judged by immunofluorescence staining (Fig. 4D). In addition, the lethality of the Df(3L)ssk strain was not rescued by Ssk re-expression, and the protrusion of the apical membrane of anterior midgut epithelial cells was still observed (Fig. 5D). Accumulation of yolk proteins within the embryonic midgut was also seen in Df(3L)ssk, but the phenotype was not rescued in Df(3L)ssk with da>ssk, (data not shown). However, the parallel plasma membranes of adjacent midgut epithelial cells and a series of septa-like structures were often observed by electron microscopy within the intercellular space in Df(3L)ssk and da>ssk embryos at late stage 17 (Fig. 5H). Taken together, we conclude that Ssk is required for sSJ formation. However, the phenotype of lethality, yolk accumulation and apical protrusion of the midgut epithelial cells, which were not rescued in Df(3L)ssk with da>ssk, cannot be rigorously assigned to the loss of Ssk by our experiments.

To firmly establish whether Ssk is truly specific for sSJs, we performed some basic characterizations of pSJs in da>ssk RNAi and Df(3L)ssk embryos. When stage 17 embryos were immunostained with the antibodies to Kune and Cora, these pSJ-associated proteins in da>ssk RNAi and Df(3L)ssk embryos were found to be similarly localized to those in WT, that is, at the apicolateral membrane of epithelial cells in ectodermally derived tissues, including the hindgut (Fig. 6A–C″). Accordingly, electron microscopy analyses of the hindgut showed that the morphology of pSJs in da>ssk RNAi and Df(3L)ssk embryos appeared quite normal with parallel plasma membranes containing clear septa (Fig. 6D–F). Furthermore, immunofluorescence staining of WT, da>ssk RNAi and Df(3L)ssk embryos at stage 16 with an antibody to Verm, a chitin-modulating protein (Luschnig et al., 2006), showed no substantial difference in the appearance of tracheal tubes or in the polarized secretion of Verm into the tracheal lumen (Fig. 6H,I). These features are known to be affected in the mutants for various pSJ component genes. These observations indicate that Ssk does not influence pSJ formation and support the notion that this protein is exclusively localized to sSJs.

It is widely believed that sSJs regulate the diffusion of solutes through the paracellular pathway (Skaer et al., 1987; Beyenbach, 2003). To investigate the role of Ssk in the barrier function of the midgut, we initially attempted to establish a barrier assay, in which we injected a fluorescent tracer into the body cavity of late stage 17 embryos. However, this assay was inadequate to demonstrate the barrier function of the midgut because tracer diffusion into the midgut lumen was often detected in WT embryos (data not shown). It is likely that sSJs are still not fully formed before hatching in Drosophila, as described previously (Tepass and Hartenstein, 1994). Therefore, WT or ssk-RNAi-expressing first-instar larvae were fed with TRITC-labeled dextran (10 kDa) and observed by using fluorescence microscopy. For this analysis, the expression of ssk-RNAi was delayed by using a combination of the ubiquitous da-GAL4 and temperature-sensitive tub-GAL80^ts, which inhibits GAL4 activity, to increase the number of hatched first-instar larvae (see scheme in Fig. 7A). In WT larvae, the midgut was well defined with the fluorescent tracer confined within the midgut (Fig. 7B,B′). By contrast, the tracer was detected in various parts of body cavity in ssk-RNAi-expressing embryos, consistent with leakage of the tracer from the lumen of the midgut resulting from Ssk deficiency (Fig. 7C,C′). These results indicate that Ssk is required for the barrier function of the midgut epithelium in Drosophila.

SJs are common intercellular junctions in invertebrates and are thought to function as paracellular diffusion barriers. Of the two types of SJs observed in arthropods, sSJs are the hallmark of endodermal epithelial cells but, so far, no molecular constituents have been shown to be specifically associated with sSJs. Here, we identified and cloned the first example of an sSJ-specific membrane protein, which we designated Ssk. We further demonstrated that Ssk is required for sSJ formation.

Like pSJs, sSJs are distinguished by parallel plasma membranes, with a spacing of 15–20 nm, and septa (Lane et al., 1994). Given that Ssk contains four transmembrane domains and each of the two predicted extracellular loops consists of only about 20 amino acids, Ssk is unlikely to be the actual septal structure bridging the intercellular space of sSJs. Interestingly, Ssk contains two positively charged amino acids in the first transmembrane domain and two negatively charged amino acids in the fourth transmembrane domain. These domains possibly associate with each other by electrostatic forces to form intramembrane protein complexes, which constitute the particle strands observed within sSJ membranes upon freeze-fracture replica electron microscopy. This kind of interaction within the plasma membrane has been reported between the T cell receptor and CD3 (Call et al., 2002). In the Ssk-deficient midgut, septa were rarely detected at sSJ regions. Therefore, Ssk might work as an adaptor to recruit sSJ-associated adhesion molecules, which form the septa. By analogy with the complex molecular structure of pSJs, sSJs might also contain other types of components, such as adhesion molecules and plaque proteins that remain unidentified. Future searches for such molecules using Ssk as a probe (e.g. in co-precipitation experiments) should lead to a better understanding of the molecular architecture of sSJs. Further analyses of these molecules will help to unravel the roles of sSJs in the morphogenesis and function of endodermal epithelial cells.

Our analyses of compromised Ssk expression in Drosophila suggest a crucial role for sSJs in epithelial barrier function. The da>ssk RNAi and Df(3L)ssk embryos were lethal at the late stage 17. In these embryos, yolk remained within the midgut, whereas it was hardly detected in WT embryos at the same stage. This suggests defects in the midgut functions of digestion and absorption before hatching, but how sSJs are involved in these functions remains elusive. Further investigation of the fluid environment in the lumen of the midgut, including pH and electrolyte concentrations, as well as the transepithelial potential difference, in these embryos is necessary to understand the role of sSJs in the functions of midgut.

Interestingly, midgut epithelial cells lacking Ssk also had an abnormal shape, with their apical membranes extending into the lumen, suggesting that sSJs are involved in the morphogenesis of midgut epithelial cells. A recent analysis of protein dynamics within Drosophila pSJs suggests that pSJ-associated proteins can be divided into two groups: SJ core components and the proteins such as Dlg, Lgl and Scrib that establish basolateral polarity (Oshima and Fehon, 2011). Mutations in the genes encoding core components of SJs do not affect cell shape in most ectodermally derived tissues, including the epidermis and salivary glands, but they do increase the length of tracheal cells (Behr et al., 2003; Genova and Fehon, 2003; Llimargas et al., 2004; Wu et al., 2004; Paul et al., 2007; Wu et al., 2007; Nelson et al., 2010). In ectodermally derived epithelial cells in Drosophila, belt-like adherens junctions (AJs) circumscribe the most apical part of the lateral membrane. By contrast, it has been reported that belt-like AJs are not observed in midgut epithelial cells, but that these cells instead have spot-like AJs (Tepass and Hartenstein, 1994; Tepass et al., 2001). Given that belt-like AJs play crucial roles in the morphogenesis of epithelial cells, including in the formation of cell–cell contacts and cell polarity in animal cells in general (Harris and Tepass, 2010), sSJs might play important roles in the morphogenesis of epithelial cells instead of belt-like AJs. Another possible explanation for the apical protrusion phenotype is that apical-basal polarity is mildly disrupted in the midgut cells lacking Ssk. In ectodermally derived epithelia, some pSJ-associated proteins play a role in maintaining basolateral identity (Bilder and Perrimon, 2000; Bilder et al., 2003; Tanentzapf and Tepass, 2003; Laprise et al., 2009). A further possibility is that disruption of the intestinal barrier function by the loss of Ssk results in the leakage of signaling molecules from the intestinal lumen to the basal side, inducing abnormal signal transduction during the morphogenesis of epithelial cells. To clarify the functions of sSJs in the morphogenesis of midgut epithelial cells, other molecular constituents of sSJs should be identified and analyzed in the context of cell polarity formation and cytoskeletal organization.

Contrary to our expectations, re-expression of Ssk in Df(3L)ssk embryos by using the da-GAL4 driver did not rescue the phenotypes of lethality, yolk accumulation or the apical protrusion of the midgut epithelial cells. One possible explanation is that the amount of exogenous Ssk provided was too low to fully rescue the function of Ssk in the midgut. Alternatively, the fact that a part of the CG42674 gene is encoded on the complementary DNA chain of ssk might have caused these results, although the function of CG42674 protein has not yet been clarified. The CG42674 gene generates five transcripts, CG42674-RA to -RE (Fig. 3A). The first and second exons for CG42674-RE containing the predicted initiation codon are deleted in Df(3L)ssk, suggesting that the expression of the protein translated from CG42674-RE is impaired in this strain. On the other hand, the DNA sequence for the dsRNA expressed in ssk-RNAi is included within the second intron for CG42674-RE. Although it was initially considered that RNAi disrupts mRNAs in the cytoplasm, additional studies have shown that pre-mRNA sequences containing introns can also be targeted (Seinen et al., 2011), implying that the expression of the protein product from CG42674-RE could be affected even in ssk-RNAi embryos. Therefore, except for the disruption of sSJ structures, currently the phenotypes observed in Df(3L)ssk and ssk-RNAi cannot be rigorously assigned to the loss of Ssk.

In contrast to TJs, the molecular mechanisms concerned with the epithelial barrier function of SJs in invertebrates have not been well elucidated. However, given that TJs are generally restricted to chordates, SJs might be the prototype of occluding junctions that occur across phyla. Some molecular components of pSJs are conserved with those of vertebrate paranodal junctions generated between neuronal and glial cells (Bhat, 2003). pSJs also contain homologs of claudins, the major integral membrane proteins of vertebrate TJs (Behr et al., 2003; Wu et al., 2004; Nelson et al., 2010). These findings imply that there is a substantial relationship between pSJs and paranodal junctions or TJs in animal evolution. However, Ssk homologs are conserved only within arthropod specie, suggesting that sSJs are arthropod-specific structures. The physiological significance of the presence of two types of SJs (i.e. pSJs and sSJs) in arthropods remains elusive. It might reflect the different functions between the ectoderm and endoderm: the ectoderm has a exoskeleton cuticle of chitin to protect the animal body against the external environment and infectious organisms, whereas the endoderm generates digestive tubes that absorb nutrients. Interestingly, various invertebrates, including coelenterates, arthropods, echinoderms and hemichordates, appear to possess multiple types of morphologically different SJs, depending on the ectodermal or endodermal origin of epithelial cells (Green and Bergquist, 1982; Lane et al., 1994). However, so far the significance of such a differences has been hardly discussed. Further investigations of the molecular architecture of sSJs, and comparisons with pSJs or SJs in other invertebrate species, will lead to a better understanding, not only of the physiological function of sSJs and pSJs in arthropods but also of the evolution of intercellular junctions across the phyla in the context of epithelial barrier function and morphogenesis.

To generate anti-Ssk (CG6981) polyclonal antibodies, a polypeptide corresponding to the C-terminal cytoplasmic domain (CIHYSKGDTDYTQ) of Ssk was synthesized and coupled via the cysteine residue to keyhole limpet hemocyanin. This peptide was injected into rabbits as an antigen. Rabbit antisera were affinity purified on nitrocellulose membranes with GST fusion proteins with the Ssk cytoplasmic domain. Anti-FasIII mouse mAb DA.1B6 (Brower et al., 1980) and anti-Verm rabbit polyclonal antibody (Luschnig et al., 2006) were kindly provided by Danny L. Brower (University of Arizona, Tucson, AZ) and Stefan Luschnig (University of Zurich, Zurich, Switzerland), respectively. Anti-Cora mouse mAb C615.16 was obtained from the Developmental Studies Hybridoma Bank (University of Iowa). Rabbit anti-Kune serum was characterized as described previously (Nelson et al., 2010). The primary antibodies used were: affinity-purified rabbit anti-Ssk (CG6981) antiserum (1:200); rabbit anti-Kune antiserum (1:1000); mouse anti-GFP (1:1000) (Roche); mouse anti-FasIII mAb DA.1B6 (1:100); mouse anti-Cora mAb C615.16 (1:20); rabbit anti-Verm pAb (1:500). Secondary antibodies were used at 1:500 (Molecular Probes and Jackson ImmunoResearch).

The Oregon R strain of Drosophila melanogaster was used as the wild type and for phenotypic analyses. Homozygous flies were identified using a TM6B ubi-GFP balancer chromosome. The fly strain w; tub-gal80^ts was obtained from the Bloomington Stock Center. The Df(3L)ssk strain was created by FLP-mediated recombination of FRT-bearing P element (P{XP}d09071 and PBac{RB}e01995), which was detected by loss of w⁺ and confirmed by genomic PCR and DNA sequence analysis (Parks et al., 2004).

The plasma membrane fraction containing sSJs was prepared from the midgut of the fifth-instar larvae of Bombyx mori. The isolated midgut from which the peritrophic membrane was removed was homogenized in 1 mM NaCO₃ containing 2 mg/ml of leupeptin (the hypotonic solution) with a loose-fitting Teflon homogenizer on ice. The homogenate was filtered with gauze and the filtrate was centrifuged at 3000 g for 30 minutes at 4°C. The precipitated fraction, mainly containing the plasma membranes and nuclei was resuspended in the hypotonic solution and then added to 81% sucrose to a final concentration of 60% (w/v). This fraction was subjected to stepwise sucrose density gradient ultracentrifugation together with 54% and 32% sucrose solutions at 100,000 g (Beckman SW28 rotor, 27,000 r.p.m.) for 90 minutes at 4°C. The membrane at the interface between 54% and 32% sucrose was collected, diluted in the hypotonic solution, and then precipitated by centrifugation. The precipitate was resuspended with the hypotonic buffer and then added with 81% sucrose to a final concentration of 60%. This fraction was again subjected to stepwise sucrose density gradient ultracentrifugation together with 50%, 42% and 36% sucrose solutions at 100,000 g (Beckman SW28 rotor, 27,000 r.p.m.) for 90 minutes at 4°C. The membranes at the interfaces of 50%–42% and 42%–36% were collected and stored at −80°C as the midgut membrane. To prepare the antigen for immunization, the midgut membrane was washed with the hypotonic solution by centrifugation and treated with 4 M guanidine chloride to extract peripheral membrane proteins, followed by washing twice with the hypotonic solution by centrifugation. Rat monoclonal antibodies were generated as described previously (Furuse et al., 1993). The hybridoma supernatants were screened by immunofluorescence staining of frozen sections of the midgut isolated from fifth-instar larvae of silkworm, which were processed as described previously (Furuse et al., 1993).

To purify the antigen protein for monoclonal antibodies, the midgut membrane was solubilized with PBS containing 1% NP-40 and incubated with monoclonal antibodies bound to Protein G beads (GE Healthcare) at 4°C for 2 hours. After washing with PBS, the bound protein and IgG were subjected to SDS-PAGE, followed by Coomassie Brilliant Blue staining or silver staining using the Silver Stain Kit (Wako). The band of precipitated protein designated Ag45 were excised from the polyacrylamide gels and processed for amino acid sequence analysis as described previously (Furuse et al., 1998). The partial amino acid sequence of Ag45 was compared with the genome sequence of B. mori, and a corresponding genome sequence was obtained. The partial cDNA fragment of Ag45 was obtained by RT-PCR with total RNA isolated from the silkworm midgut. By using this cDNA fragment as a probe, we screened a lambda ZAP-based cDNA library for the midgut isolated from fifth-instar larvae of silkworm, and obtained the cDNA containing the whole open reading frame of Ag45. The cDNA sequence of CG6981, designated Ssk, a Drosophila ortholog of Ag45, was identified by database searching. To generate the expression constructs for Ssk, the open reading frame of Ssk was amplified by PCR from Drosophila embryonic cDNA and cloned into pUAST (Brand and Perrimon, 1993) with or without an N-terminal GFP tag. For RNAi experiments, a DNA fragment containing the full open reading frame of ssk was amplified by PCR with the forward primer corresponding to the 5′ non-coding region with a NotI site (5′-gcggccgcAACGAATCATTTCGCTGTAC-3′) and the reverse primer corresponding to the 3′ non-coding region with a KpnI site (5′-ggtaccAATGTGGATGGTGTTTATTCGGC-3′) and subcloned into pGEM-T Easy Vector (Promega). The DNA fragment excised by digestion with NotI and KpnI was inserted into pUAST as a head-to-head dimer and transformed into SURE2 competent cells (Stratagene). Transgenic flies were generated by standard P-element transformation procedures.

S2 cells were grown in Schneider's Drosophila medium (Invitrogen) containing 10% fetal calf serum. Cells were maintained at 28°C under a normal atmosphere. For transient transfections, pUAST, pUAST-Ssk and pUAST-GFP-Ssk were co-transfected with an actin-Gal4-encoding plasmid using lipofectamine LTX reagent (Invitrogen). Transfected cells were cultured for 48 hours and harvested.

Embryos were collected, dechorionated with bleach, fixed with 4% paraformaldehyde and heptane for 20 minutes, and devitellinized in methanol. When late stage 17 embryos were stained in situ, the embryos were briefly sonicated to permeabilize the cuticle. Late stage 17 embryos and larval midguts were dissected in 4% paraformaldehyde in PBS, and incubated for 30–60 minutes. S2 cells were fixed with 4% paraformaldehyde in PBS for 20 minutes. After fixation, specimens were washed with PBS and blocked with 2% BSA in PBS with 0.3% Triton X-100. Thereafter, they were incubated with primary antibodies at 4°C overnight. Following five washes, they were incubated with secondary antibodies for 3 hours. After five additional washes, they were embedded and imaged with a confocal laser scanning microscope (TCS SP2; Leica) with a HCX PL FLUOTAR 10× NA 0.30 objective lens (third-instar larva anterior midgut), HC PLAN APO 20× NA 0.70 objective lens (embryos, third-instar larva middle and posterior midgut and adult midgut), and HCX PL APO 63× NA 1.40 oil immersion objective lens (late stage 17 midgut and S2 cells) at room temperature.

Flies and S2 cells were lysed in sample buffer (7 M urea, 2 M thiourea, 4% SDS/10% glycerol, 100 fmM DTT, 100 mM Tris-HCl pH 6.8) and incubated with rocking overnight at room temperature. 25 μg (fly lysate) or 1 μg (S2 cell lysate) of protein was loaded into each lane. Subsequently, proteins were separated by SDS-PAGE and electrophoretically transferred from gels onto PVDF membranes, which were then incubated with the primary antibody. Bound antibodies were detected with biotinylated secondary antibodies and streptavidin-conjugated alkaline phosphatase (GE Healthcare). Nitroblue tetrazolium and bromochloroindolyl phosphate were used as substrates for the detection of alkaline phosphatase.

Crosses were performed and embryos were collected at 2-hour intervals at 25°C. These were aged for 28 hours at 17°C and subsequently for 16 hours at 30°C. Tetramethylrhodamine-labeled dextran (MW 10,000; Molecular Probes), in a solution of 10% sucrose, was fed to first-instar larvae for 3–4 hours. After five washes with PBS with 0.3% Triton X-100, larvae were anesthetized with ether and immediately imaged with an inverted fluorescence microscope (FX81; Olympus) with a UPlan SApo 10× NA 0.40 objective lens and a cooled CCD camera (model ORCA-AG; Hamamatsu Photonics K K.) controlled by MetaMorph software (Molecular Devices) at room temperature.

Conventional electron microscopy was performed as previously described (Furuse et al., 1998), with the exception that tannic acid was not used in fixation. For immunoelectron microscopy, third-instar larvae were dissected in 4% paraformaldehyde and 0.1% glutaraldehyde in PBS and incubated for 60–90 minutes. After fixation, guts were washed three times with glycine in PBS. Subsequently, they were cryoprotected with 20% sucrose in PBS and stored in 25% sucrose in PBS at 4°C overnight. The specimens were frozen in liquid nitrogen and 10-μm thick sections were cut on a cryostat. The sections were blocked with 4% normal goat serum and 0.01% saponin in PBS for 1 hour and incubated overnight with a primary antibody at 4°C. After five washes, the sections were incubated overnight with a secondary antibody conjugated with 1.4-nm NANOGOLD particles (Nanoprobes, Inc.) at 4°C. Following five washes, the sections were fixed for 20 minutes with 1% glutaraldehyde in phosphate buffer (pH 7.4), washed once with phosphate buffer, four times with 50 mM HEPES (pH 5.8) and two times with distilled water. Signals were silver-enhanced by the HQ-silver kit (Nanoprobes Inc.). The sections were again washed with distilled water and postfixed with 0.5% osmium oxide in 0.1 M phosphate buffer for 45 minutes on ice, followed by en bloc staining, dehydration and embedding, which was performed in the same way for conventional electron microscopy. Ultra-thin sections were generated, stained with uranyl acetate and lead citrate, and observed under an electron microscope (Model JEM-1011; JEOL).

We are grateful to C. Fujiwara for her excellent technical assistance and to R. Niwa, R. Murakami, A. Nagafuchi, H. Hiroaki, Y. Fujiwara and all the members of Furuse and Uemura laboratories for helpful discussions. We also thank D. L. Brower for anti-FasIII mAb, S. Luschnig for anti-Verm pAb, and the Bloomington Stock Center and the Drosophila Genetic Resource Center at Kyoto Institute of Technology for fly stocks.