A tetraspanin regulates septate junction formation in Drosophila midgut

Epithelia separate distinct fluid compartments within the bodies of metazoans. For this epithelial function, specialized intercellular junctions, designated as occluding junctions, regulate the free diffusion of solutes through the paracellular pathway. In vertebrates, tight junctions act as occluding junctions, whereas in invertebrates, septate junctions (SJs) are the functional counterparts of tight junctions (Anderson and Van Itallie, 2009; Furuse and Tsukita, 2006; Lane et al., 1994; Tepass and Hartenstein, 1994). Septate junctions form circumferential belts around the apicolateral regions of epithelial cells. In transmission electron microscopy, septate junctions are observed between the parallel plasma membranes of adjacent cells, with ladder-like septa spanning the intermembrane space (Lane et al., 1994; Tepass and Hartenstein, 1994). Septate junctions are subdivided into several morphological types that differ among different animal phyla, and several phyla possess multiple types of septate junctions that vary among different types of epithelia (Banerjee et al., 2006; Green and Bergquist, 1982; Lane et al., 1994).

In arthropods, two types of septate junctions exist; pleated (pSJs) and smooth (sSJs) (Banerjee et al., 2006; Lane et al., 1994; Tepass and Hartenstein, 1994). pSJs are found in ectodermally derived epithelia and surface glia surrounding the nerve cord, whereas sSJs are found mainly in endodermally derived epithelia, such as the midgut and the gastric caeca (Lane et al., 1994; Tepass and Hartenstein, 1994). The outer epithelial layer of the proventriculus (OELP) and the Malpighian tubules also possess sSJs, although these epithelia are ectodermal derivatives (Lane et al., 1994; Tepass and Hartenstein, 1994). The criteria distinguishing these two types of septate junctions are the arrangement of the septa; in oblique sections of lanthanum-treated preparations, the septa of pSJs are visualized as regular undulating rows, but those in sSJs are observed as regularly spaced parallel lines (Lane et al., 1994; Lane and Swales, 1982). In freeze-fracture images, the rows of intramembrane particles in pSJs are separated from one another, whereas those in sSJs are fused into ridges (Lane et al., 1994; Lane and Swales, 1982). To date, more than 20 pSJ-related proteins, including pSJ components and regulatory proteins involved in pSJ assembly, have been identified and characterized in Drosophila (Banerjee et al., 2006; Byri et al., 2015; Deligiannaki et al., 2015; Hildebrandt et al., 2015; Izumi and Furuse, 2014; Tepass et al., 2001; Wu and Beitel, 2004). By contrast, few genetic and molecular analyses have been carried out on sSJs. Recently, two sSJ-specific membrane proteins, Ssk and Mesh, have been identified and characterized (Izumi and Furuse, 2014; Izumi et al., 2012; Yanagihashi et al., 2012). Ssk consists of 162 amino acids and has four membrane-spanning domains, two short extracellular loops, cytoplasmic N- and C-terminal domains, and a cytoplasmic loop (Yanagihashi et al., 2012). Mesh has a single-pass transmembrane domain and a large extracellular region containing a NIDO domain, an Ig-like E set domain, an AMOP domain, a vWD domain, and a sushi domain (Izumi et al., 2012). Mesh transcripts are predicted to be translated into three isoforms of which the longest isoform consists of 1454 amino acids. In western blot studies, Mesh is detected as a main ∼90 kDa band and a minor ∼200 kDa band (Izumi et al., 2012). Compromised expression of Ssk or mesh causes defects in the ultrastructure of sSJs and in the barrier function of the midgut against a 10-kDa fluorescent tracer (Izumi et al., 2012; Yanagihashi et al., 2012). Ssk and Mesh physically interact with each other and are mutually dependent for their sSJ localization (Izumi et al., 2012). Thus, Mesh and Ssk play crucial roles in the formation and barrier function of sSJs.

Tetraspanins are a family of integral membrane proteins in metazoans with four transmembrane domains, N- and C-terminal short intracellular domains, two extracellular loops and one short intracellular turn. Among several protein families with four transmembrane domains, tetraspanins are characterized especially by the structure of the second extracellular loop. It contains a highly conserved cysteine–cysteine–glycine (CCG) motif and two to four other cysteine residues. These cysteines form two or three disulfide bonds within the loop (Charrin et al., 2014,, 2009; Hemler, 2005; Yanez-Mo et al., 2009). Tetraspanins are believed to play a role in membrane compartmentalization and are involved in many biological processes, including cell migration, cell fusion and lymphocyte activation, as well as viral and parasitic infections (Charrin et al., 2014,, 2009; Hemler, 2005; Yanez-Mo et al., 2009). Several tetraspanins regulate cell–cell adhesion (Chattopadhyay et al., 2003; Ishibashi et al., 2004; Shigeta et al., 2003) but none are known to be involved in the formation of epithelial occluding junctions. In the Drosophila genome, there are 37 tetraspanin family members (Charrin et al., 2014; Fradkin et al., 2002; Hemler, 2005; Todres et al., 2000), and some have been characterized by genetic analyses. Lbm, CG10106 and CG12143 participate in synapse formation (Fradkin et al., 2002; Kopczynski et al., 1996). Sun associates with light-dependent retinal degeneration (Xu et al., 2004). TspanC8 subfamily members, including Tsp3A, Tsp86D and Tsp26D, are involved in the Notch-dependent developmental processes through the regulation of a transmembrane metalloprotease, ADAM10 (human ortholog of Kuz in Drosophila) (Dornier et al., 2012). However, the functions of most other Drosophila tetraspanins remain obscure.

Here, we identify a tetraspanin family protein, Tsp2A, as a newly identified molecular component of sSJs in Drosophila. We demonstrate that Tsp2A is required for sSJ formation and for the barrier function of Drosophila midgut. Tsp2A and two other sSJ-specific membrane proteins, Mesh and Ssk, show mutually dependent localizations at sSJs and form a complex with each other. Therefore, we conclude that Tsp2A cooperates with Mesh and Ssk to organize sSJs.

We previously identified Mesh as a molecular component of sSJs (Izumi et al., 2012). Because sSJs are located in the apicolateral region of the plasma membrane in endodermally derived epithelia, apicolateral localization of Mesh might be considered to indicate the occurrence of sSJs. To identify genes responsible for sSJ formation, we attempted to obtain Drosophila strains defective in the apicolateral accumulation of Mesh by a genetic screen of a chromosomal deficiency stock, Bloomington Deficiency Kit. Previously, it was reported that the formation of sSJs is completed at the end of embryogenesis, by late stage 17 in the midgut epithelia, and by early stage 17 in the OELP (Tepass and Hartenstein, 1994). Here, we observed apicolateral accumulation of Mesh in the stage 16 OELP, suggesting sSJ formation in the OELP commences during this stage (Fig. 1A, upper panel). Because most chromosomal deficiencies in the kit are embryonic lethal, we evaluated sSJ formation in our screen by immunofluorescence staining of the stage 16 OELP with an anti-Mesh antibody. We identified several deficiencies that caused a defect in the accumulation of Mesh in the region of sSJs in the OELP.

In the OELP of Df(1)BSC534 (deleted segment: 1D1-2A3), Mesh was distributed diffusely in the cytoplasm (Fig. 1A, lower panel). To more precisely identify the genomic region responsible for the phenotype, we investigated the Mesh distribution of other deficiencies overlapping with Df(1)BSC534. As Df(1)BSC709 (deleted segment: 1E4-2A1) and Df(1)Exel6227 (deleted segment: 1F3-2B1) OELP exhibited the phenotype of the cytoplasmic distribution of Mesh, the Df(1)BSC534 phenotype was mapped to the 1F3-2A1 interval (data not shown). Within the genomic region 1F3-2A1, we focused on the gene for Tsp2A (Fig. 1B) because a Flybase search revealed that it is highly expressed in the midgut and the Malpighian tubules (http://flybase.org/reports/FBgn0024361.html). The Tsp2A gene encodes a tetraspanin family protein (Fig. 1D) (Fradkin et al., 2002; Todres et al., 2000). To examine whether a lack of Tsp2A caused the cytoplasmic localization of Mesh, we expressed N-terminal EGFP-tagged Tsp2A (EGFP–Tsp2A) in Df(1)Exel6227 using da-GAL4 (Fig. 2C,C′, see below). Interestingly, apicolateral accumulation of Mesh in the OELP was recovered and EGFP–Tsp2A was colocalized with Mesh (Fig. 2C,C′) in EGFP–Tsp2A-induced Df(1)Exel6227, whereas Mesh remained in the cytoplasm in control Df(1)Exel6227 (Fig. 2D′). These observations strongly suggest that among the genes deleted in Df(1)Exel6227, Tsp2A is responsible for sSJ localization of Mesh, and that Tsp2A itself is a component of sSJs.

To determine the expression pattern and the subcellular localization of endogenous Tsp2A, we obtained two anti-Tsp2A antibodies raised against the C-terminal cytoplasmic region of Tsp2A. Immunofluorescence microscopic analyses revealed that both of the affinity-purified anti-Tsp2A antibodies labeled the apicolateral region of the OELP in late-stage embryos (Fig. 2A,E; Fig. S1C,G). The staining pattern of the OELP with these antibodies distinctly overlapped that of an anti-Mesh antibody (Fig. 2A,A′,E–E″; Fig. S1C–C″,G–G″). Furthermore, the immunoreactivities of these antibodies in the OELP were diminished in Df(1)Exel6227 (Fig. 2B) and in Tsp2A mutant embryos (Fig. S1D,H, see below), indicating the specificity of these anti-Tsp2A antibodies. Immunofluorescence staining of first-instar larvae with one of the anti-Tsp2A antibodies revealed strong honeycomb-like signals of Tsp2A in the midgut, OELP and Malpighian tubules, but not in the foregut and hindgut (Fig. 2F,G). At higher magnification, the staining with the anti-Tsp2A antibody overlapped well with that of the anti-Mesh antibody in apicolateral regions of the midgut epithelial cells (Fig. 2I–I″). The anti-Tsp2A antibody also labeled apicolateral regions of cell–cell contact in adult midgut epithelial cells (Fig. 2H), and coincided with staining with the anti-Mesh antibody in these regions (Fig. 2J–J″). Taken together, these observations indicate that Tsp2A is a component of sSJs in Drosophila from the embryo to adulthood.

To confirm that Tsp2A is involved in the sSJ localization of Mesh, we induced Tsp2A-RNAi (11415R-2 generated by NIG-Fly) using 48Y-GAL4 in wild-type larvae. In the Tsp2A-RNAi first-instar larvae, the levels of Tsp2A were decreased in the sSJs of the midgut epithelial cells and the OELP (Fig. 3B and data not shown). In these cells, Mesh was distributed diffusely in the cytoplasm (Fig. 3B′,F′; Fig. S2B′) and Ssk was mislocalized to the apical membrane, basolateral membrane and cytoplasm (Fig. 3F; Fig. S2B). These results indicate that Tsp2A is required for the sSJ localizations of Mesh and Ssk.

To further confirm the role of Tsp2A in the molecular organization of sSJs, we generated Tsp2A mutants using the CRISPR/Cas9 method provided by NIG-Fly (See Materials and Methods section) (Kondo and Ueda, 2013). We successfully obtained three independent Tsp2A mutant strains (Tsp2A^1-2, Tsp2A^2-9 and Tsp2A^3-3), all of which had small indel mutations encompassing the target site (Fig. 1C). These Tsp2A mutant embryos hatched into first-instar larvae but died at this stage (data not shown). Mutations in all Tsp2A mutants caused frameshifts and premature stop codons. Among these mutant strains, we mainly used Tsp2A^1-2 for further experiments. In western blot analysis, the anti-Tsp2A antibodies detected a large number of bands in the extracts of both wild-type and Tsp2A^1-2 first-instar larvae (Fig. S3A). Among them, only a ∼21 kDa band disappeared in Tsp2A^1-2 extracts (Fig. S3A), implying that the ∼21 kDa band represents Tsp2A and that the other bands commonly observed in both the wild type and the Tsp2A^1-2 mutants originated from cross-reactions of these antibodies. We also noticed that a ∼18 kDa band was detected by an anti-Tsp2A antibody (301) only in Tsp2A^1-2 extracts (Fig. S3A, left panel). The open reading frame of the Tsp2A gene has an in-frame ATG codon at the 94th nucleotide and it is likely that the ∼18 kDa band corresponds to a peptide translated from this ATG. Because the ATG codon at the 94th nucleotide is present in the coding region corresponding to the middle of the first transmembrane domain of Tsp2A, the peptide translated from this ATG might be nonfunctional because of an abnormal conformation. Consistently, anti-Tsp2A antibodies did not stain the sSJ region in first-instar larva of Tsp2A^1-2 mutants (Fig. 3C,D). In the Tsp2A^1-2 midgut and the OELP, Mesh was distributed diffusely in the cytoplasm of the epithelial cells (Fig. 3C′; Fig. S2C′) and Ssk was mislocalized to the apical and basolateral membranes and cytoplasm (Fig. 3G,H; Fig. S2C,D), as also observed in the Tsp2A-RNAi midgut (Fig. 3B′,F,F′; Fig. S2B,B′). Tsp2A^2-9 and Tsp2A^3-3 mutants showed the same phenotype (Fig. S2F–I‴). Other sSJ components, such as Lgl [also known as L(2)gl], Cora and FasIII (also known as Fas3), were also mislocalized in the Tsp2A-RNAi and Tsp2A^1-2 mutant midgut; Lgl was distributed along the basolateral membrane (Fig. S4D,G), Cora was partially localized to the apicolateral regions but also was distributed diffusely throughout the cytoplasm (Fig. S4E,H), and FasIII was observed as large aggregates in the apicolateral and apical plasma membrane regions (Fig. S4F,I). By contrast, Dlg was still localized at the apicolateral region in the RNAi and mutant midgut (Fig. 3B″,C″,D″,F″,G″,H″; Fig. S2B″,C″,D″,F″,G″,H″,I″), suggesting that Tsp2A does not play a major role in specifying the apical–basal polarity of epithelial cells. Taken together, these results indicate that Tsp2A is required for the proper localization of a number of sSJ components.

Western blot analyses revealed that the densities of the main bands of Ssk (∼15 kDa) and Mesh (∼90 kDa) were not notably changed in Tsp2A^1-2 mutant larva, compared with wild type (Fig. S3B,D). However, two minor bands of Mesh at ∼200 kDa in the wild type converged into a single band with the higher molecular mass in the Tsp2A^1-2 mutant (Fig. S3D). This suggests that some post-translational processing of Mesh is affected by the loss of Tsp2A, although it is unknown whether there are differences in the biochemical properties of the two bands of Mesh around 200 kDa.

To examine whether the role of Tsp2A is specific to sSJs, we observed the localization of the pSJ components, Kune and Cora, in the ectodermally derived epithelial cells in Tsp2A^1-2 mutants. In stage 16 embryos of both wild type and Tsp2A^1-2 mutants, Kune and Cora were found to be similarly localized within the apicolateral regions of the hindgut, epidermis and the salivary gland (Fig. S4J–K″ and data not shown), indicating that Tsp2A does not influence the organization of pSJs.

We next examined the Tsp2A distribution during the sSJ formation process by immunofluorescence staining of wild-type embryos from stage 15 to stage 16. In the OELP of stage 15 embryos, Tsp2A was localized in aggregations along the lateral membranes (Fig. S1A,A″,E,E″) and accumulated into the apicolateral plasma membrane region in the stage 16 OELP, suggesting that Tsp2A is incorporated into the sSJs at stage 16 (Fig. S1C,C″,G,G″). These signals are specific for Tsp2A, because they were diminished in Tsp2A mutants (Fig. S1B,D,F,H). By contrast, Mesh was localized along the lateral membranes with partial accumulation in the apicolateral region in the stage 15 OELP, but did not form aggregates (Fig. S1A′,A″,E′,E″). In the stage 16 OELP, Mesh accumulated in the apicolateral regions and colocalized with Tsp2A (Fig. S1C′,C″,G′,G″). Thus, the sSJ targeting process of Tsp2A seems to differ from that of Mesh.

To test whether the Tsp2A^1-2 mutation affects the assembly or maintenance of sSJs, we monitored the distribution of Mesh during sSJ maturation in wild-type and Tsp2A^1-2 embryos. In the wild-type OELP, Mesh began to localize in the apicolateral region at stage 15 (Fig. S1A′,E′), and exclusively accumulated at sSJs at stage 16 (Fig. S1C′,G′). By contrast, in the Tsp2A^1-2 OELP, Mesh was distributed diffusely in the cytoplasm from stage 15 to stage 16 (Fig. S1B′,D′,F′,H′). Taken together, these results suggest that Tsp2A is required for the initial assembly of sSJs.

To investigate the role of Tsp2A in sSJ formation, the ultrastructure of the first-instar larval midgut in Tsp2A^1-2 mutants was examined by electron microscopy of ultrathin sections. In wild-type midgut epithelial cells, sSJs were observed as parallel plasma membranes between adjacent cells with ladder-like septa in the apicolateral regions of cell–cell contacts (Fig. 4A–D). In the Tsp2A^1-2 mutant, two types of defects were observed in the ultrastructure of the apicolateral regions of cell–cell contacts. First, septa-like structures were occasionally visible but they were less electron-dense than the septa of sSJs in wild-type midgut (Fig. 4E–H), and the plasma membranes of adjacent cells were often more closely apposed than in the wild type (Fig. 4E). Second, large gaps were formed between the lateral membranes of adjacent cells (Fig. 4I–L). These results indicate that Tsp2A is required for proper sSJ formation at the ultrastructural level.

We next examined whether Tsp2A is involved in the barrier function of the midgut epithelium, as previously observed for Mesh and Ssk (Izumi et al., 2012; Yanagihashi et al., 2012). We designed a dye permeability assay, which analyzed leakage from the midgut to the body cavity of a fluorescent tracer taken with the food (Izumi et al., 2012; Yanagihashi et al., 2012). During the initial assay, we found that the Tsp2A-RNAi 11415R-2 first-instar larvae ate the food but the Tsp2A mutant larvae did not, although both animals died during the larval stage. Thus, we used the Tsp2A-RNAi 11415R-2 larvae for further experiments. Control first-instar larvae (+/UAS-Tsp2A-RNAi 11415R-2) and Tsp2A-RNAi 11415R-2 first-instar larvae (48Y-GAL4/UAS-Tsp2A-RNAi 11415R-2) were fed yeast containing fluorescently labeled dextran of 10 kDa and observed by confocal microscopy. In most control larvae, the midgut was well contrasted, with the fluorescent tracer confined within the midgut (Fig. 5A, upper panel and C). Conversely, the tracer was detected in various parts of the body cavity in a number of Tsp2A-RNAi 11415R-2 larvae (Fig. 5A, lower panel), indicating leakage of the tracer from the lumen of the midgut epithelium. In some Tsp2A-RNAi 11415R-2 larvae, the tracer was not detected in the body cavity, possibly caused by lower RNAi efficiency (Fig. 5C). To exclude off-target effects, we tested an additional RNAi line that targets a different sequence of the Tsp2A mRNA (Fig. 1B, Tsp2A IR1-2). In the Tsp2A-RNAi IR1-2 first-instar larvae (48Y-GAL4/UAS-Tsp2A-RNAi IR1-2), the levels of Tsp2A were decreased and the mislocalization of Mesh was observed in the midgut epithelial cells (Fig. S2J–J‴). Leakage of the tracer from the midgut was also observed in a number of Tsp2A-RNAi IR1-2 larvae but not in most control larvae (+/UAS-Tsp2A-RNAi IR1-2) (Fig. 5B). These observations indicate that Tsp2A is required for the barrier function of the midgut epithelium.

Because Mesh and Ssk were mislocalized in Tsp2A-deficient midgut epithelial cells, we next investigated whether the localization of Tsp2A is affected by loss of Mesh or Ssk. In most mesh^f04955 mutant and Df(3L)ssk midgut epithelial cells, Tsp2A failed to localize to the apicolateral region but was distributed diffusely and formed aggregates in the cytoplasm (Fig. 6B,D), indicating that both Mesh and Ssk are required for the sSJ localization of Tsp2A. These observations, together with our previous report that Mesh and Ssk are interdependent for their localization (Izumi et al., 2012), indicate that the three sSJ-specific components, Tsp2A, Mesh and Ssk, are mutually dependent on each other for their proper localization, and work together to organize sSJs.

We noted that, whereas Tsp2A co-localized 100% with Dlg-positive apicolateral regions in wild type (n=234), this colocalization was reduced to 7.9% (n=202) in mesh^f04955 mutant and 1.3% in Df(3L)ssk (n=225) epithelial cells (Fig. 6C-C″ and E-E″). Furthermore, we also observed a reduction in the accumulation of Mesh at the apicolateral regions where Dlg was localized from 100% in wild-type (n=202) to 3.2% in Tsp2A^1-2 (n=245) midgut epithelial cells (Fig. 3D′–D‴,H′–H‴; Fig. S2D′–D‴,G′–G‴ and I′–I‴). These observations suggest that although reduced, Tsp2A and Mesh do not completely lose the ability to accumulate in the apicolateral regions, even in the absence of other sSJ components.

The interdependency between Tsp2A, Mesh and Ssk for their localization at sSJs prompted us to examine whether Tsp2A is physically associated with Mesh and Ssk. Because we failed to immunoprecipitate endogenous Tsp2A with our anti-Tsp2A antibodies from the wild-type embryo in a range of conditions (data not shown), the embryos expressing EGFP–Tsp2A with da-GAL4 driver were subjected to immunoprecipitation with anti-GFP antibodies. When we used a lysis buffer containing NP-40 as a non-ionic detergent, EGFP–Tsp2A was immunoprecipitated with the anti-GFP antibodies, but co-precipitation of Ssk and Mesh with EGFP–Tsp2A was not detected (data not shown). Previous reports have shown that co-immunoprecipitation of tetraspanins with other tetraspanins or their partner integral membrane proteins depends on the detergents included in the lysis buffers (Charrin et al., 2009; Haining et al., 2012; Hemler, 2005; Yanez-Mo et al., 2009). Thus, the embryos expressing EGFP–Tsp2A were lysed with a buffer containing the non-ionic detergent Brij97, which has often been used in co-precipitation experiments in tetraspanin studies, and were subjected to immunoprecipitation with two anti-GFP antibodies. Western blot analysis of both of the precipitates with an anti-GFP antibody or the anti-Tsp2A antibodies specifically detected identical protein bands, indicating successful immunoprecipitation of EGFP–Tsp2A (Fig. 7A; Fig. S3E). Furthermore, co-precipitation of Mesh and Ssk with EGFP–Tsp2A was detected in these precipitated samples under these conditions (Fig. 7B). Neither Mesh nor Ssk was precipitated from EGFP–Tsp2A-expressing embryos with the control IgG, or from EGFP-expressing embryos with the anti-GFP antibody (Fig. 7B). Taken together, these results indicate that Tsp2A forms a complex with Mesh and Ssk in vivo.

We have identified a tetraspanin family protein, Tsp2A, as a newly discovered member of the sSJ-specific proteins. Tsp2A is required for sSJ formation and for the paracellular diffusion barrier in Drosophila midgut. Tsp2A forms a complex with Mesh and Ssk and cooperates with them to form sSJs. This study is the first demonstration that a tetraspanin family protein is required for the formation of an epithelial occluding junction.

Of the sSJ-specific components, Mesh is a membrane-spanning protein and has an ability to induce cell–cell adhesion, implying that it is a cell adhesion molecule and might be one of the components of the electron-dense ladder-like structures in sSJs (Izumi et al., 2012). By contrast, both Ssk and Tsp2A are unlikely to act as cell adhesion molecules in sSJs because each of the two extracellular loops of Ssk (25 and 22 amino acids, respectively) appear to be too short to bridge the 15–20-nm intercellular space of sSJs (Lane et al., 1994; Tepass and Hartenstein, 1994). Furthermore, overexpression of EGFP–Tsp2A in Drosophila S2 cells did not induce cell aggregation, which is a criterion for cell adhesion activity (data not shown). Several observations in Tsp2A mutants might provide clues for understanding the role of Tsp2A in sSJ formation. In most Tsp2A mutant midgut epithelial cells, Mesh fails to localize to the apicolateral membranes but was distributed in the cytoplasm, possibly to specific intracellular membrane compartments. To further examine where Mesh was localized in Tsp2A mutant cells, we double-stained the midgut with the anti-Mesh antibody and the antibodies against typical markers of various intracellular membrane compartments, including the Golgi apparatus (anti-GM130), early endosomes (anti-Rab5), recycling endosomes (anti-Rab11) and lysosomes (anti-LAMP1). However, we were unable to detect any overlap between staining by these markers and that of Mesh (data not shown). We found that the staining pattern in Tsp2A mutant midgut epithelial cells produced with the anti-KDEL antibody, which labels endoplasmic reticulum, was similar, although not identical, to that produced by the anti-Mesh antibody (data not shown). Interestingly, some tetraspanins are known to control the intracellular trafficking of their partners. For instance, a mammalian tetraspanin, CD81, is necessary for normal trafficking or for surface membrane stability of a phosphoglycoprotein, CD19, in lymphoid B cells (Shoham et al., 2003). The TspanC8 subgroup proteins, which all possess eight cysteine residues in their large extracellular domain, regulate the exit of a metalloproteinase, ADAM10/Kuz, from the ER and differentially control its targeting to either late endosomes or to the plasma membrane (Dornier et al., 2012; Haining et al., 2012; Prox et al., 2012). Consequently, TspanC8 proteins regulate Notch signaling through the activation of ADAM10/Kuz in mammals, Drosophila and Caenorhabditis elegans (Dornier et al., 2012; Dunn et al., 2010). If Mesh is retained in the trafficking pathway from endoplasmic reticulum to plasma membrane in Tsp2A mutant cells, Tsp2A might have an ability to promote the intracellular trafficking of Mesh in the secretory pathway. To clarify the role of Tsp2A in sSJ formation, it will be necessary to determine the intracellular membrane compartment where Mesh was localized in Tsp2A mutant cells.

We found that Tsp2A, Mesh and Ssk were mutually dependent for their localization at sSJs. Consistent with this intimate relationship, the co-immunoprecipitation experiment revealed that Tsp2A physically interacts with Mesh and Ssk in vivo. However, the amount of Ssk observed in the co-immunoprecipitation with EGFP–Tsp2A was barely enriched relative to the total amount detected in extracts of embryos expressing EGFP–Tsp2A. This was particularly striking in comparison to the degree of enrichment of Mesh in the co-immunoprecipitation with EGFP–Tsp2A. To interpret these results, we need to further clarify the detailed manner of the interaction between Tsp2A, Mesh and Ssk proteins. Many tetraspanin family proteins are known to interact with one another and with other integral membrane proteins to form a dynamic network of proteins in cellular membranes. Tetraspanins are also believed to have a role in membrane compartmentalization (Charrin et al., 2014,, 2009; Hemler, 2005; Yanez-Mo et al., 2009). Given such functional properties of tetraspanins, Tsp2A might determine the localization of sSJs at the apicolateral membrane region by membrane domain formation.

In the Tsp2A mutant midgut epithelial cells, Lgl was distributed throughout the basolateral membrane region, whereas it was localized in the apicolateral membrane region in the wild type. In view of the role of Lgl in the formation of the apical–basal polarity of ectodermally derived epithelial cells, it is of interest to consider whether this abnormal localization of Lgl in the Tsp2A mutant affects epithelial polarity. However, in the Tsp2A mutant midgut epithelial cells, Dlg still showed polarized concentration into the apicolateral membrane region and Lgl did not leak into the apical membrane domain. These observations suggest that the lack of Tsp2A does not affect the gross apical–basal polarity of the midgut epithelial cells.

Some tetraspanins have been reported to be involved in the regulation of cell–cell adhesion (Chattopadhyay et al., 2003; Ishibashi et al., 2004; Shigeta et al., 2003). A mammalian tetraspanin, CD151, regulates epithelial cell–cell adhesion through PKC- and Cdc42-dependent actin reorganization (Shigeta et al., 2003), or through complex formation with α3β1 integrin (Chattopadhyay et al., 2003). A mammalian tetraspanin, CD9, is concentrated in the axoglial paranodal region in the brain and in the peripheral nervous system, and CD9 knockout mice display defects in the formation of paranodal septate junctions and in the localization of paranodal proteins (Ishibashi et al., 2004). Paranodal septate junctions have electron-dense ladder-like structures and their molecular organization is similar to that of pSJs, but tetraspanins involved in pSJ formation have not been reported in Drosophila.

Interactions between several tetraspanins and claudins, the key integral membrane proteins involved in the organization and function of tight junctions, are also known. Claudin-11 forms a complex with OAP-1/Tspan-3 (Tiwari-Woodruff et al., 2001) and chemical crosslinking reveals a direct association between claudin-1 and CD9 (Kovalenko et al., 2007). Furthermore, the interaction between claudin-1 and CD81 is shown to be required for hepatitis C virus infectivity (Evans et al., 2007). To date, no tight junction defect has been reported in CD9 knockout mice, CD81 knockout mice, or CD9/CD81 double knockout mice (Charrin et al., 2014,, 2009; Hemler, 2005; Takeda et al., 2003; Yanez-Mo et al., 2009). Further investigation is necessary to clarify whether the interactions between tetraspanins and tight-junction proteins are involved in the formation and function of tight junctions.

The fly strains mesh^f04955, da-GAL4 and 48Y-GAL4 were obtained from the Bloomington Stock Center, and the Tsp2A-RNAi strain 11415R-2 was obtained from NIG-Fly. We also used the strain Df(3L)ssk (Yanagihashi et al., 2012). For the phenotype rescue experiment, pUAST vectors (Brand and Perrimon, 1993) containing EGFP–Tsp2A were constructed and a fly strain carrying this construct was established.

The ORF of Tsp2A, including the initiation codon, was amplified by PCR with the forward primer with a BglII site (5′-gaagatctATGGGCATCGGCTATGGAGC-3′) and the reverse primer with a KpnI site (5′-ggggtaccTCAGCGGCGGTAGTTGCTGG-3′) from Drosophila embryonic cDNA, and cloned into the BglII and KpnI sites of pUAST vector (Brand and Perrimon, 1993). To generate an expression vector for N-terminal EGFP-tagged Tsp2A, EGFP cDNA lacking the stop codon with 3′ and 5′ BglII sites was cloned into the BglII site of pUAST-Tsp2A. To generate an RNAi line (Tsp2A IR1-2), a DNA fragment containing nucleotides 381 to 725 of the Tsp2A ORF was amplified by PCR with the forward primer with an EcoRI (5′-gaattcGCAGTTCAGCACTATCAACT-3′) or a KpnI site (5′-ggtaccGCAGTTCAGCACTATCAACT-3′), and the reverse primer with a BglI site (5′-agatctTAGTTGCTGGCCTGTTCCTC-3′). The two types of DNA fragment were inserted into pUAST as a head-to-head dimer and transformed into SURE2 competent cells (200152, Agilent Technologies, Santa Clara, CA, USA). Transgenic flies were generated by standard P-element transformation procedures.

Generation of Tsp2A mutants using the CRISPR/Cas9 system was performed according to the method described by Kondo and Ueda (2013). To construct a guide RNA (gRNA) expression vector for Tsp2A, two complementary 24-bp oligonucleotides of the target sequence with 4-bp overhangs on both ends (5′-cttcGAGCAGCTGGAGAAGCAAAT-3′ and 5′-aaacATTTGCTTCTCCAGCTGCTC-3′) were annealed to generate a double-strand DNA, and cloned into BbsI-digested pBFv-U6.2 (pBFv-U6.2-Tsp2ACR1). pBFv-U6.2-Tsp2ACR1 was injected into the y¹ v¹ nos-phiC31; attP40 host (Bischof et al., 2007). Surviving G₀ males were individually crossed to y² cho² v¹ virgins. A single male transformant from each cross was mated to y² cho² v¹; Sp/CyO virgins. Offspring in which the transgene was balanced were collected to establish a stock.

Females carrying a U6.2-Tsp2ACR1 transgene were crossed to nos-Cas9 males (y² cho² v¹/Y; Sp/CyO, P{nos-Cas9}2A) to obtain founder flies that have both the U6.2-Tsp2ACR1 and the nos-Cas9 transgenes. Female founders were crossed to FM7c/Y male flies. Each female possessing a y² cho² v¹/FM7c; CyO, P{nos-Cas9}2A/+ genotype was crossed to FM7c/Y male flies and the offspring possessing a y² cho² v¹/FM7c genotype were collected to establish the lines. The embryos of the lethal lines were immunostained for Tsp2A. Genomic DNA of the Tsp2A-negative lines was extracted and analyzed for lesions on the Tsp2A gene locus.

Embryos of deficiency lines obtained from the Bloomington Stock Center and the Drosophila Genetic Resource Center (Kyoto, Japan) were stained for Mesh. The OELP in stage 16 embryos was observed to determine whether Mesh was mislocalized. In the first screen, more than 20 embryos of deficiency lines were observed and the candidate lines that showed mislocalization of Mesh were isolated. To confirm the mislocalization of Mesh observed in the candidate lines, their balancer chromosomes were replaced by β-galactosidase-expressing balancers and the β-galactosidase-negative embryos were assumed to be homozygous for the deficiency lines.

To generate anti-Tsp2A polyclonal antibodies, a polypeptide corresponding to the C-terminal cytoplasmic domain (NH₂-CAVKKEEEQASNYRR-COOH) of Tsp2A was synthesized and coupled by means of the cysteine residue to keyhole limpet hemocyanin (Eurofins Genomics, Tokyo, Japan). Polyclonal antibodies were generated in rabbits (301 and 302) by Kiwa Laboratory Animals (Wakayama, Japan). Rabbit antisera were affinity purified by Sulfolink Immobilization Kit (44999, Thermo Fisher Scientific, Waltham, Massachusetts, USA) according to the manufacturer's instructions (301AP and 302AP).

Embryos were fixed with 3.7% formaldehyde in PBS for 20 min. Larvae were dissected in Hanks' Balanced Salt Solution and fixed with 3.7% formaldehyde in PBS-0.4% Triton X-100. The following antibodies were used: rabbit anti-Tsp2A (302AP, 1:200), rabbit anti-Mesh (955-1; 1:1000) (Izumi et al., 2012), rat anti-Mesh (8002; 1:500) (Izumi et al., 2012), rabbit anti-Ssk (6981-1; 1:1000) (Yanagihashi et al., 2012), mouse anti-Dlg [4F3, Developmental Studies Hybridoma Bank (DSHB); 1:50], mouse anti-Coracle (C615.16, DSHB; 1:50), mouse anti-Fasciclin III (7G10, DSHB; 1:20), rabbit anti-Lgl (provided by F. Matsuzaki, RIKEN CDB; 1:1000), rabbit anti-Kune (1:500) (Nelson et al., 2010). Alexa-Fluor-488-conjugated (A21206, Invitrogen), and Cy3- and Cy5-conjugated (712-165-153 and 715-175-151, Jackson ImmunoResearch Laboratories, West Grove, PA, USA) secondary antibodies were used at 1:400. For the Tsp2A staining in Fig. 6, the anti-Tsp2A antibody was pre-incubated with Tsp2A^1-2 first-instar larvae overnight at 4°C. Samples were mounted in Vectashield (H-1000, Vector Laboratories, Burlingame, CA, USA). Images were acquired with a confocal microscope (model TCS-SPE; Leica Microsystems, Wetzlar, Germany) with its accompanying software using HC PLAN Apochromat 20× NA 0.7 and HCX PL Apochromat 63× NA 1.4 objective lenses (Leica Microsystems). Images were processed with Adobe Photoshop.

First-instar larvae of wild type or Tsp2A mutants were dissected and fixed overnight at 4°C with a mixture of 2.5% glutaraldehyde and 2% paraformaldehyde in 0.1 M cacodylate buffer (pH 7.4). The specimens, including the midguts, were prepared as described previously (Izumi et al., 2012). Ultrathin sections (50–100 nm) were stained doubly with 4% hafnium (IV) chloride and lead citrate, and observed with a JEM-1010 electron microscope (JEOL, Tokyo, Japan) equipped with Velete TEM CCD Camera (Olympus, Tokyo, Japan) at an accelerating voltage of 80 kV.

Fly embryos expressing EGFP–Tsp2A or GFP (da-Gal4>UAS-EGFP-Tsp2A or UAS-GFP) were mixed with a 5-fold volume of lysis buffer [30 mM Tris-HCl pH 7.5, 150 mM NaCl, 1% Brij97 (P6136, Sigma-Aldrich, St. Louis, MO, USA)] and protease inhibitor cocktail (25955-11, Nakarai Tesque, Kyoto, Japan) and homogenized using a pestle for 1.5 ml microfuge tubes. The method for immunoprecipitation was essentially the same as described previously (Izumi et al., 2012). Rabbit anti-GFP (598, MBL, Nagoya, Japan) and rat anti-GFP (RQ2, MBL and GF090R, Nakarai Tesque) antibodies were used for immunoprecipitation. Immunocomplexes and extracts of the first-instar larva were separated on SDS-polyacrylamide gels, transferred to polyvinylidene difluoride membranes and western blot analyses were performed using mouse anti-GFP (clone 7.1+13.1, Roche, Basel, Switzerland; 1:500), rabbit anti-Tsp2A (301AP and 302AP; 1:200), rabbit anti-Ssk (6981-1; 1:1000), rat anti-Mesh (8002; 1:500) and mouse anti-α-tubulin (DM-1A; 1:1000, Sigma-Aldrich) antibodies.

Embryos (1–15 h after laying) were put on yeast paste containing Alexa-Fluor-555-labeled dextran (MW 10,000; Thermo Fisher Scientific) to feed newly hatched larvae. After 10–15 h, first-instar larvae were washed with water. Images were acquired with a confocal microscope (model TCS-SPE; Leica Microsystems) and its accompanying software using an HC PLAN Apochromat 10× NA 0.3 objective lens (Leica Microsystems). Images were processed with Adobe Photoshop. Mean Alexa-Fluor-555-labeled dextran-derived fluorescence intensity in the most anterior part (area of the circle: 1232.783 µm²) of the body cavity in individual larva was quantified using Image J software (National Institutes of Health, MD, USA). Mean fluorescence intensity was taken as the read-out value. Statistical significance was evaluated by the Mann–Whitney U-test (KaleidaGraph, Synergy Software).

We are grateful to S. Yonemura, A. Nagafuchi and all the members of Furuse laboratories for helpful discussions. We also thank F. Matsuzaki for the antibody and the fly stocks, and the Bloomington Stock Center, the Drosophila Genetic Resource Center at Kyoto Institute of Technology and the fly stocks of National Institute of Genetics (NIG-Fly) for fly stock.