Pasiflora proteins are novel core components of the septate junction

The generation of distinct chemical milieus within the body is essential for metazoan development. This compartmentalization is accomplished by epithelia that impede paracellular diffusion and selectively transport substances via membrane channels and transporters. To provide a barrier, epithelia have a narrow intercellular space, which is sealed by specialized junctions, including tight junctions (TJs) in vertebrates and septate junctions (SJs) in invertebrates (Noirot-Timothée et al., 1978; Tepass and Hartenstein, 1994; Schwabe et al., 2005). SJs are the ancestral sealing junctions and are found in all invertebrates from sponges to arthropods but are also present in vertebrates (Leys and Riesgo, 2012). In electron micrographs, SJs appear as an array of regularly spaced septa, which operate by extending the travel distance for solutes through the paracellular route. SJs are found in both primary epithelia, such as epidermis, trachea and hindgut, and secondary epithelia, which develop through mesenchymal-epithelial transition, such as the blood-brain barrier (BBB) and midgut (Tepass, 1997; T. Schwabe, X. Li and U.G., unpublished). The BBB ensheaths the nervous system and is required to maintain its homeostasis. Owing to the high potassium content of the hemolymph, animals with a defective BBB die of paralysis. In Drosophila, the BBB is a squamous epithelium established late in embryogenesis by SJ-forming subperineurial glia (SPG) (Edwards et al., 1993; Schwabe et al., 2005). Here, in addition to providing a paracellular barrier, SJs also serve as a fence for the diffusion of proteins across the lateral membrane (T. Schwabe, X. Li and U.G., unpublished). Molecularly and functionally homologous SJs are found in vertebrates at the node of Ranvier, where they form the paranodal junction between axons and myelinating glia (Poliak and Peles, 2003).

The SJ consists of a large multi-protein complex. In Drosophila, more than 20 proteins have been characterized that when missing lead to disruption of SJs and loss of barrier integrity (Izumi and Furuse, 2014). Most of these are transmembrane (TM) and lipid-anchored proteins that localize at the SJ, such as the claudins Sinuous (Sinu) (Wu et al., 2004), Megatrachea (Mega; Pickel – FlyBase) (Behr et al., 2003) and Kune-kune (Kune) (Nelson et al., 2010), the cell adhesion molecules Neurexin IV (Nrx-IV) (Baumgartner et al., 1996), Contactin (Cont) (Faivre-Sarrailh et al., 2004), Neuroglian (Nrg) (Genova and Fehon, 2003), Lachesin (Lac) (Llimargas et al., 2004) and Fasciclin III (FasIII, or Fas3) (Woods et al., 1997), the sodium pump with its two subunits ATPα and Nervana 2 (Nrv2) (Genova and Fehon, 2003; Paul et al., 2003), Melanotransferrin (Transferrin 2 – FlyBase) (Tiklová et al., 2010) and Macroglobulin complement-related (Mcr) (Bätz et al., 2014; Hall et al., 2014). The complex also includes the intracellular scaffold proteins Coracle (Cora) (Fehon et al., 1994) and Varicose (Vari) (Wu et al., 2007) that interact with the cytoplasmic tails of membrane proteins and connect them to the actin cytoskeleton. A hallmark of SJ proteins is that they are interdependent for localization, and removal of one component is sufficient to destabilize the whole complex. In addition, half of the known SJ proteins can be co-immunoprecipitated from tissue extracts and detected by mass spectrometry (MS), further suggesting that they function together in a multi-protein complex (Genova and Fehon, 2003; Faivre-Sarrailh et al., 2004; Tiklová et al., 2010; Jaspers et al., 2012). Fluorescence recovery after photobleaching (FRAP) experiments have been instrumental in classifying most SJ proteins as core components based on their limited mobility after photobleaching and the observation that upon loss of function other SJ proteins diffuse rapidly into the bleached region due to impaired complex formation (Laval et al., 2008; Oshima and Fehon, 2011).

Accompanying epithelial morphogenesis, SJs are remodeled into mature junctions. At embryonic stage 12, SJ proteins accumulate along the lateral membrane of columnar epithelial cells. Subsequently, they gradually localize at more apical compartments and by stage 15 are restricted to the apicolateral membrane, basal to adherens junctions. The Ly-6 proteins Crooked (Crok), Crimpled (Crim) and Coiled (Cold) are required for SJ formation; however, they do not reside at SJs and instead localize to cytoplasmic puncta. In Ly-6 mutants, the FRAP kinetics of SJ proteins mirrors that of core complex mutants and therefore Ly-6 proteins are thought to be involved in the assembly of SJ (sub)complexes in an intracellular compartment (Nilton et al., 2010; Oshima and Fehon, 2011). The subsequent relocalization of SJs requires endocytosis from the basolateral membrane and recycling to the apicolateral compartment (Tiklová et al., 2010; Oshima and Fehon, 2011). Gliotactin (Gli) and Discs-large (Dlg; Dlg1 – FlyBase) localize at SJs (Woods and Bryant, 1991; Woods et al., 1997; Schulte et al., 2003) but, in contrast to core components and Ly-6 proteins, upon their loss of function the complex is properly formed and SJ proteins, although mislocalized, retain their restricted mobility (Oshima and Fehon, 2011). Together with a lack of physical interactions with SJ components, this result suggests that Gli and Dlg are required for complex localization rather than its assembly (Ward et al., 1998; Schulte et al., 2003, 2006).

In contrast to SJs, TJs localize apically of the zonula adherens and in electron microscopy appear as a series of fusions of adjacent membranes. Although the set of proteins that composes the TJ is different from that of the SJ, the two complexes share a key molecular component, the claudins. Claudins are a tetraspan membrane family of 20-34 kDa proteins with intracellular N- and C-termini and constitute a main component of TJs. The larger first extracellular loop contains a claudin family signature motif and bears critical residues that define TJ charge and size selectivity in a tissue-specific manner (Günzel and Yu, 2013). Claudins are part of a large protein clan, comprising the PMP22/EMP/MP20/Claudin (PF00822), MARVEL (PF01284), tetraspanin (PF00335), connexin (PF00029) and innexin (PF00876) families, which share the same overall topology but differ in size and motif composition of extracellular and intracellular domains. Many members of this clan can form homo- and heterotypic oligomers on the same and neighboring membranes and play essential roles in junctional complexes, including TJs, gap junctions and the casparian strip of plants, as well as in membrane traffic and fusion events. Claudins have been shown to interact with other tetraspan proteins such as occludins, tetraspanins and MARVEL, as well as cell adhesion proteins and receptors. Similarly, tetraspanins form microdomains in the plasma membrane, in which cell adhesion proteins, TM receptors and their signaling components are enriched and, thereby, are thought to be modulated in their activity (Sánchez-Pulido et al., 2002; Hua et al., 2003; Hemler, 2005; Cording et al., 2013; Simske, 2014; Van Itallie and Anderson, 2014; Roppolo et al., 2014).

Here we identify and characterize two new core components of the SJ, Pasiflora1 and Pasiflora2, which are part of a novel tetraspan protein family that is conserved across the prostostome-deuterostome divide and is characterized by specific sequence features. Both proteins localize at SJs, show interdependence for localization and restricted mobility with known SJ members and are required for the integrity of epithelial barriers. Our work provides new insight into the composition of the SJ and implicates a second family of tetraspan proteins in the development of these crucial cell junctions.

To identify novel genes required for BBB formation, we followed an in vivo RNAi approach using 10,450 UAS-RNAi strains (75% genome coverage) from the Vienna Drosophila Resource Center (VDRC) KK library (Dietzl et al., 2007). To efficiently phenocopy the impaired genotype, UAS-dicer2 was co-expressed in all screening steps. We initially tested the lines for adult lethality using the strong pan-glial driver repo-Gal4. The lines causing lethality or subviability were retested for impaired viability with the SPG-specific but weaker moody-Gal4. To directly examine whether the BBB is compromised in the knockdown of the genes, we performed the embryonic dye penetration assay in a selection of candidates (using repo-Gal4); in wild-type (wt), the injected dye is excluded from the CNS, but in BBB mutants, such as in Nrx-IV embryos, it rapidly diffuses into the nervous system (Fig. 1C,F). To quickly quantify dye accumulation in a systematic fashion, we developed an automated analysis script using Definiens (Fig. 1B).

Among the candidates identified, the lines CG7713¹⁰²²²³ and CG8121¹⁰⁵⁸⁰⁶ caused complete adult lethality with repo-Gal4 and adult subviability with moody-Gal4 (23% and 17% survivors for 102223 and 105806, respectively; 51% for negative controls). Pan-glial knockdown of both genes resulted in leaky BBB (Fig. 1D-F) and late embryonic lethality (1% hatch; wt, 99%). Interestingly, CG7713 and CG8121 belong to one family (With et al., 2003). Inspired by the paralysis resulting from the BBB defect, we named the genes pasiflora1 (pasi1, CG7713) and pasiflora2 (pasi2, CG8121) from the Greek mythological goddess who induced paralysis in her victims. Our results suggest that the family members pasiflora1 and pasiflora2 are novel genes required for BBB formation.

To analyze the phenotypes in complete loss of function, we sought to generate genomic mutants. The viable line P{EP}G4182 carries a P-element insertion 219 bp upstream of the pasiflora1 5′UTR. We created imprecise excisions and isolated a line, pasiflora1^Δ, that deletes the entire pasiflora1 locus and 59 bp of the CG7379 3′UTR (Fig. 1A). pasiflora1^Δ die as late embryos (0% hatch) and have a permeable BBB. A similarly leaky BBB is observed in embryos transheterozygous for pasiflora1^Δ and the deficiency chromosomes Df(3R)BSC566 and Df(3R)ED5785, which uncover the locus. The dye leakage is severe, but weaker than that of the amorphic Nrx-IV⁴³⁰⁴ SJ mutant (Fig. 1E,F; data not shown). However, Nrx-IV is only zygotically expressed (Baumgartner et al., 1996), whereas pasiflora1 is also maternally provided [see Fig. 3A; BDGP website (http://insitu.fruitfly.org)] (Graveley et al., 2011). To ultimately prove that the glial loss of pasiflora1 is causing the leaky BBB, we sought to rescue the dye penetration of pasiflora1^Δ. Pan-glial expression of pasiflora1 restores BBB function (Fig. 1E,F), demonstrating that neither the neighboring CG7379 nor other mutations on the chromosome contribute to BBB breakdown, and indicating that pasiflora1 is cell-autonomously required.

In the vicinity of the pasiflora2 locus, no P-element insertions were available. Since the gene belongs to the same family, we decided to pursue pasiflora2 using RNAi. Moreover, the KK line is very potent as it causes strong BBB permeability (Fig. 1D,F) and embryonic lethality with repo-Gal4 (1% hatch). The fact that an impaired BBB is observed in the glial-specific knockdown suggests that pasiflora2 is also cell-autonomously required. To exclude off-target effects, we tested two additional RNAi lines that target different sequences of the mRNA. With pan-glial expression, the VDRC line GD43952 and an shRNAi line that we generated (TRiP design; Ni et al., 2011) show qualitatively similar defects of dye penetration and embryonic lethality, with milder defects observed in the GD line (Fig. 1D; data not shown). For all our experiments with pasiflora2, we used the KK RNAi line.

To exclude the possibility that the leaky BBB is a result of earlier defects in glia specification and/or migration, we analyzed the number and positions of SPG (Ito et al., 1995; Beckervordersandforth et al., 2008). We detect the full set of SPG in both pasiflora1^Δ and repoGal4;UAS-pasiflora2-RNAi embryos, with somewhat variable positions of nuclei in both control and mutant embryos (Fig. 1G). In summary, our results show that pasiflora1 and pasiflora2 are novel genes with a specific role in BBB formation.

We noticed that the tracheal tubes of pasiflora1^Δ do not fill with air (data not shown), indicating that the tracheal barrier is also compromised. To confirm this observation, we performed the dye penetration assay and visualized the dorsal trunks. In wt, the dye is excluded from the tracheal lumen, but it rapidly diffuses into the tubes of pasiflora1^Δ homozygous, transheterozygous pasiflora1^Δ over the two deficiency chromosomes and in embryos with ubiquitous knockdown of pasiflora2 (tubulin-Gal4) (Fig. 2B; data not shown). Both mutants also show excessively elongated and convoluted dorsal trunks, a result that was confirmed by staining stage 16 embryos with the 2A12 antibody that recognizes the luminal protein Gasp (Fig. 2A). Overelongated dorsal trunks are observed in the majority of SJ mutants and are believed to be due, at least in part, to the role of SJs in the transcytosis of chitin deacetylases that terminate tube elongation (Luschnig et al., 2006; Wang et al., 2006; Dong et al., 2014). Knocking down pasiflora2 with the more trachea-specific breathless-Gal4 leads to qualitatively similar phenotypes but with lower penetrance (Fig. 2B). As expected, the tracheal defects are not restored in our rescue experiment with the glial driver, further supporting that pasiflora1 is cell-autonomously required (Fig. 2B). Thus, pasiflora1 and pasiflora2 are required for tracheal barrier function and tube size control.

To characterize the expression pattern of pasiflora1 and pasiflora2, we performed RNA in situ hybridization in wt embryos. The genes show identical expression patterns throughout embryogenesis (Fig. 3A). Ubiquitous weak expression is first detected at stages 1-4, suggestive of maternal contribution. Zygotic transcripts are detected in epithelial tissues from stage 10 onwards. The tracheal placodes are labeled at stage 10 and the anterior hindgut at stage 11; expression persists in these tissues throughout development. During stages 14-16, trachea, foregut, hindgut, epidermis and salivary glands are marked. At stage 16, we detect weak staining in the nervous system and labeling of some cells that, based on their position, are likely to be exit and/or peripheral glia. A clearer in situ for pasiflora1 showing similar expression is displayed on the BDGP website (http://insitu.fruitfly.org). Therefore, both genes are specifically expressed in embryonic epithelia and insulating glia – all tissues that form SJs.

Several attempts to generate specific antibodies recognizing the two proteins were unsuccessful. Overexpression of the highly hydrophobic pasiflora proteins was toxic to the bacteria. We therefore raised antibodies against a mixture of two peptides (Fig. 3B; see supplementary Materials and Methods for epitopes), but unfortunately neither sera nor affinity-purified antibodies showed specific labeling in embryos (data not shown).

Pasiflora1 and Pasiflora2 are small proteins of 169 and 258 amino acids, respectively, with four TM domains but no signal peptide. Their predicted topology is very similar, with intracellular N- and C-termini and a larger first extracellular loop (Fig. 3B). To examine whether the proteins localize at the plasma membrane or some intracellular membrane compartment, we analyzed their subcellular localization in vivo and in cell culture. We tagged both proteins with GFP and FLAG, attached to an alanine-rich linker (Fig. 3B), and expressed them in Schneider cells (S2) as well as glia, which we imaged in stage 16 embryos (repo-Gal4) and third instar CNS (moody-Gal4). We find that pasiflora proteins localize at the plasma membrane in vivo and in S2 cells (Fig. 3C). Importantly, C-terminal tagging with GFP does not seem to affect protein function, as pan-glial expression of pasiflora1-GFP rescues the leaky BBB of pasiflora1^Δ at a level similar to that of untagged pasiflora1 (Fig. 1E,F).

Although the protein topology of pasiflora proteins resembles that of claudins, they show no sequence similarity to this or other tetraspan families. However, Pasiflora1 orthologs are readily identified in all arthropods, including insects, arachnidae and crustacea, as well as in molluscs, echinoderms and hemichordates, suggesting that the protein predates the protostome-deuterostome divide. Alignment of the best protein matches from species within these different phyla indicates conservation along the entire length of the protein, with two absolutely conserved motifs: a PW motif at the beginning of TM3 and a VxSQYQ motif that straddles the boundary between TM4 and the C-terminal intracellular domain. Pasiflora2 orthologs are found within arthropods but not beyond; they share the PW motif in TM3 and a shortened VxS motif at the boundary of TM4. Interestingly, the sequence comparison shows that Pasiflora1 is more closely related to its orthologs in other phyla than to Pasiflora2, with a sequence similarity of 15%, suggesting that the separation of the two family members is ancient (Fig. 3D). Consistent with this significant sequence divergence, we find that pasiflora proteins act non-redundantly, since pan-glial expression of pasiflora1, but not pasiflora2, rescues the BBB phenotype of pasiflora1^Δ (Fig. 1E; supplementary material Fig. S1).

Pasiflora proteins were previously shown to belong to a family of otherwise uncharacterized tetraspan proteins with similar length and topology in Drosophila and Anopheles (With et al., 2003). Based on our current analysis, we find seven additional members in this wider family (CG13288, CG13747, CG15098, CG12825, CG10311, CG42288 and Fire exit); however, with the exception of CG13288, which closely resembles Pasiflora1 (23% identity), these proteins are slightly more diverged and only share the PW motif in TM3 (Fig. 3D). Interestingly, the original founding member Fire exit, which is the most strongly diverged, is expressed in exit and peripheral glia but no molecular or biological function has been demonstrated (With et al., 2003). Whereas pan-glial knockdown of Fire exit causes adult subviability (19% survivors; 51% for negative control), knockdown of the other family members did not impair viability (CG10311, CG12825, CG13747, CG15098) or was not performed owing to a lack of RNAi strains in the collection (CG13288, CG42288).

Different lines of evidence indicate that pasiflora1 and pasiflora2 are co-expressed. First, RNA in situ hybridization showed that the genes are similarly expressed in embryonic epithelia (Fig. 3A). Second, both genes were identified as differentially expressed in embryonic glia based on microarray transcriptome profiling (U.G., unpublished). Third, based on developmental RNA-seq, the genes are part of co-expression clusters with SJ genes (kune, cold; sinu, Nrx-IV, Mcr, Gli, crok, cold, crim) (Graveley et al., 2011). The notion that pasiflora1 and pasiflora2 expression is tightly co-regulated is also supported by the observation that both genes, together with more than half of the known SJ component-encoding mRNAs, are predicted targets of miR-184 (Hong et al., 2009; Iovino et al., 2009).

Based on the phenotypic analysis, the expression patterns and the targeting by miR-184, we hypothesized that pasiflora proteins are either SJ components themselves or play a role in complex assembly and/or trafficking. However, neither of the proteins was found in an MS-based proteomic analysis of Mega complexes that succeeded in identifying at least ten known SJ components, possibly because of their small size (Jaspers et al., 2012). Notably, the claudins Sinu and Kune, which are of similar size and structure to the pasiflora proteins, were also not detected in the MS analysis.

To confirm that pasiflora1 and pasiflora2 play a role in SJ development, we analyzed the morphology and subcellular localization of SJs in the mutants. We first visualized the embryonic BBB using the endogenously expressed live-imaging markers Nrg-GFP and Lac-GFP (Morin et al., 2001). In wt late stage 17 embryos, both markers label SJs and trace the outlines of SPG, which make continuous contacts with their neighbors to seal the CNS. In pasiflora1^Δ and repo-Gal4;UAS-pasiflora2-RNAi embryos, SJs appear discontinuous and severely disorganized (Fig. 4A), demonstrating that both genes are required for SJ formation in the embryonic BBB.

SPG are very large but thin cells, complicating the visualization of SJ localization along the lateral membrane during embryonic stages. We therefore examined the hindgut and tracheal columnar epithelia. In the hindgut of wt stage 12 embryos, SJ proteins accumulate evenly along the lateral membrane, but at stage 15 are restricted to the apicolateral membrane compartment, as revealed by staining for Cora. In pasiflora1^Δ and tubulin-Gal4;UAS-pasiflora2-RNAi embryos, Cora localizes similarly to wt at stage 12, but fails to restrict apicolaterally at stage 15 and remains distributed along the lateral membrane. At stage 15 we observe similar mislocalization of additional SJ markers, such as Nrg-GFP, ATPα-GFP and FasIII (Fig. 4C). Notably, in the variably penetrant phenotype of pasiflora2-RNAi, the mislocalization phenotype is more pronounced for Cora and FasIII followed by Nrg, whereas ATPα is only mildly mislocalized. SJs are also mislocalized in tracheal cells. In the tracheal epithelium of wt stage 15 embryos, Cora, ATPα-GFP and Nrg-GFP accumulate in the apicolateral membrane, but spread basolaterally in the mutants (Fig. 4B). Furthermore, the intensity of SJ proteins appears somewhat lower in the mutants. Dimmer SJ immunostaining has been reported for other SJ mutants as well, but in the cases where it was tested by western blots, the protein levels were, overall, not reduced (Genova and Fehon, 2003; Paul et al., 2003). This suggests that the weaker staining is not a result of reduced transcription or increased protein degradation, but rather a consequence of the dispersed localization. In addition, pasiflora2 mutants show milder defects compared with loss of pasiflora1, which is likely to be due to incomplete knockdown of pasiflora2 with RNAi as compared with genomic knockout of pasiflora1 in the deletion mutant. In addition to w¹¹¹⁸, we also tested genetically closer controls; i.e. a viable line with precise excision of P{EP}G4182 and embryos with ubiquitous knockdown of a random RNAi (VDRC KK 108356) and confirmed that they have wt localization of SJs (data not shown). Collectively, our results show that pasiflora1 and pasiflora2 are required for the apicolateral localization of SJs in hindgut and tracheal epithelia.

SJs also play an earlier, independent role in maintaining cell polarity by restricting the size of the apical Crb domain (Laprise et al., 2009). To investigate if cell polarity is disturbed in pasiflora mutants, we analyzed the distribution of Crb in hindgut and trachea. In both tissues, Crb localizes at the apical membrane similarly to wt (Fig. 4B,C). Thus, in columnar embryonic epithelia, pasiflora1 and pasiflora2 selectively affect SJ organization but not the establishment or maintenance of cell polarity.

To determine if pasiflora proteins accumulate in a specific membrane compartment, we analyzed the localization of GFP-tagged versions in the hindgut epithelium using 69B-Gal4. Both Pasiflora1 and Pasiflora2 colocalize at the membrane with Cora, but not Crb, and display the characteristic dynamic expression of SJ proteins: at stage 12, they localize along the lateral membrane and at stage 15 become restricted apicolaterally (Fig. 5A). Occasionally, we detect a small amount of pasiflora proteins, but not Cora, at the apical and basolateral membranes. We believe that this is due to overly high protein levels under Gal4-UAS expression because we only observe it in a minority of cells and it correlates with the strength of expression in the given cell. Altogether, our results show that Pasiflora1 and Pasiflora2 are membrane proteins localizing at SJs.

Core SJ proteins are known to be interdependent for localization, and removal of one component is sufficient to destabilize the entire complex and mislocalize other SJ proteins. To address whether Pasiflora1 and Pasiflora2 localization is similarly affected, we analyzed their distribution in the hindgut of stage 15 embryos homozygous for amorphic mutations in SJ proteins. Co-staining for Cora served as a readout of SJ integrity. In kune^C309 and Nrg^I4 core component mutants, as well as in cold^f05607 and crok^KG06053a embryos, both pasiflora proteins and Cora lose their restricted localization and extend basolaterally (Fig. 5B; data not shown). In summary, our results show that pasiflora proteins localize at the SJ and their localization depends on other SJ proteins, suggesting that they are core complex components.

To further show that pasiflora proteins are core SJ components, we performed a series of FRAP experiments. In the epidermis of wt stage 15 embryos, when mature SJ complexes are established, the fluorescence of GFP-tagged core SJ proteins exhibits slow recovery after photobleaching because the stable SJ complexes are very large and move slowly within the membrane. In mutants of core components or of proteins involved in complex assembly, the SJ complex is not properly formed and the free GFP-linked proteins can diffuse rapidly to the bleached region (Oshima and Fehon, 2011).

To determine if SJ complex formation is impaired in pasiflora mutants, we performed FRAP of Nrg-GFP in the epidermis of stage 15 embryos. For our analysis, we extracted from the fitting procedure the percentages of mobile fractions (F_m) and, more importantly, the characteristic time of diffusion (τ_D) [see supplementary Materials and Methods for half time (t_1/2) and detailed analysis]. In wt, Nrg-GFP shows very slow recovery and even 10 min after bleaching only 10% of the fluorescence has recovered. Recovery has not reached a plateau, but the strong embryo movements did not allow us to systematically perform longer time-lapse recordings. For Nrg-GFP in wt we extrapolate τ_D=29.5 min and 29% mobile fraction. By contrast, in both pasiflora1^Δ and tubulin-Gal4;UAS-pasiflora2-RNAi embryos, Nrg-GFP recovers rapidly (τ_D=2.2 and 4.5 min, respectively) and has a large mobile fraction (65% and 43%, respectively) (Fig. 6A,B). Notably, fluorescence never recovers to 100% in our experiments or those of others (Laval et al., 2008; Oshima and Fehon, 2011), but the nature of this immobile fraction is currently unclear. Therefore, the behavior of Nrg-GFP in pasiflora mutants is similar to that observed in mutants of SJ core components and proteins involved in complex assembly. Together with their localization at SJs, these results argue that pasiflora proteins are core components required for the formation of SJ complexes.

To further show that pasiflora proteins are integral SJ components, we analyzed their mobility within the membrane in the epidermis of stage 15 embryos. We used paired-Gal4 and expressed GFP-tagged pasiflora proteins in epidermal stripes. As a control, we used membrane-tagged mCD8-GFP and imaged embryos at stage 14, at a time when SJs are not yet mature and diffusion within the plasma membrane should not be impeded. mCD8-GFP recovers remarkably quickly (τ_D=33 s) and its mobile fraction is 40%. By contrast, the recovery of Pasiflora1-GFP and Pasiflora2-GFP is significantly slower (τ_D=4.5 and 6.7 min, respectively), with their mobile fractions being 50% (Fig. 6C). The faster recovery of Pasiflora1-GFP and Pasiflora2-GFP proteins compared with Nrg-GFP could be due to the overexpression conditions. Therefore, pasiflora proteins are more immobile than other TM proteins, suggesting that they are part of a membrane complex. To ultimately show that this is the SJ complex, we analyzed the mobility of Pasiflora1-GFP and Pasiflora2-GFP proteins in epidermal cells of kune^C309 mutants, which have disrupted SJs, and observed that both proteins lose their restricted mobility and diffuse very rapidly (Pasiflora1-GFP, τ_D=47 s, F_m=59%; Pasiflora2-GFP, τ_D=43 s, F_m=67%) (Fig. 6C). Taken together, these results validate that pasiflora proteins are indeed core SJ components (Fig. 7).

We have identified two previously uncharacterized proteins, Pasiflora1 and Pasiflora2, as novel components of the Drosophila SJ. Several lines of evidence support this notion. First, pasiflora1 and pasiflora2 mutants exhibit all the characteristic phenotypes associated with disrupted SJs: breakdown of blood-brain and tracheal barriers, overelongated dorsal trunks, and SJ mislocalization in a variety of tissues. In the BBB, SJs appear severely disorganized and in columnar epithelia SJ proteins fail to localize at the apicolateral membrane and instead spread basolaterally. Second, the genes are co-expressed in embryonic epithelia that rely on SJs for their function and the proteins overlap with Cora at the apicolateral membrane. Similar to known SJ proteins, pasiflora localization depends on other complex members, as they spread basolaterally in SJ mutant backgrounds. Finally, using FRAP we demonstrate that pasiflora proteins are core SJ components. In stage 15 epidermal cells, Nrg-GFP displays limited lateral mobility after photobleaching owing to its incorporation in the large multi-protein complex. By contrast, in pasiflora mutants, Nrg-GFP diffuses rapidly, indicating that SJ complex formation is compromised. Overexpressed pasiflora proteins also move slowly within the membrane of wt cells, but diffuse rapidly in cells with disrupted SJs, showing that they are themselves associated with the SJ complex.

An emerging idea is that not all SJ proteins are as interdependent as previously thought and that distinct subcomplexes exist within the large, highly ordered, multi-protein complex. Our observations and those of others (Nelson et al., 2010; Oshima and Fehon, 2011; Hall et al., 2014) indicating that in SJ mutants the localization of other complex members is differentially affected and that the fluorescence of GFP-tagged SJ proteins does not fully recover after photobleaching support this notion.

Pasiflora proteins are conserved in arthropods and beyond and share the global topological features of the tetraspan superfamily, with short conserved sequence motifs. The ability of different tetraspan families to form ribbons based on homo- and heterotypic interactions in cis within the plasma membrane suggests that pasifloras, together with claudins, are involved in forming the highly regularly spaced septa of the SJ. Freeze-fracture experiments have shown that SJs form ribbons, with an apparent size of a single septum of 10 nm and a regular spacing of 15-20 nm. Depending on the tissue, these ribbons are either highly aligned with each other (mature ectoderm) or meandering (developing wing disc) (Fristrom, 1982; Lane and Swales, 1982; Furuse and Tsukita, 2006). In the SJ, the plasma membranes of neighboring cells are not fused but closely juxtaposed at a distance of 15 nm and there is no evidence in invertebrates that different tissues have distinct paracellular permeability. Claudins and pasifloras are therefore unlikely to create pores in trans with specific size and charge selectivity. This suggests that the small claudins and pasifloras act only in cis to form ribbons, while the single-pass membrane proteins of the complex mediate the trans interaction with the neighboring cell via their large extracellular adhesive domains. To date, the structural basis for the intermolecular interaction between tetraspan proteins has not been resolved (Krause et al., 2015). The pasiflora proteins belong to a larger family with nine members in Drosophila. We have shown that Pasiflora1 and Pasiflora2 are expressed in embryonic epithelia and glia and act non-redundantly during SJ formation. Little is known about the other family members: Fire exit is expressed in exit and peripheral glia, which also form SJs; CG15098 is expressed in the midgut, which forms structurally different, smooth SJs.

Our study reveals that the composition of the SJ complex strongly resembles that of other junctional and TM protein complexes, where adhesive or signaling receptors are embedded in a complex environment of hydrophobic tetraspan proteins of different types, in this case three different claudins and two different members of the novel pasiflora family. Membrane complexes such as the SJ are particularly refractory to biochemical and structural analysis owing to their hydrophobicity and large size. However, due to their crucial function in all invertebrates and the vertebrate paranode, it is possible, by genetic means, to identify and study the structural core components as well as the biogenesis of the complex. Given the medical importance of the paranodal SJ in particular and of tetraspan proteins in general, our discovery of pasiflora proteins opens the possibility of studying these proteins and their interactions in a highly accessible and sensitive paradigm.

For generation of transgenic lines we used the ɸC31 integrase method and inserted constructs in attP2 and attP40 docking sites (Groth et al., 2004; Markstein et al., 2008; Pfeiffer et al., 2010). Rescue constructs were generated by PCR amplification from cDNA clones RE54605 (pasiflora1) and LD42595 (pasiflora2) [Drosophila Genomics Resource Center (DGRC), Indiana, USA]. Tagged proteins were generated by fusion of Drosophila-optimized GFP (pJFRC14, Addgene) or 3×-FLAG to the C-terminus after an alanine-rich linker. For in vivo expression, pJFRC2 (10x-UAS) was used (Addgene). For S2 cell expression, the pMT vector was used (metallothionein promoter). The pasiflora2-shRNAi line was generated according to Ni et al. (2011) by inserting a 21 nt hairpin (sense strand: TACAATGTGATTATGGTGCTC) in pWalium20 [Transgenic RNAi Project (TRiP), Harvard Medical School, Boston, USA]. pasiflora1^Δ was generated by imprecise excision of P{EP}G4182 [Bloomington Drosophila Stock Center (BDSC)]; the deletion spans the region 17794826-17796435. For fly strains obtained from published sources see the supplementary Materials and Methods. For live genotyping, Kruppel-Gal4;UAS-GFP or Dfd-YFP balancers were used (BDSC). All strains were raised at 25°C.

The dye permeability assay was performed as previously described (Schwabe et al., 2005). CNS dye penetration was quantified using a custom Definiens (http://www.definiens.com) script that automatically measures pixel intensity after excluding overexposed areas. Mean pixel intensity was taken as readout value. To assess significance, one-way ANOVA was performed over all groups with Student-Newman-Keuls post-hoc test.

To measure lethality, stage 15 embryos were dechorionated, rinsed, mounted on a coverslip coated with heptane glue, covered with halocarbon oil (Huile 10S VOLTALEF), and placed on an agar plate facing a pile of yeast. Embryos were followed during late embryogenesis and larval life and the stage at which they died was scored.

Immunohistochemistry of embryos was performed following standard procedures. For antibodies used see the supplementary Materials and Methods. Live imaging of embryos was performed as described (Schwabe et al., 2005). Dissected third instar CNS was mounted in PBS and imaged directly. S2 cells were transfected with pMT-Pasiflora-FLAG, induced with 0.2-0.5 mM CuSO₄ 24 h post-transfection, and fixed 24 h post-induction. All confocal images were acquired using an LSM 710 system and ZEN acquisition software (Carl Zeiss). Image analysis was performed using ImageJ (NIH).

Whole-mount in situ hybridization on embryos was performed as previously described (Lehmann and Tautz, 1994) with the following modifications: the post-fix step between embryo rehydration and proteinase K treatment was removed, and incubation with anti-DIG antibodies was overnight at 4°C. Antisense probes were generated by in vitro transcription from RE54605 (pasiflora1) and LD42595 (pasiflora2).

Embryos were dechorionated, rinsed, mounted on coverslips with glue, and covered with halocarbon oil. Imaging and photobleaching were performed with a c-Apochromat 40×/1.20 W Korr M27 objective. Two images were acquired before photobleaching and GFP was bleached using maximal output power of a 488 nm laser. The bleached membrane was located in the lateral epidermis and was approximately 3 μm in length. A time series of images was started immediately after photobleaching, with one image every 30 s for 10 min, except for paired-Gal4;UAS-mCD8-GFP and kune^C309;paired-Gal4,UAS-pasiflora1/2-GFP for which images were captured every 4 s for 3 min. A home-written Definiens script was used for correction of embryo movements and a second script for extraction and normalization of fluorescence intensity of photobleached membranes at each time point. Data were fitted to an equation for one-dimensional free diffusion; characteristic time of diffusion and percentages of mobile fractions were then extracted. For detailed analysis of FRAP data see the supplementary Materials and Methods.

PSI-BLAST and manual inspection were performed to identify orthologs and Drosophila paralogs of pasiflora proteins (for a list, see the supplementary Materials and Methods). Protein sequences were aligned and a phylogenetic tree was constructed using the ClustalW algorithm as implemented in Vector NTI 11.5 (Life Technologies). Protein topologies were verified using SMART-EMBL.

We thank E. Arama, V. Auld, W. Chia, L. Luo, the Vienna Drosophila Resource Center, the Bloomington Drosophila Stock Center, the TRiP at Harvard Medical School [NIH/NIGMS R01-GM084947], the Drosophila Genomics Resource Center and the Developmental Studies Hybridoma Bank for providing fly strains, constructs and antibodies. Special thanks to C. Ludwig for help with generating transgenic lines, U. Unnerstall for phylogenetic analysis and all lab members for helpful comments on the manuscript.