Crooked, Coiled and Crimpled are three Ly6-like proteins required for proper localization of septate junction components

The multiprotein complexes of cellular junctions consist of membrane-spanning proteins, cytoplasmic scaffolding proteins and extracellular membrane-associated proteins that colocalize at specific sites along the lateral cell surface. Various mechanisms are presumably needed to ensure correct complex formation, including the attachment of cell adhesion molecules to cytoskeletal scaffolding proteins, membrane clustering of junction proteins by lateral interactions, and selective trafficking of vesicles to and from the site of junction assembly.

In Drosophila, the most prominent junctions are the pleated septate junctions (SJs). These are rows of protein assemblies found at the lateral cell surface (basal to the zonula adherens) of ectodermal-derived epithelia and between glial cells of the nervous system, and they appear as ladder-like intermembrane bands in electron micrographs (Stork et al., 2008; Tepass et al., 2001). An important function of SJs is to provide a barrier to paracellular diffusion. Epithelial SJ formation commences at mid-embryogenesis, after cell polarization and formation of the adherens junctions, and initially appears as single or small groups of septa. Additional septa are then added until the ladder-like bands cover one- to two-thirds of the lateral cell surface at stage 16 (Tepass and Hartenstein, 1994).

Several proteins required for SJ formation have been identified, and these show a high degree of interdependence for their junctional localization. A core SJ complex consists of the three cell adhesion molecules Neurexin IV (Nrx-IV), Contactin (Cont) and Neuroglian (Nrg), and interacts via the Nrx-IV cytoplasmic tail with the cytoplasmic FERM domain protein Coracle (Cora) (Faivre-Sarrailh et al., 2004; Genova and Fehon, 2003; Tepass et al., 2001; Ward et al., 1998). An analogous protein complex, consisting of the cell adhesion molecules NCP1, NF-155, Contactin and the scaffolding protein 4.1B, is required for formation of the vertebrate paranodal septate junction, providing evidence for a molecular and functional conservation of these junctions (Banerjee et al., 2006; Girault and Peles, 2002; Tepass et al., 2001). Drosophila SJs also include two claudin-related tetraspan proteins, Sinuous (Sinu) and Megatrachea (Mega; Pickel – FlyBase) (Behr et al., 2003; Wu et al., 2004), the homophilic cell adhesion protein Lachesin (Llimargas et al., 2004), the Na,K-ATPase (Genova and Fehon, 2003; Paul et al., 2003) and a cytoplasmic MAGUK protein called Varicose (Vari) (Wu et al., 2007). However, the composition of the SJs is only partially defined, and how the components come together to form arrays of septa with paracellular fence function is not clear.

Many SJ components have been identified through studies of the developing embryonic trachea (respiratory organ), a network of epithelial tubes that spans the entire organism. In embryos with disrupted SJ components, the main tracheal tubes, the dorsal trunks, become excessively elongated (Wu and Beitel, 2004). The tube size defect could be ascribed to a need for SJs in the apical secretion of proteins that have functions in tube size regulation (Moussian et al., 2006; Wang et al., 2006).

In search of components required for SJ formation, we screened for new tracheal tube size genes and identified crooked (crok), a Ly6-like gene essential for SJ formation. A recent study has identified another member of the Ly6 gene family, boudin (bou), which is involved in SJ formation (Hijazi et al., 2009). Bou accumulates in the cytoplasm, but is also secreted and can support SJ assembly in a non-cell-autonomous manner (Hijazi et al., 2009). We show that Crok is needed for septa formation, the membrane accumulation of SJ components and paracellular barrier functions. Crok accumulates in cytoplasmic puncta that probably represent intracellular vesicles but, unlike Bou, Crok is required tissue-autonomously for SJ formation. The finding that two Ly6 proteins are required for SJ assembly, prompted us to survey additional Ly6 genes for a role in SJ formation. Functional analyses revealed two genes, crimpled (crim) and coiled (cold), that are expressed in the ectoderm and are required tissue-autonomously for SJ assembly, demonstrating that four Ly6-like proteins have selective and non-redundant functions in SJ assembly. A detailed analysis of Cold shows that this protein also accumulates in intracellular vesicles. Moreover, we find that Crok and Cold affect the intracellular trafficking of Nrx-IV, and propose that they have specific roles in the assembly or maintenance of the core Nrx-IV-containing SJ complex.

crok^KG06053 was obtained from Bloomington Drosophila Stock Center (BDSC, Indiana, USA) and is reported to carry two P-element insertions: KG06053a within CG17218 (corresponding to C²⁹ of Crok) and KG06053b within the pipsqueak (psq) gene. We could not detect KG06053b by PCR amplification of genomic DNA, and the tracheal phenotype and lethality was retained when crok^KG06053 was placed in trans to Df(2L)esc-P3-0, which uncovers crok, but not when it was placed in trans to Df(2R)stan1, which uncovers psq. Excision of KG06053a yielded five lethal crok mutant alleles and three lines in which lethality and tracheal phenotypes were rescued.

P{Mae-UAS.6.11}bou^GG01077, l(1)air7²⁸/Dp(1;Y)ct⁺y⁺/FM7 (which uncovers bou), PBac{WH1}CG2813^f05607 and PBac{WH1}CG2813^f02238 (which are located in the first intron of CG2813), and the deficiency lines listed in Fig. 5 were obtained from BDSC. UAS-RNAi lines were obtained from the Vienna Drosophila RNAi Center (VDRC; Vienna, Austria): 44197 (CG2813), 44198 (CG2813), 44880 (CG6583), 11746 (CG6583), 30379 (CG6038) and 5443 (CG14430). Their expression was driven in tracheal cells from stage 11 using btl-Gal4.

The UAS-crok and UAS-cold constructs contain the crok (CG17218) or cold (CG2813) open reading frame from RE26906 or LD16147, respectively. In UAS-crok-GFP, the GFP coding sequence was inserted after position I²⁴ in the Crok protein. The two UAS-crok-FLAG lines contain the epitope sequence DYKDDDDK after K²⁵ (with the endogenous IK duplicated on either side) or after W²⁷ (with the endogenous IKCW duplicated on either side), and the FLAG was flanked by single glycine residues. In the UAS-cold-FLAG construct, the DYKDDDDK sequence was inserted after position F¹²¹. Germline transformations were carried out by BestGene or the Duke University Model Systems Genomics facility.

The GFP-tagged Nrx-IV line (Carnegie protein trap collection) contains a GFP-encoding exon inserted within an intron in the Nrx-IV locus (Buszczak et al., 2007). This insertion line is homozygous viable, suggesting that the tagged protein is fully functional.

Embryos were fixed with 4% formaldehyde for 20 minutes, except when using anti-Arm, anti-Baz and anti-Sinu, for which heat-methanol fixation was used. Primary antibodies used were: mouse monoclonal IgM 2A12 (1:10, Developmental Studies Hybridoma Bank, DSHB), rabbit anti-GFP (1:500; Molecular Probes), guinea pig anti-Cora (1:2000) (Fehon et al., 1994), mouse IgG anti-Atpα (1:10, DSHB), rabbit anti-Nrx-IV (1:500, from S. Baumgartner, Lund University, Sweden) mouse IgG1 monoclonal anti-Crb (1:10; DSHB), mouse monoclonal IgG2a anti-Fas3 (1:10; DSHB), rabbit anti-Sinu (1:500, from G. Beitel, Northwestern University, IL, USA), mouse anti-Arm (1:10, DSHB), rabbit anti-Baz (1:1000, from A. Wodarz, Georg-August-University, Germany) and rabbit anti-Knk (from B. Moussian, Tübingen University, Germany). Alexa-conjugated secondary antibodies (Molecular Probes or Jackson ImmunoResearch) were used at dilutions of 1:500 to 1:1000. Labelling with the Fluorescein-conjugated Chitin-binding Probe (1:500, New England Biolabs) was performed according to the manufacturer's recommendations.

UAS-cold-FLAG was transfected into S2 cells with dimethyldioctadecylammonium bromide (DDAB) (Han, 1996). Two days after transfection, cells were transferred onto Con-A-coated coverslips and cultured for 15 minutes. After adding a 10-kDa dextran conjugated with a tetramethylrhodamine dye (Molecular Probes) to the medium (final concentration 100 μg/ml), cells were incubated for 1.5 hours, fixed with 4% paraformaldehyde for 10 minutes and stained with mouse anti-FLAG (M2, 1:10,000, Sigma-Aldrich).

Crok antisera were generated by Inbio (Tallinn, Estonia) by co-injecting two peptides, CRKIRQKVHGEWRYFRS (amino acids 68-84) and CAYMGEPGIEGDERF (amino acids 85-99), into rabbits. The antisera were affinity-purified on columns covalently linked to either peptide, and were used at a dilution of 1:100.

The dextran dye permeability assay was performed as previously described (Lamb et al., 1998). Embryos were analyzed 20 minutes after injection. To label the endocytic and recycling vesicles, embryonic epidermis was observed one and half hours after injection of the fluorescent dextran dye. Confocal imaging was performed using either a Bio Rad Radiance 2000 system or a Zeiss LSM510 microscope. A Nikon eclipse E1000 microscope was used for obtaining wide-field fluorescence images. Early-stage 16 wild-type and crok mutant embryos were prepared for TEM as previously described (Tonning et al., 2005).

Twenty stage 15 crok/crok or crok/CyO embryos were crushed in 20 μl sample buffer, the protein separated on a 15% SDS-PAGE gel and blotted onto a PVDF-membrane. The primary antisera, anti-Crok (1:1000), anti-Nrx-IV (1:5000), anti-ATPα (1:100) and anti-α-tubulin (Sigma, 1:10,000), were detected with HRP-conjugated secondary antibodies.

Whole-mount in situ hybridization was performed with digoxigenin-labelled RNA probes as previously described (Tonning et al., 2005). RNA probes were generated from genomic DNA templates amplified with primers containing T7 and T3 polymerase sequences (see Table S1 in the supplementary material), except for crok, when cDNA RE26906 was used as template for in vitro transcription. The crok probe was hydrolyzed to generate 0.5 kb fragments.

We searched for new tracheal tube size mutants by analysing available transposon-insertions in genes expressed in the embryonic trachea (according to the Berkeley Drosophila Genome Project, BDGP) (Tomancak et al., 2002). Using this strategy, we identified a mutant that we called crooked (crok), owing to its excessively long and convoluted tracheal dorsal trunks (Fig. 1A,B). crok mutants are reported to carry a P-element insertion allele, CG17218^KG06053, in the second protein-coding exon of CG17218, and by genomic sequencing we confirmed the position of the CG17218^KG06053 P-element insertion. The tracheal tube size defects and lethality are also detected in embryos transheterozygous for CG17218^KG06053 and Df(2L)esc-P3-0, which uncovers CG17218 (Fig. 1C). Moreover, excision of the CG17218^KG06053 P-element yielded viable lines with normal trachea, indicating that the P-element insertion causes the tracheal phenotype and lethality of crok mutants. To confirm that disruption of CG17218 brings about the crok mutant tracheal defects, we used the UAS/GAL4 system to express CG17218 in the ectoderm of crok mutants, driven by the 69B GAL4 line (ect>crok). Such expression restored the tracheal tube size defects of crok mutants (Fig. 1D) and produced healthy larvae, providing evidence that CG17218/crok is a new tracheal tube size gene.

The convoluted dorsal trunks in crok mutants are typical for embryos with disrupted SJ components (Wu and Beitel, 2004). Additional characteristics of such SJ mutants are disorganized tracheal luminal chitin matrix and impaired apical localization of the Knickkopf (Knk) protein, which is required for correct chitin matrix assembly (Moussian et al., 2006; Tonning et al., 2005; Wang et al., 2006). Indeed, when staining for chitin and Knk, we found that crok mutants also exhibit a similar disorganized luminal chitin matrix, associated with a failure to restrict Knk to the apical surface (Fig. 1E,F).

To directly test whether crok is required for SJ functions, we analyzed the paracellular diffusion barrier. A 10-kDa dextran conjugated to a Rhodamine dye was injected into the embryonic body cavity of crok mutant and wild-type embryos at stage 16, and the passage of the dye across the tracheal epithelium was assessed. In wild-type embryos, the dye is excluded from the tracheal lumen when analyzed 20 minutes after injection (Fig. 1G) (Lamb et al., 1998). In crok mutant embryos, the dye rapidly diffused into the tracheal lumen (Fig. 1H). The dye also leaked into the lumen cavity of the mutant salivary glands (Fig. 1I,J), suggesting that crok is required for paracellular barrier functions in both tracheal and salivary gland epithelia. Consistent with the impaired paracellular barrier, transmission electron microscopy (TEM) analyses of stage 16 crok mutant embryos revealed reduced and irregular SJs. Although wild-type SJs are seen as ladder-like structures of electron-dense septa between adjacent cell membranes, the mutant tracheal cells formed sparse and less electron-dense septa (Fig. 1K,L). The tracheal cells of crok mutants were also seen to contain large unusual vesicle-like structures (Fig. 1M,N), which have previously been observed upon loss of the SJ component Sinu (Wu et al., 2004). Crok is therefore required for paracellular barrier functions and septa formation.

To further understand the basis for the reduced septa in crok mutant embryos, we analyzed the subcellular distribution of SJ components in the mutants. The embryos were stained for Nrx-IV, ATPα, Cora, Sinu and the homophilic cell adhesion molecule Fasciclin 3 (Fas3), which is a sensitive marker for SJ integrity (Fig. 2). SJ components begin to accumulate along the apicolateral plasma membrane at stage 14 in the wild-type trachea (Laprise et al., 2009). In crok mutants, tracheal cells show defects in plasma membrane accumulation of SJ components already at early stage 15, but the components appear present in the cytoplasm (Fig. 2A). Consistently, comparison of the levels of Nrx-IV and ATPα in extracts from stage 15 wild-type and crok mutant embryos on immunoblots shows that crok mutants produce these SJ components (Fig. 2D). At stage 16, the analyzed SJ components appear at low levels along the entire lateral surface of crok mutant tracheal cells with little, if any, apical concentration (Fig. 2B). Crok is therefore required for efficient incorporation of SJ components in the apicolateral part of the plasma membrane.

We also analyzed the hindgut epithelium, as SJ components normally show clear accumulation within the apical third of the lateral membrane in these columnar cells (Fig. 2C). Again, the apical staining for SJ components is reduced in crok mutant embryos. The strongest effects were observed for Sinu and Fas3, which show almost no apical accumulation in the crok mutant hindgut. Intriguingly, only small changes in the distribution of SJ components were seen in the crok mutant salivary gland (Fig. 3E-G; data not shown), despite it being leaky to the 10-kDa dextran dye, suggesting that SJ assembly in the salivary glands might involve partly redundant mechanisms. As no other effects on epithelial integrity were observed in crok mutants, as assessed by staining for the apical determinants Crumbs (Crb) and Bazooka (Baz), and the adherens junction protein Armadillo (Arm) (Fig. 3A-D; data not shown), we conclude that the requirement for Crok in junction formation is specific to the SJs.

Crok is predicted to be a small protein of 151 amino acid residues (Fig. 4A). A signal peptide and a short C-terminal transmembrane region together with an omega-site at asparagine¹²³ suggest that the protein is GPI-linked (bigPI predictor) (Eisenhaber et al., 1998; Eisenhaber et al., 1999). Structural homology detection using the Phyre web server to search the SCOP database of known 3D structures (Kelley and Sternberg, 2009) predicts that the processed Crok protein contains a three-fingered protein (TFP) fold (Fig. 4B). The TFP fold, originally discovered for snake neurotoxins, consists of three adjacent loops emerging from a hydrophobic palm, and is stabilized by 3-6 disulphide bonds (Galat et al., 2008). The TFP fold is present in several receptors that are tethered to the membrane via GPI-anchorage, like the urokinase/plasminogen activator receptor (uPAR) and the Ly6 protein family. Ly6 proteins contain only a single TFP fold that usually is stabilized by five or six disulfide bridges, also called a LU (Ly-6 antigen/uPA receptor-like) domain. As the TFP fold occurs singly in Crok and contains 10 cysteines, Crok is related to the GPI-linked proteins of the Ly6 family.

crok (CG17218) is expressed in ectodermal tissues from stage 11/12 (BDGP). To better understand the pattern of crok expression, we performed in situ hybridization. Expression of crok is evident in tracheal cells from stage 11, as the cells invaginate from the ectoderm, and ceases in the main dorsal trunks (DT) during stage 15 while remaining in the visceral branches (Fig. 4C,D). crok transcripts are also prominent in the epidermis, foregut and hindgut from stage 12 to 15, and thus exhibit a general ectodermal distribution at the time of SJ assembly (Fig. 4D). To address the subcellular localization of Crok, we raised a Crok antiserum. The antiserum appeared to be specific to Crok on western blots (Fig. 4K) and stained wild-type embryos weakly in a pattern reminiscent of the crok mRNA distribution (Fig. 4E-G). This expression pattern was not detected in crok mutant embryos (Fig. 4H), and the antiserum strongly stained embryos that overexpress Crok in the trachea (Fig. 4I), thereby supporting that it has specificity for Crok. Confocal imaging of wild-type tracheal cells revealed a punctate Crok-staining in the cytoplasm, and co-staining for KDEL to mark the endoplasmic reticulum showed that Crok distribution overlaps with, but is not restricted to, this subcellular compartment (Fig. 4J). Intriguingly, no accumulation of Crok was seen at the SJs, as shown by co-staining for Crok and ATPα (Fig. 4L), even when Crok was overexpressed in the ectoderm (Fig. 4M). The detection of Crok outside the ER was too weak to allow examination of possible colocalization with other vesicle proteins. In further attempts to address Crok subcellular localization, we generated flies that express either FLAG- or GFP-tagged Crok variants, in which the tag was inserted after the N-terminal signal sequence. These proteins displayed a largely perinuclear localization when expressed in epithelial tissues (not shown) similar to endogenously expressed Crok, but unfortunately they could not substitute for wild-type Crok in genetic rescue tests. Together, the results suggest that Crok resides in intracellular membrane compartments, including the ER, presumably assisting the assembly or trafficking of SJ components to or from the plasma membrane.

In our screen for new tube size mutants, we found a second mutant with convoluted tracheal tubes and defective SJs (see Fig. S1 in the supplementary material). This mutant carries a P-element (CG14430^GG01077) in the 3′UTR of CG14430, another gene encoding a small GPI-linked protein with 10 cysteines, similar to Crok (see Fig. S1 in the supplementary material). CG14430 was recently reported to encode a Ly6-like protein required for SJ formation, called Boudin (Bou) (Hijazi et al., 2009). Bou was shown to be secreted and to function non-cell-autonomously in SJ formation. Thus, expression of bou in the trachea, driven by the btl-GAL4 driver, not only rescued the bou mutant tracheal phenotype, but also the hindgut defects (Hijazi et al., 2009). By contrast, when Crok was expressed in the trachea of crok mutant embryos (tra>crok), the tracheal mutant defects were rescued, but no rescuing effects were seen in the hindgut (Fig. 3H-J). Crok expression in the entire ectoderm does, however, also rescue the hindgut defects of crok mutants (Fig. 3K), suggesting that Crok functions tissue-autonomously to support SJ assembly.

Vertebrate Ly6-like proteins exhibit diverse substrate specificity and tissue expression, but groups of Ly6 proteins have been found to participate in similar biological processes (Bamezai, 2004). Because Crok and Bou are required for SJ formation, we investigated whether additional Drosophila Ly6-like proteins might be expressed in SJ-forming tissues with potential roles in SJ assembly. Using position-specific iterative BLAST (PSI-BLAST), including only proteins of 130 to 200 amino acids, we obtained 23 Ly6-like protein sequences apart from Crok and Bou. Two of these, encoded by CG4363 and CG15347, contain 20 cysteines and appear to harbour two TFP domains. By screening the remaining 21 proteins for the presence of a signal peptide and a potential GPI-anchor signal, we found 16 proteins with features similar to those of Crok and Bou (Fig. 5A). All 16 proteins are predicted to form a single TFP fold (Kelley and Sternberg, 2009), and represent a subset of the recently reported Ly6 domain-containing proteins (Hijazi et al., 2009). A multiple sequence alignment of the Ly6-like proteins revealed little primary sequence homology (Fig. 5B). Apart from the ten cysteines (marked by asterisks in Fig. 5B) and a few flanking amino acids, the proteins showed similarity mainly at their N- and C-termini, which is consistent with extensive interactions between the N- and C-terminal sequences seen in TFP domains (Galat et al., 2008).

Embryonic RNA expression patterns for nine of the 16 genes are provided by BDGP. We used in situ hybridization to verify the expression of some of these and to establish the expression of the remaining seven genes (see Fig. S2 in the supplementary material), and found that many Drosophila Ly6-like genes can be grouped according to their pattern of expression (summarized in Fig. 5A). The three genes, CG2813, CG6038 and CG6583, are intriguing, as they are expressed in the ectoderm in a similar pattern to that of crok and bou. Another three genes show selective expression in the central nervous system (CNS), which forms SJs between glial cells to provide a blood-brain barrier (Stork et al., 2008). Other prominent gene expression domains include the midgut and the peripheral nervous system (PNS) (Fig. 5A; see also Fig. S2 in the supplementary material; the BDGP).

In order to test whether additional Ly6 genes might be required for ectodermal SJ formation, we analyzed embryos homozygous for chromosomal deficiencies that uncover the gene of interest. Deficiency lines for nearly all genes were included in this survey (listed in Fig. 5A) to cover for misinterpretations of the expression data and possible non-cell-autonomous gene functions (like bou). The deficiency lines were stained with 2A12 to reveal tracheal tube size defects, and with anti-Fas3 to assess SJ integrity.

Two deficiency lines, Df(2L)ED49 and Df(2L)ED4470 (deficient for the ectodermally expressed CG2813 and CG6038, respectively) exhibited elongated and convoluted tracheal dorsal trunks in homozygous embryos at stage 16 (see Fig. S3 in the supplementary material). The embryos also show reduced and mislocalized Fas3 staining in ectodermal tissues and failed to prevent the 10-kDa dextran from diffusing across the tracheal epithelium (see Fig. S3 in the supplementary material). A third gene expressed in the ectoderm that also removes crok, CG6583, was uncovered by Df(2L)ED775. However, the tracheal tube size defects of Df(2L)ED775 were rescued by tracheal expression of crok (not shown), suggesting that CG6583 is not essential for SJ formation. None of the other deficiencies caused tracheal tube size defects.

The requirement for CG2813 in SJ formation was confirmed through analysis of a lethal piggyBac insertion located within the first intron of CG2813 (CG2813^f05607), by gene knockdown using RNA interference (RNAi), and by genetic rescue using a UAS-CG2813 cDNA transgene. We named the gene coiled (cold), reflecting its tracheal phenotype. Embryos carrying cold^f05607 in trans to Df(2L)ED49 developed elongated tracheal dorsal trunks (Fig. 6A), failed to correctly localize Fas3 in the hindgut, trachea and salivary glands (Fig. 6B-D), and showed a compromised tracheal barrier function upon dextran-dye injection (Fig. 6E). RNAi-mediated knockdown of cold in the trachea caused similar elongated dorsal trunks, reduced and mislocalized Fas3 in tracheal cells, and a defective tracheal paracellular barrier (Fig. 6G-I).

No known mutant alleles for CG6038 were available, and instead we obtained an RNAi line for CG6038 that is reported to have no off targets. RNAi knockdown of CG6038 in tracheal cells, using btl-Gal4, caused tracheal tube size defects, and we called the gene crimpled (crim) with reference to the tracheal phenotype (Fig. 6J). Tracheal expression of the RNAi line (crim^RNAi) also caused Fas3 mislocalization in tracheal cells and a compromised paracellular barrier (Fig. 6K,L), suggesting that crim is yet another Ly6 protein required for SJ formation. Because the restricted expression of cold^RNAi and crim^RNAi in the trachea was able to impair tracheal SJ functions, it appears that the tracheal requirement for Cold and Crim cannot be rescued by secretion of these proteins in other tissues. In support of the notion that cold functions autonomously, we found that expression of UAS-cold in the dorsal hindgut epithelium of cold mutant embryos under the engrailed-Gal4 driver (en>cold) rescued the SJ defects in dorsal cells, but not in cells of the ventral hindgut epithelium (Fig. 7A,B). We conclude that SJ formation requires the non-redundant function of four Ly6-like proteins of which three, Crok, Cold and Crim, function tissue-autonomously.

Our results indicating that Crok is associated with intracellular vesicles, and that SJ components are produced but fail to accumulate at the SJs in crok mutants, suggested that Ly6 proteins might have a role in the preassembly or trafficking of SJ components. We therefore examined whether Cold localizes to cytoplasmic puncta, as we had observed for Crok. Unfortunately, attempts to generate anti-Cold antibodies that recognize endogenously expressed protein were unsuccessful. As an alternative approach, we expressed epitope-tagged Cold in cultured S2 cells. Flag-tagged Cold protein accumulated in cytoplasmic puncta that partially colocalized with fluid phase-endocyosed dextran (Marchetti et al., 2009) (Fig. 7C,D), suggesting that Cold associates with a vesicular compartment in these cells.

To further explore the effect of Ly6 proteins on SJ assembly, we examined the distribution of Nrx-IV-GFP, expressed under its endogenous promoter, in live embryos. Nrx-IV is a core SJ component that is conserved between flies and mammals (Banerjee et al., 2006). Strikingly, we observed that Nrx-IV-GFP accumulates in cytoplasmic puncta in the embryonic epidermis of crok or cold mutants (Fig. 8A-C,K), suggesting a defect in some aspect of Nrx-IV trafficking. By contrast, such puncta were only rarely observed in wild-type embryonic epithelial cells (Fig. 8A,K).

To determine whether the Nrx-IV-GFP cytoplasmic puncta observed in crok and cold mutant embryos are associated with a vesicular compartment, we performed dextran-labelling experiments. Depending on the length of incubation, fluorescent dextran can label multiple compartments, including endosomes, lysosomes and recycling vesicles (Marchetti et al., 2009). Dextran injection into cold and crok mutant embryos expressing Nrx-IV-GFP revealed an obvious, although not complete, correlation between the Nrx-IV puncta and dextran-labelled vesicles (Fig. 8G,H,L). A similar correlation was seen in cora mutant embryos (Fig. 8I,L). By contrast, the rare Nrx-IV-GFP puncta seen in wild-type embryos were much less frequently associated with dextran, suggesting that Nrx-IV trafficking is severely altered in these mutant backgrounds. The large Nrx-IV-GFP puncta in crok and cold mutants might correspond to the unusual vacuolar structures seen by TEM in embryos mutant for crok (Fig. 1N), and could represent late endosomal compartments or lysosomes.

To test whether the vesicular accumulation of Nrx-IV-GFP is a general feature in embryos with defective SJs, we analysed additional SJ mutants. Accumulation of Nrx-IV-GFP in cytoplasmic puncta was evident also in cora and nervana 2 (nrv2, the beta subunit of Na⁺/K⁺ ATPase) embryos, but not in Gliotactin (Gli) mutants (Fig. 8D-F,K). Moreover, loss of cold function did not have a similar effect on either ATPα or Neuroglian (data not shown), suggesting that the puncta observed in cold mutants do not contain fully formed SJ protein complexes. However, we did observe Cora staining in these puncta (data not shown), consistent with previous findings that Cora interacts directly with the cytoplasmic tail of Nrx-IV (Ward et al., 1998) and suggesting the presence of a partial complex. Taken together with our observation that they localize to cytoplasmic puncta rather than to the plasma membrane, these results suggest that Crok and Cold have a selective role in assembly or trafficking of the Nrx-IV complex.

In this study, we have identified a group of Drosophila Ly6-like proteins with specific functions in SJ formation. On the basis of our analyses of these proteins, we propose that they are not integral SJ components, but intracellular proteins that facilitate the correct assembly and incorporation of core SJ components to form stable junctional complexes.

Ly6 proteins constitute large protein families in both vertebrates and insects. In mammals, they are expressed in cells of hematopoetic origin, the brain, vascular epithelium, kidney tubular epithelium, lung, keratinocytes, stomach, testis and prostate (Bamezai, 2004). Reflecting their differential expression, Ly6 proteins are used in diverse biological processes. Apart from acting as GPI-linked cell accessory proteins of the immune system, vertebrate Ly6 proteins function in the modulation of nicotinic acetylcholine receptors (Choo et al., 2008; Darvas et al., 2009; Grando, 2008; Levitin et al., 2008; Liu et al., 2009), remodelling of the extracellular matrix during skeletal muscle regeneration (Kafadar et al., 2009), self-renewal of erythroid progenitors (Bresson-Mazet et al., 2008), and lipolytic processing of triglyceride-rich lipoproteins by binding lipoprotein lipase (Beigneux et al., 2009). The presence of a GPI-anchor in Ly6 molecules, a lipid anchor that tethers the proteins on the outer leaflet of the membrane, also suggests that Ly6 proteins can aggregate in lipid rafts to alter the activity of associated proteins (Bamezai, 2004; Halova et al., 2002).

We find that the Drosophila Ly6-like genes also exhibit a diverse tissue-specific distribution during development, with subsets of genes showing similar tissue expression, suggesting that they participate in similar biological processes. Indeed, five of the 18 Drosophila Ly6-like genes are expressed in a similar pattern in the developing ectoderm, and at least four are required for SJ formation, including bou (Hijazi et al., 2009), crok, crim and cold (this study). The only other GPI-linked Drosophila Ly6 protein studied so far is quiver (qvr; also known as sleepless), which is required for sleep and appears to affect the levels of the voltage-dependent potassium channel Shaker (Koh et al., 2008).

Our phenotypic analyses of crok mutants show that Crok is required for plasma membrane accumulation of SJ components. As SJ components were detected in the tracheal cytoplasm of stage 15 crok mutant embryos and protein analyses on immunoblots show that ATPα and Nrx-VI are present in crok mutants, we conclude that Crok is not required for the synthesis of SJ components. Moreover, we found elevated levels of ATPα and Nrx-VI in mutant embryonic extracts, despite an apparent reduction of immunofluorescence staining for these proteins, suggesting that these components are more easily solubilized in the mutant embryos. It thus appears that Crok is required for the formation of SJ complexes at the correct plasma membrane domain.

The function of both Crok and Cold appears to be necessary for the efficient incorporation of Nrx-IV into a stable SJ-associated complex. Upon loss of Crok and Cold, Nrx-IV-GFP accumulates in large intracellular vesicles. A similar situation occurs upon loss of Cora, which interacts with the cytoplasmic tail of Nrx-IV. The presence of fluid-phase dextran in these vesicles suggests that the Nrx-IV-GFP puncta in crok and cold mutants represent endocytosed protein, possibly on route to degradation in lysosomal compartments. Consistent with this idea, we observed (data not shown) that subsets of the Nrx-IV vesicles in cold mutants colocalize with endocytic markers, including Rab5 (early endosomes), Rab11 (recycling endosomes), Hrs (late endosomes) and Dor (lysosomes). As other known SJ components were not detected in these vesicles, it appears that they contain SJ subcomplexes that include Nrx-IV and Cora. This specificity could suggest either that Nrx-IV and interacting proteins are more sensitive than are other components to disruption in SJ assembly, or that Crok and Cold are specifically involved in the stable integration of Nrx-IV and interacting proteins into the SJ complex. Further experiments will be required to discern these alternative possibilities.

Bou, Crok and Cold all appear to accumulate in intracellular membrane compartments. Previous studies have shown that uPAR, an extensively studied GPI-linked protein with two LU domains, is endocytosed and recycled, and that this trafficking is essential for its function (Sturge et al., 2003). Consistent with this idea, we found that Cold colocalizes with dextran-labelled vesicles in cultured cells. In addition, although Bou appears to accumulate in the perinuclear cytoplasm (Hijazi et al., 2009), the non-autonomous behaviour of Bou suggests that it travels to the plasma membrane. Like Bou, Crok shows an apparent association with internal membrane compartments, particularly the ER. The Crok antiserum does not detect endogenous Crok at high levels, and it is possible that Crok in addition associates transiently with the cell surface. We have so far not been able to conclusively address a possible colocalizaiton of Nrx-IV with Crok or Cold, or with markers for endosomal compartments, and it is currently unclear whether the Ly6 proteins act in the sorting, trafficking or pre-assembly of SJ components prior to their transport to the site of junction assembly, or in an endocytic recycling of the components.

The non-redundant requirement for four Ly6-like proteins in SJ assembly is intriguing and coherent with the need for multiple Ly6-like proteins in the allosteric modulation of nicotinic acetylcholine receptor (nAChR) functions. These Ly6-like proteins include Lynx1 and possibly Lypd6 in neurons, and the secreted proteins Slurp1 and Slurp2 in the ectoderm (Arredondo et al., 2005; Darvas et al., 2009; Hogg et al., 2005; Lester et al., 2009). It has further been suggested that the lynx-nAChR interactions occur during receptor biosynthesis and maturation in the endoplasmic reticulum, a main site for assembly of the multi-subunit membrane proteins of nAChRs (Lester et al., 2009). In analogy, the Drosophila Ly6-like proteins might assume roles as allosteric modulators of multiprotein SJ complexes to promote their functional association.

Together, the analyses of Crok, Crim, Cold and Bou highlight a central task for Ly6 proteins in SJ formation that might also be essential for vertebrate paranodal junction assembly. Identification of their ligands and the subcellular site of action should contribute further understanding of how highly ordered, multi-protein complexes form along precise subdomains of the plasma membrane.

We thank S. Baumgartner, G. Beitel, S. Luschnig, B. Moussian and A. Wodarz for providing antisera; S. Katz for technical help; members of the R. Fehon and I. Rebay laboratories for comments; and the Centre for Cellular Imaging at the Sahlgrenska Academy, University of Gothenburg for the use of confocal microscopy. The antibodies obtained from DHSB developed under the auspices of the NICHD and maintained by the University of Iowa were developed by M. Krasnow, M. Douglas, E. Wieschaus, E. Knust and C. Goodman. This work was funded by the Swedish Research Council and the Royal Swedish Academy of Sciences to A.E.U., and by the NIH (GM074063) to R.G.F. Deposited in PMC for release after 12 months.