Scribble and Discs Large mediate tricellular junction formation

Permeability barriers are crucial to prevent solute flow across epithelia and other barrier tissues, and are formed by tight junctions in vertebrates and septate junctions (SJs) in invertebrates (Tsukita et al., 2001). Another key junction, the tricellular junction (TCJ), is formed at the corners of epithelial cells at the point of convergence of three tight junctions or SJs. In Drosophila, the TCJ is created by a complex of three SJs and a regularly spaced series of plugs or diaphragms that fill the gap at the corners of cells to create the permeability barrier at the corners (Fristrom, 1982; Noirot-Timothée et al., 1982). SJs consist of a ladder-like array below the adherens junctions in epithelial cells (Tepass and Hartenstein, 1994; Tsukita et al., 2001). At the TCJ, the SJs turn 90° and run parallel to the cell corner and are tightly linked to the plugs that line the TCJ (Fristrom, 1982; Noirot-Timothée et al., 1982). SJs are large protein complexes with core components that include Neurexin IV (Nrx-IV) (Baumgartner et al., 1996), Coracle (Cora) (Fehon et al., 1994), the Na⁺/K⁺ ATPase [the alpha subunit (ATPα) and the β subunit, Nervana 2 (Nrv2)] and Neuroglian (Nrg) (Genova and Fehon, 2003; Paul et al., 2003). Two core transmembrane proteins of the Drosophila TCJ have been identified. Gliotactin (Gli), a choline esterase-like protein and a member of the Neuroligin family, is essential for TCJ formation and maturation of the neighboring SJ (Schulte et al., 2003). Bark Beetle (Bark) (also known as Aka, Anakonda) (Byri et al., 2015; Hildebrandt et al., 2015), is a transmembrane protein with a large extracellular triple-repeat domain that is proposed to form the tricellular plug (Byri et al., 2015). Loss of Bark results in barrier defects, disruption of the TCJ and loss of Gli from the TCJ. Loss of Gli does not disrupt the targeting of Bark to the TCJ, supporting a role for Bark in targeting of Gli to the TCJ (Byri et al., 2015). However, how Bark and Gli interact and how these proteins are recruited to the TCJ is not known. Similarly, how the TCJ complex interfaces with the three SJs that converge at the TCJ is not known.

Bark and Gli both have highly conserved PDZ (PSD95, Dlg, ZO1) binding motifs at the C-terminal end suggesting that TCJ formation or function might be mediated by PDZ domain proteins. Gli colocalizes with the PDZ scaffolding protein Discs Large (Dlg; Dlg1) in a range of tissues, including imaginal discs, glia and trachea (Schulte et al., 2003), and forms a protein complex with Dlg (Padash-Barmchi et al., 2013; Schulte et al., 2006). Dlg contains three PDZ domains, an SH3 domain and a GUK domain, but the role of Dlg at the TCJ is not fully understood. Dlg in multiple cellular contexts interacts with another scaffolding protein, Scribble (Scrib), which contains four PDZ domains and a leucine-rich repeat (LRR) domain. Scrib has roles in cell polarity, cell proliferation and trafficking in epithelial cells and scaffolding at the neuromuscular junction (Dow et al., 2007; Humbert et al., 2003; Mathew et al., 2002). The polarity complex containing Scrib and Dlg is crucial for apicobasal polarity establishment by excluding apical proteins from the basolateral surface in epithelial cells (Bilder, 2004; Bilder et al., 2000; Jacob et al., 1987; Woods and Bryant, 1991). Loss of Scrib or Dlg leads to disruption of epithelial organization, loss of polarity, and disc overgrowth. In polarized epithelia, both Scrib and Dlg are localized to the SJ but their role at the TCJ has not been determined. Here, we show that Scrib and Dlg are key components of the TCJ. Scrib and Dlg are in close proximity with the TCJ proteins Gli and Bark, and both scaffolding proteins are key to the integrity of the TCJ.

We hypothesized that there is a scaffolding protein complex at the TCJ, which mediates TCJ formation and function and interacts with Gli and Bark. Gli and Bark both contain conserved PDZ binding motifs. Although these motifs are not necessary for protein localization (Byri et al., 2015; Schulte et al., 2006), we wondered whether the TCJ Gli and Bark proteins interact with PDZ scaffold proteins, such as Scrib and Dlg, and if this interaction facilitates the formation of the TCJ. To test this hypothesis, we first checked the distribution of the TCJ proteins and the scaffolding proteins. We used Bark tagged with GFP (Bark::GFP), a tagged transgene that does not disrupt the PDZ motif, protein function or protein localization to the TCJ (Byri et al., 2015). As expected, both Gli and Bark are restricted to the TCJ within the epithelia of the wing imaginal disc, including the large ribbon-like TCJ of the peripodial epithelia and the smaller TCJ of the columnar epithelia (Fig. 1A-D‴). We used a proximity ligation assay (PLA) to confirm that Gli and Bark are in close proximity at the TCJ; PLA can detect proteins in proximities of less than 40 nm (Wang et al., 2015). PLA puncta are generated by annealing of oligonucleotides conjugated to secondary antibodies in close proximity, priming synthesis of a long DNA chain by rolling circle amplification. Therefore, positive PLA puncta can appear diffuse and offset from the origin of the interacting protein pair and associated subcellular structures. We detected Gli+Bark PLA puncta in the peripodial (Fig. 1E‴, arrow) and columnar epithelia (Fig. 1F‴, arrow, and 1G-G″) compared with controls lacking either one of the antibodies (Fig. S1). To further test the specificity of the PLA, we used apterous-GAL4 to drive Bark-RNAi in the dorsal side of the wing imaginal disc. Knockdown of Bark led to loss of the Gli+Bark PLA (Fig. 1H-H″). Thus, our result confirms that Gli and Bark are in close proximity at the TCJ. Using Scrib endogenously tagged with GFP (Scrib::GFP, a homozygous viable insertion) or immunolabeling, we observed that Scrib is localized throughout the bicellular junction and concentrated at the TCJ with Gli and Bark in both the peripodial (Fig. 1A-B‴; arrows) and columnar (Fig. 1C-D‴; arrows) epithelia of the wing imaginal disc. These observations suggest that Scrib, Bark and Gli could form a complex at the TCJ.

To identify how Scrib interacts with the two TCJ proteins Bark and Gli, we used PLA in the wing imaginal disc. We found that Scrib is in close proximity with Gli (Fig. 2A-D″) and Bark (Fig. 2E-H″) in both peripodial (Fig. 2A‴,E‴, arrows) and columnar (Fig. 2B‴,F‴, arrows) epithelia. We noted that the Bark+Scrib PLA puncta appeared to be more extensive within the peripodial cells and extended beyond the TCJ (Fig. 2E‴) into intracellular vesicles, compared with the Gli+Scrib PLA puncta (Fig. 2A‴). We confirmed the specificity of this close proximity using controls lacking either antibody (Fig. S1) and through RNAi-mediated knockdown. Knockdown of Bark using apterous-GAL4 reduced the Bark+Scrib PLA puncta (Fig. 2H-H″). Knockdown of Bark also led to a reduction in the Gli+Scrib PLA puncta (Fig. 2D-D″), which was predicted given that Bark is necessary for recruitment of Gli to the TCJ (Byri et al., 2015). Unexpectedly, knockdown of Gli caused a reduction in the Bark+Scrib PLA puncta (Fig. 2I-I″). With the Gli knockdown, Dlg, Bark and Scrib were still retained at the membrane (Fig. 4E-I″); thus, the reduction of PLA puncta is not due to the loss of the entire TCJ complex from the membrane. These findings suggest that the presence of Bark and Gli is required for the formation of a complex with Scrib at the TCJ and suggest that recruitment of Scrib is dependent on the presence of Gli.

We next wanted to identify whether the PDZ scaffolding protein Dlg is also part of the TCJ complex. Gli has previously been shown to colocalize and form a complex with Dlg at the TCJ (Schulte et al., 2006). Using PLA we determined that Scrib is in close proximity with Dlg throughout the SJ and TCJ domains (Fig. 3A-C″) in both peripodial (Fig. 3A‴, arrow) and columnar (Fig. 3B‴, arrow) epithelia. We confirmed that Gli and Dlg are in close proximity at the TCJ of both peripodial (Fig. 3D‴, arrows) and columnar epithelia (Fig. 3E‴, arrow). We also found that Bark and Dlg are in close proximity at the TCJ (Fig. 3G-I″) of both peripodial (Fig. 3G‴, arrow) and columnar (Fig. 3H‴, arrow) epithelia. We used apterous-GAL4 to knock down Gli in the dorsal side of the wing disc, and observed a reduction in the Bark+Dlg PLA puncta (Fig. 3J-J″) and Gli+Dlg PLA puncta (Fig. 3K-K″). As a control, we also tested for PLA between Gli or Dlg and Talin (Rhea), a component of the focal adhesions at the basal-most side of the columnar epithelia. We used Talin endogenously tagged with mCherry (Talin::mCh), which does not disrupt function or localization (Venken et al., 2011; Klapholz et al., 2015) and we found no Gli+Talin or Dlg+Talin PLA puncta (Fig. S1F-J″).

Quantification revealed that the knockdown of Gli led to a significant reduction in both the Bark+Dlg and Bark+Scrib PLA puncta compared with control (Fig. 3L). Overall, our PLA results suggest that there is a novel TCJ complex between the scaffolding proteins Scrib and Dlg with the membrane proteins Gli and Bark (Table 1). The loss of this complex in the absence of Bark is expected given that Bark is necessary to recruit Gli to the TCJ. However, the reduction of the positive PLA between Bark+Dlg and Bark+Scrib in the absence of Gli suggests that the complex requires the presence of Gli to recruit Scrib and Dlg effectively.

To test the interactions between Scrib and the other components of the TCJ, we carried out RNAi-mediated knockdown experiments. Bark mutants result in the loss of Gli from the TCJ of the embryonic epidermis or from mutant clones in the wing disc (Byri et al., 2015). RNAi-mediated knockdown of Bark using apterous-GAL4 in the dorsal side of the wing imaginal disc led to a basolateral spread of Scrib (Fig. 4A-D‴) and Dlg (Fig. 4A-B‴), and loss of Gli from the TCJ and cell membrane (Fig. 4A-D‴). Quantification revealed that the basolateral spread was significant for both TCJ and SJ components compared with control (Fig. 4J). Using apterous-GAL4 with Gli-RNAi, we found that Gli downregulation did not disrupt the distribution of Bark at the TCJ (Fig. 4E-E‴) or change the distribution of Scrib (Fig. 4E-E‴) or Dlg (Fig. 4G-G‴) at the TCJ and SJ. However, there was a basolateral spread of Scrib, Dlg and Bark (Fig. 4F-F‴,H-I‴), similar to Nrv2 and Cora (Fig. 4I-I‴). Quantification revealed that the basolateral spread was significant for both TCJ and SJ components compared with control (Fig. 4K). These findings suggest that Bark is necessary for accumulation of Gli at the TCJ, and that loss of both Bark and Gli from the TCJ results in a basolateral spreading of Dlg and Scrib along with other SJ components. This suggests that the presence of both TCJ proteins Bark and Gli are needed to hold the SJ in the normal location just basal to the adherens junction. Given that Bark is present when Gli is knocked down, this also suggests that Gli may play a key role in mediating the interface between the SJ and TCJ.

Our results suggested a link between the SJ and TCJ domains. To further test the link between the SJ and TCJ, we knocked down the two core SJ components, Nrx-IV and the alpha subunit of the Na/K ATPase pump (ATPα). Using apterous-GAL4, ATPα RNAi-mediated knockdown led to the basolateral spread of Scrib, Gli (Fig. 5A-C‴,M; Fig. S2D-E″) and Bark (Fig. 5D-F‴,M; Fig. S2B-C″). Surprisingly, we found that Scrib, Gli and Bark remained concentrated at the corner of the cells even while the complex spread into more basal regions (Fig. 5C-C‴,F-F‴; Fig. S2C-C″,E-E″), and seemed to reflect an extension of the TCJ complex in the basal direction. Similar to ATPα, we knocked down Nrx-IV or Varicose (Vari, a PDZ domain SJ protein; Bachmann et al., 2008) separately in the dorsal side of the wing imaginal disc. Reduction of Nrx-IV or Vari (data not shown) resulted in the basolateral spread of Scrib, Gli (Fig. 5G-I‴,N; Fig. S2F-G″) and Bark (Fig. 5J-N, Fig. S2H-I″), and again the TCJ protein complex continued to be localized at the cell corners (Fig. 5I-I‴,L-L‴; Fig. S2G-G″,I-I″). Quantification revealed that the TCJ components extended consistently three times the normal length of the TCJ/SJ domain (Fig. 5M,N). Our findings suggest that, even in the absence of SJ components, TCJ proteins are targeted to the corner of the epithelial cells but the tight concentration of the TCJ just under the zona adherens was disrupted and the entire complex extended basally down the cell corners.

To test the interactions between Scrib, Dlg, Gli and Bark at the TCJ, we used RNAi-mediated knockdown of Scrib and Dlg. We regulated the degree of RNAi expression using temperature shifts to 29°C for 24 or 48 h to knock down Scrib to levels that resulted in loss of Scrib from the SJ domain without disrupting epithelial polarity. Loss of Scrib immunolabeling from the entire SJ and TCJ region resulted in loss of Gli (Fig. 6A-C‴) and Bark (Fig. 6D-F‴) from the TCJ. Dlg-RNAi has previously been shown to downregulate Gli at the TCJ (Padash-Barmchi et al., 2013). We then used RNAi to knockdown Dlg to a level that did not interfere with polarity but did remove Dlg immunolabeling from the TCJ and SJ domains. Loss of Dlg from the SJ led to downregulation of Scrib (Fig. 6G-H‴), and loss of Bark (Fig. 6I-K‴) from the TCJ. In attenuating the RNAi, we observed examples in which Dlg and Scrib were lost from the SJ but still retained at the TCJ in patches of cells (Fig. 6C-C‴, arrows). In these instances, when Scrib or Dlg were still concentrated at the TCJ (but not the SJ domain), Gli, Bark and Dlg or Scrib were also retained (Fig. 6C-C‴,F-F‴,K-K‴, arrows). These results suggest that the formation of the TCJ protein complex is dependent of the presence of both Dlg and Scrib at the cell corners and point to a key role for both Dlg and Scrib in mediating the recruitment of Bark and Gli to the TCJ.

To identify the role of Scrib and Dlg in the control of SJs, we assayed the SJ protein Nrv2, the beta subunit of the Na/K ATPase, and a core component of the SJ (Genova and Fehon, 2003; Paul et al., 2003). Loss of Scrib and Dlg caused the reduction and basolateral spread of Nrv2 (Fig. 7). The expression and localization of the adherens junction protein E-cad (Shotgun) was unaffected (Fig. 7A‴,B‴,D‴,E‴). With attenuated Scrib or Dlg knockdown, as before, Scrib or Dlg were retained at the TCJ (and immunolabeling lost from the SJ). In these regions, Nrv2 was still localized to the SJ domain around the columnar epithelia (Fig. 7C″,F″, arrows). Even with the variability of the RNAi-mediated knockdown, our results suggest that the presence of Scrib and Dlg at the TCJ is required for the correct localization of the SJ complex just below (basal to) the adherens junction.

Scrib and Dlg contain a range of conserved protein domains known to scaffold protein complexes. Scrib contains a series of LRR domains in the N-terminal half of the protein and four PDZ domains in the C-terminal half (Bilder and Perrimon, 2000) (Fig. 8G). To determine which domains of Scrib might be important for control of the TCJ complex, we utilized previously characterized scrib alleles that lack one or more of the PDZ domains (Zeitler et al., 2004). We tested the scrib[dt6] mutant, which retains the LRR domains but lacks all four PDZ domains, and the scrib[dt12] mutant, which lacks the third and fourth PDZ domains and retains the first and second PDZ domains (Fig. 8G) (Zeitler et al., 2004). Both mutants, when placed over the scrib[673] null allele, generate wing discs with different degrees of epithelial disruption (Zeitler et al., 2004). In the scrib[dt6]/scrib[673] discs, the distribution of the TCJ protein complex, including Dlg, Gli (Fig. 8A-C‴) and Bark (Fig. 8A,A″), was disrupted along with the AJ protein E-cad (Fig. 8B-C‴). However, in the scrib[dt12]/scrib[673] discs, Dlg and Gli (Fig. 8D-F‴) were correctly localized at the TCJ and Mcr, an SJ protein, localized normally (Fig. 8E-E‴). However, Bark (Fig. 8D,D″) was localized to both the bicellular SJ and the TCJ in the scrib[dt12]/scrib[673] discs. Loss of the Scrib PDZ2 domain leads to a cytoplasmic distribution suggesting that PDZ2 is required for Scrib recruitment to the lateral membrane and SJs (Albertson et al., 2004). The scrib[dt6] mutant protein is localized along the basolateral membrane and lacks polarization in wing epithelia, whereas the scrib[dt12] protein lacking the PDZ3-4 domains is concentrated at the SJ (Zeitler et al., 2004). Our findings suggest that the PDZ1-2 domains are essential for polarity, and that the disruption of the SJ and TCJ complex formation is likely due to a disruption in polarity or a failure to recruit Scrib to the SJ domain. Our results also point to a differential role of the PDZ3-4 domains in the recruitment of Bark compared with Gli.

Dlg is also a scaffolding protein and contains three PDZ, a Src homology 3 (SH3), and guanylate kinase (GUK) domain (Fig. 9H) (Woods and Bryant, 1991). Loss or disruption of the PDZ domains leads to a complete disruption of epithelia polarity (Hough et al., 1997; Woods and Bryant, 1991) and thus it was not possible to assess SJs in wing discs of these mutants. Both the SH3 and GUK domains can interact with other protein complexes and the interaction between the SH3 and GUK domains regulates the accessibility of the GUK domain to binding with other proteins (Newman and Prehoda, 2009; Shin et al., 2000). To investigate the role of the Dlg domains, we used two dlg1 mutants. dlg1[12] (also known as dlg1[m30]) changes a highly conserved Leu to Pro in the SH3 domain (Tejedor et al., 1997). dlg1[6] (also known as dlg1[v59]) lacks the last two-thirds of the GUK domain (Woods and Bryant, 1991). Both mutant proteins are stable (Mendoza et al., 2003; Woods et al., 1996) and retain function at the neuromuscular junction in terms of clustering of Shaker channels (Tejedor et al., 1997). In the dlg1[12] wing discs in both peripodial and columnar epithelia, Gli and Bark were concentrated at the cell membrane, and Dlg was also recruited to the membrane (Fig. 9A-A‴). Gli and Bark were present in those regions where Scrib was concentrated at what appeared to be junctions (Fig. 9A-B‴, arrows); however, it was difficult to detect whether these regions were intact TCJs as polarity was severely disrupted. In the dlg1[6] discs, Scrib was recruited to the bicellular junctions at the apical side of the columnar epithelia (Fig. 9D″), and the peripodial epithelia (Fig. 9C″) similar to the mutant Dlg protein (Fig. 9C‴, arrow). Scrib was concentrated in the apical domain and not found in more basal regions (Fig. 9E″). Bark was localized throughout the epithelium but not at the membrane and concentrated in intracellular puncta (Fig. 9C-E‴). We used E-cad to define the cell membrane and observed that the Bark puncta corresponded to intracellular vesicles (Fig. 9C-E‴). In the dlg1[6] mutant, Gli was not colocalized with Bark and was concentrated in smaller intracellular vesicles (Fig. 9F′,F‴). Gli localization to the TCJ is controlled by endocytosis and lysosome-mediated degradation (Padash-Barmchi et al., 2010) and so we tested for the distribution of Gli and Bark with endosomes. In dlg1[6] wing discs, Gli vesicles frequently colocalized with the late endosomal marker, Rab7 (Fig. 9F″,F‴; arrows) but the Bark intracellular vesicles did not (Fig. 9F′,F″). To confirm these observations, we knocked down Dlg (using RNAi) and this also led to the presence of Bark in intracellular vesicles (Fig. 9G′, arrows) that were distinct from the Gli-positive vesicles. Our results suggest that in the absence of Bark and/or Dlg, Gli is not retained at the TCJ, possibly due to increased endocytosis of Gli similar to the effects observed with the loss of the Gli PDZ binding motif (Padash-Barmchi et al., 2013). Overall, our results suggest that it is the collective interactions of Scrib and Dlg that result in the presence of Bark and Gli at the TCJ and that this protein complex requires the Scrib PDZ domains and the presence of the Dlg GUK domain.

We have identified a complex at the TCJ of imaginal disc epithelia that consists of Scrib, Dlg and the two TCJ proteins Gli and Bark. Scrib and Dlg are in close proximity with Gli, Bark and Dlg at the TCJ, and loss of any component from the TCJ disrupts the integrity of the complex. Although Bark is crucial for recruitment of Gli to the TCJ, Gli also plays a role in complex formation, as loss of Gli changed the proximity between Bark+Scrib and Bark+Dlg. We found that loss of Scrib and Dlg proteins from the TCJ (but not the SJ) led to the loss of Bark, Gli, and Scrib or Dlg. The last region where Scrib and Dlg were detected before entire loss from the cell membrane was at the TCJ. Only when Scrib or Dlg was lost from the TCJ was the TCJ complex disrupted. This leads us to suggest a model in which the TCJ complex is interdependent whereby all four components (Bark, Gli, Scrib and Dlg) are required for formation of a stable TCJ complex.

The TCJ complex also regulates the organization of the bicellular SJs. Knockdown of both TCJ proteins Gli and Bark results in the basolateral spread of Scrib, Dlg, and other SJ components, and knockdown of ATPα and Nrx-IV, two main SJ proteins, leads to the spread of Gli, Bark and Scrib at the TCJ. This supports prior work where the loss of Gli and Dlg in embryonic epithelia leads to the basolateral spread and mislocalization of the SJ components within the membrane (Oshima and Fehon, 2011; Schulte et al., 2006). In our model, we propose that Gli is key to linking the SJ with the TCJ as loss of Gli does not affect the recruitment of Bark to the cell corners yet leads to the spread of the SJ proteins. Our results suggest that the formation of the Scrib/Dlg/Gli/Bark complex at the cell corners is independent of the presence of the SJ. However, there is a strong relationship between the SJ and TCJ as the tight localization of each of these junctions just basal to the zona adherens requires the presence of the other. These results mirror previously suggested models of the TCJ based on scanning and transmission electron microscopic analyses (Noirot-Timothée et al., 1982; Fristrom, 1982), in which the distribution and periodicity of the plugs of the TCJ are matched with the descending SJ strands, suggesting an intimate association of both junctional complexes.

Scrib, in partnership with Dlg, has been found to be key to complex formation in multiple tissues. Scrib and Dlg are components of the Scrib polarity complex in polarized epithelia. Dlg is required to stabilize the Scrib association with the membrane and Scrib is essential to recruit Lethal giant larvae [Lgl (l(2)gl] to the cortical domain (Bilder et al., 2000). Scrib forms a complex with Dlg at the neuromuscular junction, where the Dlg GUK domain binds the GUK-holder protein, which in turn binds to the Scrib PDZ2 domain (Albertson et al., 2004; Albertson and Doe, 2003; Mathew et al., 2002). In vertebrate epithelia, Scrib is a regulator of tight junction function; knockdown of Scrib disrupts the epithelial barrier and inhibits tight junction reassembly leading to discontinuous patterns of ZO-1 (Tjp1) and occludin at cell-cell contacts (Ivanov et al., 2010). Thus, Scrib is important for tight junction assembly and can partner with other proteins to control junctional integrity. However, the role of Scrib at the tricellular tight junction and its interaction with the vertebrate TCJ proteins tricellulin (Marveld2) or lipolysis-stimulated lipoprotein receptor (LSR) has not been tested.

Both Scrib and Dlg contain protein-binding domains known to scaffold a wide range of proteins. In our study, we found that the PDZ domains of Scrib plus the GUK domain of Dlg play an important role in the formation of the TCJ. In scrib mutants lacking the PDZ domains, apical proteins and adherens junction markers can form, albeit in the context of disorganized epithelial structure (Zeitler et al., 2004). We found that loss of all four Scrib PDZ domains (scrib[dt6]) disrupts Gli, Bark and Dlg distribution whereas in a scrib allele retaining the first and second PDZ domains (scrib[dt12]), Gli and Dlg proteins are localized to the TCJ. This mirrors current work showing that Dlg localization is reduced in scrib[dt6] somatic clones, leading to the suggestion that Dlg localization requires the presence of the first and second PDZ domain of Scrib (Nakajima et al., 2019). Of note is the localization of Bark throughout the bicellular and TCJ domain in the scrib[dt12] allele, suggesting a further role for the PDZ3-4 domains in the specific localization to the TCJ. For Dlg, we determined that the GUK domain is important in formation of the TCJ. The SH3 and GUK domains interact with a range of binding proteins including an intramolecular SH3-GUK association. Mutations in SH3-GUK interaction disrupt protein complex assembly (McGee and Bredt, 1999; Nix et al., 2000), and dysregulate the binding of proteins to the GUK domain (Qian and Prehoda, 2006). However, SH3-GUK intramolecular interaction is not required for the localization of Scrib and Dlg to the SJ (Newman and Prehoda, 2009). Similar to the previous studies, we found that in dlg[12], which interferes with the SH3 and GUK interaction, Dlg and Scrib are localized to the membrane. However, Gli and Bark are mislocalized through the bicellular SJ, possibly due to the dysregulation of complex assembly at the TCJ. Our results suggest that Scrib PDZ1-2 and the Dlg GUK domains are essential for recruitment of the TCJ protein complex; however, it is unknown whether the interaction with Dlg and the two TCJ membrane proteins is direct or indirect through intermediary proteins.

In summary, we have determined that the scaffolding proteins Scrib and Dlg play an important role in TCJ formation, in recruiting Bark and Gli to the junction, and that this complex is mutually dependent for stability at the corners of polarized epithelia. At the TCJ, the three converging SJs turn 90° to run parallel to the TCJ, where both complexes form extensive and stereotypic contacts (Fristrom, 1982; Noirot-Timothée et al., 1982). We propose that the TCJ complex and the SJ complex are linked at the corners likely through the actions of Gli and that each complex is mutually dependent to ensure the tight localization of both just basal to the AJ.

The following fly strains were used: Bark[L200], Bark::GFP or Df(2L)Excel6009, Bark::GFP (Byri et al., 2015), scrib[dt6], scrib[dt12], scrib[673] (Zeitler et al., 2004). UAS-Scrib-RNAi (v105412), UAS-Dlg-RNAi (v41134), UAS-Bark-RNAi (v52608), UAS-NrxIV-RNAi (v8353) and UAS-ATPα-RNAi (v12330) were from the Vienna Drosophila Resource Center (VDRC). UAS-Gli-RNAi (NIG3903R-3) and dlg1[6] (DGRC106785) were from the Kyoto Drosophila Stock Center. UAS-Gli-RNAi (Bl:31869), dlg1[12] (Voelker et al., 1985), Scrib::GFP, Dlg::GFP (Morin et al., 2001), Talin::mCh (Venken et al., 2011) and apterous-GAL4 were from the Bloomington Drosophila Stock Center.

We used apterous-GAL4 drivers to express an RNAi line known to effectively knock down Scrib or Dlg. To regulate the degree of Scrib knockdown, the ap-GAL4; Scrib-RNAi larvae were moved from 25°C to 29°C for 24-48 h. This temperature shift resulted in loss of Scrib from the SJ without disrupting epithelial polarity. For Dlg knockdown, we used Dlg RNAi (VDRC v41134) without Dicer-2 at 25°C as previously described (Padash-Barmchi et al., 2013). Discs were assessed that had lost SJ localization but retained patches of Scrib or Dlg at the TCJ.

All statistical tests were calculated and presented graphically with GraphPad Prism 6. Immunofluorescence levels and distance spread were quantified using ImageJ (Schneider et al., 2012). For controls, apterous-GAL4 alone was used. The fluorescence intensity was measured and calculated in a single z-slice at the plane of the SJ domain. The mean intensity was calculated using the same fixed-size square in two areas: the apterous side (RNAi) versus the non-apterous side (WT) (Fig. 3). For each box, the mean intensity was calculated using ImageJ and the ratio of apterous/non-apterous calculated. The length of the basolateral spread was measured for the apterous side (RNAi) compared with the non-apterous side (WT) using side projections. Side projections were obtained using Softworx and a width of one cell diameter. Using ImageJ, the lower limit for each immunolabeled protein was determined (Figs 4, 5, arrows), a line drawn and the distance to the top of the SJ domain measured. The distance spread was calculated as a ratio (RNAi side over non-apterous side). For all ratios, the significance was determined using one-way ANOVA with Tukey's post-hoc test. For all statistics, the mean plus standard deviation is indicated.

Third instar larval imaginal discs were immunolabeled as described previously (Schulte et al., 2006). Primary antibodies were: mouse anti-Gli 1F6.3 at 1:200 (Auld et al., 1995), rabbit anti-Gli at 1:300 (Venema et al., 2004), mouse anti-Dlg 4F3 at 1:200 (Developmental Studies Hybridoma Bank) (Parnas et al., 2001), rat anti-DE-Cadherin at 1:50 (Developmental Studies Hybridoma Bank) (Oda et al., 1994), guinea pig anti-Mcr at 1:800 (Hall et al., 2014), rabbit anti-Scrib at 1:2500 (Albertson and Doe, 2003), rabbit anti-Nrv2.1 at 1:1000 (Sigma-Aldrich, SAB2700166), rabbit anti-GFP at 1:500 (ThermoFisher, A-6455), mouse anti-GFP at 1:300 (ThermoFisher, A-11120), rabbit anti-mCherry at 1:500 (Abcam, ab167453), rabbit anti-Rab7 at 1:2000 (Tanaka and Nakamura, 2008). DAPI was used at 1:1000 (Thermo Fisher Scientific). Secondary antibodies were used at 1:300: goat anti-rabbit (conjugated to Alexa Fluor 647 or Alexa Fluor 568), goat anti-mouse (conjugated to Alexa Fluor 647, Alexa Fluor 488 or Alexa Fluor 568), goat anti-rat (conjugated to Alexa Fluor 647 or Alexa Fluor 568) (ThermoFisher, A-32723, A-11031, A-11036, A-11077, A-32733, A-32728 and A-21247). Discs were equilibrated with Vectashield (Vector Laboratories) for 24-48 h at 4°C before mounting and imaging.

PLA was carried out using Duolink In Situ PLA reagents (Sigma-Aldrich) as described previously (Samarasekera and Auld, 2018) on imaginal wing discs fixed and incubated overnight with the following antibody pairs: rabbit polyclonal (Rb) anti-GFP+mouse monoclonal (mAb) Gli; mAb GFP+Rb anti-Gli; mAb GFP+Rb anti-Scrib; Rb anti-GFP+mAb Dlg. Control PLAs were carried out with Talin endogenously tagged with mCherry plus Gli (Rb anti-mCherry+mAb Gli) or Dlg (Rb anti-mCherry+mAb Dlg). Negative control PLA reactions were carried out with single primary antibodies alone (Rb anti-GFP, or Rb anti-Gli, or mAb Dlg) or no primary antibodies. Samples were equilibrated in Vectashield (Vector Laboratories) for 2-3 days at 4°C before mounting and imaging.

Image stacks were collected with a DeltaVision Spectris microscope (Applied Precision) with a 20× air- or 60× oil-immersion lens (NA1.4) and CoolSnap HQ digital camera. Deconvolution of 0.2 µm z-sections with Softworx (Applied Precision) used a point-spread function measured from 0.2-μm beads conjugated with Alexa dyes (Molecular Probes) mounted in Vectashield (Vector Laboratories). Side projections were created using the Softworx program. Images were exported to Photoshop CS (Adobe Systems) for compilation.

We would like to thank Drs Bilder, Budnik, Doe, Luschnig and Ward for generous sharing of antibodies and fly lines. The Developmental Studies Hybridoma Bank (DSHB) created by the NICHD of the NIH and maintained at The University of Iowa, Department of Biology, Iowa City, IA 52242 provided monoclonal antibodies used in this study. Stocks obtained from the Bloomington Drosophila Stock Center (NIH P40OD018537) were used in this study.