Scribble and Discs-large direct initial assembly and positioning of adherens junctions during the establishment of apical-basal polarity

Cell polarity is a fundamental property of cells, from yeast to neurons. Epithelial apical-basal polarity provides an example (Campanale et al., 2017). Precisely positioning polarity and junctional proteins at the plasma membrane allows cells to create domains with distinct biochemical properties, allowing, for example, intestinal cells to position glucose importers apically and glucose exporters basally. Cell-cell adherens junctions (AJs) reside at the boundary between apical and basolateral domains, and are key polarity landmarks. Once established, mutually exclusive apical and basal domains are maintained and elaborated by recruitment or antagonism between apical and basolateral polarity complexes, with remarkable conservation of function across species, although sometimes acting in different combinations.

Drosophila embryos provide a superb model for defining the mechanistic basis of apical-basal polarity establishment and maintenance (Harris, 2012). Development begins with 13 rounds of nuclear division without cytokinesis. The last four occur at the egg cortex, which provides a key polarity landmark to polarize the mitotic divisions. The plasma membrane then moves down around each nucleus, creating ∼6000 cells. During cellularization, cell polarity is initially established, with cadherin-catenin complexes assembled into spot AJs (SAJs) at the boundary of what will become apical and basolateral domains. AJ proteins are not essential for syncytial divisions (Grevengoed et al., 2003), but in their absence cell adhesion and polarity are lost when gastrulation begins (Cox et al., 1996; Müller and Wieschaus, 1996; Sarpal et al., 2012; Tepass et al., 1996).

Focus then turned to defining mechanisms required to properly position and assemble AJs. The polarity regulator Bazooka (Baz; i.e. fly Par3) colocalizes with cadherin-catenin complexes and is required for their apical positioning and supermolecular assembly (Harris and Peifer, 2004; Müller and Wieschaus, 1996). An apical actin-based scaffold thought to anchor Baz and AJs apically and dynein-driven apical transport also play roles (Harris and Peifer, 2005). The actin-junction crosslinker Canoe (Cno; Afadin in fly) and its regulator, the GTPase Rap1, act upstream of both Baz and AJ positioning (Bonello et al., 2018; Choi et al., 2013). Apical activated Rap1 positions Cno at nascent SAJs where Cno directs Baz and AJ positioning.

Polarity is then maintained and increasingly elaborated via recruitment or antagonism between a complex network of apical and basolateral polarity complexes (Tepass, 2012). The apical domain is initially defined by the Par complex (aPKC/Par-6), which is recruited apically in a Baz- and cdc42-dependent fashion (e.g. Bilder et al., 2003; Harris and Peifer, 2005, 2007; Hutterer et al., 2004), and then acts in opposition to Par-1 to focus and position belt AJs (McKinley and Harris, 2012; Wang et al., 2012). The Crumbs complex is then localized and promotes apical membrane identity (reviewed by Bazellieres et al., 2018). The Par and Crumbs complexes antagonize apical localization of basolateral polarity proteins, while the basolateral Scribble (Scrib)/Discs-Large (Dlg)/Lethal giant larvae (Lgl) module and Yurt group act in parallel to prevent basolateral spread of apical and junctional proteins (Bilder et al., 2003; Laprise et al., 2006, 2009; Tanentzapf and Tepass, 2003). Much of this is mediated by phosphorylation – aPKC phosphorylation excludes basal proteins from the apical domain (reviewed by Hong, 2018) while Par-1 phosphorylation excludes apical proteins from the basolateral domain.

Here, we focus on the Scrib module (Stephens et al., 2018). Dlg and Lgl regulate tissue growth in imaginal discs, which are precursors of the adult epidermis (Gateff and Schneiderman, 1974; Woods et al., 1996). Scrib promotes embryonic epithelial integrity (Bilder and Perrimon, 2000) and works with Dlg and Lgl in both epithelial polarity and growth regulation (Bilder et al., 2000). They are mutually required for one another's localizations, but whether they form a protein complex or act in parallel remains unclear. In the mature ectoderm, the Scrib-module antagonizes apical domain expansion and maintains AJs, with loss resulting in ectopic expansion of apical and AJ proteins along the basolateral axis (Bilder et al., 2000, 2003; Bilder and Perrimon, 2000; Tanentzapf and Tepass, 2003). Interestingly, later the Scrib module colocalizes with and is required for the assembly of core septate junction (SJ) proteins into functional SJs (Woods et al., 1996; Zeitler et al., 2004), which, like mammalian tight junctions, provide epithelial barrier function. Assembly of core SJ proteins is accompanied by enhanced immobilization on the cortex, a property not disrupted by Dlg loss (Oshima and Fehon, 2011). Thus, the Scrib module is required for SJ positioning and supermolecular assembly but not for SJ core complex formation. Dlg and its binding partner Strabismus also have a reported role in cellularization (Lee et al., 2003). Scrib and Dlg also regulate assembly/localization of other supermolecular complexes like neural synapses. Parallel work in C. elegans has revealed essential roles for the Scrib and Dlg orthologs for junctional function and polarity maintenance (Bossinger et al., 2001; Caria et al., 2018; Firestein and Rongo, 2001; Köppen et al., 2001; Legouis et al., 2000; Lockwood et al., 2008; Mathew et al., 2002; McMahon et al., 2001).

Work in cultured mammalian cells revealed parallel but distinct roles for Scrib in cell adhesion and polarity, regulating AJ formation/dynamics. Scrib localizes to the basolateral membrane, thus overlapping AJs though its localization relative to tight junctions remains less clear (e.g. Métais et al., 2005; Navarro et al., 2005). In MDCK cells, Scrib knockdown reduces Ecad-based adhesion and Ecad retention at AJs, and delays polarization, potentially via effects on p120-Ecad interactions and the Rho-GEF SGEF (Awadia et al., 2019; Lohia et al., 2012; Qin et al., 2005). Scrib also regulates polarization of MCF7 cells in 3D cysts (Hendrick et al., 2016). However, whereas mouse Scrib mutants have defects in neural tube and other tissues (Murdoch et al., 2003; Pearson et al., 2011; Yates et al., 2013; Zarbalis et al., 2004), often via effects on planar cell polarity (reviewed by Bonello and Peifer, 2019), they do not phenocopy loss of Ecad. Work from the Troyanovsky lab published while this paper was under review may provide an explanation. DLD1 cells knocked down for all three LAP family proteins – Scrib, Erbin and Lano – exhibit much more severe defects in assembly of apical junctions (Choi et al., 2019) and apical-basal polarity.

Our goal is to define the mechanisms that direct polarity establishment, with the ultimate objective being to define the full protein network involved. While the data above are sometimes portrayed as a simple linear pathway – Rap1GEFs→Rap1→Cno→Baz→AJs→proteins involved in polarity elaboration – things are more complex. For example, Cno is ‘upstream’ of Baz and aPKC, but their loss perturbs Cno supermolecular assembly (Choi et al., 2013). Wiring of the polarity system also varies significantly between different tissues. Here, we explore the roles of Scrib and Dlg. In contrast to the idea that they are primarily involved in polarity elaboration and maintenance, we find they play a crucial role in polarity establishment, acting at the top of the known hierarchy.

Polarity establishment begins at cellularization (stage 5), at the end of which Baz and AJs are positioned apically. At gastrulation onset (stage 6), additional polarity regulators come into play, successively elaborating polarity. In the canonical model, apical Crumbs, Yurt and Par6/aPKC complexes, and the basolateral Scrib module and Par-1 reinforce and elaborate polarity by mutual antagonism. In the mature ectoderm, Scrib and Dlg localize basolateral to AJs (Fig. 1A, quantified in B), where they are known to overlap with SJs. Core SJ proteins such as Coracle and Neurexin do not localize until mid-embryogenesis (Baumgartner et al., 1996; Fehon et al., 1994). If the roles of Scrib/Dlg were restricted to polarity elaboration and maintenance, one might expect they would not localize until gastrulation onset, when polarity elaboration begins, and that they then would be basolateral to AJs. However, Dlg localizes to the lateral membrane prior to SJ formation, and in fact is found all along the lateral domain as early as cellularization when polarity is established, overlapping the forming AJs (Woods and Bryant, 1991; Lee et al. 2003; Harris and Peifer, 2004; Sokac and Wieschaus, 2008). We therefore compared localization of Scrib to that of Dlg at this stage.

As cellularization proceeds, cadherin and the catenins localize to an apicolateral position in punctate SAJs (Fig. 1C, cyan arrows). AJ proteins also localize to smaller puncta along the lateral membrane, and are enriched in basal junctions (BJs; Fig. 1C, magenta arrows). Cno localization is largely restricted to nascent SAJs (Fig. 1E). In contrast to later Scrib and Dlg basolateral SJ localization, in cellularizing embryos both extended the entire length of the lateral membrane (Fig. 1C″,E″), although neither was enriched with F-actin/myosin at the furrow front. More surprising, Dlg and Scrib extended apically to overlap nascent SAJs, and by late cellularization there was enrichment of both Dlg (Fig. 1C″, cyan arrow) and Scrib (Fig. 1E″, cyan arrow) in a restricted region overlapping the nascent SAJs. Maximum intensity projections (MIPs) of multiple cross-sectional planes further emphasized this (Fig. 1C‴,E‴). The degree of enrichment was revealed by traces of intensity along the apical basal axis (Fig. 1H) or by quantification of pixel intensity in the region of the nascent SAJs versus the basolateral region (Fig. 1I; see Fig. S1A,B for details). En face views revealed that, while AJ proteins and Cno localize discontinuously to SAJs (Fig. 1F′; Fig. S2A′), Scrib and Dlg localize relatively uniformly around the circumference of the cell (Fig. 1F″; Fig. S2A″,B″). While Cno and other AJ proteins are enriched at tricellular junctions (Bonello et al., 2018), Dlg (Fig. S2C,D) and Scrib (Fig. 1G,N) are not. Apical Dlg and Scrib enrichment became even more pronounced as gastrulation began (stage 6; Fig. 1D-D″,L-L″, cyan arrows). Once again, MIPs (Fig. 1M) and pixel quantification (Fig. 1J) emphasized this. By stage 7, AJ proteins and Cno move apically and tighten in their localization, thus initiating separation from the site where Scrib is enriched (Fig. 1K). These data prompted us to consider the possibility that Scrib/Dlg might already have roles during polarity establishment rather than exclusively during polarity maintenance.

The close spatial relationship of SAJs, Dlg and Scrib during cellularization prompted us to investigate their functional interdependency. We first considered the hypothesis that proteins that position AJs are necessary for apical enrichment of Dlg and Scrib. Baz and Rap1/Cno regulate AJ positioning (Choi et al., 2013; Harris and Peifer, 2004). We thus examined whether they regulate apical Dlg enrichment. We assessed two enrichment parameters, quantifying cortical enrichment in the SAJ region versus the basolateral region (Fig. S1B), and also examining effects on the tight enrichment of Dlg at the level of the SAJs. In wild type, tight apical enrichment at the level of SAJs was consistently observed (Fig. 2A,D, arrows; seen in 21/22 embryo cross-sections) and Dlg mean fluorescence intensity was 1.60 times higher in the apical cortex (Fig. 2G). A previously validated Rap1 RNAi construct (Bonello et al., 2018) significantly reduced but did not eliminate Dlg apical enrichment: tight apical enrichment was seen in only 5/8 embryo cross-sections (Fig. 2B, arrow, versus C) and Dlg fluorescence intensity was only 1.44 times higher in the apical cortex (Fig. 2G). baz RNAi (validated in Fig. S2E) resulted in an even more pronounced change in Dlg apical enrichment: tight apical enrichment was lost (Fig. 2E; 0/9 embryo cross-sections observed), even at stage 6 (Fig. 2F), while Dlg fluorescence intensity was only 1.47 times higher in the apical cortex (Fig. 2G). Thus, both Rap1 and Baz help regulate Dlg apical enrichment, with Baz being particularly important for the strong elevation at the level of nascent SAJs.

We next examined whether AJ proteins are required for Dlg apical enrichment. To do so, we knocked down maternal and zygotic Arm (Drosophila β-catenin) using RNAi (Fig. S2F). β-Catenin is required to traffic E-cad to the plasma membrane in mammals and Drosophila (Chen et al., 1999; Harris and Peifer, 2004). arm-RNAi or mutation substantially reduces plasma membrane DE-cad (Harris and Peifer, 2004); DE-cad is lost from SAJs and BJs, and localizes to puncta aligned along the basolateral membrane and in an apical compartment above the nucleus (Fig. 2A versus H), consistent with a trafficking defect. This continues during early gastrulation (Fig. 2J). However, tight apical Dlg enrichment at the level of SAJs was not perturbed (Fig. 2H″,J″, observed in 12/13 embryos), and enrichment of Dlg in the apical cortex was unchanged (Fig. 2N). This is consistent with previous work showing that Scrib is not lost from the plasma membrane in arm maternal/zygotic mutants later in development (Bilder et al., 2003). Interestingly, Cno and Baz positioning also do not require AJ function (Fig. 2L versus M; Harris and Peifer, 2004; Sawyer et al., 2009). These data suggest AJ assembly is not essential for Dlg apical enrichment at polarity establishment, consistent with the possibility that Scrib/Dlg act upstream.

Polarity establishment begins with SAJ positioning and supermolecular assembly. Small cadherin-catenin clusters originating from the apical surface must be correctly positioned at the interface between apical and basolateral domains. Next, they organize into larger supermolecular assemblies (McGill et al., 2009). These are distributed around the circumference, with some enrichment at TCJs. AJ positioning requires Baz, an intact actin cytoskeleton, dynein-directed microtubule transport, and the small GTPase Rap1 and its effector Cno.

The surprising enrichment of Scrib/Dlg near nascent SAJs led us to examine whether they play roles during early polarization. To test this, we maternally expressed small hairpin RNAs against scrib and dlg to reduce both maternal and zygotic transcripts. We found effective shRNAs for each, which reduced protein levels below the detection threshold of immunoblotting (Fig. S2G-J; interestingly, knockdown of one did not alter levels of the other), and replicated known cuticle phenotypes (Fig. S3A-D). These included two scrib shRNAs targeting different regions of the mRNA, reducing likelihood of off-target effects.

In wild type, cadherin-catenin assembly into apical SAJs begins early in cellularization (Fig. 3A, red arrow), and by mid-late cellularization AJs proteins are strongly enriched there (Fig. 3A′,A″, red arrows; quantified in G,H). Our standard heat-fixation procedure emphasizes stably associated junctional proteins over the diffuse cytoplasmic pool, making this clearer. This enrichment becomes even clearer in MIPs of multiple cross-sectional planes (Fig. 3A‴, red arrows). Cadherin-catenin complexes also accumulate in smaller puncta along the lateral membrane (Fig. 3A-A‴, cyan arrows), and are enriched just apical to the furrow front in BJs (Fig. 3A-A‴, yellow arrows; Hunter and Wieschaus, 2000). Both scrib-RNAi (Fig. 3B-B‴) and dlg-RNAi (Fig. 3C-C‴) caused pronounced AJ protein mislocalization; Arm localized to small puncta distributed nearly evenly along the apical-basal axis. To quantify this, we measured pixel intensities along the apical-basal axis from MIPs and displayed these as heat maps (Fig. 3G). We also calculated the ratio of pixel intensity of Arm in the SAJs versus in the basolateral region (see Fig. S1A for details). This revealed that although Arm is roughly twofold enriched in the SAJ relative to the basolateral cortex in wild type, this enrichment is essentially lost after scrib-RNAi (Fig. 3H). Interestingly, AJ puncta not only spread basally (Fig. 3B-B‴, C-C‴, cyan arrows, D″ versus E″,F″) but also mislocalize into the apical domain (Fig. 3A‴ versus B‴ and C‴, brackets, D versus E,F).

A second junctional pool forms near the furrow tip (Fig. 3A-A‴, yellow arrows). BJs contain core AJ proteins, but unlike SAJs do not contain Cno or Baz. BJs are regulated by the cellularization-specific protein Nullo, and disassemble at gastrulation onset (Hunter and Wieschaus, 2000). We were surprised to find that after scrib-RNAi or dlg-RNAi, BJs also failed to form (Fig. 3A-A‴ versus B-B‴ and C-C‴, yellow arrows). Unlike punctate SAJs, in BJs Arm/DE-cad form linear arrays along bicellular contacts and are excluded from TCJs (Fig. 3D‴). After scrib-RNAi or dlg-RNAi, these arrays were disrupted (Fig. 3E‴,F‴).

As gastrulation initiates, cadherin-catenin complexes in SAJs move apically and focus (Fig. 3I arrow, quantified in L) as BJs disassemble. After scrib-RNAi, AJ mislocalization became even more pronounced at gastrulation onset – small cadherin-catenin complexes localized along the lateral membrane with little apical enrichment (Fig. 3J, quantified in L). Similar defects were seen after dlg-RNAi (Fig. 3K). Thus Scrib and Dlg play important roles in positioning and promoting supermolecular assembly of both SAJs and BJs.

Apical restriction of both Baz and AJs requires Cno and its regulator Rap1. We thus examined whether Scrib/Dlg are required to apically position Cno. Cno colocalizes with Arm and Baz in SAJs from mid-late cellularization (Fig. 4A,B, brackets), and is especially enriched at TCJs (Fig. 4G′, arrows), where it forms cable-like structures (Bonello et al., 2018). Unlike core AJ proteins, Cno puncta are not found basolateral to SAJs or in BJs (Fig. 4B″, red arrow, quantified in I,J). scrib-RNAi disrupted this tight apical restriction during mid- (Fig. 4A versus C, brackets) and late cellularization (Fig. 4B versus D, brackets). Cno puncta extended both apical (Fig. 4B″ versus D″, yellow arrows, G versus H) and basal to the usual position of SAJs (Fig. 4B″ versus D″, red arrows, G″ versus H″), though remnant apical enrichment in the top half of the cortex remained (quantified in Fig. 4I,J; Fig. S1A). Cno organization into apical supercellular cables was also lost (Fig. 4B″ versus D″, brackets), and strong TCJ enrichment was substantially reduced but not lost (Fig. 4G′ versus H′). dlg-RNAi had very similar effects (Fig. 4E,F).

We next asked whether misplaced AJ proteins remained associated or became randomly distributed with respect to one another. In wild-type, Cno and Arm colocalize in apical SAJs (Fig. 4K, yellow arrows), though Cno is more enriched at TCJs and is not included in BJs (Fig. 4K, red arrows). After scrib-RNAi, mislocalized Cno and Arm puncta showed clear colocalization in puncta along the entire length of the membrane (Fig. 4L, arrows), although the relative intensity of Arm is higher basally. This suggests that Scrib or Dlg loss does not affect basic molecular assembly of AJ complexes, but rather supermolecular assembly and retention at the apicolateral boundary.

In wild type, Cno and AJs move apically at gastrulation onset and tighten into belt AJs (Fig. 4M, arrow). Effects of scrib-RNAi on Cno became even more pronounced at gastrulation onset. Tight enrichment at apical AJs was completely lost, and instead Cno localized along the apical-basal axis (Fig. 4N, quantified in Fig. 4P). Some embryos retained residual apical enrichment (Fig. 4P) – these may be embryos receiving one rather than two copies of the shRNA construct. dlg-RNAi similarly disrupted Cno apical enrichment (Fig. 4O). Thus, Scrib/Dlg are essential for proper Cno assembly into apical AJs and its retention there during gastrulation.

AJ positioning requires Baz function. In its absence, SAJs fail to assemble and cadherin-catenin complexes localize along the lateral membrane. Cno and localized Rap1 activity are required to effectively polarize Baz. Like Cno, Baz localizes in nascent SAJs, and, unlike cadherin-catenin complexes, does not localize further basally (Fig. 5A, brackets). We therefore asked whether Scrib was essential for directing Baz localization. scrib-RNAi led to substantial reduction in cortical Baz (Fig. 5A versus B, intensity matched from same experiment; quantified in C; Fig. S1C). In contrast, immunoblotting revealed that total Baz levels were at most modestly reduced (Fig. S2E). When we elevated the Baz signal to visualize the remaining Baz, we found its tight confinement to SAJs was also reduced after scrib-RNAi (Fig. 5D″ versus E″, quantified in D‴ versus E‴,G, Fig. S1C), with cortical Baz puncta now seen in the basolateral region (Fig. 5E″, blue arrows). However, Baz depolarization was more limited than that seen after Rap1 knockdown (Fig. 5F-G), in which Cno cortical localization is completely lost (Bonello et al., 2018). Thus, the modest pool of Cno retained apicolaterally after scrib-RNAi may be sufficient to partially support apical Baz enrichment, albeit at reduced levels and with impaired clustering. Disruption of Baz localization intensified as gastrulation began. Junctional maturation and regulation differ dorsally and ventrally in wild type (e.g. Wang et al., 2012), and effects of scrib-RNAi on Baz also differ, with puncta spread along the lateral membrane dorsally (Fig. 5H versus I) and lacking tight apical localization ventrally (Fig. 5J versus K). Even more striking, the remaining Baz puncta were not arrayed around the lateral membrane, as in wild type (Fig. 5L), but formed irregular clusters (Fig. 5M). These were often on dorsal/ventral cell boundaries (Fig. 5M, arrows), potentially reflecting the normal planar polarization of Baz (Fig. 5L, arrows). Thus, Scrib is essential for correctly assembling Baz into nascent SAJs and retaining it in intact junctions as gastrulation begins.

In the absence of Scrib, epithelial integrity is strongly disrupted (Bilder and Perrimon, 2000), but how rapidly this occurs remains unclear. We thus followed scrib-RNAi embryos through gastrulation. In some mutants, including cno and Rap1 (Choi et al., 2013), other mechanisms are activated at gastrulation onset to largely restore apical restriction of AJs and Baz. There was no polarity rescue after Scrib/Dlg knockdown. In wild type, by stage 9 Arm, Cno and Baz colocalize in apical AJs (Fig. 6A,C,G, arrows), and only small Arm puncta are seen more basally (Fig. 6D). In contrast, scrib-RNAi cells began to separate apically (Fig. 6E, magenta arrow), AJs were fragmented but Arm and Cno largely colocalized in these fragments (Fig. 6E,F, cyan arrows). Baz was even more disrupted, with strong localization limited to larger junctional fragments (Fig. 6H,I, cyan versus magenta arrows). After scrib-RNAi, cortical actin was altered apically, with F-actin-rich protrusions near the apical surface (Fig. 6J,K), a phenotype also seen in embryos with reduced activity of the formin Diaphanous (Homem and Peifer, 2008) or expressing dominant-negative Rab11 (Roeth et al., 2009). The common phenotype after different perturbations suggests that correctly organized AJs restrain actin-driven protrusive behavior. When we followed scrib-RNAi embryos to much later stages (e.g. stage 13/14), we observed partial rescue of polarity, with return of Arm and Cno to apical junctions of epithelial balls and folded fragments (Fig. S3I), consistent with earlier work (Laprise et al., 2009).

The cytoskeleton of cellularizing embryos is distinctly polarized (Mavrakis et al., 2009; Schmidt and Grosshans, 2018), reflecting establishment of cytoskeletal polarity during syncytial divisions. Actin is organized into distinct pools: actin-rich microvilli are seen apically, actin lines the basolateral cortex (Fig. S4A′) and actin (Fig. S4A, arrow, A″) and myosin (Fig. S4G,I,K,M, arrows) are strongly enriched at the furrow front. Some probes suggest a pool of actin is enriched at apical TCJs (e.g. Sawyer et al., 2009). Microtubules form inverted baskets over each nucleus (Fig. S4D-D″), with nucleation from apical centrosomes. Baz/AJ positioning during cellularization requires both an apical actin scaffold and dynein-mediated retrograde transport (Harris and Peifer, 2005). One mechanism by which Scrib/Dlg knockdown could affect Baz and AJ localization is by disrupting this polarized cytoskeletal organization. However, after scrib-RNAi or dlg-RNAi, F-actin (Fig. S4B,C), myosin (Fig. S4H,J,L,N) and α-tubulin (Fig. S4E,F) localization appeared unperturbed during cellularization, although, as noted above, actin organization is altered after gastrulation. Thus, Scrib/Dlg do not affect initial AJ positioning by disrupting overall cytoskeletal organization during cellularization.

Scrib regulates protein trafficking in imaginal discs and other contexts (reviewed by Bonello and Peifer, 2019), dynein-dependent trafficking maintains AJ/Baz positioning during cellularization (Harris and Peifer, 2005), and during later stages DE-cad can traffic through early endosomes (Roeth et al., 2009). We thus examined the hypothesis that Scrib acts to organize apical AJs by trafficking DE-cad. The early endosome marker Rab5 and the recycling endosome marker Rab11 are enriched in an apical compartment above nuclei (Pelissier et al., 2003; Fabrowski et al., 2013; Fig. S5A,C, white arrows); Rab5 also accumulates in small puncta laterally, and at the cellularization front (Fig. S5A,C, blue arrows), a place where DE-cad also accumulates after Arm knockdown (Harris and Peifer, 2004). However, scrib-RNAi does not eliminate apical enrichment of the Rab5 endocytic compartment (Fig. S5B,D, white arrows) nor does DE-cad accumulate there after scrib-RNAi. To directly test the hypothesis that Scrib/Dlg regulates AJ localization and organization via a Rab5-dependent mechanism, we expressed dominant-negative Rab5 (Rab5DN; Zhang et al., 2007; this led to highly penetrant embryonic lethality). Embryos maternally expressing Rab5DN completed cellularization normally and Cno remained tightly localized and correctly clustered in SAJs. However, SAJs were shifted downward along the apical-basal axis (Fig. S5E versus F) and mispositioning continued into early gastrulation (Fig. S5G versus H; quantified in K). This effect gradually disappeared as gastrulation proceeded, with Cno focusing apically as usual (Fig. S5I versus J). This unexpected phenotype is consistent with the idea that SAJ apical-basal polarization is regulated by at least two regulatory inputs – one determining precise SAJ localization along the apical-basal axis and others controlling supermolecular organization. It will be important to follow this up using Rab5 loss-of-function mutants or RNAi. One possibility is that the effect we observed is due to Rab5 reducing the size of the apical domain, as was suggested by Pelissier et al. (2003). However, these data do not support the hypothesis that Scrib/Dlg act by promoting Rab5-regulated AJ trafficking, although this does not rule out a role for Scrib/Dlg more generally in trafficking of junctional proteins.

Par-1 is a basolateral polarity protein that helps restrict junctional and apical protein localization. Phosphorylation by Par-1 excludes Baz from the basolateral domain, and manipulating the balance of aPKC and Par-1 activity can re-position AJs (Bayraktar et al., 2006; Benton and St Johnston, 2003; Wang et al., 2012). However, the effects of Par-1 loss on DEcad and Baz localization during polarity establishment (Bayraktar et al., 2006) are largely rescued at gastrulation onset, owing to functional redundancy with dynein-based apical transport (McKinley and Harris, 2012). To investigate Par-1 in our own model, we first validated the Par-1 antibody and par-1 RNAi tool by immunofluorescence (Fig. S6A versus B,C versus D). During cellularization, Par-1 extends all along the membrane (Bayraktar et al., 2006; Fig. 7A), with highest cortical levels basally (Fig. 7B3, arrows in A indicate section planes) and gradually decreasing levels and less tight cortical localization as it extends apically to overlap SAJs and beyond (Fig. 7B2,B1). Par-1 is gradually cleared from the apical domain during early gastrulation, concentrating basolateral to AJs (Fig. S6A close-up, E). We first asked whether Scrib/Dlg regulate Par-1 localization during cellularization. After dlg-RNAi, cortical Par-1 levels were strongly reduced but not eliminated (Fig. 7A versus C). Interestingly, the strongest reduction of Par-1 cortical signal was in the basolateral region (Fig. 7B3 versus D3) rather than near the SAJs (Fig. 7B1 versus D1). To quantify this, we measured signal intensity at the basolateral membrane across the cell in cross-sections – in wild type, clear periodic peaks were observed at the cortex, and these were substantially reduced after dlg-RNAi (Fig. 7E; Fig. S1D; Fig. S6H). We assessed maximum peak height over the baseline across these transects and this revealed a significant drop in cortical intensity after dlg-RNAi (Fig. 7F). scrib-RNAi similarly reduced cortical Par-1 (Fig. S6F versus G; quantified in I,J).

We then examined whether par-1-RNAi recapitulated Scrib/Dlg knockdown, using a previously characterized shRNA. par-1-RNAi led to strong defects in blastoderm integrity due to partial loss of furrows in syncytial embryos, as previously reported (McKinley and Harris, 2012). This alone suggests Scrib/Dlg knockdown does not completely disable Par-1, as we do not observe this after Scrib/Dlg knockdown. We next examined effects of par-1-RNAi on AJs. Although par-1-RNAi altered polarity establishment, there were both differences and similarities between Par-1 and Scrib/Dlg knockdown. First, BJs were largely unaffected by par-1-RNAi (Fig. 7G,K,M versus I,L,N, cyan arrows, H versus J). However, the effects of Par-1 and Scrib knockdown on SAJs were quite similar. The normally tight enrichment of Arm in SAJs was lost (Fig. 7K′ versus L′,M′ versus N′, magenta arrows) – quantification revealed that the degree of loss of enrichment was similar to that seen after scrib-RNAi (Fig. 7Q versus 3H). Cno and Arm puncta also mislocalized up into the apical domain (Fig. 7K′ versus L′,M′ versus N′, yellow arrows), and the regular SAJ organization seen en face was also lost (Fig. 7O versus P), although this may reflect in part general disorganization of the apical membrane. In strong contrast to Scrib/Dlg knockdown, after par-1-RNAi there was partial rescue of AJ localization at gastrulation onset, with Arm and Cno cleared from the basolateral domain and AJs tightening (Fig. 7R versus S). This is consistent with Baz repolarization observed at this stage (McKinley and Harris, 2012). However, AJs in the dorsal and lateral ectoderm remained abnormally shifted basally (Fig. 7R versus S,T versus U, arrows), as was previously observed (Wang et al., 2012). Thus, Scrib/Dlg are required for Par-1 localization, and the partial overlap in phenotypes between scrib-RNAi and par-1-RNAi is consistent with the idea that altered Par-1 regulation may account in part for the effect of their loss. However, distinctions in phenotype suggest Par-1 regulation does not account for the full suite of polarity defects observed after Scrib/Dlg knockdown.

Scrib and Dlg are both multi-domain scaffolding proteins with diverse sets of binding partners (Stephens et al., 2018). The overlap between Scrib/Dlg and SAJs is potentially consistent with a simple hypothesis – Scrib and/or Dlg act as scaffolds for AJ proteins during polarity establishment. Scrib contains an N-terminal leucine-rich repeat (LRR) and four C-terminal PDZ domains. Binding partners were identified for each domain, and mutational analyses begun to assess their individual roles. In Drosophila, C. elegans and mammals, the LRRs are required for cortical localization and are essential for almost all known functions (reviewed by Bonello and Peifer, 2019). To examine LRR function in this context, we used a particularly useful missense mutant in the LRRs of Scrib. In scrib¹, leucine 223 in the 10th LRR is changed to glutamine (Zeitler et al., 2004). This residue is conserved in all LAP proteins, and structural predictions suggest it is on the surface of the horseshoe-shaped LRR. Expression of a transgenic version in Drosophila suggest it encodes a stable protein, although it does not localize to the cortex. In previous assays, scrib¹ behaved as a null allele in most assays, including its effects on polarity in follicle cells, polarity maintenance and epithelial integrity in embryos, and loss of polarity and growth regulation in imaginal discs (though transheterozygous interactions with weak alleles suggest it may retain a small amount of function; Bilder et al., 2000; Bilder and Perrimon, 2000; Zeitler et al., 2004).

We thus generated females with germlines mutant for scrib¹ and crossed them to scrib¹ heterozygous males –50% of progeny are maternally and zygotically mutant, while 50% have a wild-type zygotic copy. The cuticle phenotype was as previously observed – maternal/zygotic mutants had the ‘scribbled’ cuticle phenotype (Fig. S3D,E). We confirmed by western blotting that the mutant protein was stable but accumulated at reduced levels (∼50% wild type; Fig. S7A,B), and, consistent with work in imaginal discs (Zeitler et al., 2004), it did not localize to the cortex (Fig. S7C versus D). scrib¹ mutants recapitulated effects of scrib-RNAi on initial polarity establishment. AJ protein enrichment was lost in both SAJs (Fig. 8A,B versus E,F, cyan arrows) and BJs (Fig. 8A′,B versus E′,F, magenta arrows), and the normally tight Cno (Fig. 8A″,B′, cyan arrows) and Baz localization to SAJs (Fig. 8C′,D, cyan arrows) was reduced with puncta now found basolaterally (Fig. 8E″,F′,H, magenta arrows). Thus, the LRRs of Scrib are essential for its function in polarity establishment. As scrib¹ embryos gastrulated, polarity disruption was enhanced, with Cno and Arm spreading basolaterally at stage 7 (Fig. S7E versus G), and Baz puncta losing apical localization (Fig. S7F versus H). By stage 9, AJs were fragmented in scrib¹ (Fig. S7I versus J; Bilder and Perrimon, 2000). As we observed after scrib-RNAi, Arm and Cno colocalized in junctional fragments (Fig. S7K versus M, arrows), while Baz was restricted to brighter junctional fragments (Fig. S7L versus N, cyan versus magenta arrows). Thus, mutating this crucial residue in the LRRs essentially eliminated Scrib activity during cellularization and gastrulation, consistent with what was observed in other tissues (Bilder et al., 2000; Bilder and Perrimon, 2000; Zeitler et al., 2004). As we observed after scrib-RNAi, there was partial rescue of junctional polarity much later (stage 13; Fig. S3J).

Intriguingly, while LRR mutations largely or completely inactivate Scrib, the PDZs are not essential for all functions. scrib⁴, which encodes a truncated Scrib lacking all four PDZs, has been very informative. This mutant protein accumulates as a stable protein when expressed from a transgene (Zeitler et al., 2004), and we verified that this transgenic protein can accumulate in early embryos (Fig. S7T; it is not recognized by existing anti-Scrib antibodies). scrib⁴ rescues apical-basal polarity in later embryos and imaginal discs, but does not rescue SJ assembly or fully rescue growth regulation in imaginal discs (Zeitler et al., 2004). Its embryonic cuticle phenotype illustrates this – while the cuticle is reduced to fragments in scrib-null mutants, it remains intact in scrib⁴ mutants, although head involution and dorsal closure are disrupted (Fig. S3F,H).

Use of scrib⁴ allowed us to selectively disrupt a different subset of protein interactions with Scrib and examine their effects. Consistent with previous reports, scrib⁴ maternal/zygotic mutants retained significant ectodermal integrity at the end of gastrulation (Fig. S7Q; Zeitler et al., 2004). We next looked during cellularization, to examine whether polarity establishment proceeded normally. To our surprise, scrib⁴ embryos had defects in SAJ positioning during cellularization, although these differed from those seen after scrib-RNAi or in scrib¹ mutants. In wild type, Arm is tightly enriched in both apical SAJs and BJs (Fig. 8I,J, cyan and magenta arrows, respectively), with lower levels along the basolateral membranes, and Cno (Fig. 8I″,J′) and Baz (Fig. 8K′,L) are tightly restricted to SAJs. SAJs occupy roughly the apical 20% of the cell length (quantified in Fig. 8Q,R). In scrib⁴, SAJ-enriched Arm and Cno expanded apically and basolaterally (Fig. 8I versus M,J versus N, yellow arrows, J′ versus N′, magenta arrows; Arm quantified in Q,R), such that they now spanned more than 40% of the cell length (see Fig. S1E for methodology). This differed from total loss of apical enrichment of Arm seen in scrib¹ mutants (Fig. 8F). Arm enrichment at basal junctions was also lost in scrib⁴ (Fig. 8J versus N, magenta arrows). Baz was still enriched at the SAJ level (Fig. 8L versus P, cyan arrows), but Baz puncta were also seen more basally in scrib⁴ mutants (Fig. 8L versus P, magenta arrows). These data are consistent with a role for the PDZ domains of Scrib in initial polarity establishment, but the AJ defects were distinct from those of scrib¹ (Fig. 8A-H) or scrib-RNAi. It will be important in the future to examine in detail the localization and accumulation level of a Scrib protein lacking the PDZs – analysis in imaginal discs suggests that this protein can localize to the cortex but may be defective in polarized accumulation (Zeitler et al., 2004).

We then followed scrib⁴ mutants to see when and how polarity was re-established. In wild type, AJs move apically and become even more focused by late stage 6 (Fig. 8S,U,V, yellow arrows), although some Arm remains in BJs (Fig. 8S,U, magenta arrows). In contrast, in scrib⁴mutants, rather than focusing, SAJs, as marked by both Arm and Cno, continued to be substantially broadened at gastrulation onset – both extended further basolaterally in both the dorsal (Fig. 8U versus W, yellow arrows) and lateral epidermis (Fig. 8V versus X, yellow arrows). The transition from SAJs to belt AJs in scrib⁴ was also delayed (Fig. S7O versus P). Remarkably, however, AJ repolarization was complete by the end of germband extension (Fig. S7Q, z-stack reconstructions R versus S), unlike what we observed after scrib-RNAi. Thus, the PDZ domains of Scrib are required for many aspects of proper polarity establishment, but PDZ domain-independent mechanisms appear during gastrulation and restore polarity.

The data above suggest that Scrib localizes in a way that overlaps the AJ during polarity establishment and is important for AJ positioning. As detailed in the Introduction, previous data have suggested a role for Scrib in cell-cell adhesion in MDCK cells. Strikingly, while our manuscript was under review, the Troyanovsky group found that the three mammalian LAP proteins Scribble, Erbin and Lano play redundant roles in epithelial polarity in cultured mammalian cells (Choi et al., 2019). We wondered whether the overlap in localization observed in early fly embryos was also a feature of mammalian cells. The BioID method, in which a promiscuous biotin ligase (BirA-R118G) is added to a protein of interest, provides a means of assessing whether two proteins are in close proximity in living cells (most estimates are tens of nanometers; Roux et al., 2012; Sears et al., 2019). We had generated for other experiments doxycycline-inducible versions of mammalian Afadin, the Cno homolog, tagged with BirA at either the N- or C-terminus. Using the well-characterized epithelial cell line MDCK, we generated stable cell lines expressing each fusion, and generated a control cell line expressing BirA alone. In both experimental cell lines, Afadin and Scribble overlapped in localization at cell-cell junctions, as assessed by immunofluorescence (Fig. 9A-D; we used two different pairs of anti-Scribble and Afadin antibodies, as well as antibodies to the Myc-epitope to confirm this), but they do not strictly colocalize, reminiscent of what we observed in Drosophila. Cells were induced to express the fusion proteins, incubated with biotin for ∼28 h and biotin-labelled proteins were then pulled down using streptavidin-coated beads. Input, unbound and bound fractions were probed for Afadin and Scrib. Afadin pulldown in cells expressing either the N-terminal or C-terminal BirA fusions was highly effective, as visualized by blotting with antibodies to Afadin (Fig. 9E, top row) or to the Myc-epitope tag (Fig. 9E, second and third rows), presumably owing to self-biotinylation. However, Afadin was not pulled down in the negative control cell line expressing BirA alone. Similarly, BirA-Myc expressed alone was also effectively self-biotinylated and pulled down with streptavidin (Fig. 9E, fourth row). The negative control, tubulin, was not pulled down in any of the cell lines (Fig. 9E, bottom row). Strikingly, Scrib was pulled down from cells expressing either BirA-Afadin or Afadin-BirA, but not in cells expressing BirA alone (Fig. 9E, fifth and sixth rows; quantified in F). These data are consistent with the overlap in localization of Scrib and Canoe/Afadin being a conserved feature of animal epithelia.

Identifying the earliest symmetry-breaking events that initially position AJs, thereby setting the boundary between apical and basolateral domains, is a key aspect of understanding how polarity is established. Here, we report that Scrib/Dlg, which are best known for their roles as basolateral determinants during polarity maintenance, play a separate and surprising role in organizing AJs during polarity establishment, positioning them near the top of the polarity network.

Scrib and Dlg are multidomain proteins with many partners, allowing them to serve diverse biological functions, from synaptogenesis to oriented cell division. Our data reveal that they play distinct roles during polarity establishment and polarity maintenance, likely engaging very different sets of binding partners. This is supported by the evolving localization pattern of Scrib/Dlg on the plasma membrane, with sequential colocalization with and roles in positioning AJ versus SJ proteins, suggesting the capacity to engage with and position distinct junctional and polarity proteins. Our analyses also begin to dissect the underlying molecular basis. The PDZ domains of Scrib are important for the precision of initial polarity establishment but are redundant with other mechanisms for polarity maintenance after gastrulation, although they regulate SJ positioning.

AJs play a key role at the boundary between apical and basolateral domains, and building a functional AJ is a multistep process. This includes assembling a stable core cadherin-catenin complex, positioning it and supermolecular assembly. Assembly of the core complex into small puncta occurs before cellularization. As cellularization proceeds, these are captured at the apicolateral interface in a process requiring Baz, Cno and an intact actin cytoskeleton, where they coalesce into SAJs, with ∼1500 cadherin-catenin complexes and 200 Baz proteins (McGill et al., 2009). Cadherin-catenin complexes form independently of either Baz or Cno, but AJ positioning and full supermolecular assembly depend on both. We found that Scrib/Dlg are also key for AJ apicolateral retention and supermolecular assembly, although Arm and Cno remain associated in misplaced puncta, and thus core AJ complexes remain intact. Furthermore, a second junctional complex that arises during polarity establishment, the BJ, also requires Scrib/Dlg for its supermolecular organization. Unlike AJs, BJ organization is not dependent on other polarity determinants, including Cno, Rap1 or Par-1. It will be of interest to examine whether Scrib/Dlg act via known regulators of cadherin clustering, considering both intrinsic (e.g. cis- and trans-interactions of cadherins) and extrinsic factors (e.g. local actin regulation, endocytosis; Truong Quang et al., 2013). It is also important to note that our analyses generally use heat fixation, which emphasizes the most stably associated junctional proteins and reduces the diffuse cytoplasmic pool – it may be that Scrib/Dlg normally stabilize junctional localization of proteins such as Baz, rather than being absolutely essential for their recruitment there. We also note that others have reported a potential role for Dlg in ‘plasma membrane formation’ during cellularization (Lee et al., 2003) – we did not observe any effects on cellularization per se in our dlg knockdown conditions, nor have defects been reported in earlier examinations of dlg maternal/zygotic mutants (Bilder et al., 2000; Blankenship et al., 2007)

Our ultimate goal is to define molecular mechanisms underlying polarity establishment. Our new data place Scrib/Dlg in a crucial position near the top of the network, but also suggest they act via multiple effectors. Perhaps the strongest evidence for multiple roles with distinct effectors comes from analysis of scrib⁴. Our data reveal that supermolecular organization of both SAJs and BJ must involve interactions with specific partners via the PDZ domains – one speculative possibility is that these include core AJ proteins, as β-catenin can co-immunoprecipitate with Scribble and interacts with PDZ domains 1 and 4 (Ivarsson et al., 2014; Zhang et al., 2006). Testing this idea will be an important future direction. The initial role of Scrib may also involve modulating Par-1. During cellularization, Scrib/Dlg and Par-1 localize in ‘inverse gradients’: Scrib and Dlg enriched at the SAJ level, and Par-1 with higher cortical intensity basolaterally. Scrib/Dlg play a role in effective membrane recruitment of Par-1 at this stage. The qualitative and quantitative effects of par-1-RNAi on SAJ protein localization during cellularization are similar to those of scrib-RNAi, although they also differed in important ways – par-1-RNAi did not disrupt BJs and its effects on polarity were partially rescued as gastrulation began. It will be interesting to test whether and to what degree overexpressing Par-1 or targeting it to the membrane can rescue loss of Scrib/Dlg. These data are consistent with the idea that regulating Par-1 may be one of several mechanisms by which Scrib/Dlg act.

Scrib then plays a second PDZ-independent role as gastrulation begins, ensuring focusing of cadherin-catenin complexes and Baz into apical belt AJs. This requires the N-terminal LRRs, as was observed in other times and places, but not the PDZs. Positioning Baz at this stage involves at least two inputs that are redundant with one another, one via Par-1 and one via an apical transport mechanism (McKinley and Harris, 2012). One speculative possibility is that Scrib/Dlg also regulate protein trafficking, a role they have in other contexts. However, disrupting Scrib/Dlg function has very different consequences from disrupting Rab5-dependent trafficking using a dominant negative, suggesting they are not likely to act by activating Rab5. aPKC also provides important cues at this stage – perhaps Scrib/Dlg regulate aPKC localization or function. It will be important to further explore the nature of this second role.

As outlined in the Introduction, work in cultured mammalian cells supports a role for Scribble in cadherin-based adhesion and apical-basal polarity, suggesting that this is a conserved function. However, effects of Scribble knockout in the mouse suggested it is not crucial in all tissues. An important paper appeared while this work was under review that may help resolve this. Mammals have three LAP family proteins – Scrib, Erbin and Lano – raising the possibility of functional overlap. Consistent with this, triple knockdown of all three proteins in DLD1 cells led to much more substantial defects in apical junctional assembly and polarity (Choi et al., 2019). It will be of interest to compare mechanisms.

Our initial goal more than a decade ago was to define roles of AJs in polarity establishment. However, it rapidly became apparent AJs are not at the top of the hierarchy. Cno, Rap1 and Baz act upstream of AJ positioning and supermolecular assembly. Our new data and earlier work also suggest that polarity establishment is not a linear pathway, but instead involves a network of interactions with multiple inputs and also feedback regulation. The data here uncover a new crucial link in the network, revealing a key role for Scrib/Dlg in regulating AJ positioning and assembly. However, they also emphasize that the process is not a simple linear pathway, and raise new questions. Loss of Scrib or Dlg almost completely disrupts AJs during cellularization. However, effects on Baz and Cno, both of which also localize to nascent AJs, are less complete – their supermolecular assembly is affected, but they are largely retained in the apical half of the membrane. This suggests that there are multiple parallel inputs into AJ assembly, an idea also supported by the relatively uniform distribution of Scrib/Dlg at the level of nascent AJs, as opposed to the punctate assembly of spot AJ proteins. The role of Scrib/Dlg in both BJs and SAJs, which have distinct inputs and a very different supermolecular architecture, also support the idea that Scrib/Dlg are one of several inputs. Finally, Rap1 and Baz are required for proper apical enrichment of Dlg near SAJs during cellularization, while assembly of AJs is not, suggesting a non-linear network with feedback regulation. A similar model with multiple upstream inputs and complex interdependencies was proposed for polarization of the follicle cell epithelium, in which AJ position requires inputs from Dlg, Arm/β-catenin and Baz/Par3 (Franz and Reichmann, 2010).

The ultimate polarizing cue during syncytial development is the oocyte membrane, which then directs cytoskeletal polarization. Cytoskeletal cues regulate Cno localization. Although our data rule out a role for Scrib/Dlg in establishing basic cytoskeletal polarity, they do not rule out more subtle roles, for example, in localizing a special ‘type’ of actin cytoskeleton in the apical domain. Retention of Cno at the membrane after Scrib/Dlg knockdown suggests that basal Rap1 activity remains intact. Changes to early Par-1 and Baz cortical localization with loss of Scrib/Dlg, also raise the possibility that lipid-based regulation is impaired (Kullmann and Krahn, 2018; McKinley et al., 2012). At this time, we do not know what cues regulate Scrib/Dlg apical enrichment but AJs do not appear to direct this, nor are they essential for polarizing Cno or Baz, which do play a role in Dlg apical enrichment. Continued characterization of the full protein network and molecular mechanisms governing polarity establishment will keep the field busy for years to come.

Fly stocks used in this study are listed in Table S1. Mutations are described at FlyBase (flybase.org). Wild type was yellow white. All experiments were carried out at 25°C, with the exception of driving arm RNAi, which was carried out at 27°C. Knockdown by RNAi was achieved by crossing double copy maternal GAL4 females (Staller et al., 2013) with males carrying the UAS hairpin construct. Female progeny were then crossed back to males carrying the UAS hairpin construct. For scrib RNAi, most experiments were carried out with the valium 22 line, but key conclusions were verified with the valium 20 line. For driving arm RNAi expression, the GAL4 was provided by the males (Ni et al., 2011). Maternal expression of dominant-negative Rab5 was carried out by crossing female double maternal GAL4 flies with males carrying UAS-YFP.Rab5.S43N. To create germline clones, scrib¹ or scrib⁴ females were crossed with males carrying hsFLP;;FRT82B ovoD1. Wandering larvae were heat shocked for 2 h in a 37°C water bath on 2 consecutive days.

Staining for Nrt, Arm, Cno, Baz, Scrib, Rab5 and Zip was performed using the heat-fix method. Dechorionated embryos were fixed in boiling Triton salt solution (0.03% Triton X-100, 68 mM NaCl) for 10 s followed by fast cooling on ice and devitellinized by vigorous shaking in 1:1 heptane:methanol. Embryos were stored in 95% methanol/5% EGTA for at least 48 h at −20°C prior to staining. For staining all other antigens, embryos were fixed for 20 min with 9% formaldehyde and devitellinized in 1:1 heptane:methanol, except for staining with phalloidin, where embryos were devitellinized in 1:1 heptane:90% ethanol. Prior to staining, embryos were washed with 0.1% Tween-20 in PBS (heat-fixed embryos) or 0.1% Triton X-100 in PBS (formaldehyde-fixed embryos), and blocked in 1% normal goat serum in PBS-T for 1 h. Primary and secondary antibody dilutions are listed in Table S1.

Fixed embryos were mounted in Aqua-poly/Mount and imaged on a confocal laser-scanning microscope (LSM 880;40×/NA 1.3 Plan-Apochromat oil objective; Zeiss). Images were processed using ZEN 2009 software. Photoshop CS6 (Adobe) was used to adjust input levels so that the signal spanned the entire output grayscale and to adjust brightness and contrast. Maximum intensity projections (MIPs) were generated by acquiring z-stacks through the embryo with a 0.3 µm step size and digital zoom of two. ZEN 2009 software was used to crop stacks to 250×250 pixels along the xy-axis and to project xyz-stacks along the y-axis as previously described (Choi et al., 2013). The apical-basal position of puncta was determined from MIPs using ImageJ (NIH). Projections were rotated 90° counterclockwise and analyzed using the Plot Profile function in ImageJ to generate values of average fluorescence intensity along the apical-basal axis. These data were displayed as heat maps, illustrating intensity along the apical-basal axis with a color gradient. Graphs and accompanying heat maps were generated using GraphPad Prism 8.0. A complete description of all other quantification can be found in Fig. S1. Statistical analysis was performed using unpaired, two-tailed t-tests. A full list of n-values is presented in Table S2.

Cuticle preparation was performed according to Wieschaus and Nusslein-Volhard (Wieschaus and Nüsslein-Volhard 1986).

Knockdown efficiency by RNAi was evaluated by western blotting. Embryo lysates were prepared by grinding dechorionated embryos in ice-cold lysis buffer [1% NP-40, 0.5% Na deoxycholate, 0.1% SDS, 50 mM Tris (pH 8), 300 mM NaCl, 1.0 mM DTT and Halt protease and phosphatase inhibitor cocktail (Thermo Fisher Scientific)]. Lysates were cleared at 16,000 g, and protein concentration determined using the Bio-Rad Protein Assay Dye (Bio-Rad). Lysates were resolved by 8% SDS-PAGE, transferred to nitrocellulose filters and blocked for 1 h in 10% dry milk powder in PBS-T (0.1% Tween-20 in PBS). Membranes were incubated in primary antibody for 2 h (see Table S1 for antibody concentrations); washed four times for 5 min each in PBS-T and incubated with IRDye-coupled secondary antibodies for 45 min. Signal was detected using the Odyssey infrared imaging system (LI-COR Biosciences). Band densitometry was performed using ImageStudio software version 4.0.21 (LI-COR Biosciences).

The rat Afadin coding sequence was subcloned into pTRE2hyg-BirA-myc, an inducible mammalian expression vector for BioID (Van Itallie et al., 2013) at the N- (pTRE2hyg-Afadin-BirA-myc) or C-terminus of the BirA gene (pTRE2hyg-BirA-Afadin-myc) using a Gibson assembly reaction (New England Biolabs). Subconfluent MDCK T23 cells were transfected with N- or C-terminal tagged BioID plasmids and selected for the stable clones using hygromycin as a selective drug. The stable clones were verified by western blot and immunostaining for Afadin and Myc tag. Cells expressing BirA-Afadin were cultured in DMEM media containing 1 g/l glucose, 10% fetal bovine serum, 15 mM HEPES (pH 7.4) and 50 ng/ml doxycycline. For plating cells in polycarbonate Transwell inserts (Thermo Fisher Scientific), doxycycline was removed to induce the expression of transgenes.

Cells were cultured for 7 days in the Transwells without doxycycline to express BirA-Afadin and fixed in ice-cold ethanol for 1 h at −20°C. After three washes with PBS, the samples were incubated with blocking buffer (10% FBS in PBS) for 1 h at room temperature. Subsequently, the inserts were incubated with primary antibodies diluted in blocking buffer. Rabbit anti-Afadin and mouse anti-Scribble, or rabbit anti-Scribble and mouse anti-Myc antibodies were used for co-staining. Following three washes with wash buffer (1% FBS in PBS), the cells were incubated with the Alexa-conjugated secondary antibodies along with Hoechst 33343 to stain DNA. After three washes with wash buffer, the insert membrane was cropped out, mounted on a microscope slide with Prolong Diamond antifade mountant (Thermo Fisher Scientific) and cured before imaging.

MDCK cells were those used by Choi et al. (2016), and were authenticated and tested for contamination. MDCK cells stably expressing BirA transgenes were cultured without doxycycline for 5 days, following which cells were treated with 50 µM biotin for 27-28 h. To prepare lysates, cells were washed three times with ice-cold PBS and scraped into lysis buffer [1% NP-40, 0.5% deoxycholate, 0.2% SDS, 50 mM Tris (pH 8), 150 mM NaCl, 2 mM EDTA, supplemented with protease/phosphatase inhibitor]. Lysates were snap frozen on dry ice prior to storing at −80°C. To capture biotinylated proteins, lysates were thawed at 4°C, sonicated (amplitude 50%, 10 strokes performed manually) and incubated on ice for 15 min. Lysates were then spun at 15,000 g for 15 min and the protein concentration of supernatant determined using BioRad Protein Assay Dye. Equal concentrations of sample were added to pre-washed Dynabeads (MyOne Streptavidin C1) and incubated with nutation overnight at 4°C. After removing the unbound sample, beads were washed twice with buffer 1 (2% SDS) for 10 min, once with buffer 2 [0.1% deoxycholate, 1% Triton X-100, 500 mM NaCl, 1 mM EDTA and 50 mM Hepes (pH 8)] for 10 min, once with buffer 3 [0.5% deoxycholate, 0.5% NP-40, 250 mM NaCl, 1 mM EDTA and 10 mM Tris (pH 8)] for 10 min and twice with buffer 4 [50 mM NaCl and 50 mM Tris (pH 7.4)] for 10 min. To elute bound proteins, 40 µl of biotin-saturated 2× SDS sample buffer (50 µm biotin) was added to the beads and boiled for 10 min. For SDS-PAGE electrophoresis, 15 µl of eluate was run out on an 8% acrylamide gel, transferred overnight to a nitrocellulose membrane and probed with the following antibodies: anti-Scribble (MAB1820), anti-Afadin (A0224) and anti-Myc (SC-40).

We thank David Bilder and Mark Khoury for the Scribble antibody, scribble mutant stocks, advice in generating maternal-zygotic mutants and helpful discussions; Jocelyn McDonald, Jeffrey Thomas, the Bloomington Drosophila Stock Center, the Transgenic RNAi Project (TRiP) and the Developmental Studies Hybridoma Bank for reagents; Tony Perdue of the Biology Imaging Center; Clara Williams, Jonathan Deliberty and Ian Windham for technical assistance; John Poulton for reagents and thoughtful discussion; Scott Williams, Peifer lab members and the three reviewers for helpful advice and comments; and Bing He for helpful discussions.