Electrostatic plasma membrane targeting contributes to Dlg function in cell polarity and tumorigenesis

Polarity proteins play conserved and essential roles in regulating the apical-basal polarity in epithelial cells of both invertebrates and vertebrates. Among them, Discs large (Dlg1, referred to here as Dlg), Scrib and Lgl [L(2)gl] also act as tumor suppressors and share the same basolateral subcellular localization in epithelial cells (Bilder et al., 2003; Tanentzapf and Tepass, 2003). Like many polarity proteins, plasma membrane (PM)/cortical localization is essential for Dlg, Lgl and Scrib for regulating apical-basal polarity and tumorigenesis (Hough et al., 1997; Ventura et al., 2020). Recent studies have demonstrated that multiple polarity proteins contain so-called polybasic (PB) domains that are typically of 20-40 aa length and are highly positively charged due to enriched Arg and Lys residues (Hammond and Hong, 2018). Polybasic (domain) polarity proteins, such as Lgl and aPKC, can specifically target to the PM by electrostatically binding negatively charged phospholipids, in particular PI4P and PI(4,5)P₂ (PIP2), enrichment of which in the PM makes the inner surface of PM the most negatively charged membrane surface the cell (Hammond et al., 2012). Although the electrostatic PM targeting of Lgl is now well established (Bailey and Prehoda, 2015; Dong et al., 2015), the exact mechanisms targeting Dlg to the PM remain to be fully elucidated. Here, we report that Dlg, like recently characterized polybasic polarity proteins Lgl and aPKC (Dong et al., 2020), contains a polybasic domain that electrostatically targets Dlg to the PM and is necessary for Dlg to regulate polarity and tumorigenesis. Our results also suggest that Scrib specifically enhances the electrostatic PM targeting of Dlg by regulating a potential phosphorylation-dependent allosteric control mechanism.

Dlg belongs to the MAGUK protein family and contains PDZ, SH3, HOOK and guanylate kinase (GUK) domains (Fig. 1A). The GUK domain is considered ‘kinase dead’ and instead acts as a protein-interacting domain that can bind the SH3 domain either intramolecularly or intermolecularly (McGee et al., 2001) (see below). By sequence analysis, we identified a well-conserved candidate polybasic domain that spans the C-terminal half of the SH3 domain and the N-terminal half of the HOOK domain (Fig. 1A), with a basic-hydrophobic index (Brzeska et al., 2010) of 0.90, comparable to polybasic domains in Lgl (1.01), aPKC (0.96) and Numb (1.07). In liposome pull-down assays, GST fusion of Dlg polybasic domain (GST-PB) bound both PI4P- and PIP2-containing liposomes but not liposomes containing only phosphatidylserine (PS) (Fig. 1B). GST-PB-KR6Q or GST-PB-KR15Q, in which positive charges were either partially or completely eliminated by K/R->Q mutations, did not bind either liposome (Fig. 1B), supporting an electrostatic interaction between the positively charged polybasic domain and the negatively charged PI4P- or PIP2-containing membrane.

To investigate whether the polybasic domain is required for targeting Dlg to PM in vivo, we generated transgenic flies expressing Dlg::GFP, Dlg^KR6Q::GFP and Dlg^KR15Q::GFP under the ubiquitin promoter (Dong et al., 2020). Although the dlg locus apparently expresses numerous isoforms, our ubi-dlg::GFP (dlg::GFP) based on isoform-G fully rescued the null mutants of dlg^[A] and was expressed at levels similar to an endogenous GFP knock-in allele (dlg::GFP^KI) (Fig. S1A). Dlg::GFP showed typical basolateral PM/cortical localization in follicular and embryonic epithelial cells (Fig. 1C, Fig. S2A), whereas both Dlg^KR6Q and Dlg^KR15Q showed a strong reduction of PM localization (Fig. 1C, Fig. S2C) and failed to rescue dlg^[A] (Table 1). Dlg interacts with multiple proteins on the PM through its three PDZ domains (Hough et al., 1997; Qian and Prehoda, 2006), and Dlg^ΔPDZ::GFP also showed reduced PM localization in follicular cells, albeit less severe than Dlg^KR6Q and Dlg^KR15Q (Fig. 1C). Consistent with both PDZ and polybasic domains contributing to PM targeting in follicular cells, Dlg^ΔPDZ-KR6Q::GFP and Dlg^ΔPDZ-KR15Q::GFP became virtually lost from the PM (Fig. 1C).

To investigate further the electrostatic PM targeting of Dlg in vivo, we used a previously established assay that applies controlled hypoxia to acutely and reversibly deplete PIP2 and PI4P in the PM (Dong et al., 2015) (data not shown). Similar to observations for the polybasic polarity proteins Lgl (Dong et al., 2015) and aPKC (Dong et al., 2020), hypoxia also induced acute and reversible loss of Dlg::GFP^KI, as well as ubi-Dlg::GFP, from the PM in follicular cells (Fig. 1D,D′, Movies 1 and 2). However, unlike Lgl, which became completely cytosolic within 30-60 min of hypoxia, residual Dlg::GFP^KI persisted on the PM over 100 min since the start of hypoxia (Fig. 1D,D′, Fig. S1B). In contrast, hypoxia induced complete PM loss of Dlg^ΔPDZ::GFP that was identical to that observed for Lgl (Fig. 1E,E′, Fig. S1B, Movie 3). Such data suggest that in follicular cells a significant portion of Dlg electrostatically binds to the PM, whereas the rest were retained at the PM by PDZ domain-dependent interactions.

Dlg is also a component of the septate junction (SJ), which is composed of over 20 different proteins (Izumi and Furuse, 2014), making the SJ the primary target of PDZ-dependent PM localization. In early embryonic epithelia cells with undeveloped SJ, Dlg::GFP was readily lost from the PM under hypoxia, suggesting that its PM targeting is mostly electrostatic (Fig. S2A). In stage 14 or later embryos with mature SJs, PM localization of Dlg::GFP^KI was significantly enhanced (Fig. S2C,D) and became completely resistant to hypoxia (Fig. S2A). In contrast, in stage 14 embryos mutant for core SJ proteins such as Coracle (cora) (Lamb et al., 1998) or Na⁺,K⁺-ATPase α-subunit (Atpα) (Paul et al., 2003), Dlg::GFP^KI PM localization became sensitive to hypoxia (Fig. S2B). In addition, enhanced PM localization in late embryos was seen in electrostatic Dlg mutants, but not in mutants lacking PDZ domains (Fig. S2C,D), consistent with the suggestion that PDZ domains are required for Dlg localization to the SJ (Hough et al., 1997). Indeed, PM localization of Dlg^ΔPDZ::GFP in late embryos is electrostatic, based on its sensitivity to hypoxia (Fig. S2B).

In summary, with the help of hypoxia-based imaging assays we were able to specifically probe the electrostatic PM targeting of Dlg in vivo. Our data revealed that in follicular and early embryonic epithelial cells full PM localization of Dlg requires both electrostatic PM targeting by the polybasic domain and protein-interactions mediated by PDZ domains, whereas in epithelial cells with mature SJs PDZ domains alone can act redundantly to retain Dlg on the PM.

PI4P and PIP2 in the PM are the major components responsible for the electrostatic binding of polybasic proteins (Hammond, 2012). To investigate the role of PIP2 in targeting Dlg in vivo, we depleted PIP2 acutely in follicular cells, using an inducible system (Dong et al., 2020, 2015; Reversi et al., 2014) in which rapamycin addition induces dimerization between an FRB-tagged PM anchor protein (i.e. Lck-FRB-CFP) and an FKBP-tagged phosphatase (i.e. RFP-FKBP-INPP5E), resulting in acute PM recruitment of INPP5E, which rapidly depletes PM PIP2. For reasons unknown, PM levels of Dlg::GFP^KI or Dlg^ΔPDZ::GFP increased in cells expressing Lck-FRB and FKBP-INPP5E, regardless of DMSO or rapamycin treatment (Fig. 2A,A′). Nonetheless, under rapamycin but not DMSO treatment, PM localization of both Dlg::GFP^KI and Dlg^ΔPDZ::GFP showed strong reduction in cells expressing Lck-FRB and FKBP-INPP5E (Fig. 2A,A′), supporting the idea that Dlg PM targeting is at least partially dependent on PM PIP2.

At present, similar tools for acutely depleting PI4P or both PI4P and PIP2 are not available in Drosophila. We thus manipulated PI4P levels genetically by RNAi knock down of PI4KIIIα, the PtdIns-4-kinase specifically responsible for maintaining the PM PI4P levels (Bojjireddy et al., 2014; Tan et al., 2014). Wild-type cells and PI4KIIIα-RNAi cells showed similar PM levels of Dlg::GFP^KI and Dlg^ΔPDZ::GFP and rates of PM delocalization under hypoxia (Fig. 2B,B′, Movies 4 and 5). However, consistent with the suggestion that PI4KIIIα is likely necessary for replenishing PM PI4P and PIP2 (Bojjireddy et al., 2014), PM recoveries of Dlg::GFP^KI and Dlg^ΔPDZ::GFP were significantly delayed in PI4KIIIα-RNAi cells (Fig. 2B,B′). Overall, our data suggest that PI4P and PIP2 act redundantly to bind Dlg electrostatically to the PM.

Polybasic domains are often regulated by direct phosphorylations, which inhibit PM binding by neutralizing the positive charges (Hong, 2018), such as in the cases of Lgl, Numb and Miranda (Bailey and Prehoda, 2015; Dong et al., 2015). To investigate whether the Dlg polybasic domain is also regulated by phosphorylation events, we mutated four well-conserved S/T sites within its polybasic domain (Fig. 1A), of which the first two are the previously characterized phosphorylation sites of aPKC and PKCα (Golub et al., 2017; O'Neill et al., 2011). Both phosphomimetic Dlg^S4D::GFP and non-phosphorylatable Dlg^S4A::GFP were significantly reduced in the PM (Fig. 3A), suggesting that these potential phosphorylation events are unlikely to regulate the polybasic domain through charge neutralization. Interestingly, PM localization of Dlg^S4A::GFP, albeit reduced, became resistant to hypoxia (Fig. S3A, Movie 6), whereas removing PDZ domains in Dlg^S4A::GFP nearly abolished the PM localization of Dlg^ΔPDZ-S4A::GFP (Fig. 3A), suggesting that the residual PM localization of Dlg^S4A::GFP is non-electrostatic but PDZ dependent.

Our recent studies also showed that the polybasic domain in aPKC, which is a pseudosubstrate region (PSr) with no characterized phosphorylation sites, is allosterically regulated (Dong et al., 2020). The PSr in aPKC binds and autoinhibits the kinase domain, which in turn prevents PSr from electrostatically binding to PM, while binding of Par-6 to aPKC induces conformation changes that expose the PSr to allow PM binding. Given that the SH3 and GUK domains in Dlg could also bind each other, we postulate that the loss of potential phosphorylation events in Dlg^S4A::GFP keeps the SH3-GUK in a closed conformation that prevents the polybasic domain from binding to the PM. Removing the GUK domain, however, resulted in Dlg^ΔGUK::GFP localizing both on the PM and in nuclei (Fig. 3A), likely due to the fact that the Arg/Lyn-rich feature of the polybasic domain is also similar to the nuclear localization signal (NLS) (Dong et al., 2015). Consistent with the observation that removing the GUK domain exposes the polybasic domain, Dlg^ΔGUK::GFP PM localization is partially electrostatic just like Dlg::GFP (Fig. S2B, Movie 7). In addition, the moderate nuclear localization of Dlg^ΔGUK::GFP can be overcome by its interaction with the SJ, as Dlg^ΔGUK::GFP was localized at the SJ in larval wing disc epithelia, which develop strong SJs (Fig. S2E) (Hough et al., 1997). In contrast, Dlg^S4A-ΔGUK::GFP showed overwhelmingly high nuclear localization both in follicular cells (Fig. 3A) and in wing imaginal epithelial cells (Fig. S2E), suggesting that removing the GUK domain likely also exposes the polybasic domain in Dlg^S4A, but the lack of potential phosphorylation events strongly enhanced the nuclear localization property of the polybasic domain.

Although additional studies are needed to confirm the phosphorylation events of Dlg in vivo, our results are consistent with a model in which the polybasic domain in Dlg is regulated by phosphorylation events that allosterically relieve its blockage by the GUK domain and promote its targeting to the PM over nuclear localization (Fig. 4D).

Scrib and Lgl colocalize with Dlg at the basolateral PM and play similar functions in cell polarity and tumorigenesis, but it is unknown how Scrib and Lgl specifically regulate the electrostatic PM targeting of Dlg. Consistent with recent studies demonstrating that Scrib, but not Lgl, appears to interact physically with Dlg (Khoury and Bilder, 2020; Ventura et al., 2020), knocking down Scrib by RNAi induced a partial loss of Dlg from the PM (Fig. 3B). More importantly, Scrib appears to specifically enhance the electrostatic PM targeting of Dlg, as Dlg^ΔPDZ, which can only bind the PM electrostatically, was dramatically reduced from the PM in scrib-RNAi cells (Fig. 3B). In contrast, the residual PM localization of the electrostatic mutant Dlg^KR6Q was not affected by scrib-RNAi (Fig. 3B). Scrib interaction depends on the SH3 domain in Dlg and a single point mutation in the SH3 domain in the dlg^m30 mutant (Fig. 1A) abolishes such interaction (Khoury and Bilder, 2020). Indeed, similar to electrostatic mutants such as Dlg^KR6Q, Dlg^S4A and Dlg^ΔGUK, PM localization of Dlg^m30::GFP was also reduced and became resistant to scrib-RNAi (Fig. 3B), supporting the hypothesis that interactions between Scrib and Dlg are necessary for Scrib to enhance Dlg electrostatic PM targeting. More importantly, PM localization of Dlg^S4A or Dlg^ΔGUK, two mutants defective in the potential phosphorylation-dependent allosteric control of the polybasic domain, was also resistant to scrib-RNAi (Fig. 3B), suggesting that Scrib may act by facilitating such a phosphorylation-allosteric mechanism (Fig. 4D).

In agreement with the recent findings that Lgl does not appear to associate physically with Dlg and Scrib (Khoury and Bilder, 2020; Ventura et al., 2020), lgl-RNAi did not affect the PM localization of Dlg, Dlg^KR6Q, Dlg^S4A, Dlg^m30 and Dlg^ΔGUK (Fig. 3C, Fig. S4C″″). However, PM localization of Dlg^ΔPDZ was mildly reduced in lgl-RNAi cells (Fig. 3C). Thus, Lgl could also play a moderate role in enhancing Dlg electrostatic PM targeting, through mechanisms that are presently unclear. Finally, knocking down GukH, which interacts with the GUK domain and is required for aPKC-regulation of Dlg in neuroblasts (Golub et al., 2017), did not affect the electrostatic PM targeting of Dlg^ΔPDZ (Fig. S4C′), consistent with the suggestion that GukH appears to be dispensable for epithelial polarity (Caria et al., 2018).

We also investigated how electrostatic PM targeting may specifically contribute to Dlg function in polarity and tumorigenesis. Follicular cells of dlg^−/− mutants showed dramatic disruption of cell polarity evidenced by the mislocalization of apical polarity proteins such as aPKC (Fig. 4A). This phenotype was fully rescued by ectopic expression of Dlg and Dlg^ΔPDZ, but not electrostatic mutant Dlg^KR6Q or Dlg^ΔPDZ-KR6Q (Fig. 4A), suggesting that electrostatic PM targeting is essential for Dlg to regulate apical-basal polarity in follicular cells. Surprisingly, dlg^[A]/Y; dlg^S4A::GFP and dlg^[A]/Y; dlg^ΔGUK::GFP males are viable, although sterile (Table 1); thus, the potential mechanisms regulating the polybasic domain may be largely dispensable for Dlg function in normal development.

We then used a well-established tumor model (Brumby et al., 2011) to investigate how electrostatic PM targeting is required for the tumor suppressor function of Dlg. Overexpression of constitutively active Ras^V12 in larval eye disc produced a moderate rough eye phenotype (Fig. 4B). Although dlg-RNAi alone in the eye did not produce obvious phenotypes (Caria et al., 2018), combining Ras^V12 overexpression and dlg-RNAi (Ras^V12/dlg-RNAi) produced massive eye tumors, which resulted in strong larval/pupal lethality, with very few adult escapers with reduced eyes size and dark tumor tissues (Fig. 4B). We made transgenic stocks that express wild-type or mutant Dlg proteins from cDNA sequences modified to be resistant to dlg-RNAi (Fig. S4C, Table S1). Each of the ΔPDZ mutants is also resistant to dlg-RNAi as the RNAi targets the sequence near the N-terminal coding region. As expected, expression of RNAi-resistant dlg::GFP (Dlg^R::GFP) fully rescued Ras^V12/dlg-RNAi lethality and eye morphology (Fig. 4B,C). Both Dlg^ΔPDZ and the polybasic mutant Dlg^KR6Q-R rescued the lethality of Ras^V12/dlg-RNAi, but eyes in dlg^KR6Q-R flies still showed strong tissue overproliferation (Fig. 4B). In contrast, Dlg^ΔPDZ-KR6Q failed to rescue either lethality or eye morphology and tumorigenesis (Fig. 4B). Thus, both electrostatic and PDZ-dependent PM targeting are required for the tumor suppressor function of Dlg and they act redundantly in rescuing the Ras^V12/dlg-RNAi tumor lethality, although electrostatic PM targeting appears to be more specifically required for inhibiting the overproliferation of Ras^V12/dlg-RNAi tumor cells.

Notably, Dlg^m30-ΔGUK, but not Dlg^m30 or Dlg^m30-ΔPDZ, partially rescued the polarity defects in dlg^−/− cells (Fig. 4A). RNAi-resistant Dlg^m30-R only moderately rescued the lethality caused by Ras^V12/dlg-RNAi, and survivors still showed strong overproliferation in eyes (Fig. 4B). Consistent with the polarity rescue results in Fig. 4A, removing GUK but not PDZ domains in Dlg^m30 significantly enhanced the rescue of the overproliferation phenotype in Ras^V12/dlg-RNAi eyes (Fig. 4B). Such data are consistent with the observation that eliminating the potential allosteric inhibition of the polybasic domain in Dlg partially compensated for the loss of Scrib-dependent enhancement electrostatic PM targeting. Interestingly, although both Dlg^S4A and Dlg^m30 were defective in electrostatic PM targeting, Dlg^S4A showed strong rescue activity, whereas Dlg^m30 did not (Table 1), suggesting that Scrib/Dlg interaction is crucial for Dlg function besides regulating the electrostatic PM targeting of Dlg. In addition, unlike Dlg^S4A or Dlg^ΔGUK, Dlg^S4AΔGUK showed no rescue activity (Table 1), probably because its strong nuclear concentration made protein levels at the PM too low to rescue (Table 1).

The electrostatic nature of Dlg PM targeting makes Dlg a new member of the polybasic polarity protein family that includes at least Lgl, aPKC, Numb and Miranda. Our results also make it increasingly clear that electrostatic PM targeting is a key molecular mechanism widely used by polarity proteins for achieving controlled subcellular localization and for regulating cell polarity. It will be of great interest for future studies to integrate the electrostatic PM targeting mechanism into the regulatory network of polarity protein and their interacting partners, by uncovering the essential molecular mechanisms regulating this simple but elegant physical interaction between polarity proteins and the PM.

Flies carrying transgenic ubi-dlg::GFP or ubi-dlg::GFP mutant (ubi-dlg**::GFP) alleles were generated by a phiC31-mediated integration protocol (Huang et al., 2009). attP^VK00022 (Bloomington Drosophila Stock Center, BL#24868) stock was used to integrate ubi-dlg:GFP and ubi-dlg**::GFP constructs to the 2nd chromosome. Transgenic alleles of ubi-dlg::GFP and ubi-dlg**::GFP were further recombined with y w dlg^[A] FRT19A/FM7c (Bloomington Drosophila Stock Center, BL#57086). A summary of the ubi-dlg**::GFP alleles is given in Table S1.

cora⁵/CyO dfdGMR-YFP and Atpα^DTS2A3/TM6 dfdGMR-YFP were gifts from Dr Greg Beitel, (Northwestern University, USA). w UASp>mRFP::FKBP-5Ptase (“FKBP-INPP5E”) and w; ; UASp>Lck-FRB::CFP were gifts from Dr Stefano De Renzis (EMBL Heidelberg, Germany) (Reversi et al., 2014). ey-Gal4, UAS-Ras^V12 stocks were a gift from Dr Helena Richardson (University of Melbourne, Australia). w P[w[+mC]=PTT-GC]dlg1^[YC0005] (dlg::GFP^KI, BL#50859), UAS-PI4KIIIα-RNAi (BL#35256), UAS-scrib-RNAi (BL#29552), UAS-lgl-RNAi (BL#38989), UAS-dlg-RNAi (II) (BL#39035), UAS-dlg-RNAi (III) (BL#34854) and gukh-RNAi (BL#42486) (Caria et al., 2018) were from the Bloomington Drosophila Stock Center. w; lgl::mCherry knock-in and w; lgl::GFP knock-in stocks were previously published (Dong et al., 2015).

Drosophila cultures and genetic crosses were carried out at 25°C.

The genotypes of Drosophila samples shown in figures are given in supplementary Materials and Methods.

To make ubi-dlg::GFP, the ubiquitin promoter (1872 bp) was PCR amplified from plasmid pWUM6 (a gift from Dr Jeff Sekelsky, University of North Carolina at Chapel Hill, NC, USA) using primers 5′-AGTGTCGAATTCCGCGCAGATCGCCGATGGGC and 5′-CTGGACGCGGCCGCGGTGGATTATTCTGCGGG and inserted into the pGE-attB vector (Huang et al., 2009) to generate the vector pGU. DNA fragments encoding Dlg::GFP were then inserted into the pGU vector. More details about DNA constructs used in this report are given in Table S1. The sequence of the Dlg isoform used in this study can be found in the NCBI RefSeq database (ID NP_996405.1).

Liposomal binding assays were carried out as described (Kim et al., 2008). A lipid mixture of 37.5% PC (840051C), 10% PS (840032C), 37.5% PE (840021C), 10% cholesterol (700000P) and 5% PI(4,5)P₂ (840046X) or PI4P (840045X; all lipids were purchased from Avanti Polar Lipids) was dried and resuspended to a final concentration of 1 mg/ml of total phospholipids in HEPES buffer and subjected to 30 min sonication. Formed liposomes were harvested at 16,000 g for 10 min and resuspended in binding buffer (20 mM HEPES, pH 7.4, 120 mM KCl, 20 mM NaCl, 1 mM EGTA, 1 mM MgCl₂, 1 mg/ml bovine serum albumin). Approximately 0.1 μg of purified protein or protein complex was mixed with 50 μl of liposome suspension in each liposome-binding assay. Liposomes were pelleted at 16,000 g for 10 min after 15 min incubation at room temperature, and were analyzed by western blot to detect co-sediment of target protein(s).

Mutant follicular cell clones of dlg^[A] were generated by standard FLP/FRT techniques (St Johnston, 2012). Young females were heat-shocked at 37°C for 1 h and their ovaries were dissected 3 days later.

Embryos and dissected ovaries were imaged according to previously published protocols (Dong et al., 2015; Huang et al., 2011). The embryos were staged by timing and kept in 25°C for 2 h before imaging. Ovaries from adult females of 2 days old were dissected in halocarbon oil (#95). Follicular cells containing overexpressing or RNAi clones were generated by heat-shocking the young females of the correct genotype at 37°C for 15-30 min and ovaries were dissected 3 days later. To ensure sufficient air exchange to samples during the imaging session, dechorionated embryos or dissected ovaries were mounted in halocarbon oil on an air-permeable membrane (YSI Membrane Model #5793, YSI) sealed by vacuum grease on a custom-made plastic slide over a 10×10 mm² cut-through window. The slide was then mounted in a custom-made air-tight micro chamber for live imaging under confocal microscope. Oxygen levels inside the chamber were controlled by flow of either air or custom O₂/N₂ gas mixture at a rate of approximately 1-5 cc/s. Images were captured at room temperature (25°C) on an Olympus FV1000 confocal microscope (60× Uplan FL N oil objective, NA=1.3) with Olympus FV10-ASW software, or on a Nikon A1 confocal microscope (Plan Fluo 60× oil objective, NA=1.3) with NIS-Elements AR software. Images were further processed in ImageJ and Adobe Photoshop.

Young females of w UASp>mRFP::FKBP-5Ptase/dlg::GFP;;hs-FLP Act5C(FRT.CD2)-Gal4 UAS-RFP/UASp>Lck-FRB::CFP or w UASp>mRFP::FKBP-5Ptase/+; ubi-dlg^ΔPDZ::GFP/+; hs-FLP Act5C(FRT.CD2)-Gal4 UAS-RFP/UASp>Lck-FRB::CFP were heat-shocked at 37°C for 15 min. Ovaries were dissected 3 days later in Schneider's medium, mounted in a drop of 20 μl Schneider's medium containing 10 μM rapamycin or DMSO on a gas-permeable slide, and imaged live as previously described (Dong et al., 2015; Huang et al., 2011).

Immunostaining of follicular cells and embryos was carried out as previously described (Huang et al., 2009). Primary antibodies were: chicken anti-GFP (Aves Lab, GFP-1010; 1:5000); mouse anti-Dlg (Developmental Studies Hybridoma Bank, 4F3; 1:50); rabbit anti-aPKC (Santa Cruz Biotechnology, Sc-216; 1:1000. Secondary antibodies were: Cy2-, Cy3- or Cy5-conjugated goat anti-rabbit IgG, anti-mouse IgG and anti-chicken IgG (Jackson ImmunoResearch, 111-225-003, 115-165-003 and 106-175-003; all at 1:400). Images were collected on Olympus FV1000 confocal microscopes (Center for Biologic Imaging, University of Pittsburgh Medical School) and processed in Adobe Photoshop for compositions.

Time-lapse movies were first stabilized using the HyperStackReg plug-in in ImageJ. Images or movies containing excessive noisy channels were denoised using the PureDenoise plugin in ImageJ prior to quantification. PM localization of GFP or RFP in images or movies were measured in ImageJ by custom macro scripts. In each image or the first frame of the movie, regions of interest (ROIs) approximately 20-40 µm² were drawn across selected cell junctions. In most cases, custom macros was used to automatically generate PM masks by threshold-segmentation that was based on the mean pixel value of the ROI. Custom macros were then used to automatically measure PM and cytosolic intensities of each fluorescent protein in ROIs in an image or throughout all the frames of a movie. Backgrounds were manually measured based on the minimal pixel value of the whole image or the first frame of the movie. The PM localization index for each fluorescent protein was auto-calculated by the macro as the ratio of (PM – background)/(cytosol – background). In live-imaging experiments, the ‘Normalized PM Index’ was calculated by normalizing (PM Index −1) over the period of recording against the (PM Index −1) at 0 min. Data were further processed in Excel, visualized and analyzed in Graphpad Prism.

RNAi-resistance wild-type or mutant ubi-dlg::GFP (dlg^R**::GFP) transgenic alleles were crossed with ey-Gal4 UAS-Ras^V12/CyO-Gal80; UAS-dlg-RNAi/TM6B and crosses were transferred to a new vial every 3-4 days. F1 progenies from each cross were scored into two groups based on their genotypes: dlg^R**::GFP /ey-Gal UAS-Ras^V12; UAS-dlg-RNAi/+ (group#1) or dlg^R**::GFP /ey-Gal UAS-Ras^V12; TM6B/+ (group#2). The ratio between groups #1 and #2 was calculated as the ‘Rescue’ index for each dlg^R**::GFP allele. Representative eyes of flies from group#1 were imaged.

In all quantifications, a two-tailed Student's t-test was carried out to calculate the P value. Differences with P>0.05 are considered to be statistically insignificant.

We are grateful to Drs David Bilder, Helena Richardson, Greg Beitel, Richard Fehon, Ulli Tepass and Stefano De Renzis for reagents and fly stocks; Dr Simon Watkins and University of Pittsburgh Medical School Center for Biologic Imaging for generous imaging and microscopy support; Bloomington Stock Center for fly stocks; and Developmental Studies Hybridoma Bank (DSHB) for antibodies. The University of Pittsburgh Medical School Center for Biologic Imaging is supported by grant 1S10OD019973-01 from the National Institutes of Health.

The peer review history is available online at https://journals.biologists.com/dev/article-lookup/148/7/dev196956/