ABSTRACT
Apical-basal polarity underpins the formation of epithelial barriers that are crucial for metazoan physiology. Although apical-basal polarity is long known to require the basolateral determinants Lethal Giant Larvae (Lgl), Discs Large (Dlg) and Scribble (Scrib), mechanistic understanding of their function is limited. Lgl plays a role as an aPKC inhibitor, but it remains unclear whether Lgl also forms complexes with Dlg or Scrib. Using fluorescence recovery after photobleaching, we show that Lgl does not form immobile complexes at the lateral domain of Drosophila follicle cells. Optogenetic depletion of plasma membrane PIP2 or dlg mutants accelerate Lgl cortical dynamics. However, Dlg and Scrib are required only for Lgl localization and dynamic behavior in the presence of aPKC function. Furthermore, light-induced oligomerization of basolateral proteins indicates that Lgl is not part of the Scrib-Dlg complex in the follicular epithelium. Thus, Scrib and Dlg are necessary to repress aPKC activity in the lateral domain but do not provide cortical binding sites for Lgl. Our work therefore highlights that Lgl does not act in a complex but in parallel with Scrib-Dlg to antagonize apical determinants.
INTRODUCTION
Cell polarity is a fundamental property that enables epithelia to act as selective barriers between different environments. Epithelial architecture relies on the asymmetric segregation of evolutionary conserved polarity proteins that define the apical and lateral domains, and correctly position intercellular junctions (Flores-Benitez and Knust, 2016). Lgl is one such regulator of apical-basal polarization (Bilder et al., 2000) that is also necessary for polarization in a range of different contexts, including stem cell division (Ohshiro et al., 2000; Peng et al., 2000), front-rear polarization (Dahan et al., 2014), and anterior-posterior polarization of Drosophila oocytes (Fichelson et al., 2010; Tian and Deng, 2008) and C. elegans embryos (Beatty et al., 2010; Hoege et al., 2010).
Lgl localizes below the adherens junction to promote basolateral identity together with Scrib and Dlg. These genes were originally described in Drosophila, where homozygous mutants display epithelial disorganization and neoplastic growth (Bilder et al., 2000; Gateff, 1978; Woods and Bryant, 1989). Further analysis placed these genes in a basolateral polarity module due to their mutually interdependent localization and antagonistic genetic interactions with the apical Crumbs and aPKC complexes (Bilder et al., 2003; Tanentzapf and Tepass, 2003). Although Scrib, Lgl and Dlg accumulate exclusively at the occluding septate junction in the adult Drosophila midgut, where they are dispensable for apical-basal polarity (Chen et al., 2018), their cortical localization plays evolutionarily conserved roles in polarity and adhesion (Chalmers et al., 2005; Dow et al., 2003; Grifoni et al., 2007; Legouis et al., 2003; McMahon et al., 2001; Müsch et al., 2002; Raman et al., 2016; Russ et al., 2012; Sripathy et al., 2011; Yamanaka et al., 2006). Loss-of-function phenotypes are, however, masked in mammalian epithelia by the presence of multiple paralogues, or, in the case of Scrib, by the compensatory function of proteins with a similar leucine-rich repeat and PDZ protein (LAP) domain (Choi et al., 2019).
The molecular basis for the function of basolateral proteins is still unclear. The predominant hypothesis is that Scrib and Dlg ensure Lgl localization at the lateral cortex (Bilder et al., 2000; Kallay et al., 2006; Zeitler et al., 2004). In turn, Lgl prevents the extension of apical determinants (Hutterer et al., 2004), likely by inhibiting the aPKC-Par-6 complex (Atwood and Prehoda, 2009; Betschinger et al., 2003; Yamanaka et al., 2003, 2006), and promoting Crumbs recycling from the lateral cortex (Fletcher et al., 2012). X-ray crystallography revealed an interaction between mammalian Lgl2 and the guanylate kinase domain of Dlg4 (Zhu et al., 2014). Moreover, Lgl may bind to the LAP unique region of Scrib (Kallay et al., 2006), in particular to a small region termed LAPSDa (Choi et al., 2019). Nevertheless, it remains unknown whether Scrib and Dlg bind Lgl in vivo to control its function and localization in the lateral domain. Dlg and Scrib are also likely to play roles independently of Lgl. Recent studies uncovered specific phenotypes for dlg and scrib mutants on adherens junction and tricellular junction formation in Drosophila embryonic and wing disk epithelia, respectively (Bonello et al., 2019; Sharifkhodaei et al., 2019). Furthermore, whereas Scrib and Dlg cooperate to orient the mitotic spindle (Bergstralh et al., 2013; Nakajima et al., 2013, 2019), Lgl release from the cortex promotes planar spindle orientation (Bell et al., 2015; Carvalho et al., 2015).
Lgl cortical localization is inhibited by phosphorylation, which is controlled by aPKC, Aurora A and Protein Phosphatase 1 (Bell et al., 2015; Betschinger et al., 2003; Carvalho et al., 2015; Moreira et al., 2019). Early studies highlighted the association of Lgl with the actomyosin cytoskeleton (Betschinger et al., 2005; Strand et al., 1994), whereas recent work uncovered that Lgl cortical localization is mediated by interactions between its basic and hydrophobic domain and membrane phosphoinositides, primarily phosphatidylinositol-4,5-biphosphate (PIP2) (Bailey and Prehoda, 2015; Dong et al., 2015; Visco et al., 2016). These binding partners could aid Scrib and Dlg in forming multivalent interactions that regulate Lgl binding at the lateral cortex, but the contribution of each putative interactor has not been defined. Here, we have characterized the mechanisms that regulate Lgl cortical dynamics in the Drosophila follicular epithelium. Lgl dynamics are dependent on both PIP2 and Dlg. However, whereas PIP2 restricts Lgl cortical mobility, Dlg and Scrib regulate Lgl by repressing aPKC ectopic localization. Moreover, we detect the presence of Dlg-Scrib complexes that do not contain Lgl, reinforcing that Lgl antagonizes apical proteins independently of Scrib and Dlg.
RESULTS AND DISCUSSION
Lgl displays membrane-diffusion behavior and faster dynamics than Dlg and Scrib
To characterize Lgl dynamics in the lateral domain of epithelial cells, we performed fluorescence recovery after photobleaching (FRAP) in post-mitotic stages 7 to 9 of the Drosophila follicular epithelium (Fig. S1). We first compared the cortical dynamics of Lgl with Scrib and Dlg using GFP-tagged proteins expressed at endogenous levels. Lgl is significantly more dynamic than Dlg and Scrib (Lglt1/2 ∼10 s; Dlgt1/2 ∼40 s; Scribt1/2 ∼60 s; Fig. 1A,B, Movie 1), likely reflecting their distinct binding domains and individual functions (Bonello and Peifer, 2019; Stephens et al., 2018). Moreover, rebleaching a previously bleached region leads to an identical plateau of Lgl fluorescence recovery (Fig. 1C). Thus, Lgl does not form a significant subpopulation of immobile complexes at the cortex and plasma membrane.
Lgl cortical dynamics are faster than Scrib and Dlg, and are coupled to membrane-diffusion. (A) Fluorescence recovery (±95% CI) of endogenously tagged Lgl-GFP (n=43), Dlg-GFP (n=18) and Scrib-GFP (n=22). (B) Pseudo-colored images of FRAP in the lateral domain of the follicular epithelium (Movie 1). Kymographs depict fluorescent recovery in the photobleached region. Scale bar: 5 μm. (C) Recovery plots of consecutive FRAP experiments in the same ROI (n=11). (D) Lgl-GFP mean intensity (±95% CI, n=23) in the interface adjacent (red) to the bleached region normalized to the mean intensity of distant interfaces (blue). (E) Half-times (t1/2; ±s.d.) for the fast and slow components of Lgl-GFP dynamics were obtained with a bi-exponential fit of recovery curves for varying photobleaching size [1.5 µm (n=9), 3 µm (n=11)]. Data are mean±s.d. P-values calculated with a Mann–Whitney U-test.
Lgl cortical dynamics are faster than Scrib and Dlg, and are coupled to membrane-diffusion. (A) Fluorescence recovery (±95% CI) of endogenously tagged Lgl-GFP (n=43), Dlg-GFP (n=18) and Scrib-GFP (n=22). (B) Pseudo-colored images of FRAP in the lateral domain of the follicular epithelium (Movie 1). Kymographs depict fluorescent recovery in the photobleached region. Scale bar: 5 μm. (C) Recovery plots of consecutive FRAP experiments in the same ROI (n=11). (D) Lgl-GFP mean intensity (±95% CI, n=23) in the interface adjacent (red) to the bleached region normalized to the mean intensity of distant interfaces (blue). (E) Half-times (t1/2; ±s.d.) for the fast and slow components of Lgl-GFP dynamics were obtained with a bi-exponential fit of recovery curves for varying photobleaching size [1.5 µm (n=9), 3 µm (n=11)]. Data are mean±s.d. P-values calculated with a Mann–Whitney U-test.
FRAP kymographs show that fluorescence recovery initiates at the edges of the photobleached region (Fig. 1B, arrow), suggesting that Lgl diffuses laterally along the membrane. Consistent with this, fluorescence decreases in neighboring membranes immediately after bleaching (Fig. 1D). Similar behavior is observed for the GFP tagged pleckstrin homology (PH) domain of Phospholipase Cδ, which binds PI(4,5)P2 (Fig. S2A,B). These results emphasize the contribution of membrane diffusion to Lgl mobility, but do not exclude the significance of membrane-cytoplasm exchange. In fact, both types of mobility contribute for the dynamic behavior of proteins that bind PIP2 (Goehring et al., 2010; Hammond et al., 2009). Fluorescence recovery via diffusion scales with the size of the photobleached region (Fritzsche and Charras, 2015; Sprague and McNally, 2005). Thus, we reduced the size of the photobleached region to verify whether bi-exponential fitting could separate the contribution of membrane diffusion. However, the reduced distance that unbleached molecules must diffuse decreases both t1/2fast and t1/2slow, indicating that membrane diffusion contributes for both half-times (Fig. 1E). Therefore, we subsequently compared the mobility of Lgl under different experimental conditions through a single half-time of fluorescence recovery (Table S1, see Materials and Methods).
Lgl dynamics in the lateral domain are restrained by PIP2 and independent of aPKC
Interactions between Lgl and putative binding partners should restrict Lgl mobility in the lateral cortex. We used FRAP to determine the relative contributions of known Lgl interactors, PIP2 and the actin cortex. Lgl remains in the cortex after disruption of the actin cytoskeleton with Latrunculin A (LatA) in the embryonic epithelium (Dong et al., 2015). LatA also does not affect the localization of either Lgl, Dlg or Scrib, suggesting that follicle cells remain polarized upon acute actin cytoskeleton disruption (Fig. 2A, Fig. S3). However, LatA induced a minor but significant acceleration of Lgl-GFP fluorescence recovery (Fig. 2B). This does not result from pleotropic effects on membrane dynamics as PH-GFP fluorescence recovery is unchanged by LatA (Fig. S2C). Thus, the actin cytoskeleton restrains Lgl mobility, possibly through Lgl binding to myosin and other actin cortex-associated proteins (Dahan et al., 2014; Strand et al., 1994).
Lgl dynamics rely on PIP2 and are independent of aPKC in the lateral cortex. (A) Egg chambers expressing UAS-driven Lgl-GFP were treated with Latrunculin A (LatA). Phalloidin labels F-actin. (B) Scatter plots show Lgl-GFP recovery half-time in control (n=16) and LatA-treated (n=24) egg chambers (mean± s.d.). (C) PIP2 was depleted by light-dependent recruitment of CRY2 fused with the inositol polyphosphate 5-phosphatase OCRL to CIBN anchored to the plasma membrane. Blue light releases PH-ChFP from the membrane (Movie 2). (D) Half-time distribution (mean±s.d.) of UAS-Lgl-GFP in control (n=29) and PIP2 depletion alone (OCRL, n=21) or together with LatA (n=16). (E) Follicle cells expressing UAS-Lgl-GFP stained for aPKC. Co-expressing UAS-aPKCΔN releases Lgl from the cortex. (F,G) Scatter plots represent half-time of fluorescence recovery for (F) UAS-Lgl-GFP in control (n=29, same dataset as D) and aPKCΔN-overexpressing cells (n=10), and for (G) endogenously tagged Lgl-GFP (n=43, control), aPKC RNAi (n=26) and LglS5A-GFP (homozygous, n=35) (mean±s.d.). (H,I) Midsagittal cross-sections of control and aPKC-depleted egg chambers stained for (H) DAPI and aPKC, or (I) expressing endogenous Lgl-GFP mislocalized to the apical side in aPKC RNAi (arrows). Arrowheads indicate multilayered tissue. Scale bars: 10 µm (E); 25 µm (A,H,I). P-values calculated using the Mann–Whitney U-test.
Lgl dynamics rely on PIP2 and are independent of aPKC in the lateral cortex. (A) Egg chambers expressing UAS-driven Lgl-GFP were treated with Latrunculin A (LatA). Phalloidin labels F-actin. (B) Scatter plots show Lgl-GFP recovery half-time in control (n=16) and LatA-treated (n=24) egg chambers (mean± s.d.). (C) PIP2 was depleted by light-dependent recruitment of CRY2 fused with the inositol polyphosphate 5-phosphatase OCRL to CIBN anchored to the plasma membrane. Blue light releases PH-ChFP from the membrane (Movie 2). (D) Half-time distribution (mean±s.d.) of UAS-Lgl-GFP in control (n=29) and PIP2 depletion alone (OCRL, n=21) or together with LatA (n=16). (E) Follicle cells expressing UAS-Lgl-GFP stained for aPKC. Co-expressing UAS-aPKCΔN releases Lgl from the cortex. (F,G) Scatter plots represent half-time of fluorescence recovery for (F) UAS-Lgl-GFP in control (n=29, same dataset as D) and aPKCΔN-overexpressing cells (n=10), and for (G) endogenously tagged Lgl-GFP (n=43, control), aPKC RNAi (n=26) and LglS5A-GFP (homozygous, n=35) (mean±s.d.). (H,I) Midsagittal cross-sections of control and aPKC-depleted egg chambers stained for (H) DAPI and aPKC, or (I) expressing endogenous Lgl-GFP mislocalized to the apical side in aPKC RNAi (arrows). Arrowheads indicate multilayered tissue. Scale bars: 10 µm (E); 25 µm (A,H,I). P-values calculated using the Mann–Whitney U-test.
Mutants for PIP2 biosynthesis maintain the distribution of apical-basal polarity proteins in the follicular epithelium (Devergne et al., 2014; Khoury and Bilder, 2020), but they can cause defects in epithelial architecture (Claret et al., 2014; Devergne et al., 2014). We therefore depleted PIP2 with high temporal control through light-induced recruitment of the 5-phosphatase OCRL to the plasma membrane, where it converts PI(4,5)P2 to PI(4)P (Guglielmi et al., 2015). This approach triggers efficient removal of PIP2 sensor PH-ChFP from the plasma membrane, while maintaining cortically localized UAS-driven Lgl-GFP for FRAP analysis (Fig. 2C, Fig. S4). PI(4)P may be sufficient to maintain overexpressed Lgl at the plasma membrane, as previous work shows that co-depletion of PI(4)P with PI(4,5)P2 is required for complete membrane removal (Dong et al., 2015). Importantly, PIP2 optogenetic depletion accelerates fluorescence recovery in the lateral cortex, supporting the idea that PIP2 is a major binding partner that restrains Lgl cortical mobility (Fig. 2D). Consistent with this, LatA treatment does not increase Lgl-GFP mobility further when PIP2 is simultaneously depleted (Fig. 2D).
aPKC-mediated phosphorylation dissociates Lgl from the apical cortex in epithelia (Betschinger et al., 2003; Plant et al., 2003; Yamanaka et al., 2003). aPKC activity at the lateral cortex should promote membrane-cytoplasm exchange by releasing Lgl from plasma membrane phosphoinositides and cortical interactors. Accordingly, ectopic expression of constitutively active aPKCΔN (Betschinger et al., 2003) induces cytoplasmic accumulation (Fig. 2E, Fig. S4) and strongly accelerates Lgl-GFP fluorescence recovery at the lateral cortex (Fig. 2F, Fig. S4C). However, LglS5A-GFP, a knock-in allele in which all serines phosphorylatable by aPKC are mutated (Dong et al., 2015), shows identical fluorescence recovery to Lgl-GFP (Fig. 2G). Moreover, although aPKC RNAi enables apical accumulation of Lgl, its dynamics are preserved in the lateral cortex (Fig. 2G-I). Thus, aPKC inhibits Lgl accumulation in the apical domain, but has no impact in the lateral cortex, suggesting that aPKC activity is efficiently limited to the apical domain of follicle cells.
Dlg-Scrib regulate Lgl dynamics and localization by preventing aPKC ectopic activity
Scrib and Lgl cortical localization is disrupted in dlg mutant follicle cells (Fig. 3A,B, Fig. S5), similar to the Drosophila embryonic epithelium (Bilder et al., 2000). However, it is unclear how Dlg and Scrib promote Lgl cortical localization. Loss of Dlg does not interfere with PIP2-binding sites, as PH-GFP remains unchanged in dlg mutant cells (Fig. 3C). Alternatively, mislocalization of aPKC activity in dlg cells ((Bilder et al., 2000; Franz and Riechmann, 2010), Fig. 3D) may displace Lgl from the plasma membrane. Accordingly, LglS5A-GFP localizes in the cortex of dlg mutant clones (Fig. 3E,F, Fig. S5C,D), and an aPKC mutation restores Lgl cortical localization in dlg mutants (Fig. 3G). This suggests that Dlg is not required for Lgl cortical recruitment in the absence of aPKC phosphorylation. Long-term disruption of aPKC could enable constitutive membrane binding of unphosphorylated Lgl and mask a role for Dlg in Lgl cortical recruitment. We therefore inactivated aPKC acutely using an analogue-sensitive aPKC allele (Hannaford et al., 2019). Live imaging of Lgl-GFP shows that aPKC inhibition induces quick reallocation of Lgl from the cytoplasm to the cortex in dlg mutant cells (Fig. 3H, Movie 3), indicating that Dlg binding is dispensable for cortical recruitment.
Dlg and Scrib control Lgl localization and dynamics by preventing ectopic aPKC activity. (A-F) dlgA mutant clones [absence of nls-RFP (magenta)] stained for (A) Dlg, (B) Lgl or (D) aPKC, or expressing (C) PLCδPH-GFP, (E) Lgl-GFP or (F) LglS5A-GFP. (G) Clonal analysis shows that Lgl cortical localization is disrupted in dlgA mutants (green, absence of nlsRFP), but not in double dlgA, aPKCK06403 mutants (asterisk). (H) Time-lapse images of egg chambers expressing the analog-sensitive allele aPKCas4 treated with 1NA-PP1 (dlgA clone marked by absence of nls-RFP). Intensity plots show Lgl-GFP fluorescence (in arbitrary units) across the indicated interfaces of aPKCas4/aPKCK0640 cells that were wild type or dlg mutant. (I-K) Scatter plots of recovery half-time (t1/2). (I) Accelerated UAS-Lgl-GFP cortical dynamics in dlgA (n=25) in comparison with control (n=29). (J) UAS-Lgl3A-GFP dynamics are unchanged in dlgA mutants (n=22) and accelerated by PIP2 optogenetic depletion (n=20) in relation to control (n=22). (K) LglS5A-GFP dynamics are maintained in dlgA (n=21) and scrib2 (n=26) in relation to control (n=18), but are accelerated upon PIP2 depletion (n=12). Data are mean±s.d.. Scale bars: 10 μm. P-value calculated with Mann–Whitney U-test.
Dlg and Scrib control Lgl localization and dynamics by preventing ectopic aPKC activity. (A-F) dlgA mutant clones [absence of nls-RFP (magenta)] stained for (A) Dlg, (B) Lgl or (D) aPKC, or expressing (C) PLCδPH-GFP, (E) Lgl-GFP or (F) LglS5A-GFP. (G) Clonal analysis shows that Lgl cortical localization is disrupted in dlgA mutants (green, absence of nlsRFP), but not in double dlgA, aPKCK06403 mutants (asterisk). (H) Time-lapse images of egg chambers expressing the analog-sensitive allele aPKCas4 treated with 1NA-PP1 (dlgA clone marked by absence of nls-RFP). Intensity plots show Lgl-GFP fluorescence (in arbitrary units) across the indicated interfaces of aPKCas4/aPKCK0640 cells that were wild type or dlg mutant. (I-K) Scatter plots of recovery half-time (t1/2). (I) Accelerated UAS-Lgl-GFP cortical dynamics in dlgA (n=25) in comparison with control (n=29). (J) UAS-Lgl3A-GFP dynamics are unchanged in dlgA mutants (n=22) and accelerated by PIP2 optogenetic depletion (n=20) in relation to control (n=22). (K) LglS5A-GFP dynamics are maintained in dlgA (n=21) and scrib2 (n=26) in relation to control (n=18), but are accelerated upon PIP2 depletion (n=12). Data are mean±s.d.. Scale bars: 10 μm. P-value calculated with Mann–Whitney U-test.
We used FRAP to further test Dlg as a Lgl binding partner. UAS-driven Lgl-GFP enabled us to specifically monitor Lgl cortical dynamics by restoring Lgl cortical localization in dlg mutants (Figs S4A,B and S5E). This is consistent with Lgl inhibition of aPKC activity and titration of aPKC activity by overexpression of its substrates (Holly and Prehoda, 2019). UAS-driven Lgl-GFP displays faster fluorescence recovery in the absence of Dlg (Fig. 3I), possibly owing to mislocalized aPKC activity. However, given that Dlg is necessary to localize Scrib, which is also required for Lgl localization (Fig. S5F), this result would also be compatible with Lgl binding to Dlg or Scrib. To determine the specific contribution of Dlg and Scrib as Lgl cortical binding sites versus their impact on the extension of aPKC activity, we measured the dynamics of aPKC-insensitive versions of Lgl in both dlg and scrib mutants. The dynamic behavior of nonphosphorylatable Lgl is not significantly different in either mutant background (Fig. 3J,K). In contrast, PIP2 depletion accelerates the fluorescence recovery of nonphosphorylatable Lgl (Fig. 3J,K). We conclude that, whereas Lgl binds to PIP2 at the plasma membrane, Lgl cortical dynamics are independent of protein-protein interactions with the Scrib-Dlg module.
Dlg-Scrib complex does not contain Lgl in the follicular epithelium
Despite the widespread notion that Dlg, Lgl and Scrib act together, there are no in vivo evidences for a tripartite complex. We have recently repurposed light-induced protein clustering by LARIAT [light-activated reversible inhibition by assembled trap (Lee et al., 2014; Qin et al., 2017)] to probe for protein-protein interactions in S2 cells (Osswald et al., 2019). By combining CRY2 tagged with a VHH-GFP nanobody and GFP knock-in lines for Lgl, Dlg and Scrib, we probed for interactions in the follicular epithelium by evaluating co-recruitment to multimeric clusters (Fig. 4A). Light stimulation induced the formation of Scrib-GFP clusters, which accumulate below the adherens junction (Fig. 4B) and contain Dlg (Fig. 4C).
In vivo detection of Scrib-Dlg complexes that lack Lgl. (A) Detection of protein-protein interactions through light-induced clustering. The GFP-tagged protein (bait) binds CRY2-VHH, which interacts with CIBN-MP upon light exposure to generate multimeric complexes and trap interactors (prey). (B) Light-induced protein clusters of Scribble-GFP localize below Armadillo-stained adherens junctions. (C-E) Dlg immunofluorescence shows co-aggregation with (C) Scribble-GFP, with (D) ScribΔPDZ-GFP but not with Scrib-ΔLRR-GFP or (E) Lgl-GFP. (F) Scribble-GFP and Dlg-GFP do not co-cluster with Lgl-mCherry. (B-F) Pixel intensity profiles were measured along the indicated membranes (arrows) and were normalized to maximum intensity. Scale bars: 5 μm (enlargements); 10 μm (egg chambers in C and E).
In vivo detection of Scrib-Dlg complexes that lack Lgl. (A) Detection of protein-protein interactions through light-induced clustering. The GFP-tagged protein (bait) binds CRY2-VHH, which interacts with CIBN-MP upon light exposure to generate multimeric complexes and trap interactors (prey). (B) Light-induced protein clusters of Scribble-GFP localize below Armadillo-stained adherens junctions. (C-E) Dlg immunofluorescence shows co-aggregation with (C) Scribble-GFP, with (D) ScribΔPDZ-GFP but not with Scrib-ΔLRR-GFP or (E) Lgl-GFP. (F) Scribble-GFP and Dlg-GFP do not co-cluster with Lgl-mCherry. (B-F) Pixel intensity profiles were measured along the indicated membranes (arrows) and were normalized to maximum intensity. Scale bars: 5 μm (enlargements); 10 μm (egg chambers in C and E).
To identify how Dlg binds Scrib, we induced clustering of Scrib versions lacking the leucine-rich repeats (ΔLRR-GFP) or the PDZ domains (ΔPDZ-GFP), which bind the Dlg-binding protein GUK-holder (Caria et al., 2018). Both proteins are largely cytoplasmic and become enriched at the cortex upon clustering, but whereas Dlg is absent from ΔLRR-GFP clusters, Dlg co-clusters with ΔPDZ-GFP (Fig. 4D). Thus, the identified Scrib-Dlg complex is formed by interactions with the LRR-LAPSD domains, which are both essential and partially sufficient for apical-basal polarization and proliferation control in mammalian and Drosophila tissues (Choi et al., 2019; Zeitler et al., 2004). We next tested whether Lgl is also included in Dlg-Scrib complexes. In contrast to Scrib-GFP clusters, Dlg is absent from Lgl-GFP clusters (Fig. 4C,E). Conversely, Lgl-mCherry does not colocalize with Scrib-GFP or Dlg-GFP light-induced clusters (Fig. 4F), indicating that Lgl is not part of the Scrib-Dlg complex present in the lateral domain of follicle cells.
Conclusions
Understanding how basolateral proteins maintain apical-basal polarity requires refined approaches to investigate their dynamic behavior, which is intimately connected to function. Dlg and Scrib mobility has previously been studied in septate junctions (Babatz et al., 2018; Oshima and Fehon, 2011). Our study compared the dynamics of the basolateral determinants in the lateral domain and dissected how Lgl associates to the cortex. Lgl dynamics are tightly linked to membrane diffusion, as previously observed for the PIP2-binding domain PLCδPH (Goehring et al., 2010; Hammond et al., 2009). PIP2 proteins are indeed major binding sites for Lgl, as previously suggested (Bailey and Prehoda, 2015; Dong et al., 2015). Moreover, even though aPKC phosphorylation induces cortical detachment, we show that aPKC inactivation does not alter Lgl dynamic behavior in the lateral cortex. Thus, there is an efficient restriction of aPKC activity to the apical domain in the steady state of polarized follicle cells. This resembles the maintenance phase of C. elegans PAR polarization, where the dynamics of PAR proteins are only altered in the region enriched in the opposing species (Goehring et al., 2011).
Although we cannot exclude the possibility that the formation of short-lived protein complexes makes a minor contribution to Lgl dynamics, this study shows that Dlg-Scrib are not major binding partners for Lgl. Instead they regulate cortical Lgl by preventing aPKC ectopic localization and activity in the lateral domain of follicle cells. Our findings are consistent with recent work, which shows that Dlg and Scrib protect Lgl from aPKC in the follicular and embryonic epithelium, possibly by regulation of specific phosphorylation sites (Khoury and Bilder, 2020). This could promote Lgl accumulation at the cortex, and concomitantly prevent aPKC lateral extension by inhibiting aPKC-mediated stabilization of apical complexes in the lateral cortex (Fletcher et al., 2012). Moreover, Dlg and Scrib may repress the extension of apical proteins by directing the assembly and location of adherens junctions, as was recently described in the embryonic ectoderm (Bonello et al., 2019). Although further work is necessary to characterize how Dlg-Scrib block apical determination, the separation of function between Scrib-Dlg and Lgl may be conserved as mammalian Scrib orthologues also have Lgl independent roles in epithelial polarity (Choi et al., 2019). In conclusion, this study indicates that Lgl does not act as part of the Scrib-Dlg complex, suggesting individual parallel mechanisms to antagonize apical identity in the follicular epithelium.
MATERIALS AND METHODS
Drosophila stocks and genetics
Stocks used were as follows: DlgCC01936-GFP, ScribCA07683-GFP (Buszczak et al., 2007), Lgl-GFP, LglS5A-GFP and Lgl-mCherry (Dong et al., 2015) were used for endogenous expression of tagged polarity proteins; UAS-Lgl3A-GFP and UAS-Lgl-GFP (Betschinger et al., 2003) for overexpression conditions; UAS-pmCIBN and UAS-CRY2-OCRL for optogenetic depletion of PIP2 in the plasma membrane (Guglielmi et al., 2015; kindly provided by Stephano de Renzis, EMBL, Heidelberg, Germany); UAS-LARIAT (VhH-SNAPtag-CRY2(PHR)-P2A-CIB1MP) for light-induced clustering of GFP-tagged proteins (Qin et al., 2017; kindly provided by Xiaobo Wang, CBI, Toulouse, France), including UAS-ScribΔLRR-GFP (BDSC: #59084) and UAS-ScribΔPDZ-GFP (BDSC: #59083); UAS- PLCδPH-GFP (BDSC:#39693), ubi-PLCγPH-eGFP or ubi-PLCγPH-mCherry (PH-ChFP) (Herszterg et al., 2013; kindly provided by Yohanns Bellaiche, Institut Curie, Paris, France) was used to label PIP2 with the pleckstrin homology (PH) domain of PLCγ; UAS-aPKC RNAi (BDSC: #25946); UAS aPKCΔN (Betschinger et al., 2003); aPKCK06403 (Wodarz et al., 2000); apkcas4, an analogue-sensitive allele used for acute inhibition of aPKC (Hannaford et al., 2019; kindly provided by Jens Januschke, University of Dundee, UK); and scrib2 (Bilder and Perrimon, 2000). We used a recently generated dlgA mutant allele, which was produced with low concentrations of ethyl methylsulphonate, to diminish secondary alterations (Haelterman et al., 2014; Yamamoto et al., 2014). dlgA encodes a stop codon on the third PDZ domain of Dlg and behaves as a protein null allele. Mitotic clones were generated with the FLP/FRT-mediated mitotic recombination system and were induced by heat shock at 37°C (Xu and Rubin, 1993).
Drosophila melanogaster stocks were reared on standard cornmeal/agar/molasses/yeast media at 18°C or 25°C. To boost UAS-RNAi or UAS-LARIAT expression, flies were raised at 29°C overnight. Flies were protected from the light before light induction in the optogenetic experiments. tj-GAL4 and GR1-GAL4 were used to drive the expression of UAS transgenes in the follicular epithelium.
Genotypes
FRAP analysis of UAS-Lgl-GFP
FRAP analysis of PLCδPH-GFP
tj-GAL4/+; UAS-PLCδPH-GFP/+ (Fig. S2)
LglS5A homozygous
hs-FLP; LglS5A-GFP FRT40A/nls-RFP FRT40A (Fig. 2G)
Lgl-GFP in aPKC RNAi
tj-GAL4, Lgl-GFP/, Lgl-GFP; UAS aPKC RNAi/ (Fig. 2G,I)
aPKC RNAi
tj-GAL4/+; UAS aPKC RNAi/+ (Fig. 2H)
Optogenetic PIP2 depletion
dlg mutant clones
dlgA FRT19A/hs-FLP nls-RFP FRT19A (Fig. 3A,B,D)
dlgA FRT19A/hs-FLP nls-RFP FRT19A;; ubi-PH-mCherry (Fig. 3C)
dlgA FRT19A/hs-FLP nls-RFP FRT19A; tj-GAL4/+; UAS-Lgl-GFP/+ (Fig. 3I and Fig. S4A,B)
dlgA FRT19A/hs-FLP nls-RFP FRT19A; tj-GAL4/+; UAS-Lgl3A-GFP/+ (Fig. 3J)
dlgA FRT19A/hs-FLP nls-RFP FRT19A; Lgl5A-GFP/+ (Fig. 3F,K, Fig. S5D)
dlgA FRT19A/hs-FLP nls-RFP FRT19A; Lgl-GFP (Fig. 3E, Fig. S5C)
dlgA FRT19/hs-FLP nls-RFP FRT19A;; Scrib-GFP (Fig. S5A,B)
Double dlg,aPKC mutants
scrib mutant clones:
LARIAT optogenetic clustering experiments
tj-GAL4 or UAS-LARIAT/Cyo; Scrib-GFP/TM6 (control) (Fig. 4B)
tj-GAL4/UAS-LARIAT; Scrib-GFP/TM6 (Fig. 4B)
tj-GAL4 or UAS-LARIAT/Cyo; Scrib-GFP/Scrib-GFP (control) (Fig. 4C)
tj-GAL4/UAS-LARIAT; Scrib-GFP/Scrib-GFP (Fig. 4C)
UAS-ScribΔPDZ-GFP/Cyo; UAS-LARIAT GR1-GAL4/+ (Fig. 4D)
UAS-ScribΔLRR-GFP/UAS-LARIAT GR1-GAL4 (Fig. 4D)
tj-GAL4/ Lgl-GFP; UAS-LARIAT GR1-GAL4/+ (Fig. 4E)
Lgl-mCherry/Cyo; UAS-LARIAT/ Scrib-GFP (control) (Fig. 4F)
tj-GAL4/Lgl-mCherry; UAS-LARIAT/Scrib-GFP (Fig. 4F)
Dlg-GFP/+; tj-GAL4/Lgl-mCherry; TM6/+ (Fig. 4F)
Dlg-GFP/+; tj-GAL4/Lgl-mCherry; UAS-LARIAT/+ (Fig. 4F)
Immunofluorescence
Ovaries of well-fed Drosophila females were fixed in 4% paraformaldehyde (in PBS) for 20 min, washed 3×10 min in PBT (PBS with 0.05% of Tween 20), blocked with PBT-10 (PBT supplemented with 10% BSA) and then incubated overnight with primary antibodies in PBT supplemented with 1% BSA. After 4×30 min washes in PBT, ovaries were incubated with a secondary antibody for 2 h, washed three times for 10 min with PBT and mounted in Vectashield with DAPI (Vector Laboratories). The following primary antibodies were used: rabbit anti-Lgl (1:100, d-300, Santa Cruz Biotechnology), rabbit anti-aPKC (1:500, c-20, Santa Cruz Biotechnology) and mouse anti-Dlg (1:200, F3 Developmental Studies Hybridoma Bank). Fixed tissue was imaged using a 1.1 numerical aperture/40× water or 1.30 numerical aperture/63× objectives on an inverted laser scanning confocal microscope Leica TCS SP5 II (Leica Microsystems) or with a 1.20 numerical aperture/63× objective on an inverted laser scanning confocal microscope Leica TCS SP8 (Leica Microsystems).
Drug treatments
Actin cytoskeleton disruption was achieved by treating dissected ovarioles for at least 1 h with 5 µg/ml of Latrunculin A (Sigma-Aldrich) added to the Imaging Medium [Schneider Insect Cell Medium (Gibco) supplemented with 10% FBS (ThermoFisher) and 200 μg/µl of Insulin (Sigma)]. Temporal inhibition of aPKC activity in dlg mutant cells was achieved by using flies carrying simultaneously the apkcas allele by addition of 1NA-PP1 (Calbiochem) to the imaging media at a final concentration of 100 µM immediately before imaging.
Optogenetic experiments
Flies were exposed for 24 h to direct blue LED light (472 nm) prior to ovary dissection and fixation to trigger GFP-tagged protein clustering. For optogenetic depletion of PIP2 by the inositol polyphosphate 5-phosphatase OCRL, ovaries were dissected using a 593 nm LED light and exposed to a 488 nm laser during the live imaging process. FRAP experiments were performed between ∼5 min and 1 h after optogenetic activation.
Fluorescence recovery after photobleaching in the Drosophila follicular epithelium
Drosophila egg chambers from stage 7 to stage 9 were imaged in culture dishes (MatTek) with either Imaging Medium (Prasad et al., 2007) or with 10S Voltalef oil (VWR Chemicals). Imaging Medium was used for experiments with overexpressed GFP-tagged proteins. We resorted to the better optical properties of 10S Voltalef to improve the fluorescence signal-to-noise ratio in the experiments using endogenously tagged-GFP proteins. FRAP experiments were performed using a 1.1 numerical aperture/40× water objective on a Leica TCS SP5 II (Leica Microsystems) confocal microscope and using the FRAP Wizard application of the LAS Advance Fluorescence (AF) 2.6. software. The imaging plane was focused at the surface of the egg chamber on the lateral cortex of the follicular epithelium (Fig. S1A), where a cortical section of two neighboring cells was selected for bleaching using either a rectangular ROI of 3 μm×1.0 μm. A 1.5 μm×1 μm ROI during the bleaching process was used to determine whether diffusion played a role in Lgl recovery (Fig. 1E).
For all the FRAP experiments performed, a three-step protocol was developed: (1) a pre-bleaching acquisition phase for GFP equilibration; (2) a photoperturbation step performed using a single bleaching pulse with the 405 laser line (50 mW) at maximum laser power using the Zoom In option of the FRAP Wizard; and (3) post-bleaching phases to measure the fluorescence recovery. Pre- and post-bleach imaging of GFP was performed using the 488 nm excitation line with 20%-35% laser intensity. The multi-step FRAP protocol was optimized for each basolateral polarity protein (Fig. 1A), ensuring that sufficient data points were collected to accurately obtain the half-time of recovery while minimizing photobleaching during acquisition: pre-bleaching, six frames acquired for equilibration of GFP signal (0.543 s between frames); post-bleaching 1 (1 s/frame) Lgl, 25 frames; Dlg, 50 frames; Scrib, 60 frames; PH; 30 frames; post-bleaching 2 (3 s/frame) Lgl, 25 frames; Dlg, 50 frames; Scrib, 60 frames; PH; 10 frames; and post-bleaching 3 (10 s/frame) Lgl, 10 frames; Dlg, 30 frames; Scrib, 35 frames.
Image processing and fluorescence recovery curve normalization
Image processing and fluorescence intensity recovery measuring were performed using ImageJ (Schindelin et al., 2012) (http://imagej.nih.gov/ij/) through a pipeline composed of three main steps.
(1) Registering the timelapse images
Image drift during acquisition was corrected using the ‘Stack Reg’ plug-in to align time lapse images.
(2) Measuring the raw fluorescence intensity values from the corrected images
Measurements were collected in four different regions: (1) 3 μm×1 μm ROI containing the photobleached region (Bl in Fig. S1A); (2 and 3) two 3 μm×1.5 μm ROIs in cortical regions (NB in Fig. S1A) non-adjacent to the photobleached region to correct for photobleaching during the acquisition of post-bleaching steps; and (4) a 2 μm diameter circular ROI placed in the prospective nucleus, used for background subtraction as none of the proteins of interest is present in the nucleus (B in Fig. S1A). We note that 1 µm diameter cytoplasmic ROIs were used to determine cytoplasmic intensity of Lgl.
(3) Subtracting the background, correcting for photobleaching during acquisition and normalizing recovery data
Data analysis and statistics
The corrected and normalized data was plotted in recovery curves using GraphPad Prism 8.2.0. Recovery plots represent the average normalized fluorescence intensity and error bars indicate 95% confidence intervals (CI). We performed an extra sum-of-squares F test to assess the preferred model (bi-exponential versus single-exponential) of each individual curve obtained for Lgl-GFP fluorescence recovery. Although bi-exponential fitting was predominantly the best model for the individual Lgl-GFP fluorescence recovery curves (79%, n=43), the bi-exponential model produced open-ended 95% confidence intervals for half-times or a plateau for a large fraction of these curves (28%). Moreover, one-exponential was the preferred model for another fraction of the curves (21%). For Fig. 1E, values were fitted to a bi-exponentional function: , where Spanfast and Spanslow are the amplitudes of the fast and slow fractions of recovery (constrained to a shared value between the individual curves of a single dataset), and Kslow and Kfast are rate constants expressed in reciprocal of the time. This function allows to calculate the fast half-time [tfast=ln(2)/Kfast] and the slow half-time [tslow=ln(2)/Kslow]. Because we concluded that we could not separate membrane diffusion from membrane-cytoplasm exchange using the double exponential equation, we used a single exponential function for individual curve fitting to compare the recovery half-times of Lgl dynamics in different experimental settings:
, where Kd represents the rate of the unbinding reaction expressed in reciprocal of the time. This function allows us to extract a simple half-time value (t1/2) for each curve representing one cell [t1/2=ln(2)/kd]. We then obtained and compared half-time dispersion plots for the different biological conditions. The statistical significance between them was assessed using the non-parametric Mann–Whitney U-test.
Individual photobleaching experiments were excluded from the experimental datasets if they met one of the following criteria: (1) inefficient photobleaching step, such that reduction of intensity in the photobleached region after the bleaching pulse was below 50% (the recommended minimum for quantitative analysis of FRAP curves; McNally, 2008); (2) low plateau of fluorescence recovery [cutoff at 65% (60% for LglS5A-GFP+pmOCRL and UAS-Lgl-GFP+aPKCΔN to correct for lower plateau in the average of the whole dataset]; (3) values with low R2 [cutoff at 0.75 (0.65 for LglS5A-GFP+pmOCRL and UAS-Lgl-GFP+aPKCΔN)] (Mean values of the collected data are shown in Table S1.); (4) the half-time of recovery is an outlier (outliers were removed using the Extreme Studentized Method (ESD Method) with a significance level of 0.05 using a GraphPad online application (http://www.graphpad.com/quickcalcs/Grubbs1.cfm).
Acknowledgements
We thank J. Januschke, J. Knoblich, S. De Renzis, Y. Bellaiche, X. Wang, D. St Johnston and the Bloomington stock center for fly stocks. We thank H. Richardson for detailed information on the Scrib-GFP deletion constructs, Y. Bellaiche for comments on the manuscript and C. Sunkel for assistance with supervision.
Footnotes
Author contributions
Conceptualization: G.V., S.M., E.M.-d-S.; Methodology: G.V., S.M., A.B.-C., M.O., E.M.-d-S.; Formal analysis: G.V., S.M., A.B.-C., E.M.-d-S.; Investigation: G.V., S.M., A.B.-C., M.O.; Writing - original draft: E.M.-d-S.; Writing - review & editing: G.V., S.M., M.O., E.M.-d-S.; Supervision: E.M.-d-S.; Project administration: E.M.-d-S.; Funding acquisition: E.M.-d-S.
Funding
This work was supported by Fundação para a Ciência e a Tecnologia (PTDC/BEX-BCM/0432/2014).
Peer review history
The peer review history is available online at https://dev.biologists.org/lookup/doi/10.1242/dev.186593.reviewer-comments.pdf
References
Competing interests
The authors declare no competing or financial interests.