A polybasic domain in aPKC mediates Par6-dependent control of membrane targeting and kinase activity

Polarity proteins play conserved and essential roles in regulation of cell polarity, and for the majority of them, achieving polarized plasma membrane (PM)/cortical localization is essential for their functions (Hong, 2018; Rodriguez-Boulan and Macara, 2014). Mechanisms mediating the PM/cortical association of these polarity proteins had often been assumed to be based primarily on the intricate protein–protein interactions among polarity proteins and their regulators (Rodriguez-Boulan and Macara, 2014). Recent studies, however, discovered that several polarity proteins such as Lethal giant larvae (Lgl), Numb, and Miranda contain so-called polybasic (i.e., basic-hydrophobic) domains (Bailey and Prehoda, 2015; Dong et al., 2015), which are short but highly positively charged owing to their enrichment of basic Arg/Lys residues. Polybasic domains can specifically bind to PM, as the inner surface of PM is the most negatively charged membrane surface, as a result of membrane phosphatidylserine (Yeung et al., 2008) and the unique enrichment of polyphosphoinositides phosphatidylinositol-4-phosphate (PI4P) and phosphatidylinositol 4,5-bisphosphate (PIP₂; Hammond et al., 2012). Moreover, polybasic domains in Lgl, Numb, and Miranda also contain conserved serine residues that can be phosphorylated by another key polarity protein, atypical PKC (aPKC; Bailey and Prehoda, 2015; Dong et al., 2015). Similar to the phosphorylation of the polybasic ED domain in myristoylated alanine rich C-kinase substrate (MARCKS) by PKC (Arbuzova et al., 2002), aPKC phosphorylation neutralizes the positive charges in a polybasic domain, therefore inhibiting its electrostatic binding to the PM (Bailey and Prehoda, 2015; Dong et al., 2015). Such aPKC-dependent inhibition serves as an elegant mechanism to polarize the PM targeting of polybasic polarity proteins, allowing apically localized aPKC to limit Lgl, Numb, and Miranda to basolateral PM. Potential aPKC-phosphorylatable polybasic domains have been found in hundreds of proteins in metazoan genomes (Bailey and Prehoda, 2015; Y. Hong, unpublished data), suggesting that aPKC plays a critical role in regulating the PM targeting of many polybasic proteins.

To date, however, the exact molecular mechanism governing the PM/cortical localization of aPKC itself remains unclear. Unlike conventional PKC (cPKC) and novel PKC (nPKC) that bind DAG, phospholipids, and calcium for their kinase activation and PM targeting, aPKC has no well-defined PM or calcium binding domains and has not been demonstrated or proposed to directly bind the PM (Rosse et al., 2010; Garg et al., 2014). In fact, it is considered a unique feature of aPKC that its kinase activity and subcellular localizations appear to be exclusively regulated by protein–protein interactions other than lipids and/or calcium (Rosse et al., 2010). aPKC has a PB1 domain that binds to the PB1 domain in another polarity protein, partitioning defective 6 (Par-6), which consistently colocalizes with aPKC in many polarized cell types, including epithelial cells and Caenorhabditis elegans one-cell embryos (Hong, 2018; Suzuki and Ohno, 2006). PM/cortical localization of aPKC/Par-6 complex has been assumed mainly based on protein interactions with other polarity proteins such as Bazooka (Baz; Izumi et al., 1998; Krahn et al., 2010; Morais-de-Sá et al., 2010), Crumbs (Crb; Sotillos et al., 2004), Stardust (Sdt/Pals1; Wang et al., 2004), tight junction-associated PDZ protein Patj (Hurd et al., 2003), or Cdc42 (Joberty et al., 2000; Lin et al., 2000; Qiu et al., 2000). Recent studies have delineated some exciting detail mechanisms by which Par-3 and Cdc42 coordinate the spatial and temporal control of aPKC kinase activity during the anterior-posterior (A-P) polarization of worm one-cell embryos (Dickinson et al., 2017; Rodriguez et al., 2017), but how aPKC PM localization and kinase activity are regulated during apical-basal polarization is much less clear. Moreover, in vitro kinase assays yielded conflicting results regarding whether binding of Par-6 to aPKC inhibits or activates its kinase activity (Hong, 2018).

Here we report that the pseudosubstrate region (PSr) of aPKC also functions as a polybasic domain that directly binds to the PM through electrostatic interaction with PM phosphoinositides PI4P and PIP₂. Interestingly, unlike the polybasic domains in Lgl, Numb, and Miranda, PSr in aPKC has not been shown to be phosphorylatable. Instead we show that the PM binding of PSr appears to be allosterically controlled by protein interactions between aPKC and Par-6. Our data suggest a model in which Par-6–dependent allosteric regulation of polybasic PSr in aPKC couples the PM targeting of aPKC and the spatially restricted activation of aPKC kinase during apical-basal polarization.

Our previous studies showed that acute hypoxia induces loss of PM phosphoinositides, which in turn disrupts the PM localization of polybasic domain proteins such as Lgl and Numb in live fly tissues and cultured cells (Dong et al., 2015). Curiously, in the same studies, we found that endogenous DaPKC also showed dramatic loss of PM localization in epithelial cells under hypoxia (Dong et al., 2015), suggesting that PM targeting of DaPKC may be based on direct binding to PM phospholipids. Similar to cPKC and nPKC, aPKC contains a PSr that binds the kinase domain (KD) to induce autoinhibition (Rosse et al., 2010). We identified that the PSr in DaPKC is in fact highly polybasic (Fig. 1 A): the 17-aa PSr contains eight Arg and Lys residues plus one Leu and one Trp residue, and the adjacent 13-aa sequence in C1 domain contains an additional four Arg and Lys residues plus two Phe residues (Heo et al., 2006). Besides the enriched Arg and Lys residues, the presence of Trp, Phe, and Leu is highly indicative of a polybasic domain, as the bulky or long hydrophobic side chains of these residues enhance the PM binding (Heo et al., 2006; McLaughlin and Murray, 2005; Yeung et al., 2006). The PSr is extremely well conserved between DaPKC and mammalian aPKC isoforms PKCζ and PKCι (Fig. 1 A), with a basic-hydrophobic index (Brzeska et al., 2010) of 0.92, comparable to polybasic domains in Lgl (1.00) and Numb (0.89).

Similar to the polybasic domain in Lgl (Dong et al., 2015), PSr in aPKC is capable of direct binding to PI4P and PIP₂ in vitro in liposome pull-down assays. Purified GST fusion of PSr from PKCζ bound PI4P- and PIP₂-containing liposomes, whereas nonpolybasic GST-PSr-KR8Q, with all eight Arg/Lys residues mutated to Gln, did not (Fig. 1 B). PSr is also required for PM targeting of DaPKC in vivo: although ubiquitously expressed aPKC::GFP localized to apical PM in both wild-type and aPKC^−/− epithelial cells, nonpolybasic aPKC^KR8Q::GFP stayed in the cytosol (Figs. 1 C and S1 A).

In vivo, DaPKC::GFP also showed acute and reversible loss from PM, similar to Lgl and Numb in live Drosophila follicular and embryonic epithelial cells under hypoxia (Fig. 1 D and Fig. S1 B; and Video 1), consistent with electrostatic binding to phosphoinositides as the primary mechanism targeting DaPKC to the PM. Furthermore, Par-6 is an essential regulatory partner of aPKC and forms a robust complex with aPKC (Hong, 2018; Suzuki and Ohno, 2006). Both proteins are mutually dependent on each other for proper localization and function during cell polarization processes. Consistently, hypoxia in live Drosophila follicular and embryonic epithelial cells also induced acute and reversible loss of PM targeting of Par-6::GFP (Fig. 1 D, Fig. S1 B, and Video 2). The loss of DaPKC/Par-6 complex from PM under hypoxia is not due to the concurrent loss of Lgl from PM. Par-6::GFP in lgl^−/− mutant follicular cells showed expanded localization to basolateral PM but responded to hypoxia identically to Par-6::GFP in wild-type cells (Fig. S1 C). Overall, our data suggest that the polybasic PSr in DaPKC is essential for localizing DaPKC/Par-6 complex to the PM in Drosophila epithelia in vivo, likely through direct interaction with PM phosphoinositides such as PI4P and PIP₂ (see below).

Given the well-documented requirement of Par-6 in aPKC subcellular localization in many cell types (Chen et al., 2018; Hong, 2018; Suzuki and Ohno, 2006), we further investigated the role of Par-6 in regulating the polybasic PSr-mediated PM targeting of aPKC. For this purpose, we took the overexpression approach in nonpolarized HEK293 cells to investigate the potential interactions between aPKC and Par-6 or other exogenously introduced proteins. To keep results consistent, all cultured cell experiments reported here used mammalian proteins of Lgl, aPKC, Par-6, Crb, and Cdc42. The mammalian aPKC family has two isoforms, PKCζ and PKCι, and both contain PSr domains nearly identical to DaPKC’s (Fig. 1 A). A PKCζ-specific antibody (Stross et al., 2009) showed that HEK293 cells express no detectable endogenous PKCζ (Fig. 2, A and B) and thus provide a relatively clean background for investigating the PM targeting of the exogenously expressed PKCζ. Surprisingly, we found that PKCζ::GFP was completely cytosolic when expressed alone (Fig. 2 A). However, although Par-6::RFP was also cytosolic when expressed alone, coexpression of PKCζ::GFP and Par-6::RFP resulted in strong and robust PM localization of both proteins (Fig. 2 A). We also made a bicistronic construct that expresses a fusion protein of PKCζ::RFP and Par-6::iRFP connected by a 2A peptide, which self-cleaves during translation (Chan et al., 2011) to produce PKCζ (PKCζ::RFP^-2A) and Par-6 (^2A-Par-6::iRFP) at a constant 1:1 ratio. The fusion protein appeared to be efficiently cleaved (Fig. 2 C), and both PKCζ::RFP^-2A and ^2A-Par-6::iRFP showed strong PM localization, as expected (Fig. 2 A).

We confirmed that such colocalization to PM requires direct physical interaction between Par-6 and PKCζ, as a C107Y mutant of PKCζ that specifically abolishes the physical interaction with Par-6 (Kim et al., 2009) failed to localize to PM when coexpressed with Par-6 (Fig. 2, D and E). Furthermore, PM targeting of PKCζ/Par-6 complex requires the polybasic PSr in PKCζ. Coexpression of Par-6 with nonpolybasic PKCζ^KR8Q resulted in no PM localization of either protein (Fig. 2 D), even though the physical interaction between Par-6 and PKCζ^KR8Q remained intact (Fig. 2 E). The loss of PM targeting of PKCζ^KR8Q is not due to potential misfolding, as PKCζ^KR8Q phosphorylates Lgl similarly to PKCζ in HEK293 cells (Fig. 6, B and C). Our findings are not unique to HEK293 cells, as COS7 and MCF7 breast cancer cells also showed the same Par-6–dependent PM targeting of PKCζ (Fig. S2, A and B). In fully polarized MDCK cells, overexpressed PKCζ alone was cytosolic, although Par-6 alone was partially PM-localized (Fig. S2 C). Nonetheless, coexpression of PKCζ and Par-6 resulted in robust PM localization of both proteins (Fig. S2 C). Interestingly, Par-6 became cytosolic when coexpressed with PKCζ^KR8Q (Fig. S2 C), which could indicate a dominant-negative effect of PKCζ^KR8Q due to the formation of PKCζ^KR8Q/Par-6 that is incapable of PM targeting. Thus, in polarized MDCK cells, additional mechanisms exist to at least partially target Par-6 to PM, but PM targeting of PKCζ remains dependent on both polybasic PSr and Par-6.

Cdc42 has been shown to be essential for aPKC/Par-6’s localization to anterior PM in C. elegans one-cell embryos and has well-characterized physical interactions with Par-6. We thus investigated whether aPKC/Par-6 PM targeting in HEK293 cells is also Cdc42 dependent. In HEK293 cells expressing constitutively active Cdc42^CA, Par-6 was readily recruited to the PM, as expected (Fig. 3 A). Furthermore, nonpolybasic PKCζ^KR8Q was cytosolic in cells expressing either Cdc42^CA (Fig. 3 A) or Par-6 (Fig. 2 D), but became PM-localized in cells coexpressing both Cdc42^CA and Par-6 (Fig. 3 A). Thus, in HEK293 cells, the interaction between Cdc42^CA and Par-6 is sufficient to recruit Par-6/aPKC complex to the PM, and such PM recruitment can be independent of the polybasic PSr in aPKC. However, the role of Cdc42^CA in targeting aPKC/Par-6 is likely context dependent, since in Drosophila follicular cells, Cdc42^CA overexpression failed to recruit DaPKC^KR8Q::GFP to the PM (Fig. 3 B).

In contrast, in HEK293 cells, overexpression of dominant-negative Cdc42^DN did not inhibit the PM targeting of aPKC/Par-6 (Fig. 3 A). In Drosophila follicular cells subjected to cdc42-RNAi, DaPKC::GFP was reduced from PM and enriched at intracellular puncta (Fig. 3 C), likely because of the interaction between aPKC/Par-6 and polarity protein Baz, which was also enriched at the same puncta (Fletcher et al., 2012). To further clarify the role of Cdc42 in our cell-based assays, we made two Par-6 mutants: PB1^Par-6, which removes the CRIB and PDZ domains, thus retaining only the N-terminal PB1 domain, and Par-6^ΔPB1, which carries the reciprocal truncation. PB1^Par-6 was fully capable of targeting PKCζ to PM, but Par-6^ΔPB1 was not (Fig. 5 B), suggesting that the physical interaction between the N-terminal PB1 domains in aPKC and Par-6 is sufficient to induce PSr-dependent PM targeting of aPKC. Given that PB1^Par-6 does not contain any known Cdc42-interacting domain, PM targeting of PKCζ by PB1^Par-6 should be Cdc42 independent. In summary, our results do not exclude the role of Cdc42 in regulating aPKC/Par-6 PM targeting under certain cellular contexts but show that PM targeting of aPKC/Par-6 can be mechanistically independent of Cdc42.

Phosphoinositides PI4P and PIP₂ are uniquely enriched in PM and are considered the major negatively charged phospholipids responsible for electrostatically binding polybasic domains (Hammond, 2012). To confirm that PM targeting of aPKC/Par-6 complex in cells is indeed mediated by PI4P and PIP₂, we used a well-established system to acutely and selectively deplete PI4P, PIP₂, or both in HEK293. In this inducible system, addition of rapamycin induces dimerization between FK506 binding protein (FKBP)-tagged phosphatases and FKBP-rapamycin binding (FRB)-tagged PM anchor protein Lyn11-FRB-CFP, resulting in acute PM recruitment of phosphatase and concurrent depletion of target phospholipids (Hammond, 2012). In particular, PM recruitment of a chimeric lipid phosphatase Pseudojanin (PJ) rapidly converts both PI4P and PIP₂ in PM to phosphatidylinositol (PI; Hammond, 2012; Fig. 4 A). In HEK293 cells expressing PKCζ::GFP and Par-6::iRFP together with RFP-FKPB-PJ and Lyn11-FRB-CFP, addition of rapamycin induced acute PM localization of PJ and concurrent loss of both PKCζ::GFP and Par-6::iRFP from PM (Fig. 4 B). Depleting either PIP₂ using FKBP-INPP5E or PI4P using FKBP-PJ-Sac induced much milder loss of PM targeting of both proteins (Fig. 4, A and B).

Our data suggest that PI4P and PIP₂ likely act redundantly to bind PKCζ/Par-6 complex to PM. This is in contrast to Lgl, which appears to rely more on PIP₂ for its PM targeting (Dong et al., 2015). Consistently, when we used the similar rapamycin-inducible system to acutely deplete PIP₂ in Drosophila follicular epithelial cells, Par-6::GFP remained on the PM (Fig. S3), whereas Lgl::GFP showed significant loss from PM in PIP₂-depleted cells (Dong et al., 2015). We could not carry out assays by inducible depletion of PI4P or both PI4P and PIP₂, as corresponding genetic tools are currently unavailable in Drosophila.

Finally, to further confirm the direct binding between aPKC/Par-6 complex and membrane PI4P and PIP₂, we performed liposome binding assays using purified PKCζ, PKCζ/Par-6 complex, or PKCζ^KR8Q/Par-6 complex. As shown in Fig. 4 C, purified PKCζ protein alone did not bind PI4P or PIP₂ liposomes, whereas purified PKCζ/Par-6 complex cosedimented with PI4P/PIP₂ liposomes. Furthermore, neither PKCζ^KR8Q nor Par-6 showed binding to PI4P/PIP₂ liposomes when purified PKCζ^KR8Q/Par-6 was used (Fig. 4 C). Such in vitro data demonstrate that only aPKC/Par-6 complex, but not aPKC alone, can directly bind membrane PI4P and PIP₂ in a polybasic PSr-dependent manner.

Why is PKCζ alone incapable of binding to PM? Previous studies suggested that the PSr in PKCζ binds the kinase domain (KD) to autoinhibit its kinase activity, and that binding of Par-6 likely induces an allosteric conformation change in PKCζ displacing the PSr from the KD (Graybill et al., 2012). We therefore postulate that in unbound aPKC, its PSr is occluded by the KD from binding to the PM but is allosterically exposed once Par-6 binds aPKC. To test this hypothesis, we generated two KD-deletion mutants, PKCζ-ΔKD and PKCζ^KR8Q-ΔKD. In contrast to full-length PKCζ, PKCζ-ΔKD localizes to PM in the absence of Par-6, but its PM localization remains dependent on the positive charges of PSr, as nonpolybasic PKC^KR8Q-ΔKD is cytosolic regardless the presence of Par-6 (Fig. 5 A). In addition, merely reducing the interaction between PSr and KD is not sufficient to make PSr accessible to PM binding, as two mutants, PKCζ^A119D and PKCζ^AADAA, carrying previously characterized mutations (Fig. 1 A) shown to reduce the autoinhibition of PSr to the KD (Graybill et al., 2012), were all cytosolic (Fig. 5 A). When coexpressed with Par-6, PKCζ^A119D strongly localized to PM (Fig. 5 A), whereas PKCζ^AADAA showed barely detectable PM localization, likely because of significantly reduced positive charges of PSr by four Arg→Ala mutations (Fig. 1 A). Finally, expressing Par-6 PB1 domain (PB1^Par-6), but not Par-6^ΔPB1, induced PM localization of both PKCζ and PB1^Par-6 (Fig. 5 B), suggesting that interaction between PKCζ and Par-6 PB1 domain alone is both sufficient and necessary to target PKCζ to PM. Although the lack of full protein structures of aPKC and Par-6 makes it difficult to conduct comprehensive structure-based experiments to further confirm the allosteric regulation of PSr by Par-6, our data strongly support a model that PB1/PB1 interaction between aPKC and Par-6 is both sufficient and necessary to allosterically displace the polybasic PSr from KD in aPKC, exposing PSr to electrostatic binding to the PM.

Binding of Par-6 is considered an essential step in regulating the kinase activity of aPKC, although whether Par-6 activates or inhibits aPKC remains unsettled and may well depend on additional regulators presented in different cell types (Hong, 2018). To investigate how Par-6 regulates PKCζ activity in vivo, we established aPKC kinase activity assays in HEK293 cells based on the loss of Lgl PM localization that serves as a sensitive, robust, and quantifiable readout for measuring aPKC phosphorylation of Lgl in live cells. When expressed alone in HEK293 cells, mammalian Lgl::GFP showed consistently strong and robust PM localization (Fig. 6 A) that was strongly reduced in cells expressing PKCζ but not kinase-dead PKCζ^K281W (Fig. 6 A). An anti-phospho-Lgl antibody confirmed the phosphorylation of Lgl in HEK293 cells coexpressing PKCζ (Fig. 6 C), suggesting that overexpressed PKCζ alone contains basal kinase activity sufficient to phosphorylate Lgl, consistent with the fact that in vitro purified aPKC has ∼10% of activated kinase activity (Zhang et al., 2014). To further confirm that the loss of Lgl PM localization is due to the phosphorylation by PKCζ, we generated a nonphosphorylatable Lgl^S6A::GFP in which all six conserved phospho-serines were mutated to alanine. As expected, Lgl^S6A::GFP remained on the PM when coexpressed with PKCζ (Fig. 6 A).

We then tested the kinase activity of PM-targeted aPKC/Par-6 complex. Strikingly, in HEK293 cells coexpressing Lgl::GFP, PKCζ::RFP, and Par-6::iRFP, all three proteins were strongly PM-localized (Fig. 6 A), and anti-phospho-Lgl antibody failed to detect phosphorylation on Lgl (Fig. 6 C). Similar results were also seen in cells expressing Lgl::GFP and PKCζ::RFP^-2A-Par-6::iRFP (Fig. 6 A). Thus, binding of Par-6 not only targets PKCζ to PM, but also appears to strongly inhibit its kinase activity. This apparent inhibition of PKCζ kinase activity by Par-6 is not due to the PM localization of PKCζ alone, as we made a Lyn11-PKCζ that contains the constitutive PM-binding domain Lyn11 and found that it also strongly delocalized Lgl::GFP to the cytosol and was efficiently inhibited by Par-6 (Fig. 6, B and C). We also compared the Par-6 inhibition on cytosolic PKCζ^KR8Q and PM-bound Lyn11- PKCζ^KR8Q. Both kinases strongly reduced the PM localization of Lgl and were similarly inhibited by Par-6 (Fig. 6, B and C), suggesting that PM localization of PKCζ is not required for the kinase inhibition by Par-6.

Curiously, in polarized MDCK cells, overexpressed PKCζ failed to dislocalize Lgl::GFP from PM (Fig. S4 B), suggesting that Lgl phosphorylation by aPKC is tightly controlled by additional mechanisms. Nonetheless, our results in nonpolarized HEK293 cells suggest that binding of Par-6 not only targets aPKC to the PM but also inhibits its kinase activity.

The PB1 domain of Par-6 alone (PB1^Par-6) was capable of targeting PKCζ to PM (Fig. 5 B), but it did not inhibit the phosphorylation of Lgl::GFP by PKCζ (Fig. 7 A), nor did Par-6ΔPDZ, in which the C-terminal PDZ domain was deleted (Fig. 7 A). Such results suggest that the PDZ domains or the C-terminus of Par-6 could be specifically required for inhibiting PKCζ kinase activity on Lgl (Fig. 7 A), which is consistent with the finding that overexpression of Par-6 C-terminus inhibits PKCɩ/λ activity in MDCK cells (Kim et al., 2007). The C-terminus of Par-6 interacts with multiple proteins, including activated Cdc42, which binds the Par-6 CRIB domain and moderately activates kinase activity of aPKC/Par-6 complex in vitro (Yamanaka et al., 2001). However, PM localization of Lgl::GFP remained high in HEK293 cells expressing Cdc42^CA, PKCζ, and Par-6 (Fig. 7 B), suggesting that Cdc42^CA is not sufficient to activate aPKC in our cell-based assays. Our results are consistent with genetic evidences that Drosophila Cdc42 is not required for aPKC to phosphorylate Lgl (Hutterer et al., 2004) and Baz (Walther and Pichaud, 2010) in vivo.

Besides Cdc42, apical polarity protein Crb also interacts with Par-6. Crb is a transmembrane protein, and the C-terminus of its intracellular domain is a PDZ-binding domain (PBD) that can bind the PDZ domain in Par-6 (Kempkens et al., 2006). We therefore investigated whether Crb could directly activate the kinase activity of aPKC/Par-6 complex through its interaction with the Par-6 PDZ domain. Par-6 became localized to the PM in HEK293 cells expressing membrane-bound Crb intracellular domain (Crb-intra) but not Crb-intra^ΔERLI, in which the C-terminal PBD was deleted, confirming that Par-6 binds to the PBD of Crb-intra (Fig. S4 A). Moreover, in HEK293 cells expressing aPKC and Par-6 together with Crb-intra but not Crb-intra^ΔERLI, Lgl::GFP was strongly reduced from the PM (Fig. 7 B). The loss of Lgl::GFP from PM in cells expressing Crb-intra, aPKC, and Par-6 is phosphorylation dependent, as nonphosphorylatable Lgl^S6A::GFP remained on PM in these cells (Fig. 7 B). Lgl::GFP also maintained PM localization in cells expressing Crb-intra, Par-6, and kinase-dead aPKC^K281W (Fig. 7 B). These results support a model in which direct interaction between Crb-intra and Par-6 in a PM-bound aPKC/Par-6 complex activates aPKC kinase activity.

Consistent with our cell-based assays, overexpression of Crb in Drosophila follicular cells and embryonic epithelial cells expanded Crb and DaPKC into basolateral PM and strongly delocalized Lgl::GFP, but not nonphosphorylatable Lgl^S5A::GFP, from basolateral PM (Fig. 8, A–C). The loss of Lgl::GFP from PM in Crb-overexpressing cells was completely reversed when DaPKC or Par-6 was also knocked down by RNAi (Fig. 8, E and F). Furthermore, the loss of PM Lgl::GFP in Crb-overexpressing cells is comparable to cells overexpressing DaPKC-ΔN, which is considered fully activated owing to the deletion of the N-terminus including PSr (Betschinger et al., 2003; Fig. 8 G). Thus, in Crb-overexpressing follicular cells, DaPKC is the key kinase phosphorylating Lgl and is likely highly activated. In contrast, in Drosophila crb^−/− embryos, DaPKC also extended to the basolateral PM but Lgl remained on PM (Fig. 8 D), suggesting that Crb is necessary to promote DaPKC phosphorylation on Lgl in embryonic epithelial cells in vivo.

In this study, we show that the PSr in aPKC is a typical polybasic domain capable of directly targeting aPKC to PM via its electrostatic interaction with negatively charged phosphoinositides PI4P and PIP₂ in PM. This is in contrast to the assumption that protein–protein interactions are solely responsible for localizing aPKC to PM or cell cortex. In addition, unlike phosphorylatable polybasic domains in Lgl, Numb, and Miranda, the polybasic PSr in aPKC has not been shown to be phosphorylatable. Instead, we report a novel example of a potential allosteric regulation of a polybasic domain for PM binding. As illustrated in Fig. 9, our data support a model in which the polybasic PSr in unbound aPKC is occluded by the KD from binding PM, whereas the binding of Par-6 to aPKC via PB1 domains induces potential conformational changes in aPKC that make the polybasic PSr accessible to PM binding.

Cdc42 plays an important role in mediating the PM targeting of aPKC/Par-6 in certain cell types such as C. elegans one-cell embryos, and we found that activated Cdc42 binds Par-6, which in turn can recruit nonpolybasic aPKC^KR8Q to the PM in HEK293 cells. However, our in vitro and in vivo studies suggest that electrostatic binding to PI4P and PIP₂ alone can be sufficient to localize aPKC/Par-6 complex to the PM. Our results are consistent with findings that removing Cdc42 in C. elegans late embryos or in Drosophila pupal epithelial cells does not severely disrupt aPKC and Par-6 localization (Georgiou et al., 2008; Zilberman et al., 2017). Whereas Cdc42 directly regulates aPKC/Par-6 PM targeting under certain cellular contexts, our results demonstrate a PM-targeting mechanism of aPKC/Par-6 that is mechanistically independent of Cdc42 or other aPKC/Par-6 interacting proteins.

Our hypoxia assay suggests that, at least in Drosophila follicular and embryonic epithelial cells, binding to PI4P and PIP₂ is likely the primary mechanism localizing aPKC/Par-6 to the PM in vivo. In these cells, the Par-6–dependent electrostatic binding to PM by PSr likely functions as the first step to localize aPKC/Par-6 to the PM, followed by further enrichment to specific membrane domains such as apical PM through protein–protein interactions with, for instance, apical Crb complex. Thus, while mechanisms governing the aPKC/Par-6 subcellular localization can be heavily cell type specific and multilayered, our results highlight that the electrostatic PM-binding property of aPKC needs to be taken into consideration when studying the protein regulators of aPKC/Par-6 subcellular localization.

Our studies also suggest potential new negative regulators that may act specifically at the electrostatic interaction to prevent aPKC from binding to the PM, by mechanisms such as masking the PSr, inhibiting the allosteric changes induced by Par-6, or sequestering the phospholipids in the PM. Moreover, it is notable that a PI4P- and PIP₂-dependent mechanism makes aPKC PM targeting vulnerable to stress conditions such as hypoxia and ATP inhibition that deplete these phospholipids in the PM (Dong et al., 2015). The acute and reversible loss of PM targeting of polybasic polarity proteins such as Lgl, aPKC, and Par-6 may have profound implications in epithelial cells for maintaining and restoring their apical-basal polarity when undergoing hypoxia, ischemia, and reperfusion.

Mechanisms coupling the aPKC kinase activity with its subcellular localization are essential for aPKC to phosphorylate targets at the right place and right time (Hong, 2018), but molecular details about these mechanisms remain largely elusive. Recent studies began to reveal exciting details on how aPKC/Par-6 kinase activity and subcellular localization can be regulated by the clustering of Par-3(Baz) and diffusive interactions with Cdc42 (Dickinson et al., 2017; Rodriguez et al., 2017), but these studies so far have been limited to the process of A-P polarization in worm one-cell embryos. Here we show that, in Drosophila epithelial cells and cultured mammalian cells, the electrostatic binding of aPKC to PM alone may provide a mechanism for Par-6 to play a pivotal role in coupling the PM targeting and control of kinase activity of aPKC (Fig. 9). It is possible that conformational changes in aPKC induced by Par-6 may not only expose PSr to PM binding but also allow the C-terminus of Par-6 to simultaneously inhibit aPKC’s KD. More experiments, however, are needed to further validate this hypothesis.

Our allosteric model is based on previous studies suggesting the displacement of PSr from aPKC upon binding of Par-6 (Graybill et al., 2012). However, in that study, Par-6 appears to activate aPKC in vitro and in Drosophila S2 cells. Such a discrepancy could be because S2 cells already express proteins capable of releasing the inhibition of Par-6 on aPKC. It should be noted that studies from multiple groups yielded conflicting results on whether Par-6 inhibits or activates aPKC kinase activity (Atwood et al., 2007; Chabu and Doe, 2008; Graybill et al., 2012; Lin et al., 2000; Yamanaka et al., 2001). Most of these studies relied on immunoprecipitated or reconstituted aPKC/Par-6 complex to measure the kinase activity in solution in vitro. Given that the majority of aPKC substrates are PM-bound polybasic domain–containing proteins, it might be critical to assay aPKC/Par-6 kinase activity in its PM-bound form in live cells.

In addition, it is conceivable that binding to PM may physically shield the polybasic domain in a target protein from being accessible to aPKC, as suggested by the increased resistance of membrane-bound Lgl to aPKC phosphorylation in vitro (Visco et al., 2016). This is consistent with our finding that phosphorylation of Lgl appears to be inhibited when both Lgl and aPKC/Par-6 complex are electrostatically attached to PM. On the other hand, the transient and dynamic nature of lipid binding by polybasic domains (Hammond et al., 2009) could also effectively enrich a local cytosolic pool of target proteins near PM, which works in favor of PM-bound aPKC to encounter its substrates. Such intricate relationships between PM targeting and aPKC phosphorylation of polybasic polarity proteins remain to be further explored.

Similar to cPKC and nPKC, the electrostatic binding of PSr to PM could potentiate the kinase activity of aPKC/Par-6 complex by further preventing PSr from autoinhibiting the KD (Rosse et al., 2010; Garg et al., 2014). This is supported by in vitro studies showing that Par-6 binding to aPKC inhibits its kinase activity but also potentiates its activation upon subsequent molecular events (Yamanaka et al., 2001). PM-bound aPKC/Par-6 complexes with inhibited but potentiated kinase activity are ideal targets for additional activation to fine-tune aPKC localization and kinase activity. Curiously, we found that Cdc42 did not activate PM-bound aPKC/Par-6 to phosphorylate Lgl in our cultured cell assays, but interaction between Par-6 and apical transmembrane protein Crb reversed Par-6’s inhibition on aPKC kinase activity. Supporting the role of Crb in enriching and activating aPKC/Par-6, Crb colocalizes with aPKC/Par-6 complex and is required for enriching the aPKC/Par-6 complex to the apical PM in fly embryonic epithelial cells and in many specialized membranes in Drosophila, including apical membranes of photoreceptors (Hong et al., 2003; Pellikka et al., 2002) and lumens in scolopale cell in chondotoral organs (Y. Hong, unpublished data). However, Crb is not required for apical enrichment of aPKC in follicular cells (Sherrard and Fehon, 2015); therefore, additional proteins besides Crb must be responsible for the apical enrichment and activation of aPKC/Par-6 in different types of epithelial cells.

It is of note that our studies are primarily based on cultured cells and limited cell types of Drosophila epithelial cells, and thus present a simplified model regarding the relationships between aPKC/Par-6 and its regulators. Our data do highlight that aPKC/Par-6 activity can be regulated by multiple regulators such as PM phosphoinositides, Cdc42, and Crb under different cellular and polarity contexts. Extrapolating our study to further investigate the role of aPKC in regulating cell polarity requires experiments to take into account the more complicated cellular contexts in different cell types.

The following Drosophila stocks were used in this study: knockout alleles of lgl^GX (lgl^KO) and crb^GX (crb^KO) and knock-in alleles of lgl::GFP and lgl^S5A::GFP were previously described (Dong et al., 2015; Huang et al., 2009). w; UAS-Crb; (gift from Dr. Eli Knust, Dresden, Germany), w par-6^Δ226 FRT^9-2/FM6 (gift from Dr. Jurgern Knoblich, Research Institute of Molecular Pathology, Vienna, Austria), y w; DaPKC^k06403/CyO (BL#10622), w; His2Av::mRFP (BL#23651), y w ubi-GFP^NLS FRT^9-2 (BL#5154), w; Act5C(FRT.CD2)-Gal4, UAS-RFP/TM3, Sb (BL#30558), w; FRT^G13 DaPKC^k06403/CyO; ubi-DaPKC::GFP/TM6, w; FRT^G13 DaPKC^k06403/CyO; ubi-DaPKC^KR8Q::GFP/TM6, w par-6::GFP hs-FLP par-6^Δ226 FRT^9-2/FM7C, w; lgl::GFP hs-FLP³⁸; Act5C(FRT.CD2)-Gal4, UAS-RFP/TM3, Sb, w; lgl::GFP UAS-Crb/CyO, w; lgl^S5A::GFP UAS-Crb/CyO, w; lgl^S5A::GFP/CyO; hs-FLP Act5C(FRT.CD2)-Gal4, UAS-RFP/TM3, w hs-FLP; FRT^G13 His2Av::mRFP, w par-6::GFP hs-FLP par-6^Δ226 FRT^9-2/FM7C; lgl^GX FRT^40A/CyO, w par-6::GFP hs-FLP par-6^Δ226 FRT^9-2/FM7C; ubi-RFP^NLS FRT^40A/CyO, w; lgl::GFP UAS-aPKC-CAAX (UAS-aPKC-CAAX, gift from Dr. Sonsoles Campuzano, Univerisidad Autonoma de Madrid, Spain), w UASp>mRFP::FKBP-5Ptase (FKBP-INPP5E) and w; UASp>Lck-FRB::CFP, gifts from Dr. De Renzis, European Molecular Biology Laboratory, Heidelberg, Germany (Reversi et al., 2014), w; αTub67C-Gal4^V2H; αTub67C-Gal4^V37 (Mat-Gal4, gift from Dr. Mark Peifer, University of North Carolina at Chapel Hill, NC; Schaefer et al., 2018), w; GFP::aPKC/CyO (gift from Dr. Daniel St Johnston, The Gurdon Institute, Cambridge, UK), and w; UAS-Cdc42^CA (BL#4854), UAS-cdc42-RNAi, w; (BL#29004), y, v; UAS-DaPKC-RNAi (BL#25946), y, v; UAS-Par-6-RNAi (BL#39010), y, v; UAS-crb-RNAi (BL#38373), w; UAS-DaPKC-CAAX, w; UAS-DaPKC-CAAX, UAS-Par-6; w;; UAS-Par-6 (gifts from Dr. Tony Harris, University of Toronto, Toronto, Canada).

Transgenic flies of ubi-DaPKC::GFP, ubi-aPKC^KR8Q::GFP, and par-6::GFP were generated by phiC31-mediated integration protocol (Huang et al., 2009). attP^VK00033 (BL#24871) stock was used to integrate ubi-DaPKC::GFP and ubi-DaPKC^KR8Q::GFP constructs into the third chromosome, and attP^ZH2A (BL#24480) stock was used to integrate par-6::GFP constructs into the X chromosome. par-6::GFP was further recombined with par-6^Δ226 null allele to generate w par-6::GFP par-6^Δ226, of which homozygote females and hemizygote males are fully viable and fertile, indicating a complete rescue of par-6^Δ226 by par-6::GFP. Drosophila cultures and genetic crosses were performed at 25°C. Detailed information about the strains from Bloomington Stock Center can be found in FlyBase.

To make ubi-aPKC::GFP, ubiquitin promoter (1,872 bp) was PCR-amplified from plasmid pWUM6 (gift from Dr. Jeff Sekelsky, University of North Carolina at Chapel Hill, NC) using primers 5′-AGT​GTC​GAA​TTC​CGC​GCA​GAT​CGC​CGA​TGG​GC-3′ and 5′-CTG​GAC​GCG​GCC​GCG​GTG​GAT​TAT​TCT​GCG​GG-3′ and inserted into pGE-attB vector (Huang et al., 2009) to generate vector pGU. DNA fragments encoding aPKC::GFP and aPKC^KR8Q::GFP were then inserted into pGU vector. To make par-6::GFP, a 4.3-kb par-6 genome DNA including 1-kbp upstream and 250-bp downstream sequences was PCR-amplified from Drosophila genomic DNA using primers 5′-ATG​CGG​CCG​CGC​TCT​TCG​GCT​CTC​GGA​TAG​TCG-3′ and 5′-GAC​GCG​TGA​TTA​AGG​CCC​GGC​TAA​TG-3′, subcloned into pGE-attB vector. AvrII enzyme site was added before stop code for GFP insertion. More details about DNA constructs used in this study are in Table S1. NCBI: Par-6 (human), NP_058644.1; Lgl (mouse), NP_001152877.1. Plasmids containing PKCζ and PKCι coding sequences were gifts from Dr. Jane Wang (University of Pittsburgh, Pittsburgh, PA)

Embryos and dissected ovaries were imaged according to a previously described protocol (Dong et al., 2015; Huang et al., 2011). Embryos were collected overnight at 25°C. Ovaries from adult females several days old were dissected in halocarbon oil (#95). Dechorinated 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–mm cut-through window. After placing the coverslip on top, a membrane at the bottom ensures sufficient air exchange to samples during the imaging session. The slide was then mounted in an air-tight microchamber (custom made) for live imaging under a confocal microscope. Oxygen levels inside the chamber were controlled by flow of either air or custom O₂/N₂ gas at the rate of ∼1–5 cc/s. Images were captured at room temperature (25°C) on a Leica TCS-NT confocal microscope (PL APO 40× oil objective, NA 1.25) by Leica TCS-NT software; an Olympus FV1000 confocal microscope (40× Uplan FL N oil objective, NA 1.3) by Olympus FV10-ASW software; or a Nikon A1 confocal microscope (Plan Fluo 40′ oil objective, NA 1.3) by NIS-Elements AR software. Images were further processed in ImageJ (National Institutes of Health) and Adobe Photoshop.

GST-PKCζ (gift from Dr. Ricardo Biondi, Instituto de Investigación en Biomedicina de Buenos Aires, Buenos Aires, Argentina; Zhang et al., 2014), GST-PKCζ^KR8Q, and Par-6::Hisx6 were expressed in Expi293F cells via transfection using the ExpiFectamine Transfection Kit (Gibco). Approximately 22.5 × 10⁷ suspension-adapted Expi293F cells were diluted in Expi293 Expression Medium to a final volume of 76.5 ml and cultured in a 37°C shaker (Eppendorf New Brunswick S41i). 90 µg of plasmid DNA was mixed with 240 µl Expifectamine reagent into a total final volume of 9 ml Opti-MEM medium and incubated for 25 min. The mixture was then added to Expi293F cells to a total volume of 85.5 ml. After 18 h of incubation in the 37°C shaker, 450 µl of Transfection Enhancer I and 4.5 ml of Transfection Enhancer II were added to the cells. Cells were harvested after another 18 h of incubation, resuspended in 50 ml PBS buffer, and homogenized with an ice-cold tight-fitting Dounce homogenizer. To purify PKCζ/Par-6 or PKCζ^KR8Q/Par-6 complex, cells expressing Par-6::His6 were mixed with an equal volume of cells expressing GST-PKCζ or GST-PKCζ^KR8Q and homogenized. After centrifuging, supernatants from lysates were collected for purification by Pierce Glutathione Agarose. Bacteria BL21 was used to express and purify GST and GST fusion of aPKC PSr proteins as previously described (Dong et al., 2015).

Liposomal binding assays were performed as described (Kim et al., 2008). To prepare liposomes, a lipid mixture of 37.5% phosphatidylcholine (840051C), 10% phosphatidylserine (840032C), 37.5% phosphatidylethanolamine (840021C), 10% cholesterol (700000P), and 5% PIP₂ (840046X) or PI4P (840045X, all lipids purchased from Avanti Polar Lipids) was dried and resuspended to a final concentration of 1 mg/ml of total phospholipids in Hepes buffer. After 30-min sonication, formed liposomes were harvested at 16,000 g for 10 min and resuspended in binding buffer (Hepes 20 mM, pH 7.4, KCl 120 mM, NaCl 20 mM, EGTA 1 mM, MgCl 1 mM, and BSA 1 mg/ml). In each liposome-binding assay, ∼0.1 µg of purified protein or protein complex was mixed with 50 µl of liposome suspension. After 15-min incubation at room temperature, liposomes were pelleted at 16,000 g for 10 min and were analyzed by Western blot to detect cosediments of target proteins.

HEK293 cells were cultured in glass-bottom dishes (In Vitro Scientific) and transfected with DNA using X-treme Gene 9 DNA transfection reagent (6365787001; Sigma-Aldrich). After 24–40 h of transfection, cells were mounted and imaged on an Olympus FV1000 confocal microscope (40× Uplan FL N oil objective, NA 1.3) by Olympus FV10-ASW software or a Nikon A1 confocal microscope (Plan Fluo 40× oil objective, NA 1.3) by NIS-Elements AR software. For images to be used for quantification, parameters were carefully adjusted to ensure no or minimum overexposure. In addition, when necessary, a fluorescent PM dye (CellMask DeepRed Plasma Membrane Stain, C10046; Thermo Fisher Scientific) was added to cell culture before live imaging to help visualize the PM for later quantifications. MDCK II cells were cultured in MEMα medium containing 10% FBS (Gibco) and 1% penicillin and streptomycin (Gibco). For growing polarized monolayers, MDCK cells were cultured on 0.4-µm Transwell filters (Corning) for 3 d. Transfection assays were performed on day 4 using Lipofectamine 2000 (Thermo Fisher Scientific), and cells were imaged on day 5.

PM localization were measured in ImageJ by custom macro scripts. For each image, PM masks were generated by an à trous wavelet decomposition method (Hammond et al., 2014; Olivo-Marin, 2002) based on the channel that contained either PM-localized proteins or fluorescent PM dyes. Cytosol masks were generated by segmentation using a threshold based on the mean pixel value of the region of interest (ROI). Cells expressing all transfected fluorescent proteins were selected for measurement by drawing ROIs around each cell. Because of the use of computer-generated PM and cytosol masks, the exact shape of the ROI was not critical except that PM segments in contact with neighboring expressing cells were avoided. Custom macros were used to automatically measure PM and cytosolic intensities of each fluorescent protein in each cell marked by an ROI in the sample image. Backgrounds were autodetected by the macro based on the minimal pixel value of the whole image. The PM localization index for each fluorescent protein was calculated by the macro as the ratio of (PM − background)/(cytosol − background). Data were further processed in Excel and visualized and analyzed in GraphPad Prism.

For Lgl-phosphorylation assay, HEK293 cells were cultured in DMEM supplemented with 10% FBS. 24 h after transient transfection, cells were directly lysed in SDS loading buffer, and equal volumes of cell lysates were resolved in 12% SDS-PAGE. Proteins were detected by Western blot using antibody chicken anti-GFP, 1:5,000 (GFP-1020; Aves Lab); rabbit anti-phospho-mLgl, 1:1,000 (AP2198a; Abgent); rabbit anti-PKC, 1:5,000 (pan-aPKC antibody, Santa Cruz, Sc-216); PKCζ-specific antibody (C24E6); rabbit mAb (Cell Signaling, 9368); or mouse anti-Flag 1:5,000 (F3165; Sigma-Aldrich). For immunoprecipitation experiments, cells were lysed in radioimmunoprecipitation assay buffer (25 mM Tris-HCl, pH 7.4, 150 mM NaCl, 0.1% SDS, 0.5% sodium deoxycholate, and 1% Triton X-100). Protein G-Sepharose beads were incubated with homemade, affinity-purified rabbit anti-GFP antibody (Huang et al., 2009) for 1 h followed by incubation with the equal amount of each lysate for 1 h. The beads were washed and boiled with SDS loading buffer. Supernatants were detected by Western blot.

The procedure has been described in detail previously (Hammond et al., 2014). In brief, HEK293 cells cultured in 35-mm glass-bottom dishes (In Vitro Scientific) were transiently transfected with 1 µg total DNA, which included the Lyn11-FRB::CFP recruiter, FKBP12 phosphatases (Hammond et al., 2012), PKCζ::GFP, and Par-6::iRFP as indicated. After 22–26 h, cells were imaged in Fluoro-Brite medium (Life Technologies) using a Nikon A1R confocal laser scanning microscope though a 100×, NA/1.45 plan apochromatic objective lens. Time-lapse imaging started 2 min before bath addition of 1 µM rapamycin. CFP::Lyn11-FRB images were used to generate binary masks to define PM. PM localization of each reporter was then calculated from the ratio of fluorescence within the PM to the whole cell and was expressed relative to the average before rapamycin addition (Hammond et al., 2014).

Young females of w UASp>mRFP::FKBP-5Ptase / w par-6::GFP par-6^Δ226; hs-FLP Act5C(FRT.CD2)-Gal4 UAS-RFP/UASp>Lck-FRB::CFP or w UASp>mRFP::FKBP-5Ptase / +; lgl::GFP hs-FLP /+; Act5C(FRT.CD2)-Gal4 UAS-RFP / UASp>Lck-FRB::CFP were heat-shocked at 37°C for 1 h. Ovaries were dissected 4 d later in 1× PBS, mounted in a drop of 20 µl Schneider’s medium containing 10 µM rapamycin on a gas-permeable slide, and imaged live, as previously described (Huang et al., 2011; Dong et al., 2015). Treated lgl::GFP ovaries served as a positive control (Dong et al., 2015).

Mutant follicular cell clones of lgl^KO or aPKC^k06403 were generated by the routine FLP/FRT technique. Young females were heat-shocked at 37°C for 1 h, and their ovaries were dissected 3 d later.

Immunostaining of embryos and adult ovaries was performed as described (Huang et al., 2009). Primary antibodies: rabbit anti-GFP (Huang et al., 2009) 1:1,500; chicken anti-GFP (Aves Lab) 1:1,000; rabbit anti-Lgl (d-300, Santa Cruz) 1:200; and rabbit anti-aPKC (Santa Cruz) 1:1,000. Secondary antibodies: Cy2-, Cy3-, or Cy5-conjugated goat anti-rabbit IgG, anti-chicken IgG, goat anti-rat IgG, goat anti-mouse IgG, and goat anti-guinea pig IgG (Jackson ImmunoResearch Lab), all at 1:400. Images were collected on an Olympus FV1000 confocal microscope and processed in Adobe Photoshop and ImageJ.

Fig. 1 C: w; FRT^G13 DaPKC^k06403 / CyO; ubi-DaPKC::GFP / TM6; w; FRT^G13 DaPKC^k06403 / CyO; ubi-DaPKC^KR8Q::GFP / TM6. Fig. 1 D: w; FRT^G13 DaPKC^k06403 / CyO; ubi-DaPKC::GFP / TM6; w par-6::GFP hs-FLP par-6^Δ226 FRT^9-2. Fig. 3 B: Top: w; GFP::aPKC / UAS-Cdc42^CA; hs-FLP Act5C(FRT.CD2)-Gal4 UAS-RFP / +. Bottom: w; UAS-Cdc42^CA / +; hs-FLP Act5C(FRT.CD2)-Gal4 UAS-RFP / ubi-DaPKC^KR8Q::GFP. Fig. 3 C: UAS-cdc42-RNAi, w / w; GFP::aPKC / +; hs-FLP Act5C(FRT.CD2)-Gal4 UAS-RFP / +. Fig. 8, A and B: w; lgl::GFP hs-FLP/ lgl::GFP UAS-crb; Act5C(FRT.CD2)-Gal4 UAS-RFP / +; w; lgl^S5A::GFP UAS-Crb / +; hs-FLP Act5C(FRT.CD2)-Gal4 UAS-RFP / +. Fig. 8 C: w; lgl::GFP UAS-Crb / +; +/+; w; lgl::GFP UAS-Crb / αTub67C-Gal4^V2H; αTub67C-Gal4^V37/ +; w; lgl^S5A::GFP UAS-Crb / αTub67C-Gal4^V2H; αTub67C-Gal4^V37/ +. Fig. 8 D: w; lgl::GFP; w; lgl::GFP; crb^KO/ crb^KO. Fig. 8 E: w; lgl::GFP UAS-Crb / +; hs-FLP Act5C(FRT.CD2)-Gal4 UAS-RFP / UAS-DaPKC-RNAi. Fig. 8 F: w; lgl::GFP UAS-Crb / UAS-par6-RNAi; hs-FLP Act5C(FRT.CD2)-Gal4 UAS-RFP / +. Fig. 8 G: w; lgl::GFP hs-FLP / +; Act5C(FRT.CD2)-Gal4 UAS-RFP / UAS-DaPKCΔN. Fig. S1 A: w / w hs-FLP; FRT^G13 DaPKC^k06403/ FRT^G13 His2Av::mRFP; ubi-DaPKC::GFP/+; w / w hs-FLP; FRT^G13 DaPKC^k06403/ FRT^G13 His2Av::mRFP; ubi-DaPKC^KR8Q::GFP/+. Fig. S1 B: w par-6::GFP hs-FLP par-6^Δ226 FRT^9-2/FM7C; w; FRT^G13 DaPKC^k06403/CyO; ubi-aPKC::GFP/TM6. Fig. S1 C: w par-6::GFP hs-FLP par-6^Δ226 FRT^9-2/ w; lgl^KO FRT^40A/ ubi-RFP^NLS FRT^40A. Fig. S3, A and B: w UAS-mRFP-FKBP-5′Ptas / par-6::GFP hs-FLP par-6^Δ226 FRT^9-2; +/+; Act5C(FRT.CD2)- UAS-RFP^NLS/ UAS-lck-FRB::CFP.

Fig. S1 shows that the polybasic PSr was required for PM targeting of Drosophila aPKC, and both Par-6::GFP and DaPKC::GFP showed hypoxia-sensitive PM localization in embryonic epithelia. In addition, hypoxia-sensitive PM localization of Par-6::GFP was not affected by the loss of Lgl. Fig. S2 shows that in MCF7, COS7, and polarized MDCK cells, PM targeting of PKCζ required coexpression of Par-6. Fig. S3 shows that acute depletion of PIP₂ in Drosophila follicular cells did not strongly inhibit the PM localization of Par-6::GFP. Fig. S4 shows that Crb-intra was capable of recruiting Par-6 to PM in HEK293 cells, and that Lgl::GFP PM localization in polarized MDCK cells is resistant to the overexpression of PKCζ::RFP. Videos 1 and 2 show the acute and reversible loss of DaPKC::GFP and Par-6::GFP from PM under hypoxia in live Drosophila follicular cells. Table S1 lists the details about DNA constructs.

We are grateful to Drs. Jane Wang, Gerald Apodoca, Mark Peifer, Eli Knust, Thomas Weide, David Bilder, Tony Harris, Daniel St Johnston, Jean Wilson, and Ricardo Biondi for reagents and fly stocks; Lu Jiang and Kriti Sanghi for technical assistance; Dr. Marijn Ford for help on protein purification; Dr. Simon Watkins and University of Pittsburgh Medical School Center for Biological Imaging for generous imaging and microscopy support; Bloomington Stock Center for fly stocks; and Developmental Studies Hybridoma Bank for antibodies.

This work was supported by National Institutes of Health, National Center for Research Resources (R21RR024869 to Y. Hong), National Institutes of Health, National Institute of General Medical Sciences (R01GM086423 and R01GM121534 to Y. Hong), and National Institutes of Health (1R35GM119412-01 to G.R. Hammond). The University of Pittsburgh Medical School Center for Biological Imaging is supported by National Institutes of Health grant 1S10OD019973-01.

The authors declare no competing financial interests.

Author contributions: Conceptualization: Y. Hong, W. Dong, G.R. Hammond; Investigation: W. Dong, J. Lu, X. Zhang, Y. Wu, K. Lettieri, G.R. Hammond, Y. Hong; Writing–Review & Editing: Y. Hong, W. Dong, G.R. Hammond; Funding acquisition: Y. Hong, G.R. Hammond; Supervision: Y. Hong.