The F-box protein Slmb restricts the activity of aPKC to polarize epithelial cells

Polarization is a fundamental feature of animal cells, from newly fertilized zygotes to dividing stem cells to homeostatic epithelia. This common feature is controlled by a conserved set of regulators, which segregate the single plasma membrane into several discrete domains. The most broadly used polarity regulators comprise the Par complex, consisting of the PDZ-containing scaffolds Par-3 and Par-6, which associate with Cdc42-GTP and the atypical protein kinase (aPKC) (Goldstein and Macara, 2007; St Johnston and Ahringer, 2010). In the C. elegans zygote and the Drosophila oocyte, these proteins localize to and specify the anterior cortex. In most epithelial cells and neural stem cells, they localize to and specify the apical plasma membrane, and in migrating cells they define and act at the leading edge. The Par complex thus serves as a ‘master regulator’ for many types of cell polarity.

To achieve and maintain polarity, the Par complex must be restrained to distinguish a complementary membrane domain. In contrast to the pre-eminent role of the Par complex, multiple protein modules that limit Par activity have been identified in different contexts (Tepass, 2012). In the C. elegans zygote, the protein kinases PAR-1 and PAR-4 act downstream of the RING finger protein PAR-2 to antagonize Par localization and define the posterior cortex (St Johnston and Ahringer, 2010; Zonies et al., 2010). Par-1 and Par-4 (Lkb1 – FlyBase) are also key regulators of fly oocyte polarization, but often have less central roles in other polarized cell types (Haack et al., 2013; Partanen et al., 2013). Instead, in many of these tissues a second group of proteins, the Scribble (Scrib) module, acts to restrict the Par complex. In the Scrib module, Scrib and Dlg are basolaterally localized PDZ-containing scaffolds that regulate Lgl, a syntaxin- and myosin-binding protein that can directly antagonize aPKC (Bilder, 2004; Elsum et al., 2012). Yet another module, the Yurt/Coracle (Cor) complex, specifies the basolateral domain in mid-stage Drosophila embryos and zebrafish photoreceptors (Laprise and Tepass, 2011). Rac and PI3 kinase also play a role at this stage (Chartier et al., 2011). Further, in fly epithelia but not neuroblasts, AP-2-mediated endocytosis restricts apical polarity regulators to their appropriate surface; endocytosis also plays a crucial role in polarization of the worm zygote (Halbsgut et al., 2011; Shivas et al., 2010). The mechanisms by which this diverse set of proteins – which we will call Par or apical antagonists – negatively regulate the Par complex is an active field of investigation. In none of these cases is the mechanism well understood, nor how they coordinate with each other.

Our incomplete knowledge of the mechanisms of Par antagonists raises the possibility that additional regulators of basolateral polarity remain to be identified. Here we report that strong mutations in the F-box protein Slmb, a substrate adaptor for SCF E3 ubiquitin ligases, result in excess Par complex activity in Drosophila imaginal discs, thereby expanding the apical membrane domain. Our results indicate that Slmb-mediated protein degradation acts in parallel to the Scrib module to oppose aPKC activity and thus specify the epithelial basolateral membrane.

To identify new regulators of basolateral polarity, we analyzed mutants isolated in a genetic screen for Drosophila tumor suppressor genes (TSGs). The screen utilized mitotic recombination to generate imaginal discs predominantly populated by homozygous mutant cells growing in an otherwise heterozygous larva. Mutations in a small set of genes cause larval or pupal lethality in this context; many of these show a set of tumor-like phenotypes collectively called ‘neoplastic’ (Menut et al., 2007). Discs mutant for one uncharacterized complementation group, MENE(3R)-B, show multiple hallmarks of neoplastic transformation. Monolayered organization is lost, disc size is dysregulated, F-actin levels are elevated and differentiation is prevented (Fig. 1A-F). In addition, Matrix metalloproteinase 1 (Mmp1), a mediator of tissue invasion, is upregulated (Fig. 1A,B). These phenotypes closely resemble those of the Scrib module, suggesting that MENE(3R)-B identifies a gene with similar function.

MENE(3R)-B alleles fail to complement mutants in slmb, which encodes an F-box and WD40 repeat protein homologous to vertebrate β-TrCP (Btrc) that functions as a specificity factor in a Skp–Cullin–F-box (SCF) E3 ubiquitin ligase complex (Frescas and Pagano, 2008; Jiang and Struhl, 1998; Theodosiou et al., 1998). Sequencing identified coding region lesions, including early and late truncations as well as missense mutations, in all six alleles (Fig. 1G), and a slmb transgene rescues the neoplastic phenotype (supplementary material Fig. S1). The existing alleles slmb¹ and slmb² have been widely used, and epithelial organization phenotypes have not been reported. Sequencing revealed that these contain missense mutations in the fifth and seventh WD40 domains (Fig. 1G), indicating that both may be hypomorphic. We confirmed that slmb¹ and slmb² mutant discs show no and only a limited degree of neoplastic transformation, respectively (supplementary material Fig. S1). However, discs predominantly mutant for the deletion allele slmb⁸ (Milétich and Limbourg-Bouchon, 2000) show neoplasia, confirming that this phenotype is induced only by strong alleles. Null mutations in Roc1a, which encodes a frequent component of the SCF^Slmb complex (Noureddine et al., 2002), also show neoplasia (supplementary material Fig. S1). These data demonstrate that slmb functions as a new neoplastic TSG, and suggest that it does so via its role in the SCF^Slmb E3 ligase.

Known neoplastic TSGs regulate epithelial polarity. We therefore analyzed slmb tissue with markers for polarized membrane domains. In wild-type (WT) imaginal epithelia, the transmembrane proteins Cadherin 87 (Cad87) and Fasciclin III (FasIII) occupy complementary apical and basolateral membrane domains. In slmb cells, Cad87 is distributed ectopically around the cell circumference in discontinuous domains that sometimes overlap with FasIII (Fig. 2A,B). Similar effects are seen with the polarized peripheral membrane proteins Bazooka (Baz; Drosophila Par-3) and Cor (Fig. 2C,D). Expanded apical domains of slmb tissue resemble those of Scrib module mutants (Fig. 2E), and indicate that Slmb also acts as an apical antagonist.

To explore the breadth of Slmb function in cell polarity, we attempted to generate clones of strong alleles in other tissues. We were generally unable to recover clones in the follicle cell epithelium, and females carrying germline clones failed to produce eggs, preventing analysis of embryonic epithelia (data not shown). Clones in the larval central nervous system showed defective optic lobe organization. Intriguingly, clones derived from type I neuroblasts frequently contained multiple neuroblast-like cells, suggesting that Slmb might also be required for asymmetric cell division (supplementary material Fig. S2) (Li et al., 2014).

Of the known Par antagonists, only the Scrib module and endocytic components have been shown to strongly regulate imaginal basolateral polarity, raising the possibility that slmb might act primarily via one of these pathways. We first asked whether either Scrib or endocytic components promote Slmb-mediated protein degradation. However, Armadillo (Arm) levels (Jiang and Struhl, 1998) are not increased in scrib or AP-2σ depleted tissue (Fig. 3A-C), nor was there evidence for misregulation of other Slmb targets (supplementary material Fig. S3). We next asked whether Slmb might act through either the Scrib module or endocytic regulators to restrain apical polarity. To test whether Slmb interferes with AP-2-mediated endocytosis, we analyzed Notch trafficking and used a lysosomal inhibitor to monitor accumulation in the endolysosomal pathway (Windler and Bilder, 2010). Endocytic mutants prevent this accumulation (Fig. 3E), but in slmb, as in WT and Scrib module mutant discs, Notch is internalized and trafficked appropriately to endolysosomal compartments (Fig. 3D,F,G). Consistent with a lack of involvement in endocytic regulation, heterozygosity for slmb does not enhance a weak avalanche (Syntaxin 7) RNAi phenotype, which is sensitive to the dosage of endocytic regulators of polarity (Morrison et al., 2008). Taken together, these data fail to support a general endocytic role for Slmb.

To test whether Slmb might directly influence Scrib module activity, we examined protein localization. A distinctive feature of scrib and dlg mutants is that, although many proteins are mispolarized, Dlg is specifically lost from the plasma membrane of scrib mutants, as is Scrib in dlg mutants (Bilder et al., 2000). By contrast, examination of slmb mutant cells reveals that both Scrib and Dlg retain tight, albeit dysregulated, cortical localization (Fig. 3H-M). Additionally, heterozygosity for slmb does not enhance weak lgl-RNAi nor lgl or dlg hypomorphic phenotypes, all of which are sensitive to the dosage of Scrib module components (H. A. Morrison and B. D. Bunker, PhD theses, University of California Berkeley, 2010). Thus, despite the many phenotypic similarities, the failure to control Scrib and Dlg membrane recruitment and the lack of genetic interaction suggest that Slmb regulates polarity in parallel to the Scrib module.

The strong neoplastic phenotype seen in slmb tissue points to the existence of a polarity-regulating substrate, the levels of which must be controlled. We therefore examined known Slmb substrates to see whether any could account for this phenotype (supplementary material Fig. S4). Both Arm and Cubitus interruptus (Ci) are subject to proteolytic regulation by Slmb. Cells individually expressing or co-expressing active, non-degradable forms of Arm and Ci display a degree of hyperplastic overgrowth, consistent with known roles in the imaginal disc, but retain normal polarity and tissue architecture and do not upregulate Mmp1 (supplementary material Fig. S4A). Overexpression of stabilized Plk4 (Sak kinase – FlyBase) and Cap-H2 (Buster et al., 2013; Rogers et al., 2009) caused no growth or polarity phenotypes (supplementary material Fig. S4B,C). Overexpression of a stabilized form of the polarity kinase Par-1, recently shown to be an Slmb substrate at Drosophila synapses (Lee et al., 2012), also failed to phenocopy slmb loss. This suggests that an unidentified Slmb substrate normally regulates epithelial organization in imaginal discs.

An attractive candidate target of Slmb-mediated polarity regulation is the Par complex component aPKC. Overexpression of activated aPKC is sufficient to expand the apical domain and confer neoplastic phenotypes similar to those of slmb tissue (Fig. 4O) (Eder et al., 2005). Intriguingly, one predicted isoform of aPKC (aPKC-G) contains two Slmb-binding degrons and, when expressed in S2 cells, is degraded in an slmb-dependent manner (supplementary material Fig. S4). However, aPKC-G transcripts are present at very low levels in L3 discs, and expression of a degron-lacking aPKC-G had no effect on disc polarity or growth (supplementary material Fig. S4). We then used an antibody directed against a shared protein region to analyze total aPKC and found that it is mispolarized in slmb tissue and more widely distributed around the plasma membrane (Fig. 4A,B). However, aPKC levels are not obviously elevated, as assessed by immunohistochemistry, when compared with neighboring WT cells, in contrast to the evident elevation seen with Arm (Fig. 3A). We quantitated western blots of disc lysates which indicated a modest but not significant elevation of aPKC in slmb versus WT (Fig. 4C). These data suggest that, although aPKC is mislocalized in slmb cells, it is unlikely to be a target of slmb-mediated degradation.

Despite the absence of elevated levels of aPKC, we found multiple signs of increased aPKC activity in slmb discs. Excessive aPKC activity drives neoplastic overgrowth in part through upregulation of JAK-STAT pathway ligands, mediated by the Yorkie (Yki) transcription factor (Doggett et al., 2011; Robinson and Moberg, 2011; Sun and Irvine, 2011). slmb discs show robust activation of a STAT signaling reporter, and their neoplastic phenotype is sensitive to the dosage of yki (Fig. 4D-G). A second aPKC-regulated process is seen upon overexpression of the Crb intracellular domain in photoreceptors (Fig. 4H,I) (Tanentzapf and Tepass, 2003); the resulting morphogenetic defects are dependent on aPKC (Fig. 4J). Although heterozygosity for slmb does not enhance endocytic or Scrib module phenotypes, it does robustly enhance Crb overexpression (Fig. 4K). Finally, elevated aPKC activity is sufficient to induce trafficking defects of the retromer-dependent transmembrane cargo Wntless (Eaton, 2008), leading to a distinctive subcortical trapping phenotype (Fig. 4L,N; plasma membrane localization, 73±12% for WT versus 52±13% for aPKC^Act) (de Vreede et al., 2014); slmb mutant tissue phenocopies this trapping (Fig. 4M; plasma membrane localization, 44±25%; P<2×10⁻⁵ versus WT, P=0.14 versus aPKC^Act).

To directly test the functional involvement of hyperactive aPKC, we reduced aPKC activity in slmb cells using a weak dominant-negative construct (aPKC^DN) that has no effect on WT cells but can suppress phenotypes driven by elevated aPKC activity (Sotillos et al., 2004). Strikingly, expression of aPKC^DN in slmb clones strongly suppressed tumorous growth (Fig. 4P-S). The resultant clones were smaller than WT clones, suggesting that aPKC activity also promotes slmb survival in this context, perhaps because tumorous slmb cells are ‘addicted’ to oncogenic aPKC, the excess activity of which allows them to survive in the presence of other misregulated slmb substrates. This reliance demonstrates a specific requirement for aPKC in slmb tissue and, along with the above results, reveal that Slmb acts as a negative regulator of aPKC activity.

Here, we extend the mechanisms involved in epithelial polarity to include a new function: targeted protein degradation. Targeted degradation can create spatial asymmetries in protein distributions, and there is precedent for roles of E3 ubiquitin ligases, including SCF^Slmb (Li et al., 2014; Morais-de-Sá et al., 2013), in polarizing different aspects of cells. The involvement of Slmb in Drosophila apicobasal polarity has gone unnoticed due to the previous use of hypomorphic alleles. The strong alleles described here display potent expansion of the apical pole of imaginal epithelia, demonstrating that Slmb is a new polarity regulator that functions to restrict the apical domain.

Loss of Slmb phenocopies the polarity defects associated with mutations in two classes of ‘apical antagonists’: the Scrib module of core polarity regulators, and endocytic regulators that control trafficking through the early endosome. Despite the similar polarity defects, slmb mutations do not alter endolysosomal cargo traffic, nor do they display protein recruitment defects characteristic of Scrib module mutants; furthermore, no genetic interactions are seen with either pathway. Nevertheless, the downstream consequences of polarity misregulation – including tumor-like transformation and the upregulation of specific target genes – are again shared between slmb and the other apical antagonists, and, moreover, slmb and Scrib module mutant cells share a distinctive trafficking defect associated with elevated aPKC activity. We therefore suggest that Slmb acts in parallel to the Scrib module to antagonize the Par complex and other apical regulators.

The role for Slmb defined here points to the existence of an apical polarity-regulating protein substrate, the levels of which must be controlled. We have ruled out a number of validated Slmb substrates as the relevant target. Bioinformatic scans of Drosophila proteins for Slmb degron sequences suggest other candidates, including Expanded (Ex), but overexpression of Ex is not sufficient to induce polarity defects resembling those of slmb (Blaumueller and Mlodzik, 2000; Fernández et al., 2011). Although we cannot rule out a contribution from the elevation of multiple substrates, slmb-like polarity phenotypes can be induced by the elevated activity of individual proteins, including Crb or aPKC. Despite evidence that aPKC undergoes ubiquitin-mediated degradation in embryos (Colosimo et al., 2009), neither aPKC nor Crb levels appear to be controlled by Slmb-mediated degradation in imaginal discs. Nevertheless, our data together suggest that the substrate of Slmb in polarity regulation will function as a positive regulator of aPKC-driven outcomes.

Our demonstration that Slmb limits aPKC activity to distinguish the epithelial basolateral domain reveals intriguing parallels to polarization of the worm zygote. In this context, Par-2 is the primary antagonist that restricts aPKC/Par activity, while Lgl homologs function in a parallel, redundant role. Par-2 contains a RING finger domain that is characteristic of single-subunit E3 ligases, but Par-2 homologs have not been identified outside of nematodes, Par-2 does not affect aPKC/Par levels, and a degraded substrate in polarity regulation has yet to be identified (St Johnston and Ahringer, 2010; Zonies et al., 2010). The discovery of a Drosophila E3 ligase with a similar function to Par-2 raises the possibility of a conserved molecular logic to polarity in these two paradigmatic systems; determination of the relevant substrate will shed further light on this question.

Predominantly mutant imaginal discs were generated as described previously (Menut et al., 2007). slmb mutant images represent slmb^UU11 unless otherwise specified. For further details of fly stocks and genetics and the generation of mutants and clones, see supplementary Materials and Methods.

Western blots loaded with equivalent protein concentrations were probed with anti-β-tubulin (E7, Developmental Studies Hybridoma Bank) and anti-aPKC (sc-216, Santa Cruz Biotechnology). Two biological replicates for each of six WT and four mutant technical replicates were quantitated.

Wandering L3 imaginal discs, larval CNS and ovaries were dissected in PBS on ice, fixed in 4% formaldehyde in PBS for 20 min at room temperature and stained using standard procedures (Bilder and Perrimon, 2000). Primary and secondary antibodies and stains are described in the supplementary Materials and Methods. Fluorescent tissues were mounted in SlowFade (Molecular Probes). Images are single sections taken with Leica TCS or Zeiss 700 confocal microscopes except for adult heads, which were frozen, mounted in agar and photographed using a Leica Z16 APO microscope and DFC300 FX camera. Images were assembled with Adobe Photoshop or Illustrator. Image quantitation used Fiji (Schindelin et al., 2012): cortical fluorescence intensity plots measured gray values along single-cell tracings; Wls cortical localization measured correlation coefficients with phalloidin staining.

We thank E. Knust, X. Lin, B. Lu and G. Bosco for providing reagents; N. Rusan for unpublished fly lines; S. Siegrist for shared expertise; X. Li for excellent technical assistance; and the D.B. lab for manuscript comments.