Clustering and Negative Feedback by Endocytosis in Planar Cell Polarity Signaling Is Modulated by Ubiquitinylation of Prickle

The core components of the planar cell polarity (PCP) signaling system, including both transmembrane and peripheral membrane associated proteins, form asymmetric complexes that bridge apical intercellular junctions. While these can assemble in either orientation, coordinated cell polarization requires the enrichment of complexes of a given orientation at specific junctions. This might occur by both positive and negative feedback between oppositely oriented complexes, and requires the peripheral membrane associated PCP components. However, the molecular mechanisms underlying feedback are not understood. We find that the E3 ubiquitin ligase complex Cullin1(Cul1)/SkpA/Supernumerary limbs(Slimb) regulates the stability of one of the peripheral membrane components, Prickle (Pk). Excess Pk disrupts PCP feedback and prevents asymmetry. We show that Pk participates in negative feedback by mediating internalization of PCP complexes containing the transmembrane components Van Gogh (Vang) and Flamingo (Fmi), and that internalization is activated by oppositely oriented complexes within clusters. Pk also participates in positive feedback through an unknown mechanism promoting clustering. Our results therefore identify a molecular mechanism underlying generation of asymmetry in PCP signaling.

Introduction PCP is the tissue-level organization of cells in the plane of an epithelium, resulting from the coordinated acquisition of cellular polarity orthogonal to the apical-basal axis. PCP signaling controls the polarity of numerous epithelial cells in both Drosophila and vertebrates. In Drosophila, the most thoroughly studied planar polarized tissue is the fly wing, in which each cell produces a trichome ("hair"), that in wild type, emerges from the distal side of the cell and points distally [1]. PCP mutants cause disruption of this pattern. In vertebrates, many of the PCP signaling components identified in flies are conserved, and function together with additional regulators not present in flies. Defects in vertebrate PCP result in a range of developmental anomalies and diseases including open neural tube defects, conotruncal heart defects, and disruption of sensory hair cell polarity. PCP is also believed to underlie the directed migration of malignant cells during invasion and metastasis (reviewed in [2,3]). Despite increasing study, understanding of PCP signaling mechanisms remains limited.
A key feature of PCP signaling is the generation of subcellular asymmetry in which critical signaling components segregate to form complexes on opposite sides of the cell. Components of the Drosophila PCP signaling mechanism may be divided into three functional module types including a core module, global directional modules and a suite of tissue specific effector modules that execute morphological polarization in individual tissues [4]. The core module acts both to amplify asymmetry, and to coordinate polarization between neighboring cells, producing a local alignment of polarity. Proteins in the core signaling module, including the serpentine receptor Frizzled (Fz) [5,6], the multi-domain protein Dishevelled (Dsh) [7,8], the Ankryin repeat protein Diego (Dgo) [9], the 4-pass transmembrane protein Van Gogh (Vang; a.k.a. Strabismus) [10,11], the Lim domain protein Prickle (Pk) [12], the seven-transmembrane atypical cadherin Flamingo (Fmi; a.k.a. Starry night) [13,14], and perhaps others [15,16], adopt asymmetric subcellular localizations that predict the hair polarity pattern (reviewed in [17]). These proteins communicate at cell boundaries, recruiting one group to the distal side of cells, and the other to the proximal side, thereby aligning the polarity of adjacent cells [18,19].
Insight into this mechanism comes from studies of clones either not expressing or overexpressing core PCP components. These clones display characteristic perturbations (or lack thereof) of cells in nearby wing tissue (referred to as domineering non-autonomy) [5,10,20,21] that have been exploited in conjunction with mathematical modeling to better understand the signaling mechanisms (reviewed in [22]).
Several global modules have been proposed to provide tissue-level directional information to the core module, aligning polarization to the tissue axes. These include the Fat/Dachsous/ Four-jointed module [23], Wnt4/Wg [24], and other undefined signals [25]. The Ft/Ds/Fj module is thought to orient core signaling by organizing polarized microtubule-dependent vesicular trafficking of distal core proteins [26][27][28].
An important unanswered question is how asymmetric subcellular localization and amplification of core components is achieved. It is proposed that an input bias from one or more of the global modules is amplified by feedback mechanisms, eventually producing strong asymmetric localization of core components. Theoretical models indicate that polarization requires a combination of a short range cooperative (positive) feedback and a long range inhibitory (negative) signal [29]. Polarization of isolated cells requires that the long range signal be intracellular and diffusible. However, coordinated polarization within sheets of cells allows the possibility that the long range negative signal, as well as the short range positive signal, might operate through contact-mediated intercellular mechanisms.
Cell biological and biophysical analyses have indicated that the transmembrane core components Fz, Fmi and Vang can assemble into stable intercellular complexes mediated by Fmi homodimerization and perhaps direct interaction between Fz and Vang ([Fz-Fmi]- [Fmi-Vang], where square brackets indicate complexed proteins in adjacent cell membranes) [30,31]. These complexes may be regulated by internalization and recycling [32]. Additional analyses suggest that the peripheral membrane associated core proteins Dsh, Pk and Dgo are required for amplification of core asymmetry [32,33]. These proteins induce stabilization of intercellular complexes and clustering into discrete puncta, of which presence and size correlate with amplification of asymmetry [32]. Clustering may be a mechanism for positive feedback. In addition, it has been proposed that oppositely oriented intercellular complexes antagonize each other, a process we refer to as mutual antagonism [18,19]. This competitive inhibition could be a form of negative long range intercellular feedback. Molecular mechanisms for these events are not known.
Regulation of levels of each of the core proteins is required for correct core PCP function, such that clonal loss or gain of function perturbs normal polarization (summarized in [18]). Ubiquitinylation has been identified as a regulator of core PCP function. E3 ligases covalently attach ubiquitin to target proteins to control protein levels and intracellular trafficking of membrane proteins [34,35]. Poly-ubiquitinylation often works as a proteasomal degradation signal, while mono-ubiquitinylation of membrane proteins typically drives targets toward vesicular trafficking pathways. Mammalian Prickle-like proteins undergo Dvl (Dsh)-dependent ubiquitinylation by Smurf E3 ubiquitin ligases and degradation at the cell surface that is required for cellular asymmetry and neural tube closure [36]. In Drosophila, the Cullin 3 (Cul3) E3 ligase complex regulates apicolateral Dsh levels, which in turn, regulates accumulation of other polarity proteins [37]. Disruption of this mechanism causes only mild PCP phenotypes. Recently, RNA interference (RNAi) of SkpA (a Cul1 complex component) was shown by both immunofluorescence and Western blot to produce accumulation of Pk in pupal wings [38]. No PCP phenotype was demonstrated or consequences characterized, nor was ubiquitinylation of Pk demonstrated.
To characterize ubiquitinylation pathways for PCP control, we carried out an RNAi screen for Drosophila E3 ligases and found Cul1 E3 ligase complex components as regulators of PCP in Drosophila wings. We provide evidence suggesting that the Cul1 complex directly regulates Pk levels and, indirectly, accumulation of other core PCP proteins. Blocking this mechanism disrupts asymmetric subcellular localization of core PCP proteins and correct hair orientation. Furthermore, our results lead us to describe a mechanism for feedback regulation required to generate asymmetry. We provide evidence for an intercellular long range inhibitory mechanism (mutual exclusion), in which Pk, triggered by Fz complexes, antagonizes the accumulation of Pk-Vang-Fmi complexes by promoting internalization. These results show that Cul1 complex-mediated control of Pk is required to ensure amplification and molecular polarization of the core module.

Results
Identification of the Cul1 E3 ligase complex as a regulator of trichome polarity patterning in the wing To identify ubiquitinylation pathways regulating PCP, RNAi constructs knocking down E3 ubiquitin ligases were screened for wing hair polarity defects. In a previous genome-wide in vivo RNAi screen, pnr-GAL4 (and MS1096GAL4) driven expression of RNAi's targeting several E3s showed notal bristle polarity defects suggesting possible PCP defects [39]. However, many E3 RNAi's caused lethality, and for these polarity could not be assessed. To overcome this limitation, we re-examined E3 RNAi constructs that caused notal bristle polarity defects and those that caused lethality by clonal over-expression using the FLP-out technique. Two independent RNAi lines knocking down each of cul1 and skpA showed wing hair polarity defects including swirling patterns and multiple hairs. This phenotype was also observed upon induction of cul1 mutant clones (Fig 1A, 1C, and 1E). Pre-hairs within cul1 and skpA knock-down clones showed abnormal polarity, and delayed emergence, as is frequently seen in PCP mutants ( Fig  1B, 1D, and 1F) [1]. Furthermore, pre-hairs in neighboring wild-type cells, up to five to six cells away, grew toward mutant clones (Fig 1B, 1D, and 1F), suggesting that the Cul1 complex affects a PCP process that acts both cell autonomously and non-autonomously. Cul1 and SkpA are components of the SCF E3 ligase complex, known to regulate the cell cycle and various signaling pathways including the Wnt and Hedgehog pathways [40,41]. Our screen suggests that they may also work together to regulate hair polarity. F-box protein components of the SCFs sequester substrates to the complex for ubiquitinylation, thereby determining substrate specificity of the E3 complex [42]. Similar to cul1 and skpA, knock-down clones of the Fbox protein Slimb produce retarded and abnormally polarized trichome emergence, and showed domineering non-autonomy (Fig 1G), suggesting that Slimb is the F-box protein mediating the function of the Cul1 complex in trichome polarization. Notably, Slimb has recently been implicated in regulating the Par-3/Par-6/aPKC complex to control polarity of the Drosophila oocyte, follicle cells, and imaginal disc cells [43,44].
Clones mutant for cell-cycle regulators show multiple hairs that correlate with irregular cell shape and size, and that are reminiscent of those seen in Cul1 complex mutant clones [45,46]. However, these phenotypes are not closely correlated in SCF complex knock-down or mutant clones, such that even clones with relatively normal cell shape and size show profound polarity defects. Furthermore, retardation of hair growth and non-autonomous hair polarity defects were not observed for cell-cycle regulator mutants. Polarity defects caused by Cul1 complex mutation are therefore not likely to result from cell cycle-dependent alterations of cell size and shape.
The Cul1 complex regulates junctional levels and asymmetric localization of core PCP proteins The nature of the cul1 clonal non-autonomy is more similar to that seen with core module components as compared to global module components. We therefore explored whether the Cul1 complex could regulate a core protein(s). Fmi, Fz, Dsh, Vang, and Pk, all accumulated to elevated levels at the apical membrane in Cul1 complex knock-down and mutant clonal wings (Fig 2A-2E and S1), suggesting that the Cul1 complex indeed regulates core proteins.

Cul1-mediated effects on core PCP proteins require Pk
To identify the potential target of Cul1, we compared the profile of effects of cul1 knock-down or mutant clones on hair polarity and the levels of other core proteins to those caused by either loss-of-function or overexpression of each of the other core components (S1 Table). Like cul1 knock-down or mutant clones, only Pk overexpression reorients hairs in neighboring cells to point toward the clone and induces elevated levels of all other core factors at intercellular junctions. Furthermore, we and others have observed that, like cul1 knock-down or mutant clones, Pk over-expression induces co-clustering of Fmi, Fz, and Dsh at membrane domains where Pk accumulates [19,33]. These observations suggest that Pk is a likely target of Cul1 complex mediated degradation. These results do not rule out the possibility that other unknown factor(s) or other core proteins might also be regulated by the Cul1 complex to control PCP.
If Pk is a principle target of Cul1 in PCP signaling, the effects of cul1 knock-down or mutation should be suppressed by loss of Pk. To test this, cul1 RNAi clones were generated in pk pksple mutant wings and patterns of core proteins were analyzed ( Fig 3A). Effects of cul1 RNAi on core protein accumulation were abolished in pk pk-sple mutant wings, such that, for both Fmi and Fz, signal strength inside and outside of clones was indistinguishable ( Fig 3A). As expected, asymmetry of Fmi and Fz was disrupted in pk pk-sple wings. Pk is therefore required for the Cul1 complex to act on the core mechanism, and taken together with the full phenocopy of cul1 knock-down or mutation by Pk overexpression (sufficiency; S1 Table), the complete blockade (necessity) suggests that no other Cul1 targets play a significant role in Cul1's ability to modify core PCP activity.
Drosophila is believed to have three Pk isoforms, two of which, Pk Pk and Pk Sple , are important for epithelial planar cell polarity [12]. Hair polarity in the wing requires Pk Pk but not Pk Sple , although some Pk Sple is expressed [27]. We therefore examined whether both isoforms are regulated by the Cul1 complex. Fmi protein accumulated in cul1 RNAi clones generated in either a pk pk /pk pk-sple or pk sple /pk pk-sple background ( Fig 3B-3D), indicating that Cul1 regulates both the Pk Pk and Pk Sple isoforms. Though our antibody detects only very low levels of diffuse Sple signal in pk pk /pk pk-sple wings (signal is assumed to represent Sple since the Pk pk isoform is not expected to be expressed; Fig 3C' and 3D'), the signal is stronger in cul1 RNAi clones than in surrounding tissue. Nonetheless, total Pk Sple levels remain well below levels of the Pk pk isoform.
We next asked if Cul1 complex activity acts post-transcriptionally on Pk. Like endogenous Pk, levels of GFP::Pk expressed from a heterologous promoter are increased in 28hr APF cul1 mutant clones (Fig 3E), thus demonstrating a post-transcriptional regulation of Pk protein levels. The same was true during third instar (S2B and S2C Fig) and 24hr APF pupal wings (S2D and S2E Fig), suggesting that Cul1 is likely required throughout the course of PCP signaling. These data are consistent with Pk being a ubiquitinylation substrate of the Cul1 complex, though they do not rule out an indirect effect.

Pk is a target for Cul1 complex-mediated ubiquitinylation
To test whether Pk could be a target of the Cul1 complex, we sought evidence of a physical interaction between Pk and the Cul1 complex. As Slimb appeared to be the F-box protein providing substrate specificity to the Cul1 complex for its PCP function, we first asked whether Slimb might colocalize with Pk. Consistent with this possibility, in wildtype tissue, ubiquitously expressed The Cul1 complex regulates accumulation of core PCP components. Apical patterns of core PCP components bearing cul1 knock-down (RFP positive in A-C,). and mutant (D; homozygous for cul1 EX , RFP negative) clones at 28hr APF. A", B", C"; quantification of fold differences for respective core proteins at junctions between cul1 RNAi (+) or wildtype (-) cells. ***P<0.0001; t-test. Endogenous Fmi (green in A and D) and Pk (green in C and blue in D) were labeled with antibodies and Fz::GFP,was expressed by heterologous promoters (green in B). All core components tested show enriched apical membrane staining in knock-down or mutant clones compared to surrounding wild-type cells (see also S1 Fig). Core proteins in wild-type cells near clones are misoriented, with Fz::GFP sequestered (E') and Vang::YFP repelled (F') in neighboring cells. Absence of Fz::GFP (E) and Vang::YFP (F) signal labels Myc::Slimb protein was detected at the apical cell boundary, with a very subtle asymmetric localization that appears to be on the proximal side of the cell (arrow heads in Figs 4A and S3A-S3C). In pk pk-sple mutant clones, Myc::Slimb staining was diminished both at the apical cell junctions and also more basally in the absence of Pk (Fig 4A), indicating that retention of Slimb protein depends on Pk. Furthermore, in wildtype tissue near a clone that induces domineering non-autonomy, orientations of Pk and Myc::Slimb are coordinately reorganized (S3C Fig).
To further assess dependence of Slimb localization on Pk, GFP tagged Pk Pk or Pk Sple isoforms were clonally overexpressed in wings ubiquitously expressing 6XMyc::Slimb. Upon GFP::Pk or GFP::Sple overexpression, Myc::Slimb was sequestered at the apical membrane ( Previous studies showed that F-box proteins are often ubiquitinylated by their own Cul complex in the absence of other substrates, and this provides a mechanism by which they can exchange their F-box proteins and substrates [48,49]. Consistent with this, Myc::Slimb protein levels were increased in cul1 knock-down clones (S3D Fig). More importantly, these results support the possibility that both Pk and Sple isoforms are substrates of the Cul1 complex.
If Pk is a substrate for the Cul1 complex, and if ubiquitinylation targets it for degradation, then in addition to co-dependent localization, levels of Pk should be dependent on Cul1. The elevation of endogenous Pk or exogenous Myc::Pk levels upon cul1 knock-down as shown by immunofluorescence was corroborated by increased levels detected by Western blot in cul1 knockdown wing discs (Fig 5A and 5B), consistent with Cul1 dependent destabilization of Pk protein.
Furthermore, If Pk is a substrate for Cul1 complex activity that targets for proteasomal degradation, then uibquitinylated Pk should be detected when proteasome activity is impaired. We identified conditions in which expression of the dominant negative form of the proteasomal subunit, Prosbeta2' knocked down proteasome activity without causing lethality. Under these conditions, we detected slowly migrating forms of Myc::Pk, consistent with ubiquitinylation, that were diminished upon cul1 knock-down, suggesting that these are ubiquitinylated forms of Pk ( Fig 5C). Finally, we verified that the slowly migrating forms of Pk are indeed ubiquitinylated by immunoprecipitating Myc::Pk from wing discs with knocked down proteasome activity, either with or without cul1 knock-down, and probed for Ubiquitin ( Fig 5D). Ubiquitin signal was stronger without cul1 knock-down; therefore, levels of ubiquitinylated Pk depend on Cul1 activity.
These results together argue that elevated endogenous Pk or exogenous Myc::Pk levels in cul1 knock-down wing discs (Fig 5A and 5B) are caused by defects in Cul1-mediated ubiquitinylation and proteasomal degradation of Pk. We do not rule out the unlikely possibility that the Cul1 complex, acting in close proximity to Pk, acts on an intermediate substrate that in turn regulates ubiquitinylation of Pk.
doi:10.1371/journal.pgen.1005259.g002 Pk is required for post-transcriptional Cul1 complex-mediated core control. Apical patterns of Fz::GFP (green) and Fmi (blue) in pk pk-sple13 /pk pk-sple14 pupal wings bearing cul1 knock-down clones (red, RFP) (A). cul1 knock-down clones (red, RFP) were also generated in pk pk-sple14 /pk sple1 (B) and pk pk-sple14 /pk pk30 (C, D) wings. Fmi (green, B-D), Pk Pk (blue, B) and PK Sple (blue, C and D) isoforms were labeled by antibody staining (the anti-Pk antibody recognizes both isoforms; in C' and D', signal was assumed to reflect Pk Sple since the Pk Pk isoform is not expressed). The presence of either single isoform enables apical core protein enrichment (compare A with B, C, and D) in cul1 knock-down clones. PK Sple staining is very weak, but slightly stronger in at 32 h APF compared to 28 h APF (compare C' with D'). Orientation of PK Sple staining corresponds to expected hair direction in pk pk /pk pk-sple mutant wings. D174GAL4 driven GFP::Pk (E), but not HA::pk dC (F), accumulates in cul1 mutant (cul1 EX ) clones (mutant clones without RFP indicated in Ea and Fa) in 28hr APF wings. In A-D and F, yellow dots are indicating cul1 knock-down or mutant cells. Scale bars: 10μm. Genotypes are (A) y, w, hsflp/+; UAS-cul1 IR108558 , Vang binding domains (S4E and S4F Fig), This truncated Pk did not accumulate in cul1 mutant clones (Fig 3F) nor did it sequester Myc::Slimb (Fig 4D), suggesting the possibility that the C-terminal domain of Pk contains one or more Slimb binding sites, though these observations may also be explained by reduced membrane recruitment or failure to interact with Vang. Of note, USPX9, the human de-ubiquitinylating ortholog of Fat Facets (Faf), interacts with the Cterminus of vertebrate Pk1 and Pk2, suggesting C-terminal ubiquitinylation of these proteins [50]. Additional work will be required to more precisely define the regions of Pk required for Cul1 ubiquitinylation Pk requires Vang to accumulate and cluster core PCP proteins Levels of each of the core PCP proteins must be regulated to achieve normal polarization. Our results thus far indicate that the Cul1 complex regulates Pk levels. Previous studies also showed that farnesylation and association with Vang are required to regulate Pk levels [33,38]. However, the function of Pk in core PCP function is not well understood, and these results fail to illuminate how excess Pk perturbs normal core PCP function.
Core PCP proteins assemble into asymmetric complexes, with Fmi homodimers spanning intercellular junctions, and the proximal proteins Vang and Pk associated with one side, and the distal proteins Fz, Dsh and Dgo associated with the other side of the complex ([Dgo-Dsh-Fz-Fmi]-[Fmi-Vang-Pk]). Asymmetric subcellular localization represents the preferential accumulation of this complex oriented in one direction and exclusion of the oppositely oriented complex. Pk has been suggested to mediate a mutual exclusion to facilitate this function via feedback [19]. Pk has also been implicated in stabilizing [Fz-Fmi]-[Fmi-Vang] complexes [19,33,47].
Though the mechanism by which Pk carries out its function is unknown, it has been known for some time that overexpression of Pk induces co-clustering of core proteins Fmi, Dsh and Vang at junctional domains [19,33]. Clustering of Fmi does not require Fz [33]. As Pk interacts with Vang, it has been assumed that Pk acts via Vang to stabilize intercellular bridges. Consistent with this idea, we observed that upon Pk overexpression, co-clustering of Pk and Fmi does not depend on Fz, but does indeed depend on Vang (Fig 6A-6D). Similarly, Pk co-clustering with Fmi at apical junctions in cul1 knock-down clones is observed in fz mutant wings (S2A Fig). These results confirm that Pk can induce clustering of Fmi complexes through Vang, but independent of the distal core protein, Fz ([Fmi]-[Fmi-Vang-Pk]). We note that while Pk overexpression can induce accumulation of Vang in the absence of Fmi, the characteristic clustering is not observed (S4A and S4B Fig).
To determine whether the Pk dependent puncta are vesicles resulting from an endocytic process, FM4-64 dye uptake assays were carried out. As FM4-64 cannot penetrate the plasma membrane of live cells, FM4-64 positive puncta indicate endocytic vesicles from internalization of plasma membrane [51]. Live pupae with exposed wings were briefly bathed in FM4-64 and live wings directly subjected to confocal microscopy. To assay for internalization of apical [Fmi-Vang-Pk] complexes, apical and subapical puncta were analyzed. Of GFP::Pk (driven by ptc-GAL4) positive puncta, 36 to 53% (five wings; Fig 8A) were positive for FM4-64. Similarly, when Pk was clonally overexpressed in the presence of Vang::YFP 55 to 74% of Vang::YFP positive puncta were positive for FM4-64 (Fig 8B;five wings tested). This fraction was increased compared to regions not overexpressing Pk (22 to 41%; five wings tested). Note that these percentages are expected to be lower bounds of the true values due to brief exposure to the dye. These results support the idea that Pk mediates endocytosis of Vang. We then asked whether endogenous levels of Vang vesicles depend on endogenous Pk. Consistent with our conclusion, in the absence of Pk (pk pk-sple mutant clones), the number of Vang::YFP positive puncta was reduced to 77% of that in wildtype cells (77± 9%, p = 0.0025; Vang::YFP puncta counted from 412 mutant and 1000 wildtype cells in pk pk-sple mutant clones in five wings). Taken together, these results demonstrate that Vang (together with Fmi) is endocytosed with Pk, and that increased Pk stimulates this internalization. Not surprisingly, these data also indicate that a fraction of Vang::YFP positive vesicles are independent of Pk, perhaps representing distinct recycling and/or biosynthetic pathways.
In the absence of Vang, GFP::Pk overexpression still produced GFP::Pk positive vesicles, but Fmi positive internalization was greatly reduced (Figs 6C and 7C), These results are consistent with the model that Pk dependent internalization of Fmi requires Vang. Conversely, depletion of Fmi does not affect Pk dependent internalization of Vang (Figs 7E and S4B for fmi knock-down efficiency). Based on these results, we hypothesize that Pk overexpression induces the coordinate internalization of Vang and Fmi through an endocytic pathway, and that internalization of Fmi requires interaction of Pk with Vang. Unlike internalized vesicles containing Fz and Dsh, a substantial fraction of which remain in the apical part of the cell and is seen to transcytose in a microtubule-dependent fashion [26][27][28], these vesicles do not transcytose, and are likely either recycled or targeted for degradation. While we have examined these vesicles in the context of manipulating Pk levels directly, we also see excess vesicle formation in cul1 EX mutant clones (S5A Fig), reinforcing the role of Cul1 dependent ubiquitinylation in regulating this function.

Amplification of asymmetry occurs in clusters
In contrast to clustering, Pk-dependent Fmi vesicle formation was substantially diminished in fz mutant tissue (Figs 6D and 7D). The requirement of Fz for Pk to stimulate Fmi-Vang-Pk vesicle formation suggests that it is the interaction between full length complexes, most likely within clusters of complexes in opposite orientations that promotes internalization (Fig 9G). This conclusion is further supported by the observation that clones overexpressing Pk, which induce recruitment of Fz to the neighboring cell boundaries and therefore hairs to point toward the clone, simultaneously induce Vang vesicle formation in the adjacent cells (S6 Fig). We note that in fz mutant clones, apicolateral Fmi, the pool from which vesicles may be drawn, is modestly reduced, perhaps by a factor of two [47]. However, there is a greater than 10 fold reduction in vesicle formation, suggesting reduced apicolateral Fmi alone cannot explain the decrease in vesicles. The same is true for vang clones.
Fz-dependent internalization of Fmi-Vang-Pk vesicles is likely to be involved in the generation of asymmetry, as competition between oppositely oriented complexes leading to dose-dependent removal of the less abundant complex would amplify asymmetry. Additional evidence for the idea that clustering might be associated with internalization and amplification comes from visualization of asymmetric localization. Asymmetric localization can be visualized by generating twin clones expressing PCP proteins tagged with different fluorophores, and examining shared cell boundaries. This experiment was performed at 16 h APF, when asymmetric localization is detectable, but incomplete, as assessed by traditional methods. As expected, labeling one clone with a distal protein, Fz::ECFP or ECFP::Dgo, and the adjacent clone with the proximal protein Vang::EYFP, shows overlapping accumulation of distal and proximal proteins, with some clustering in puncta (Fig 9A, 9B, 9D, and 9E). In contrast, when both twin clones are labeled with the same protein, in this case Vang::ECFP and Vang::EYFP, asymmetry would be reflected by unequal amounts on the distal and proximal sides of the boundary. Notably, we observed unequal amounts of proximal and distal Vang in puncta, but not in the more diffusely localized protein (Fig 9C, 9F, and 9J). This observation indicates that the generation of asymmetry in core PCP protein complexes is occurring within clusters.

Pk is required for membrane exclusion of Vang-Fmi complexes
Pk is not needed for assembly of complexes including [Fz-Fmi]-[Fmi-Vang], but does play a role in stabilizing these complexes, perhaps via clustering [32,33]. However, Pk is also needed for efficient generation and propagation of polarity, a function hypothesized to occur by negative feedback [19] that appears to occur among clustered complexes. We initially proposed that negative feedback involves the Vang dependent exclusion of [Dsh-Fz-Fmi] complexes [18,19]. However, our current observations suggest that negative feedback might also occur in the opposite direction, in the form of Fz-dependent removal of proximal complexes (Fmi-Vang-Pk) by endocytosis (Figs 6D, 7D, and 9G). We therefore further examined the requirement for Pk in the feedback-dependent amplification/mutual exclusion observed during intercellular PCP communication.
was probed as loading controls (A-C). (D) Myc::Pk from wing discs with genotypes as in (C) was immunoprecipitated and detected with anti-Ub, anti-Myc and anti-Pk antibodies. More Ub signal was detected in the absence as compared to the presence of cul1 knock-down. Genotypes are (A) Lane 1: D174GAL4/+(Y); pk pk-sple13 / pk pk-sple14 , Lane 2: D174GAL4/+(Y), Lane 3: D174GAL4/+(Y); UAS-cul1 IR108558 /+, (B) Lane 1: D174GAL4/+(Y), Lane 2: D174GAL4/+(Y); +/+; actP-6XMyc::pk/+, Lane 3: D174GAL4/+(Y); UAS-cul1 IR108558 / +; actP-6XMyc::pk/+, (C) Lane 1: D174GAL4/UAS-lacZi; +/+; actP-6XMyc::pk/UAS-prosbeta2'; Lane 2: D174GAL4/+(Y); UAS-cul1 IR108558 /+; actP-6XMyc::pk/UAS-prosbeta2', (D) same as (C).    Fig 6A-6D. Overexpression of GFP::pk, or pk, induces Fmi, or Vang::YFP, positive puncta (A, B). Most puncta positive for Fmi (red, A) are also positive for GFP::Pk and all Pk positive puncta are also Vang::YFP positive (B) Fmi-positive puncta were not induced or co-labelled with GFP::Pk-positive puncta in the absence of vang (C). In the absence of fz, GFP::Pk-positive puncta are less frequent, yet those seen co-label with Fmi (D). Without fz, Fmi-positive puncta are less frequent or less bright with or without pk overexpression (see Fmi staining in-and outside of pk overexpressing clones in D and compare with A). (E) Vang::YFP puncta are induced by ptc-GAL4 driven pk overexpression when fmi is simultaneously knocked down, showing that Fmi is not absolutely required for Pk dependent Vang puncta formation (Pk, red; Vang::YFP, green; Fmi, blue). Clones are outlined and magnified areas are indicated by white squares. Scale bars: 10μm. Genotypes for A-D are same as those in Fig 6. Genotype for E is, UAS-Dcr2/+; ptc-GAL4/UAS-pk; actP-vang::YFP/UAS-fmi IRFBst0026022 . Overexpression of Fz causes hairs in neighboring cells to point away from the clone; this is associated with recruitment of Vang to the boundary of neighboring cells, and at the same time repulsion of Fz from that boundary [30]. Conversely, we observe that overexpressing Vang causes hairs in neighboring cells to point toward the clone, and this is associated with recruitment of Fz and exclusion of Vang from the neighboring cell border (Fig 9I). Because Pk appears to act via Vang, we tested whether Pk might be required for this exclusion of Vang in the wildtype neighbor. We observe that whereas in wildtype cells adjacent to a Vang overexpressing clone, Vang is excluded from the cell boundary, in pk mutant cells Vang exclusion was not observed (Fig 9K 9L, and 9M). In contrast, Pk is not required for recruitment of Vang adjacent to cells that express Fz but not Vang (vang mutant clone) (S7 Fig). These results support the idea that Pk regulates Vang removal.

Ubiquitinylation of Pk by the Cul1 E3 ligase complex
In this study, we have shown that Cul1 complex-mediated ubiquitinylation of Pk is required for correct function of the core PCP signaling module, thereby ensuring proper alignment of hairs on the Drosophila wing. Ubiquitinylation by the Cul1 complex targets Pk for  proteasome-dependent degradation, and in its absence, excess Pk accumulates, resulting in disruption of core PCP function. In several previous reports, ubiquitinylation has been recognized to regulate PCP signaling. In a mouse model, Smurf E3 ligases were shown to regulate PCP signaling by modulating Pk levels [36]. However, mutation of Drosophila smurf failed to show PCP defects [37]. In Drosophila, Cul3 E3 ligase-BTB protein-mediated regulation of Dsh ubiquitinylation modulates PCP signaling, as does the de-ubiqutinylating enzyme Faf, possibly acting on or upstream of Fmi [37], or more recently proposed to act on Pk [50]. Loss of either activity shows subtle effects on final PCP outcomes in Drosophila. In no case is there a demonstrated mechanism for how these events impact the characteristic asymmetric subcellular localization of PCP proteins that underlies cell polarization.
We find that Slimb is the F-box protein that mediates Pk and Cul1 complex association in vivo. It appears likely that the motif that mediates interaction between Pk and Slimb resides in the C-terminal half of the protein, as do the Vang interaction domain and the farnesylation (CaaX) motif [52]. Of note, the amount of Slimb protein in the cell was also dependent on Pk (Fig 4A). In previous cell culture studies, F-box proteins themselves were targeted for ubiquitinylation by their own Cul complexes when not bound by other substrates [48], and this appears to be the case here, as Slimb levels are increased in cul1 knock-down clones (S3D Fig). Furthermore, this result supports the idea that Pk is the major target of the Cul1 complex during pupal wing development.
If the Cul1-SkpA-Slimb complex targets Pk for degradation, why do Slimb and Pk accumulate together on the proximal side of wildtype cells? Pk is known to bind to Vang, and to localize with it in the proximal complex. Slimb adapts the Cul1 complex to Pk and is seen to colocalize with Vang on the proximal side, as well as with overexpressed Pk (Fig 4). However, this suggests that the Pk in this location is resistant to Cul1 complex-dependent degradation. Pk levels have long been known to be limited by a Vang-dependent activity [33]. Recently, Strutt et al. showed that farnesylation of Pk is required for Pk to interact with Vang and promote its degradation, and that levels of Pk also depend on SkpA, leading to the suggestion that farnesylation-dependent Pk-Vang interaction results in SkpA-dependent Pk degradation [38]. We provide evidence suggesting that the Cul1-SkpA-Slimb E3 complex directly targets Pk for destruction, but in contrast, our finding that Pk with deleted CaaX domain accumulates to elevated levels in cul1 knock-down cells (S4C and S4D Fig) indicates that Cul1/SkpA/Slimb-dependent degradation is independent of farnesylation. Furthermore, our finding that Pk promotes internalization of Fmi-Vang-Pk during mutual exclusion of oppositely oriented core PCP complexes leads to a model, described below, that is consistent with our shared observation that Pk associated with stable intercellular complexes ([Dsh-Fz-Fmi]-[Fmi-Vang-Pk]) is protected from degradation.

Generation of cell polarity
In theory, generation of cell polarity requires the combination of a local self-enhancement of a cell polarity factor and a long range inhibition of the same factor [29]. In isolated cells, likely the evolutionarily more ancient mechanism, intracellular local self-enhancement can arise through cooperativity among P proteins (Fig 10A). Intracellular long range inhibition is most easily accomplished by limiting amounts of a component of the P complex, such that aggregation of P complexes in one location decreases the probability of aggregation elsewhere by depletion of that component.
Cell polarization within a multicellular system introduces additional possible intercellular mechanisms for both the local self-enhancement and the long range inhibition (Fig 10B and  10C). If two polarity complexes, P and Q, exist, and can interact at junctions between adjacent cell boundaries, then both the local and long range effects can be mediated through these intercellular interactions. If P complexes recruit Q complexes to opposing sides of junctions, and if mutual antagonism between P and Q occurs, then long range inhibition can occur by P recruiting Q to the neighbor, where P is then excluded (Fig 10B). Similarly, exclusion of P decreases Q in that region of the original cell, enabling the accretion of more P (in effect, cooperativity) (Fig 10C).
The peripheral membrane associated core PCP proteins Pk, Dsh and Dgo appear to mediate these polarization events, but how they do so is not known. They are not required for assembly of asymmetric [Fz-Fmi]-[Fmi-Vang] complexes, but were known to share the ability to induce clustering, and are all required for the feedback amplification that results in the asymmetric subcellular localization of PCP signaling complexes. While their action somehow promotes the assortment of proximal and distal core proteins to opposite sides of the cell, how they carry out this function, and in particular whether this is through intracellular or intercellular mechanisms, is unclear.

Pk-dependent endocytosis and amplification of PCP
To understand how excess Pk resulting from mutation of the Cul1 E3 complex disrupts PCP, we further studied Pk's role in establishment of core asymmetry. pk mutation causes symmetric distribution of other core proteins without substantially diminishing or enhancing their junctional recruitment. On the other hand, Pk overexpression causes both accumulation of higher levels of all proximal and distal core proteins and induces their clustering at apical membrane domains, generating discrete puncta. A Pk induced clustering of similarly oriented core complexes, as previously proposed [32], could explain both the aggregated punctate appearance and the increased levels of accumulated proteins if one assumes a steady state relationship between free asymmetric complexes and unassembled components as asymmetric complexes are sequestered into puncta.
Our Pk over-expression study shows that Fz is not required for making Fmi clusters, but Vang is (Fig 6). This suggests an intracellular mechanism in which Pk interacts with Vang at the apical membrane to induce clustering. However, since Vang over-expression does not cause accumulation of other core proteins [33], a specific function for Pk beyond stabilization of Vang must be considered to explain the accumulation of other core proteins. Furthermore, the depletion of Fmi from the membrane achieved by the very high levels of Pk upon simultaneous Pk overexpression and Cul1 depletion (S5B, S5C, and S5D Fig) argues for a function for Pk beyond clustering.
A model for Pk in amplification of core protein asymmetric localization Pk might stimulate amplification simply by promoting clustering, with long range inhibition mediated by other mechanisms, or perhaps by limiting amounts of Pk. However, our data suggest an alternative interpretation, as we observe Pk-dependent mutual exclusion of oppositely oriented complexes: forcing local accumulation of distal proteins induced the Pk-dependent removal of proximal proteins within the same cell (Fig 9G, 9I, 9K, 9L, and 9M). Exclusion is associated with Pk mediated internalization of Pk-Vang-Fmi complexes, suggesting that this exclusion involves endocytosis (Fig 10D; Pk positive vesicle). The requirement for Vang in this internalization is consistent with a previous study showing that Vang contributes to Fmi internalization [47]. We therefore propose that Pk is involved in an intercellular long range inhibition to promote feedback amplification (Fig 10D).
Like clustering, the Pk-induced routing of Fmi into intracellular vesicles was dependent on Vang, and Pk, Vang and Fmi colocalize in vesicles both apically and more basally (Figs 6 and  7), indicating that Fmi-Vang complex trafficking is regulated by associated Pk. However, unlike clustering, it is also dependent on Fz (Figs 6D and 7D). This suggests a model for feedback inhibition in which oppositely oriented asymmetric complexes interact within clusters, leading to endocytosis and removal of Pk-Vang-Fmi. Competitive interaction between the proximal protein Pk and the distal protein Dgo for Dsh binding is known to occur [19,[52][53][54], suggesting that these interactions might result in either of two alternative outcomes, one of which would be disruption of the proximal complex, and the other disruption of the distal complex ( Fig  10D). We propose that if the distal complex 'wins,' thus remaining stable, the proximal Pk-Vang-Fmi complex becomes internalized in a Pk-dependent step. Once there is a predominance of complex in a given orientation, Vang will be enriched on one side of the intercellular boundary with relatively little Fz present. Since Pk and Slimb associate with Vang, they too will be enriched, but the absence of competitive interactions from the Fz complex allows them to remain within clusters, accounting for the accumulation of Pk and Slimb on the proximal side of wildtype wing cells. According to this model, Pk and Slimb are observed primarily at sites where they are inactive and therefore not internalized.
Modest levels of Pk overexpression both enhance accumulation of PCP protein complexes at the membrane and disrupt the normal orientation of polarization. This may be explained by enhanced feedback amplification that overwhelms the ability to interpret directional inputs. In contrast, the depletion of Fmi from the membrane observed with the very high levels of Pk induced by simultaneous Pk overexpression and Cul1 depletion (S5B-S5D Fig) suggests that sufficient Pk can induce internalization even without the competitive interactions from the Fz complex that normally stimulate internalization.
The mechanism for Pk-dependent clustering is not known. As previously proposed, clustering may result from a scaffolding effect; the possibility of decreased endocytosis accounting for clustering was previously discounted [32]. Whatever the mechanism, clustering by Pk must occur independent of Fz ( Fig 6D). Furthermore, Pk must enable the multimeric aggregation of complexes containing [Vang-Fmi]- [Fmi] or [Vang-Fmi]-[Fmi-Fz]. Induction of multimeric clustering would also provide a context for the dose-dependent competition that determines internalization of either the proximal or distal complex. Additional work will be required to determine how Pk facilitates clustering.

Function of the Cul1 complex
Since Cul1 depletion increases the amount of Pk, and excess Pk produces clustering and amplification, we now consider how Cul1 might produce the observed phenotype. The simplest possibility is that in the Cul1 mutant, excess Pk produces excess clustering and amplification that overwhelms the directionality in the system. However, because Pk is associated with Slimb and yet stable in the polarized state, and because Pk degradation is dependent on Vang, we also entertain the possibility that Cul1-dependent degradation is somehow functionally coupled to Pkmediated internalization. Additional studies will be required to distinguish these possibilities.
In summary, we propose a model in which Pk-dependent internalization of proximal complexes provides an intercellular long range inhibition that contributes to amplification of core protein asymmetric localization (Fig 10D). At the same time, Pk provides a local cooperative effect by inducing clustering and accumulation of proximal complexes. We don't know the mechanism for clustering, but a simple model is that Pk mediates closely related internalization events.
We note that a similar intercellular long range inhibition was initially discussed long ago [18][19][20], except that [Vang-Pk] was proposed to disrupt [Fz-Dsh]. This interpretation was based largely on inference. Here, we provide evidence that [Fz-Dsh] disrupts [Vang-Pk] (by promoting internalization). On theoretical grounds, either one would be sufficient to cause polarization, but we don't exclude the possibility that both may occur. Indeed, vesicles containing Fz, Dsh and Fmi have been shown to be transcytosed in a microtubule-dependent fashion with a directional bias [26][27][28], and these vesicles appear to derive from apical junctions, where they may arise by exclusion.

Smurf and Slimb
Although knock-down of smurf in flies reveals no function in PCP, the mechanism we describe is similar to that inferred for Smurf in mouse PCP [36]. Narimatsu and colleagues found that mice mutant for both Smurf1 and Smurf2 show PCP defects and lose asymmetric localization of core PCP proteins. Furthermore, biochemical evidence was provided that Smurfs, in the presence of the Dsh homolog Dvl2 (and Par6) mediated ubiquitinylation of mouse Pk1. From this, they proposed the model that proximal complexes containing Pk1, and presumably Vang and Celsr (Fmi), are disrupted upon proximity to distal complexes containing Fzd and Dvl2. This model is similar to our model of mutual exclusion, except that the mode of disruption was not directly addressed. While we propose disruption by internalization, perhaps coupled to degradation, they were only able to address degradation. Furthermore, it is not known if, in mouse, Pk1 mediates clustering, perhaps by a related mechanism, as we describe in flies.
The de-ubiquitinase USPX9 was recently identified as a regulator of Pk in the context of Pk's role in epilepsy in human, mouse, zebrafish and flies [50]. Similarly, the orthologous Drosophila de-ubiquitinase Faf modulates the pk sple dependent seizure phenotype in flies. These observations suggest that while the ubiquitinylating and de-ubiquitinylating activities of Smurf and USPX9 control the ubiquitinylation state of vertebrate Pk's, Cul1 and Faf may serve the analogous function to regulate ubiquitinylation of Drosophila Pk.
Transgenic flies were generated by BestGene Inc. (Chino Hills, CA, USA) for the C-terminal deleted pk construct and direct injection for other constructs.
FLP-on (using the actP>CD2>GAL4 construct for trans-gene expression) and FLP/FRT mitotic clones were generated by incubating third-instar (for analyzing pupal wings) or second-instar (for analyzing wing discs) larvae at 37°C for 1 hr and pupal wings (60 to 72 hr after heat shock) and third-instar larval wing discs (48 hr after heat shock) with appropriate clones were selected for analysis at indicated developmental time points.

Imaging and quantification
As mutation or knock-down of components in the Cul1 complex induces size and shape defects of cells, clonal wings with mild defects where cell shape and size are comparable with wildtype cells were selected for quantification. To measure the intensity of junctional core proteins in Fig  2 and S1 Fig, 50 P/D and 20 A/P junctions were randomly selected from cul1i clones and wildtype areas from each of five wings. Using Adobe Photoshop, mean intensity was obtained for P/ D and A/P junctions and a background intensity from the non-junctional region was subtracted. Fold differences were calculated for each wing. Average fold differences were calculated and pvalues were obtained from a t-test with values obtained from five wings for each genotype.
Vesicle numbers were quantified by imaging through several planes and counting vesicles/ cell in the relevant wildtype and mutant cells. Since this does not capture a complete apicalbasal scan, numbers were normalized against the wildtype value to allow comparisons between mutant and wildtype, and expressed as a percentage for each image stack.
To analyze adjacent clones expressing ECFP and EYFP-tagged PCP proteins in Fig 9, pupae were dissected at 16 hAPF as previously described [58]. Pupal wings were imaged using an Olympus Fluoview FV1000 confocal microscope equipped with a 63x oil immersion lens. ECFP and EYFP were excited using 458 and 514 nm laser lines, respectively. Boundary profiles were measured in Fiji using a 3px wide line and the Plot Profile function. To plot normalized boundary profiles for the adjacent Vang::ECFP and Vang::EYFP twinspots, measured boundary intensities were normalized to the average pixel intensity along cell boundaries that only touch cells belonging to the respective twinspot. All other immunofluorescence images were taken with a Leica TCS SP5 AOBS confocal microscope and processed with LAS AF (Leica) and Adobe Photoshop. Adult wings were imaged with Spot Flex camera (Model 15.2 64 MP) equipped with Nikon Eclipse E1000M.

Western blot and immunoprecipitation
Third-instar larval wing discs were dissected and lysed in protein loading buffer. Lysates from eight discs were loaded per lane for SDS-PAGE analysis. Antibodies: Guinea pig polyclonal anti-Pk (1:1000), mouse monoclonal anti-Myc (1:1000, Sigma-Aldrich), mouse monoclonal anti-γ-Tubulin (1:1000, Sigma-Aldrich). For the experiments using the temperature sensitive dominant negative form of proteasomal subunit, Prosbeta2', larvae with appropriate genotypes were incubated for 2hr 30 min at 30°C before isolating wing discs. Lysates from 15 wing discs were loaded in each lane.