A novel function for the Rab5 effector Rabenosyn-5 in planar cell polarity

Cells in epithelia exhibit a characteristic apicobasal polarity, perpendicular to the plane of the tissue. Their plasma membrane is asymmetrically organized in distinct apical and basolateral domains, differing in protein and lipid composition and in functional properties (Mostov et al., 2003). Additionally, epithelial cells can be polarized along the plane, a form of polarity called planar cell polarity (PCP). The wing of Drosophila melanogaster provides an excellent biological system in which to study both forms of epithelial polarity in vivo (Adler, 2002; Eaton, 2003). The larval wing imaginal disc is an epithelial structure that undergoes dramatic morphogenetic changes during pupal metamorphosis: irregularly shaped cells convert into a monolayer of regularly packed hexagonal cells with the acquisition of PCP (Classen et al., 2005). Each cell of the monolayer produces an actin-enriched hair or trichome at its distal-most apical edge pointing towards the distal side of the wing. All hairs in the adult wing are precisely positioned and aligned with each other. Genetic studies in Drosophila have identified a ‘core’ group of proteins that control PCP in several organs and that are highly conserved in vertebrates (Klein and Mlodzik, 2005; Strutt and Strutt, 2005). This group includes transmembrane proteins, such as Frizzled (Fz), Flamingo (Fmi; also known as Starry night) and Strabismus (Stbm; also known as Van Gogh), and peripheral membrane proteins, such as Dishevelled (Dsh), Prickle and Diego. In the pupal wing, shortly before pre-hair emergence, PCP proteins adopt an asymmetric proximodistal (PD) subcellular distribution at the apical cell boundaries of epithelial cells: Stbm and Prickle localize to the proximal cell boundaries; Fz, Dsh and Diego to the distal; whereas Fmi is present on both sides. Additionally, all of these proteins might interact both intra- and inter-cellularly at the apical cell boundaries and function interdependently. Mutations in any of them affect the apical recruitment or the PD polarity of all other PCP proteins, leading to global misalignment of hair polarity, as well as to defects in hexagonal packing geometry (for reviews, see Adler, 2002; Klein and Mlodzik, 2005; Strutt and Strutt, 2005; Zallen, 2007) (Classen et al., 2005). The asymmetric localization and the ability to form stable complexes appear to lead to asymmetric intercellular signalling specifying the site of hair initiation. Such signalling is thought to depend on long-range patterning cues along the PD wing axis, involving the upstream cadherin molecules Dachsous (Ds) and Fat, and the Golgi-protein Four-jointed (Fj) (Saburi and McNeill, 2005; Strutt and Strutt, 2009).

Over the past few years, intracellular membrane trafficking has emerged as a crucial regulator of PCP in the Drosophila wing. Functional blockade of Dynamin and Rab11, which are required for the scission of endocytic vesicles from and the recycling of endocytic cargo to the plasma membrane, respectively, disrupts PCP-dependent hexagonal repacking (Classen et al., 2005). Intracellular vesicles double-positive for Fz-GFP and Fmi moving preferentially towards distal boundaries were detected by live-cell imaging before the localization of PCP proteins to the PD cortex (Shimada et al., 2006). In fixed tissue, blockade of trafficking to late endosomes resulted in intracellular accumulation of Fmi and Fz on early endosomes (Strutt and Strutt, 2008). Therefore, endocytic trafficking has been proposed to regulate the localization and signalling asymmetries of PCP proteins and to establish PCP in the tissue.

Elucidating the molecular mechanisms underlying polarized cell trafficking is thus an important step towards the understanding of epithelial biogenesis and maintenance. In this respect, small GTPases of the Rab family play a key role. Among these proteins, Rab5 is an evolutionarily conserved, master regulator of endocytosis (Zerial and McBride, 2001). It regulates the tethering and fusion of endocytic vesicles with early endosomes and homotypic early endosome fusion (Bucci et al., 1992; Gorvel et al., 1991), the motility of endosomes along actin and microtubules (Hoepfner et al., 2005; Nielsen et al., 1999; Pal et al., 2006), and signalling (Barbieri et al., 2000; Lanzetti et al., 2000; Miaczynska et al., 2004). Rabenosyn-5 (Rbsn-5) is one of the best-characterized Rab5 effectors, highly conserved from yeast to mammals (Nielsen et al., 2000). Rbsn-5 associates with early endosomes in a Rab5- and phosphatidylinositol 3-phosphate [PtdIns(3,4,5)P₃]-dependent manner, and interacts with multiple endosomal proteins. It binds to the Sec1-like protein VPS45, which interacts with endosomal syntaxins (Bryant et al., 1998; Pevsner et al., 1996; Tellam et al., 1997). In addition to Rab5, it also binds to Rab4 [through a distinct binding site (de Renzis et al., 2002)] and to EHD1/RME-1 (Naslavsky et al., 2004), two regulators of the recycling route (Grant et al., 2001; Iwahashi et al., 2002; Lin et al., 2001; van der Sluijs et al., 1992). Consistent with these interactions, Rabenosyn-5 is required for cargo recycling to the plasma membrane (Naslavsky et al., 2004). Moreover, it also regulates sorting of cargo for lysosomal degradation (Nielsen et al., 2000). Whereas Rabenosyn-5 has been investigated in non-polarized cells, its function in polarized epithelial cells is, to date, largely unknown. To address this problem, and given the functional conservation of Rabenosyn-5 in evolution, we explored the role of this protein in the development of Drosophila wing epithelia.

Control and driver strains (Bloomington Drosophila Stock Center, Indiana University); UAS-GFPRab5 (Wucherpfennig et al., 2003), Sp/CyO;UAS-Fmi (Usui et al., 1999); fmi¹⁹²,FRT42 [T. Wolff (Rawls and Wolff, 2003)]; UAS-Fz (Adler et al., 1997).

We identified the Rbsn-5 gene (CG8506 on FlyBase) by sequence comparison to the human Rabenosyn-5 gene (Nielsen et al., 2000) and cloned it from the cDNA LD29542. However, during the preparation of this manuscript, the Drosophila Rbsn-5 gene was reported (Tanaka and Nakamura, 2008). Rbsn-5, DmVps45 (CG8228) and Past1 (CG6148) cDNAs were obtained from the Berkeley Drosophila Genome Project (BDGP).

To generate rbsn³⁴ and rbsn²²⁰ mutants, imprecise excision was performed using the P-element GE11609 (GenExel). rbsn³⁴ is a 705-bp deletion starting 80-bp upstream of the 5′-UTR and covering the ATG and 372 bp of the first exon. rbsn²²⁰ is a 1518-bp deletion starting 83-bp upstream of the 5′-UTR and covering the ATG, the first exon and 455 bp of the second exon. Both deletions do not extend into the gene CG8552, positioned 235 bp upstream of the 5′ UTR region. To test the specificity of the lethality phenotype observed, we used the stock In(1 w^m4h,y¹;Df(2L)TE29Aa-11/CyO). For rescue experiments, the GFP-tagged Rbsn-5 protein was expressed in transgenic flies under the control of the ubiquitous Tubulin promoter (Marois et al., 2006). In the rescue experiment, homozygous rbsn³⁴ or rbsn²²⁰ expressing GFPRbsn-5 (genotype w; rbsn^−/−; Tub-GFPRbsn-5/TM3Sb) were generated and developed normally until adult stage. Adult wings of rescued flies did not shown any phenotype.

Mutant clones were generated using the FLP/FRT technique (Xu and Rubin, 1993). For rbsn³⁴ clones, females of the genotype y,w,hsp70-FLP;rbsn³⁴,FRT40A/CyO were crossed to w¹¹¹⁸;UbiGFP(S65T)nls,FRT40A/CyO males. For the control cross, females of the genotype y,w,hsp70-FLP;Gbe/+ were crossed to w¹¹¹⁸;UbiGFP(S65T)nls,FRT40A/CyO males. Crosses were grown at 25°C. Then first/second instar larvae were heated for 1 hour at 37°C and transferred at 18°C.

fmi¹⁹² clones were generated as previously described (Rawls and Wolff, 2003; Usui et al., 1999) and grown at 18°C. In Fmi, Stbm and Fz overexpression experiments, pupae were dissected at 26 hours after puparium formation (APF), before the establishment of PD polarity in the whole wing, in order to detect at best Rbsn-5 recruitment to the apical cell boundaries. Staging of pupal wing development at 18°C and 25°C was performed as described (Classen et al., 2008).

Proteins were in vitro translated and labelled by ³⁵S-methionine using a TnT-coupled transcription-translation kit (Promega). Five μl of the DmVps45 or Past1 proteins were incubated with 30 μl of glutathione-Sepharose beads loaded with GST or GST-Rbsn-5. Bound proteins were eluted from the beads and analysed by SDS-PAGE and autoradiography.

GST fused to full-length Drosophila Rab5, Rab4, Rab7 and Rab11 proteins were expressed using pGEX-6P-1 (Pharmacia Biotech). GTP/GDP binding assays were performed as previously described (Christoforidis et al., 1999) by incubating 5 μl of the translated Rbsn-5 protein with 30 μl of glutathione-Sepharose beads (Pharmacia Biotech) loaded with GST-Rab5, GST-Rab4, GST-Rab7 or GST-Rab11, in presence of either GDP or GTPγS. Bound proteins were eluted from the beads, separated by SDS-PAGE and analysed by autoradiography. GST-Fmi and GST-Stan were used as previously described (Wasserscheid et al., 2007).

In all the GST pull-down experiments, the loading control (L) represents 20% of the total amount of in vitro translated proteins used for incubation.

Full-length Rbsn-5 and Rab5 were expressed in Escherichia coli as GST fusion proteins using pGEX-6P-1 (Pharmacia Biotech). The purified fusion proteins were injected into rabbit and bled out. Rabbit antisera were pre-absorbed with glutathione-Sepharose beads loaded with GST, in order to remove anti-GST antibodies, and then directly used for immunoblotting and fluorescence analysis.

A polyclonal antiserum against Rbsn-5 was generated by immunizing rabbits with the SGNPFDSDEDQSSASC synthetic peptide corresponding to the 16 N-terminal amino acid residues of Rbsn-5 (Eurogentec, Belgium) and affinity purified as described (Nielsen et al., 2000).

Specificity of the antisera was verified by western blot analysis of extracts from Drosophila S2 cells, pupae (Fig. 1D) or all developmental stages (not shown).

Antibody staining on wing imaginal discs and pupal wings was performed as previously described (Classen et al., 2005; Strigini and Cohen, 2000), by using PBS 0.01% Triton X-100. Dextran uptake was performed as previously described (Entchev et al., 2000). Confocal images were acquired at 40×, 63× and 100× magnification on a Meta Zeiss Confocal Microscope. Confocal sections were spaced 0.4 μm apart.

Primaries antibodies were: mouse anti-Fmi [1:20 (Usui et al., 1999)] and rat anti-E-Cad [1:100 (Oda et al., 1994)] from the Developmental Studies Hybridoma Bank (DSHB, University of Iowa, USA); rabbit anti-Fz (1:300) (Bastock and Strutt, 2007); rat anti-Stbm (Strutt and Strutt, 2008); rabbit anti-Stbm (1:500) (Rawls and Wolff, 2003); rat anti-Dsh (1:1000) (Yanagawa et al., 1995); guinea pig anti-Hrs (1:1000) (Lloyd et al., 2002), rabbit anti-Rab7 (1:100); mouse anti-GFP (Molecular Probes, Eugene, OR, USA). Secondary antibodies were goat anti-mouse and anti-rabbit coupled to Alexa 488, Alexa 546 (Molecular Probes) or Cy5 (Jackson ImmunoResearch Laboratories). Rhodamine-conjugated Phalloidin (Molecular Probes) was used to stain actin filaments.

A single orthologue of human (Hs) Rabenosyn-5, CG8506, has been identified (Tanaka and Nakamura, 2008). The CG8506 gene encodes a protein of 505 amino acid residues (aa), named Rbsn-5, sharing with HsRabenosyn-5 (Fig. 1A) the N-terminal C₂H₂ and the FYVE zinc fingers, responsible for endosomal targeting via the interaction with Rab5 and PtdIns(3,4,5)P₃, respectively (Nielsen et al., 2000). In addition, the C-terminus of Rbsn-5 (aa 447-491; Fig. 1A, red rectangle) presents significant homology to both the central (aa 440-503) and the C-terminal region (aa 728-784) of HsRabenosyn-5, containing a Rab4-binding site (Eathiraj et al., 2005) and a second high-affinity Rab5-binding site (de Renzis et al., 2002), respectively. Rbsn-5 has two NPF motifs in the N-terminus (aa 4-6 and 68-70) instead of the C-terminal NPF motifs (Fig. 1A, blue bars) binding EHD1/RME-1 in HsRabenosyn-5 (Naslavsky et al., 2004).

To explore the functional conservation of Rbsn-5, we tested its ability to bind the Drosophila counterparts of the human interacting proteins. First, GST-Rab5, -Rab4, -Rab7 and -Rab11 fusion proteins immobilized on glutathione-Sepharose beads were loaded with GDP or GTPγS and incubated with in vitro translated Rbsn-5 (see Materials and methods). Rbsn-5 specifically bound to only Rab5 in a GTPγS-dependent manner (see Fig. 1B, lane 7) (see also Morrison et al., 2008; Tanaka and Nakamura, 2008). Rbsn-5 was also able to bind HsRab5:GTPγS, but not HsRab4:GTPγS (data not shown). As a control, HsRabenosyn-5 was shown to bind to both Rab5:GTPγS and Rab4:GTPγS (data not shown) (de Renzis et al., 2002). Therefore, Rbsn-5 displays biochemical properties of a Rab5, but not a Rab4, effector.

Second, we tested the interaction of Rbsn-5 with the Drosophila melanogaster (Dm) orthologues of HsVps45 and EHD1/RME-1 (CG8228 and CG6148, named Past1 by the FlyBase Consortium). In vitro translated DmVps45 and Past1 bound to GST-Rbsn-5 (Fig. 1C, lanes 3 and 6) but not to GST alone (Fig. 1C, lanes 2 and 5), indicating that the interactions with components of the SNARE and recycling machinery reported for HsRabenosyn-5 are conserved in its fly orthologue.

To investigate the intracellular localization of endogenous Rbsn-5, we generated two rabbit polyclonal anti-sera (see Materials and methods) that recognized a band of the expected molecular mass (~55 kDa, Fig. 1D) and of similar mobility to in vitro translated Rbsn-5 (compare Fig. 1D with 1B) by western blot analysis. To determine whether Rbsn-5 is recruited to Rab5-positives early endosomes in vivo, we analyzed its intracellular distribution by immunofluorescence microscopy in wing imaginal discs expressing GFP-Rab5 (Entchev et al., 2000). The anti-full-length Rbsn-5 serum detected the endogenous protein, and its specificity was demonstrated by: (1) loss of staining in Rbsn-5 null mutant clones (see Fig. S1 in the supplementary material); and (2) strong signal on endosomes enlarged upon GFP-Rab-5 overexpression (Fig. 1E,a-d, arrows).

Altogether, the combination of bioinformatics, biochemical and morphological studies supports the idea that Rbsn-5 is the fly orthologue of HsRabenosyn-5.

To explore the role of Rbsn-5 during development, two distinct deletions in the Rbsn-5 CG8506 gene were generated by imprecise excision: rbsn³⁴, having the 5′-UTR, the first ATG codon and almost the entire first exon deleted; and rbsn²²⁰, with a larger deletion extending to almost the end of the second exon (Fig. 2A). The two deletions exhibited similar behaviours: zygotic rbsn³⁴ and rbsn²²⁰ mutants died during the first instar larva stage. Consistent with previous results (Tanaka and Nakamura, 2008), maternal/zygotic rbsn³⁴ or rbsn²²⁰ mutants died early during embryogenesis. Heterozygotes of rbsn³⁴ or rbsn²²⁰ over the deficiency Df(2L)TE29Aa-11, which deletes the entire Rbsn-5 gene, died early during development. The zygotic lethality of both rbsn³⁴ and rbsn²²⁰ could be rescued to viability by expressing GFP-tagged Rbsn-5 under control of the ubiquitous Tubulin promoter (data not shown). Rbsn-5 protein was totally absent in rbsn³⁴ clones, as verified by immunofluorescence analysis in pupal wings (see Fig. S1 in the supplementary material). These data suggest that both rbsn³⁴ and rbsn²²⁰ behave as null, rather than hypomorphic, alleles. In subsequent experiments we focused mainly on rbsn³⁴.

To investigate the role of Rbsn-5 in endocytic transport, we performed a Texas Red-dextran endocytosis assay on wing imaginal discs containing rbsn³⁴ clones (Fig. 2B, panels a-c). In GFP-positive wild-type cells, the fluorescent tracer labelled punctuate cytoplasmic structures (white arrows), corresponding to early and late endosomes (Entchev et al., 2000). In the rbsn³⁴ clone (defined by the absence of GFP), the number of dextran-labelled endocytic structures was strikingly reduced, confirming the requirement of Rbsn-5 for endocytosis (Tanaka and Nakamura, 2008).

Interestingly, in rbsn³⁴ clones Rab5 accumulated in several brighter and, occasionally, larger vesicles (Fig. 2B,d-f, arrows). This occurred also in the pupal wing, mainly at the apical side of the epithelium (see below, Fig. 7B,C). Although to a lesser extent, Hrs, a marker of multivesicular bodies (Raiborg et al., 2001) (Fig. 7B,g,k), and Rab7, a marker of late endosomes (Marois et al., 2006) (not shown), also accumulated in rbsn³⁴ clones. No intracellular accumulation was detected (data not shown) for the Golgi protein Lava lamp (Sisson et al., 2000). These alterations of the endocytic machinery resemble those induced upon impairment of recycling to the plasma membrane (Cormont et al., 2001; Deneka et al., 2003; Sommer et al., 2005) and are thus compatible with a role of Rbsn-5 in sorting and recycling from early endosomes, as shown in mammalian cells (de Renzis et al., 2002; Naslavsky et al., 2004; Nielsen et al., 2000).

Adult wings containing either rbsn³⁴ (Fig. 3A) or rbsn²²⁰ (not shown) clones showed patches of tissue with defects characteristic of mutations in PCP genes (Klein and Mlodzik, 2005). Instead of pointing towards the distal side of the wing, hairs were deflected from the PD axis and oriented towards either the posterior or anterior margin (Fig. 3A). Occasionally, cells within the presumed clone had doubled or shorter hairs (Fig. 3A, black arrowheads). Consistently, in pupal wing at pre-hair initiation, cells within rbsn³⁴ clones displayed emerging trichomes that were misoriented, pointing away from the clone, (Fig. 3B,a,d,g) or shorter, possibly due to an impairment/delay of hair outgrowth (Fig. 3B,g; red arrowheads). In larger clones, which we found in the distal region of the pupal wing, such impairment in hair formation and elongation was more evident (see Fig. S2A in the supplementary material). Moreover, all defects manifested predominantly in the proximal regions of large rbsn³⁴ clones (Fig. 3B,d,g; see also Fig. S2A in the supplementary material). We also observed doubled trichomes in smaller clones (in >4% mutant cells; Fig. 3B,a; red arrowheads) and, rarely (in >2% mutant cells), trichomes emerging from the cell center (Fig. 3B,i; arrowheads). Because these defects were absent in adult wings of rescued flies (data not shown) but were detectable upon overexpression of Rbsn-5 (not shown), they are most likely due to specific alterations of Rbsn-5 function.

PCP mutants alter both the orientation of hair outgrowth and the packing geometry in the pupal wing (Classen et al., 2005). In rbsn³⁴ clones, cells failed to acquire the expected hexagonal shape, and Fmi was still cortical and asymmetrically localized, but not always at the PD sites (Fig. 3B,e,h; green asterisks; see also Fig. S2B in the supplementary material). Such defects were also more pronounced in the proximal side of rbsn³⁴ clones, where sometimes Fmi was uniformly localized to all cell interfaces (Fig. 3B,h; white asterisks), and was already detectable before pre-hair initiation (see Fig. S3C,G,K,O in the supplementary material). In larger clones, Fmi accumulated along the clone border at the interface between rbsn³⁴ and wild-type cells (Fig. 3B,h; see also Fig. S3G,O in the supplementary material, arrowheads), ignoring the normal PD axis. The PD distribution of the proximal PCP protein Stbm and of the distal PCP proteins Fz and Dsh was similarly perturbed (see Fig. S3 in the supplementary material).

The overall level of Fmi at the apical cell boundaries was mainly unchanged inside rbsn³⁴ clones compared to wild-type surrounding cells. We noticed that mutant cells had a narrower apical cross section, characteristic of vein cells, in ~25% of rbsn³⁴ clones, and that such clones were usually adjacent to veins areas and positioned at the pupal wing border. Fmi, other PCP proteins (see Fig. S2B and S3B,C in the supplementary material) and the junctional protein E-Cadherin (E-Cad, not shown) accumulated at the cell boundaries. The localization of E-Cad at the cell junctions was not significantly perturbed (not shown), excluding a general disruption of apicobasal polarity. Reduced apical cross-sectional area is also characteristic of vein cells. Interestingly, adult wings containing rbsn³⁴ clones often showed patches of extra or thickened wing veins (not shown). It is possible that rbsn³⁴ cells might differentiate into vein cells.

PCP proteins have also been implicated in cell-to-cell propagation of polarity information to establish PD orientation throughout the wing tissue. Consequently, mutations in PCP proteins often show cell non-autonomous effects, i.e. disruption of hair orientation in mutant cells also affects surrounding wild-type cells (for reviews, see Adler, 2002; Klein and Mlodzik, 2005; Strutt and Strutt, 2005; Zallen, 2007). We observed a weak proximal non-autonomy in trichome polarity, such that in proximal cells nearby rbsn³⁴ clones trichomes point away from the clone (Fig. 3B,a; see also Fig. S2A in the supplementary material, green arrowheads), and hexagonal packing and PD domain orientation were also altered, with Fmi running diagonally rather than orthogonally to the PD axis (Fig. 3B,e; see also Fig. S2B in the supplementary material, red arrowheads). Multi-haired cells abutting the rbsn³⁴ clone occasionally appeared (Fig. 3B,a). In the first row of wild-type cells adjacent to the rbsn³⁴ clone, PCP proteins lost the normal PD orientation and re-aligned according to the orientation of the clone boundary (Fig. 3B,h; see also Fig. S3G in the supplementary material, cells marked with red dot). However, shorter and tiny trichomes were detected only inside the mutant clones. Based on these observations, we conclude that Rbsn-5 is necessary cell-non-autonomously for hexagonal packing, alignment of the PD domains and hair orientation, and is required cell autonomously for hair formation.

The defects observed in rbsn³⁴ clones prompted us to re-examine the localization of Rbsn-5 in pupal wings during hexagonal cell packing and PCP establishment. Rbsn-5 was always enriched in the apical side of the wing epithelium throughout pupal development (Fig. 4) but was redistributed within the plane of the membrane over time. In pre-pupal wings (Fig. 4A-C), when cells are packed irregularly and PCP proteins are not uniformly PD oriented, although adjacent cells display coordinated PD polarity (Classen et al., 2005) (Fig. 4C), Rbsn-5 exhibited a punctuate intracellular staining. However, in the course of hexagonal packing and redistribution of Fmi at the apical cell boundaries, Rbsn-5 became progressively more enriched along the cortex, although it was still present in the cytoplasm (Fig. 4D-I). Once hexagonal packing was completed and PD polarity of Fmi fully established in the whole wing, Rbsn-5 at the apical cell cortex preferentially aligned along the PD boundaries (Fig. 4J-L). Such asymmetric distribution resembled (albeit it was less evident) the zig-zag pattern on the epidermal plane characteristic of Fmi (Usui et al., 1999) and other PCP proteins (Uemura and Shimada, 2003). Indeed, Rbsn-5 retained a vesicular cytoplasmic pattern. Its cortical localization was not always apparent in all areas of the wing and was only detected for a limited period of time, before the onset of trichome outgrowth. Nevertheless, the cortical localization and PD redistribution of Rbsn-5 were specific as it was lost in rbsn³⁴ clones (see Fig. S1A-C in the supplementary material, arrowheads). Although Rab5 also localized along the cortex (Fig. 5G-I), it was not found oriented along the PD axis at this stage (not shown), whereas YFP-Rab11, YFP-Rab4, Hrs and Rab7 never accumulated at the cortex (not shown).

When Fmi lost its polarized pattern (Usui et al., 1999) and hairs emerged and initiated outgrowth, Rbsn-5 became enriched in the distal part of the cell at the base of the emerging trichomes (Fig. 4N-P) and, later on, inside the trichomes (not shown).

These data show that during pupal wing development, Rbsn-5 displays a striking similarity with PCP proteins with respect to intracellular redistribution at the apical cell boundaries. Moreover, the relocation of Rbsn-5 at the base of the emerging trichomes together with the hair defects suggests a requirement for hair formation.

During hexagonal packing and establishment of PCP, Fmi, together with Fz but not E-Cad, localizes in patches at the cortex and in endosomes (Classen et al., 2005; Shimada et al., 2006) (Fig. 4F). Interestingly, a fraction of Fmi vesicles in the apical side of the pupal wing epithelium colocalized with Rbsn-5 (Fig. 5A-C, arrowheads). Such colocalization progressively increased during hexagonal packing, with 66% of Fmi-positive vesicles being co-labelled for Rbsn-5 at 26-28 hours APF at 25°C, when the total number of Fmi intracellular vesicles detected increases (Shimada et al., 2006). At the same stage, Fmi vesicles were positive for Rab5 as well (Fig. 5G-I), suggesting that they correspond to early endosomes. These vesicles were labelled neither with E-Cad (Fig. 5J-M), nor YFP-Stbm (data not shown). In addition, Fmi also colocalized with Rbsn-5 on cortical punctuate structures, which might correspond to either endocytic vesicles or subdomains of the plasma membrane (Fig. 5D-F, arrowheads).

The colocalization with Fmi, the recruitment at apical cell boundaries concomitant with Fmi redistribution at the cortex and the PD polarity suggest that Rbsn-5 is linked to the intracellular trafficking of Fmi at the apical side of the pupal wing epithelium in the course of the establishment of PD polarity.

Fmi is required for the cortical membrane association of other PCP proteins (Axelrod, 2001; Rawls and Wolff, 2003; Shimada et al., 2001; Strutt, 2001; Usui et al., 1999). As Rbsn-5 colocalizes with Fmi, we investigated whether the cortical redistribution of Rbsn-5 is dependent on Fmi. In fmi¹⁹² null mutant clones at the time of PD polarization, Rbsn-5 no longer associated with the apical cortex, but appeared diffuse intracellularly (Fig. 6A-C). Overexpression of Fmi in pupal wings induced the accumulation of Rbsn-5 along the entire cortical domain where Fmi itself was also localized (Fig. 6D-F). Such Fmi-dependent recruitment of Rbsn-5 occurred regardless of the developmental stage and of the initial localization of Fmi and Rbsn-5 (see Fig. S4A-C in the supplementary material). The distribution of Rab5, however, was never affected by Fmi overexpression (not shown), indicating selectivity for Rbsn-5 localization at the cortex. Overexpression of other transmembrane PCP core proteins, such as Fz and Stbm, did not alter the intracellular distribution of Rbsn-5 in pupal wings (see Fig. S4D-I in the supplementary material), suggesting that only Fmi out of these proteins is rate limiting for the recruitment of Rbsn-5 at the cortex.

We next tested whether Fmi can interact directly with Rbsn-5. No interaction was observed between in vitro translated Rbsn-5 and the cytoplasmic tail of both Fmi isoforms (Fmi and Stan) so far described (Wasserscheid et al., 2007) in GST pull-down assays (data not shown). It remains possible that the recruitment of Rbsn-5 by Fmi might require an intermediate binding partner.

Given its intracellular colocalization during hexagonal packing, we reasoned that Rbsn-5 could regulate internalization and recycling of Fmi. Indeed, loss of Rbsn-5 led to intracellular accumulation of the recycling Fmi fraction. Although the total amount of Fmi at the apical cell boundaries seemed unchanged (see Fig. S5A-D in the supplementary material), a fraction of Fmi was detected below the level of apical junctions in small cytoplasmic punctuate structures within the cells of rbsn³⁴ clones, but not in the surrounding wild-type cells (Fig. 7A,a-b; arrowheads). In rbsn³⁴ clones, we observed little colocalization between Fmi and Rab5 at the apical cell boundaries where Rab5 became strongly enriched (Fig. 7B,a-d). By contrast, we detected a partial colocalization with Hrs (Fig. 7B,e-h; arrowheads), and, to a lesser extent, with Rab7 (not shown), but no colocalization with the Golgi protein Lava lamp (not shown). These data are consistent with the established role of Rbsn-5 in transport to, and recycling from, endosomes (de Renzis et al., 2002). We cannot exclude that newly synthesized protein might contribute to the fraction of cytoplasmic Fmi. Nevertheless, the increased levels of endocytic markers, together with the partial colocalization of Fmi with Hrs and Rab7, suggest that endocytic trafficking is altered in rbsn³⁴ clones and that Fmi is mis-sorted to a late Hrs-positive compartment. The failure to recycle Fmi to the apical cell boundaries could explain why Fmi levels do not increase there, despite defects in endocytosis.

Consistent with recent data showing that during hexagonal packing Fmi is trafficked with Fz (Shimada et al., 2006; Strutt and Strutt, 2008), Fz accumulates sub-apically and colocalizes with Fmi on punctuate structures in rbsn³⁴ clones (Fig. 7C,a-d). By contrast, E-Cad did not accumulate in sub-cortical structures in rbsn³⁴ clones (see Fig. S5A-I in the supplementary material), ruling out a general defect in endocytic trafficking. Additionally, although at apical cell boundaries the distribution of the PCP proteins Dsh and Stbm was impaired as previously described (see Fig. S3 in the supplementary material), none of them accumulated sub-apically (see Fig. S5J-O in the supplementary material). These results suggest that Rbsn-5 specifically mediates the trafficking of Fmi.

As Rbsn-5 is a Rab5 effector, we next tested whether alterations of Rab5 activity could perturb Rbsn-5 association with the apical cell boundaries or endocytic trafficking of Fmi. Indeed, GFP-Rab5 overexpression at 30 hours at 25°C (Classen et al., 2008) shifted Rbsn-5 from the apical cell boundaries to Rab5-positive enlarged endosomes (Fig. 8E-H). In wings observed before the completion of hexagonal repacking and PCP polarization in the whole pupal wing (the stage when intracellular trafficking of Fmi was apparent in wild-type tissue), GFP-Rab5 overexpression could be seen to cause the accumulation of Fmi in enlarged early endosomes, where it colocalized with Rbsn-5 and GFP-Rab5 (Fig. 8A-D). When wings were observed at later stages, however, Fmi was normally polarized and no longer accumulated with overexpressed Rab5 (Fig. 8E-H). Furthermore, adult wings derived from these flies did not display trichome polarity defects (Fig. 8I). Interestingly, defects in planar polarization could be uncovered in Rab5 overexpressing adult wings when apoptosis was prevented by co-expression of p35 (Fig. 8J), a baculovirus-derived apoptosis inhibitor (Hay et al., 1994). Indeed, in pupal wings, GFP-Rab5 and p35 co-expression also caused the accumulation of Fmi and Rbsn-5 in GFP-Rab5 enlarged early endosomes (see Fig. S6A-D in the supplementary material). However, at later stages hexagonal packing was dramatically affected and the PD polarity of Fmi was lost (see Fig. S6E-G in the supplementary material). These data show that Rab5 overexpression alters Fmi trafficking, and causes planar polarization defects that are masked by apoptosis.

In the present study, we uncovered a novel role of Rbsn-5 in the establishment of PCP during pupal wing development, and we further demonstrated that the PCP protein Fmi undergoes endocytic trafficking in a process that is dependent on Rbsn-5 and required for the establishment of PCP.

Rbsn-5 shares with its mammalian orthologue Rabenosyn-5 several structural features, as well as the function of molecular coordinator of endocytosis and recycling (de Renzis et al., 2002; Naslavsky et al., 2004). First, the inhibition of fluid-phase endocytosis observed in rbsn³⁴ cells is consistent with the impairment of early endocytic transport described for both Rabenosyn-5 and Vps45 in mammalian cells, C. elegans and Drosophila (Gengyo-Ando et al., 2007; Morrison et al., 2008; Nielsen et al., 2000; Tanaka and Nakamura, 2008). Second, although Rbsn-5 does not bind Rab4, it interacts with EHD/RME1, a protein that is required for recycling cargo from endosomes to the surface (Grant et al., 2001; Lin et al., 2001; Naslavsky et al., 2004; Sommer et al., 2005). Third, the formation of expanded Rab5-positive endosomes in Rbsn-5 mutant cells phenocopies the endosomal enlargement observed upon inhibition of recycling (Cormont et al., 2001; Deneka et al., 2003). Moreover, the accumulation of Fmi in late endocytic compartments, which is also consistent with the requirement of Rbsn-5 (and the yeast orthologue Vac1p) for protein sorting to the degradative pathway (Nielsen et al., 2000; Burd et al., 1997; Weisman and Wickner, 1992), resembles the phenotype previously described for sec5^E13 clones in Drosophila oocytes (Sommer et al., 2005).

We found that the function of Rbsn-5 in endocytic transport is required for the re-distribution of Fmi between endosomes and the apical cell boundaries during the establishment of PCP in the Drosophila wing. Before PD asymmetry is established in the whole tissue, endogenous Fmi is detected on Rbsn-5- and Rab5-positive early endosomes. At later stages, Fmi must recycle back to the plasma membrane because it subsequently localizes to the apical cell boundaries concomitantly with Rbsn-5. Recycling from endosomes to the cell surface is also consistent with the dependence on Fmi for the recruitment of the exocyst subunit Sec5 at the apical cell boundaries (Classen et al., 2005). Loss of Rbsn-5 causes intracellular accumulation of Fmi, which correlates with defects in PD polarity and hair orientation. Consistently, Rab5 overexpression, which influences Rbsn-5 redistribution, also alters Fmi trafficking and causes PCP defects. Our data therefore indicate that Rbsn-5-dependent trafficking of Fmi is relevant for the establishment of PCP. Clearly, our data do not exclude the possibility that other (e.g. biosynthetic) trafficking events of Fmi might contribute to this process.

Why is Fmi endocytosed and recycled during establishment of PCP in the pupal wing? It has been recently proposed that a combination of polarized secretion, Fmi endocytosis and stabilization of Fmi and Fz to the distal apical cell boundaries might underlie the establishment of cellular asymmetry (Shimada et al., 2006; Strutt and Strutt, 2008). In line with this proposal, Rbsn-5-dependent trafficking might be required to remove unstable Fmi (associating only to Fz) from the apical cell boundaries and relocate it in regions of the plasma membrane where it can be stabilized in proximal PCP complexes. Therefore, the weak proximal non-autonomy in trichome orientation observed for rbsn³⁴ clones, which resembles the phenotype of fmi clones rescued with a GFP-tagged Fmi mutant lacking the cytoplasmic domain [FmiΔintra-EGFP (Strutt and Strutt, 2008)], might be explained with the blockade of Fmi at the plasma membrane preferentially bound to Fz:Dsh complexes.

Some defects observed for rbsn³⁴ clones, such as Fmi redistribution as swirls and proximal perturbation of trichome polarity inside mutant clones, together with weak proximal non-autonomy are also reminiscent of defects in Fat and Ds mutant clones (Ma et al., 2003). Interestingly, we observed that big rbsn³⁴ clones could be found only on the distal side of the pupal wing. This might reflect a less important requirement for Rbsn-5 on this side compared with the proximal one. However, whether Rbsn-5 is also involved in the global propagation of PCP signalling via the upstream module Ds-Fat-Fj remains to be determined.

Additionally, Rbsn-5 mutant clones show defects in hair formation and elongation. As endocytosis and actin cytoskeleton remodelling are functionally connected (Lanzetti et al., 2001), these defects might be indirect consequences of endocytosis impairment. However, the specific accumulation of Rbsn-5 at the bottom of emerging hairs would be consistent with the idea that Rbsn-5 mediated endocytic/recycling trafficking might actively contribute to outgrowth of wing hairs, possibly by regulating specific membrane delivery.

While preparing this manuscript, a study on the regulation of membrane protein localization by PI3K (III) and Rabenosyn-5 in Drosophila wing cells reported (Abe et al., 2009) that loss-of-function mutation of Rbsn-5 does not affect Fmi localization and hair formation and orientation. The discrepancy with our data could be explained by the fact that, in that study, the analysis was conducted at 25°C instead of 18°C. Indeed, we noticed that rbsn³⁴ clones are less healthy and tend to be smaller when grown at higher temperatures. Under these conditions, the intracellular accumulation of Fmi might well be less noticeable and the rate of lysosomal degradation may be higher.

In conclusion, the characterization of Rbsn-5 during Drosophila wing development allowed us to discover a novel function for this Rab5 effector in vivo in a developmental context and provided evidence in favor of a role of the apical endocytic trafficking of Fmi in the establishment of PCP. Future studies will hopefully provide additional molecular links and mechanistic insights into the functional interplay between the endocytic and the PCP machineries.

We are indebted to Drs Tadashi Uemura, Jeffrey D. Axelrod, Tanja Wolff, David Strutt, Hugo Bellen, Paul N. Adler and Elisabeth Knust for generously providing antibodies, constructs and fly stocks. The Flamingo #74 monoclonal antibody developed by T. Uemura and the rat anti-E-Cadherin developed by M. Takeichi were obtained from the Developmental Studies Hybridoma Bank, developed under the auspices of the NICHD and maintained by The University of Iowa, Department of Biological Sciences, Iowa City, IA 52242, USA. We thank Sven Ssykor for excellent assistance with preparing transgenic flies, Angelika Giner for help with fly stocks, the Light Microscopy, Antibody and Protein Expression Facilities of the MPI-CBG for technical support, and Bianca Habermann for assistance in bioinformatic analysis. We thank Drs Elisabeth Knust, Stefano Bonatti, André Le Bivic and Manos Mavrakis for helpful discussions and critical reading of the manuscript. We gratefully acknowledge the financial support of the Human Frontier Science Program Organization (G.M.), the Max-Planck Society, and grants from the E.U. (Endotrack), the BMBF and the DFG (M.Z.).