Kinesin-II recruits Armadillo and Dishevelled for Wingless signaling in Drosophila

Wnt signaling plays important roles in growth and patterning. Key protein components involved in canonical Wnt signaling are evolutionarily conserved, and misregulation of this signaling pathway in humans can result in various diseases, including cancer (Clevers, 2006; Coombs et al., 2008). In Drosophila, Wg is required for a wide range of processes, such as body patterning in embryos (Bejsovec and Martinez Arias, 1991; DiNardo et al., 1994), growth control and cell fate specification during organogenesis (Siegfried and Perrimon, 1994; Cadigan and Nusse, 1997). Among these, patterning of the embryonic epidermis and the wing imaginal disc has been extensively studied to understand the mechanism of Wg signaling. In embryogenesis, Wg is expressed in a strip of cells within each segmental unit and is required for anterior-posterior patterning of the epidermis (van den Heuvel et al., 1993; Siegfried et al., 1994; McCartney et al., 1999).

In the developing wing disc, Wg is secreted from the dorsoventral (DV) compartment boundary where the wing margin structures are formed (Diaz-Benjumea and Cohen, 1995; Micchelli et al., 1997; Strigini and Cohen, 2000; Baena-Lopez et al., 2012). Wg and Cut expression in the DV boundary are induced by Notch (N) signaling. Wg secretion is known to form a gradient to elicit long-range Wg signaling. Secreted Wg induces the expression of Senseless (Sens) and Distal-less (Dll). Wg can also indirectly induce Cut expression by activating Dl, which in turn activates N in the DV boundary (de Celis and Bray, 1997).

Wg exerts its effects by regulating the transcription of target genes in the responding cells. Arm plays a key role in controlling the transcriptional outputs of the canonical Wg/Wnt pathway (Dierick and Bejsovec, 1999; Stadeli et al., 2006). β-catenin, the vertebrate homolog of Arm, also plays a pivotal role in transducing Wnt signaling in higher animals (Willert and Nusse, 1998). In the absence of a Wg ligand, a protein complex composed of Adenomatous polyposis coli (APC), Axin, GSK3β (Shaggy) and Casein kinase I (Cki) mediates phosphorylation of Arm and directs it for degradation via the proteasome pathway. Binding of Wnt to its receptor Frizzled (Fz) at the cell surface primes a signaling cascade that inhibits the function of this destruction complex via Dishevelled (Dsh), resulting in Arm stabilization. Wg target genes are activated when stabilized Arm enters the nucleus and recruits other coactivators to the target loci through TCF/Pangolin (Pan) (Mosimann et al., 2009). In culture cells, Wnt ligands induce clustering of Fz receptors and Dsh, which in turn promote Axin recruitment and stabilization of β-catenin (Bilic et al., 2007). It has also been proposed that in Drosophila embryos, Wg signaling causes relocation of Axin from intracellular vesicles to the plasma membrane, thus inactivating the Axin complex and stabilizing Arm (Cliffe et al., 2003). These studies imply the importance of intracellular trafficking of Wg signaling components for Wnt signaling.

Among the components of the canonical Wg signaling pathway, Arm is unique in that it is not only required for Wg signaling but also for the function of adherens junctions (AJs). Whereas cytosolic Arm is used for transduction of Wg signals, Arm localized at AJs together with α-catenin and cadherin homologs is essential for cell adhesion (Peifer et al., 1993). Arm contains 12 repeated domains in the middle portion of the protein called Arm repeats. The N-terminal region of Arm is crucial for its function at AJs. By contrast, its middle region consisting of the Arm repeat motifs is required for transduction of Wg signaling (Orsulic and Peifer, 1996b), although it is also necessary for proper localization of Arm (Orsulic and Peifer, 1996a; Valenta et al., 2011). Many binding partners for the Arm repeat domain have been found. These partners, including Pan and Cadherin, have overlapping binding sites, suggesting that competition between them might be important for the regulation of Wnt signaling (Pai et al., 1996; Blauwkamp et al., 2008). Despite extensive studies on Arm repeat domain interactions, it is poorly understood how intracellular trafficking of Arm/β-catenin is regulated. Thus, it is important to identify new partners that could provide clues to the mechanism of intracellular trafficking of Arm/β-catenin for Wnt signaling.

Interestingly, Klp64D, the Drosophila homolog of Kif3A of the kinesin II microtubule-based motor (Ray et al., 1999), has been shown to be required for localization of cell junction proteins during morphogenesis of photoreceptor cells in the eye (Mukhopadhyay et al., 2010). Arm is one of the proteins that are mislocalized from AJs in the absence of Klp64D. This raises the possibility that Klp64D might also be required for Wg signaling. Kinesin II motors are typically found as heterotrimeric complexes containing three subunits: two kinesin-like motor subunits, Kif3A and Kif3B, and a non-motor subunit termed Kap3 (Kinesin-associated polypeptide 3) (Cole, 1999; Hirokawa, 2000). Drosophila kinesin-II motors are known to be involved in anterograde intraflagellar transport (Tanaka et al., 1998; Ray et al., 1999), but their function in developmental signaling has not been studied. Although mammalian Kif3A has been implicated in the modulation of Wnt signaling (Jimbo et al., 2002; Corbit et al., 2008), it is unknown whether Kif3A and β-catenin play direct roles.

In this study, we show that Klp64D is required for Wg signaling during wing and embryo patterning. Our data indicate that Klp64D is crucial for Arm function by physical interactions and vesicular trafficking. We provide genetic, biochemical and immunocytochemical evidence to support the proposed novel function of the Klp64D-Arm-Dsh complex for Wnt signaling. We also demonstrate similar interaction between human KIF3A and β-catenin. Thus, the function of the kinesin II-β-catenin complex for Wnt signaling might be conserved in higher animals.

To test whether Klp64D plays an important role in wing development, we knocked down Klp64D expression in developing flies by targeted expression of double-stranded RNA (dsRNA) for RNA interference (RNAi) using the UAS/Gal4 system (Brand and Perrimon, 1993). First, we used nubbin (nub)-Gal4 driver, which activates the expression of klp64D RNAi in the entire wing pouch of the larval wing disc (Calleja et al., 2000). At 18°C, nearly all animals expressing klp64D RNAi induced by nub-Gal4 (named hereafter nub>klp64D RNAi) showed mild defects in the distal part of wings (Fig. 1A,B). Interestingly, most wing defects in this condition were notching in the wing margin and loss of margin bristles, even though klp64D RNAi was targeted to all wing pouch cells of the wing disc. As Wg signaling from the DV boundary induces distal outgrowth of wing tissues and sensory organ bristles along the wing margin (Zecca and Struhl, 2010), notching in the wing margin suggests that Klp64D might be preferentially required in the DV boundary region. As Gal4 activity depends on temperature, RNAi effects were stronger at higher temperatures (Brand and Perrimon, 1993). At 29°C, flies with nub>klp64D RNAi could almost develop to the adult but died in the pupal case. These flies showed severe shrinkage or complete loss of wing (data not shown), indicating an essential role of Klp64D for wing development.

The defective wing margins described above raised the possibility that Klp64D plays a role in the DV boundary region of the wing disc. To test this hypothesis, we examined the effects of klp64D RNAi knockdown targeted to the DV boundary region. C96-Gal4 was used to induce Gal4 expression at and near the DV boundary that includes both edge and marginal cells (Gustafson and Boulianne, 1996). RNAi knockdown of Klp64D by C96-Gal4 caused notching along the wing margin in 100% of all tested flies (Fig. 1C,D). However, C96>klp61F RNAi knockdown of a different kinesin family protein, Klp61F, did not affect wing margin development (supplementary material Fig. S1). Independent klp64D RNAi lines (see Materials and Methods) gave similar results. To further confirm the effects of klp64D RNAi lines, we generated klp64D^k1 null mutant clones in the wing by FLP-mediated mitotic recombination (Xu and Rubin, 1993). Adult wings containing mutant clones showed severe notches (Fig. 1E,F), which is consistent with the klp64D RNAi effects on the wing margin.

We then tested whether Klp64D is required for Wg signaling by examining the expression pattern of the Wg target genes senseless (sens) and Distal-less (Dll) (Gorfinkiel et al., 1997; Nolo et al., 2000). RNAi knockdown of Klp64D by C96-Gal4 caused strong reduction in the expression of both Sens and Dll-lacZ reporter in the DV boundary region (Fig. 1G-J). We also confirmed that Sens and Dll expression are strongly reduced or lost within klp64D^k1 mutant clones (Fig. 1K‴,L‴). The loss of these markers is unlikely due to cell death at this stage, as cell junction markers were normally expressed (supplementary material Fig. S2). There was also no detectable level of Caspase 3 induction by klp64D RNAi targeted to the DV boundary region or to the posterior compartment using C96-Gal4 or en-Gal4, respectively (supplementary material Fig. S3). Furthermore, mutant cells within klp64D^k1 clones showed similar Discs-large (Dlg) membrane marker staining as the wild-type cells outside the clones (supplementary material Fig. S4). Although mutant clones were also found away from the DV boundary (Fig. 1K′,L′), all adult wings with mutant clones showed notching in the wing margin but no loss of wing tissues in non-margin areas (Fig. 1F). As reduced N activity can also cause wing notching, we checked whether klp64D mutations might affect N signaling. In contrast to sens and Dll, the expression of the N target E(spl)m8 was not affected in klp64D^k1 mutant clones (supplementary material Fig. S4). Taken together, these data suggest that Klp64D is preferentially required in the DV boundary region for Wg signaling and wing margin development.

To identify the functional step of Klp64D in the Wg signaling pathway, we tested whether the klp64D RNAi phenotype could be rescued by key components of the pathway. First, we overexpressed the wg transgene by C96-Gal4. Although Wg was highly expressed (supplementary material Fig. S5), it failed to suppress the notching phenotype of C96>klp64D RNAi (Fig. 2F). This suggests that Klp64D is required for Wg signal transduction rather than Wg expression or Wg secretion, as expected from the cell-autonomous phenotype of klp64D loss-of-function clones (Fig. 1K-L). We then overexpressed Dsh, but it was also unable to rescue the klp64D RNAi phenotype (Fig. 2G). By contrast, overexpression of wild-type Arm or activated Arm (Arm^S10) strongly suppressed the C96>klp64D RNAi phenotype (Fig. 2H,I), whereas inactive Arm could not (Fig. 2J). Further, benign UAS controls (UAS-GFP and UAS-lacZ) did not rescue the klp64D RNAi phenotype (supplementary material Fig. S6), indicating that the rescue by Arm^S10 is not due to titration of Gal4 by an extra copy of UAS. The control overexpression of Wg, Dsh or Arm^S10 on the wild-type background showed little effects on the wing under the same condition (Fig. 2B-D), although C96>arm^S10 has been shown to induce some ectopic hairs at a higher temperature (Helms et al., 1999). These results suggest that Klp64D functions downstream of Wg and is required in a step between Dsh and Arm. Although activated Arm^S10 strongly suppressed the notching effects of klp64D RNAi, wing margin bristles were partially restored in the suppressed wings. Thus, it is possible that in addition to Arm, a full rescue of klp64D RNAi effects might require additional factors. Although C96>Wg did not cause any detectable phenotypes at 22°C, GFP-Wg overexpression at 25°C led to 100% early pupal lethality without detectable wing development. klp64D RNAi was able to partially rescue the lethal effects of GFP-Wg overexpression, resulting in approximately 30% survival of adult flies with small wings (Fig. 2K,L). This also supports the notion that Klp64D acts downstream of Wg.

To confirm the effect of Klp64D on Wg signaling, a TopFlash assay was performed in Drosophila S2 cell culture to measure Wg signaling activity by the ratio of firefly luciferase and Renilla luciferase reporter (Beckett et al., 2013). S2 cells transfected with UAS-HA were used as a control in the absence or presence of exogenous Wg secreted in the media. Wg-containing media was added 4 days after transfection to induce signaling. In the absence of Wg, the luciferase ratio was near zero (Fig. 2M). An addition of Wg strongly increased the luciferase ratio to 1.0. This activity was further increased to 1.8 by expressing Klp64D. By contrast, Klp64D knockdown by RNAi (Fig. 2N) inhibited Wg signaling to the basal level (0.008) (Fig. 2M). These results indicate that Wg signaling in S2 cell culture depends on Klp64D. The effects of Klp64D on Wg signaling in S2 cells, where junctional complexes are absent, supports the idea that Klp64D functions on the Wg pathway, rather than having an indirect effect on cell interaction by changing the nature of junctional complexes.

As reduced N activity can cause wing notching, we checked whether klp64D mutations might also genetically interact with N. Whereas heterozygotes for N null (N^264-40) or hypomorphic allele (N^nd3) show small notching in the distal region of the wing margin (Jack and DeLotto, 1992), this phenotype was not enhanced by either klp64D^k1/+ or arm^xk22/+. Besides, overexpression of N or the N intracellular domain (N^intra) could not rescue the phenotype caused by C96>klp64Di (supplementary material Fig. S7). These results are consistent with the observation that expression of the N target E(spl)m8 was not affected in klp64D^k1 mutant clones (supplementary material Fig. S4). Hence, our data support that Klp64D is involved in Wg signaling and is probably independent of the N pathway.

As the klp64D RNAi phenotype is suppressed by activated Arm but not by Dsh, the function of Klp64D in the wing appears to be closely related to Arm. We tested this possibility by examining genetic interactions, using mutant alleles or RNAi knockdown of these two genes. Heterozygous flies for arm^xk22, a null allele that generates a premature stop codon (Peifer and Wieschaus, 1990), have normal wings (Fig. 3A). However, nub>klp64D RNAi knockdown in the arm^xk22/+ heterozygote condition resulted in about 27% larger notching than the nub>klp64D RNAi control and/or loss of hairs along the posterior margin of the wing (supplementary material Fig. S8), indicating that the reduced arm function enhances the klp64D knockdown phenotype. A more striking genetic interaction was found between klp64D^k1/+ and arm^xk22/+ heterozygotes. Whereas heterozygous flies for either arm or klp64D mutation showed normal wings (Fig. 3A,B), more than 50% of double-heterozygous flies (arm^xk22/+; klp64D^k1/+) showed significant loss of wing margin tissue (Fig. 3C).

We then checked whether the genetic interaction between klp64D and arm mutations affects the expression pattern of Wg target genes, such as sens and Dll. As expected, wing discs of arm^xk22/+; klp64D^k1/+ double-heterozygotes showed strong reduction or loss of Sens and Dll-lacZ expression along the DV boundary region (Fig. 3E,G). This suggests that the wing margin defects shown in arm^xk22/+; klp64D^k1/+ double-heterozygous adult flies were due to defects in Wg signaling.

Wg signaling is also essential for the establishment of segment polarity during embryogenesis (DiNardo et al., 1994; Bauer and Willert, 2012). Loss of Arm or other Wg signaling components causes disruption of the anterior-posterior pattern of denticles in the ventral epidermis of embryo and larva. About 18% of klp64D^k33 homozygous first instar larvae from klp64D^k33 heterozygous parents balanced with Kr-GFP marker showed a weak segment polarity phenotype with additional ectopic denticles in the abdominal segments A4-A7 (Fig. 3I,I′; supplementary material Table S1). About 17.7% of null mutant klp64D^k1 embryos showed more severe segment polarity phenotypes than klp64D^k33 (Fig. 3J,J′). To examine the genetic interaction between arm and klp64D, arm^xk22 and klp64D mutant flies were balanced by FM7 act-GFP and TM3 Kr-GFP, respectively. arm/+; klp64D/+ double heterozygous flies were recognized by the loss of GFP. About 20% of klp64D and arm mutations showed synergistic genetic interactions in double-heterozygous embryos (Fig. 3K′,L′; supplementary material Table S1). These data suggest that klp64D functions together with arm for Wg signaling not only in wing outgrowth but also in segment polarity patterning of the embryo.

Our results for genetic interaction suggest that Klp64D and Arm may physically interact. We carried out biochemical assays to test this possibility of direct interaction. S2 cells were transfected with Myc-arm and Flag-klp64D. Cell lysates were immunoprecipitated by anti-Flag antibody-conjugated beads and analyzed by western blotting with anti-Myc antibody. In this assay, Arm was co-immunoprecipitated by Klp64D (Fig. 4C). In a similar assay, Klp61F, another kinesin family protein, did not precipitate Arm, supporting the specificity of Klp64D interaction with Arm. The interaction between Klp64D and Arm was also confirmed by co-immunoprecipitation from embryo extracts using either anti-Arm or anti-Klp64D antibody (Fig. 4D). These results indicate that Klp64D and Arm are associated in a protein complex in vivo.

To determine whether Klp64D and Arm directly interact in a protein complex, we performed pull-down experiments using purified recombinant GST-fusion proteins. GST-Klp64D fusion protein could pull down Arm (Fig. 4E), indicating direct binding of these two proteins. Next, we examined which regions of Klp64D and Arm proteins are involved in their binding. Kinesin-II consists of a N-terminal motor domain (aa 1-360), a coiled-coil (CC) domain (aa 361-581) and a C-terminal tail domain (aa 582-677) that is used for binding cargos, adaptors and scaffold proteins (Fig. 4B) (Hirokawa et al., 2009). Pull-down assays with truncated Klp64D proteins showed that Arm binds to the CC domain and the tail domain of Klp64D (Fig. 4E).

Arm protein contains a conserved Arm repeat (AR) domain flanked by N- and C-terminal regions (Fig. 4A). To identify the region of Arm involved in binding Klp64D, we tested for binding of Klp64D with three fragments of Arm: the N-terminal fragment (aa 1-140), the Arm repeat region (aa 141-670) and the C-terminal fragment (aa 671-798). These binding assays showed that Klp64D was pulled down by the Arm repeats domain, but not by the N- or C-terminal fragments (Fig. 4F).

As Dsh is required for the inhibition of Arm degradation, we tested whether Klp64D-bound Arm might be physically associated with Dsh. In pull-down assays, Dsh failed to bind Klp64D (Fig. 5A). However, Dsh was co-immunoprecipitated with both Klp64D full-length protein and the tail domain of Klp64D but not with Klp61F (Fig. 5B). The interaction between Klp64D and Dsh was also confirmed by co-immunoprecipitation from embryo extracts using either anti-Dsh or anti-Klp64D antibody (Fig. 5C). These results indicate that Dsh is indirectly associated with Klp64D in a protein complex. As Dsh does not directly bind to Klp64D, we then tested whether the association of Dsh with Klp64D complex depends on Arm. When the Arm levels were reduced by RNAi knockdown, Klp64D could not associate with Dsh (Fig. 5D,E). We also confirmed that Arm and Dsh form a complex in S2 cells and embryo extracts (Fig. 5F,G). Thus, although Klp64D does not directly bind Dsh, it forms a complex with Dsh in the presence of Arm. Furthermore, klp64D and dsh mutations showed genetic interaction in a double-heterozygote condition (dsh^v26/+; klp64D^k1/+), resulting in notched wings (Fig. 5I), as seen in arm^xk22/+; klp64D^k1/+ flies (Fig. 3C). These results suggest that the association of Dsh with the Klp64D-Arm complex is important for the Arm function in Wg signaling.

Genetic and physical interactions between Klp64D and Arm raised the possibility that these two proteins might function together in the same subcellular compartments. To test this possibility, we examined their localizations in tissues. Both an available anti-Klp64D antibody (Baqri et al., 2006) and our newly generated antibodies specifically detected Klp64D protein on a western blot, but could not recognize the protein in tissues by immunocytochemical staining. Thus, we generated a transgenic fly line that carries UAS-Klp64D-HA (Hemagglutinin). We then induced expression of klp64D-HA using actin-Gal4 and nub-Gal4, and stained with an anti-HA antibody to check the expression pattern of klp64D in the embryo and wing imaginal disc. Overexpression by actin>klp64D-HA or nub>klp64D-HA on the wild-type background did not cause any obvious phenotype. By contrast, Klp64D-HA could rescue the wing phenotype caused by klp64D RNAi (supplementary material Fig. S9). Overexpression of Klp64D-HA on act>klp64D (or nub>klp64D) RNAi background showed a similar level of klp64D-HA expression compared with that of control endogenous Klp64D (supplementary material Fig. S9). These data suggest that Klp64D-HA provides sufficient Klp64D function to rescue the RNAi wing phenotype or lethality of the null mutation. Anti-HA staining showed a distribution of Klp64D-HA at cell boundaries in blastoderm embryos. Klp64D-HA staining showed a partially overlapping expression with Arm in the embryo epithelium, suggesting that a fraction of Klp64D localizes to the AJs (Fig. 6A-A″). In the wing disc, Klp64D-HA induced by nub-Gal4 also showed a significant overlap with Arm at the cell membranes (Fig. 6D-D‴).

In addition to the cell junction staining in the apical region, more basal confocal sectioning of embryo epithelium revealed intracellular puncta co-stained for Arm, Klp64D-HA and Dsh (Fig. 6B-C″). These punctate stainings seem to represent the endogenous pattern of localization, as wild-type embryos stained with anti-Arm and anti-Dsh antibodies also showed similar punctate staining (supplementary material Fig. S10). Similar puncta were also found in wing disc cells (Fig. 6E-E″). Our data indicate that Klp64D and Arm localization partially overlaps in both AJs and intracellular punctate structures in embryo and wing disc cells. In the region basal to AJs of embryonic epithelial cells and wing discs, many Arm and Dsh-GFP puncta were found together, as previously shown (Seto and Bellen, 2004). In the basal region, those puncta were also detected together with Klp64D-HA staining (Fig. 6E-E‴). Furthermore, Klp64D-HA was co-stained with Rab5-YFP, a marker for early endosomes (Fig. 6F-F‴). These results suggest that a significant fraction of Klp64D, Arm and Dsh localize together in the same endosomal compartment. Interestingly, klp64D^k1 mutant embryos showed abnormally high levels of Arm accumulated in vesicular structures (supplementary material Fig. S11C,D). In normal embryos, only 10% of GM130 puncta overlapped with Arm. However, in klp64D mutant embryos, a much higher fraction of GM130 (93%) overlapped with Arm (supplementary material Fig. S11). This trend was also observed in wing discs. In normal wing discs, only a small subset (13%) of GM130 punctate structures overlapped with Arm puncta. By contrast, when Klp64D was reduced in wing discs by RNAi knockdown, the levels of both Arm and GM130 were strongly increased, and GM130 and Arm puncta overlapped significantly (Fig. 7C-D″). This suggests that Arm protein is abnormally accumulated in Golgi when the Klp64D function is compromised.

Protein sequence alignment shows 99% conservation between the motor domains of Klp64D and human KIF3A. It has been shown that mouse kinesin heavy chain proteins mutated in the conserved ATP binding motif can bind microtubules but cannot move (Nakata and Hirokawa, 1995). These rigor type mutants can cause dominant-negative effects. Based on this information, we generated two kinds of defective Klp64D proteins mutated in the identical motor domain (supplementary material Fig. S12A): Klp64D^ΔABD deleted in eight amino acid core residues for ATP binding (aa 107-114), and Klp64D^T114A point mutation at position T114. We tested whether these mutant proteins could show dominant-negative effects in wing discs. Overexpression of wild-type klp64D by C96-Gal4 had no effect on wing development (supplementary material Fig. S12B). On the contrary, overexpression of either klp64D^Δ107-114 or klp64D^T114A resulted in similar notching along the wing margin in 60% of flies examined (supplementary material Fig. S12D,E). Moreover, the motor domain mutants Klp64D^T114A and Klp64D^ΔABD could not rescue the C96>klp64Di phenotype (supplementary material Fig. S12F,G). This dominant notching effect by rigor mutant proteins suggests that the motor activity is required for the normal function of Klp64D in wing margin development.

Klp64D and Arm are conserved homologs of mammalian Kif3A (74% identity) and of β-catenin (72% identity), respectively. To determine whether the interaction between Klp64D and Arm is also conserved, we checked for binding between the mammalian homologs of these two proteins. As shown in Fig. 8, human KIF3A directly interacted with β-catenin in pull-down assays (Fig. 8A). Furthermore, overexpression of human KIF3A fully rescued the wing phenotypes of C96>klp64D RNAi (Fig. 8D). Hence, physical and functional interaction between Klp64D and Arm seems to be evolutionarily conserved.

Here, we provide evidence for a novel function of Klp64D, the Drosophila kinesin II subunit, in wing development. Our genetic analysis suggests that Klp64D is required for Arm function in Wg signaling. This role of Klp64D is supported by synergistic genetic interaction between klp64D and arm mutations and by physical interaction between Klp64D and Arm. Our data indicate that the klp64D RNAi phenotype can be suppressed by activated Arm (Arm^S10) but not by inactive Arm. The klp64D RNAi phenotype is unlikely due to a defect in arm expression, as, based on immunostaining, the level of Arm expression in klp64D mutant embryos is not reduced. In fact, klp64D mutant embryos show abnormally elevated levels of Arm accumulated in intracellular vesicles (supplementary material Fig. S11). This suggests that Arm is not properly targeted in klp64D mutants.

It has been shown in mammalian cells that APC directly binds to the KAP3 non-motor subunit of kinesin II (Jimbo et al., 2002). According to these studies, Wnt-3a not only induces translocation of β-catenin to the nucleus but also to the APC clusters in the membrane protrusions by kinesin II (Jimbo et al., 2002). In our study, we showed a distinct physical interaction between Klp64D and Arm. Given the function of Klp64D as a motor protein, it is possible that Klp64D might be involved in transporting Arm to AJs. As Arm is required for AJ formation and maintenance, loss of Klp64D might impair AJs by failing to transport Arm to the junctions. This is consistent with the disruption of Arm localization in klp64D mutant photoreceptor cells (Mukhopadhyay et al., 2010). However, the abnormal junctional localization of Arm in klp64D mutant cells might not account for its role in Wnt signaling described in this study, as the junctional function of Arm is probable to be separate from its role in Wnt signaling (Peifer et al., 1993). Furthermore, Fz localized in the apical region is mainly used for planar cell polarity in the wing, whereas canonical Wnt signaling occurs through Fz receptors located in the basal region (Strigini and Cohen, 2000; Wu et al., 2004). Thus, Klp64D function for Wnt signaling might not be directly related to its localization at AJs.

Interestingly, our immunostaining data indicate that Klp64D-HA is not only localized to AJs but also to intracellular punctate structures in the region basal to AJs. Thus, the role of Klp64D in Wg signaling might be related to intracellular trafficking of vesicles containing Arm. It has been shown that some components of Wg signaling, including Arrow (Arr) and Dsh, are internalized to endosomal vesicles (Seto and Bellen, 2006). Upon Wg signaling, Axin relocates from the intracellular vesicles to the plasma membrane, thereby preventing the destruction of Arm (Cliffe et al., 2003). Remarkably, we found that Klp64D-HA staining in the wing disc and embryo epithelia overlaps with Arm and Dsh at a number of intracellular puncta labeled by the early endosomal marker Rab5 (Fig. 6). Furthermore, two kinds of rigor mutant forms of Klp64D cause dominant inhibition of wing margin development, suggesting the importance of the motor activity for endosomal trafficking. As Dsh can form a complex with Klp64D only in the presence of Arm, it appears that Klp64D first binds Arm, and subsequently Arm might be stabilized, as Dsh is recruited to the Klp64D-Arm complex.

Kinesins are involved in multiple transport pathways, including from the endoplasmic reticulum to Golgi, endosomes and the plasma membrane (Hirokawa and Noda, 2008). Therefore, in the absence of kinesin function, cargo proteins might be mis-targeted or accumulated in vesicles without being transported properly. Indeed, in klp64D^k1 mutant embryos, Arm was found to be abnormally accumulated in intracellular vesicles labeled with GM130 (supplementary material Fig. S11). Similar accumulation of Arm was also found in wing discs when the level of Klp64D was reduced (Fig. 7). In addition, the number of GM130 puncta overlapping with Arm was greatly increased in klp64D^k1 mutant embryos and wing discs, which might be due to an increased processing demand in Golgi (Sengupta and Linstedt, 2011).

It has been suggested that Kif3a in ciliated mammalian cells has a negative role in Wnt signaling by inhibiting Cki phosphorylation of Dvl, thus providing a brake to the Wnt-induced β-catenin stabilization (Corbit et al., 2008). However, this role of cilia and Kif3A in Wnt signaling has not been clearly defined. Recent studies suggest that cilia have no role in Wnt signaling in early development (Goetz and Anderson, 2010). On the contrary, conditional deletion of Kif3A significantly attenuates Wnt signaling in osteoblasts by reducing Wnt-induced accumulation of β-catenin (Qiu et al., 2012), indicating the importance of Kif3A for Wnt signaling. Our study in non-ciliary Drosophila wing tissues suggests a novel function of Klp64D in direct trafficking of Arm for Wg signaling. Loss of Klp64D also affects the Wg-dependent anterior-posterior patterning of embryonic segments (Fig. 3). This suggests that Klp64D is an important component of Wg signaling in diverse developmental contexts. We also show that human KIF3A physically interacts with β-catenin and that KIF3A can replace Klp64D in vivo. Hence, similar interactions between mammalian kinesin and β-catenin might also play an important role for Wnt signaling in higher animals.

Canton-S was used as the wild-type Drosophila strain. UAS lines used were UAS-klp64D RNAi [10642R-1 and 10642R-2 from the National Institute of Genetics (NIG), Japan; 45373 and 103358 from the Vienna Drosophila Resource Center, Austria (VDRC)], UAS-klp61F RNAi (v52548 and v52559 from VDRC), UAS-klp64D (Sarpal et al., 2003), UAS-arm RNAi (Bloomington Drosophila Stock Center, USA), UAS-arm^S10 (Pai et al., 1997), UAS-arm^S2 (Kyoto 107878), UAS-GFP-Wg (Pfeiffer et al., 2002), UAS-hid (Bloomington Drosophila Stock Center) and UAS-arm^Exel (Bloomington Drosophila Stock Center). Gal4 lines used were nub-Gal4, ptc-Gal4 (Bloomington Drosophila Stock Center) and C96-Gal4 (Gustafson and Boulianne, 1996).

Klp64D-HA was cloned into a pUAST vector (Brand and Perrimon, 1993) for the generation of transgenic fly lines.

Mitotic wing clones were generated by FLP-mediated recombination (Xu and Rubin, 1993). klp64D^k1 FRT80B/TM6,Tb flies were crossed into hsflp; arm-lacZ FRT80B lines. Wing clones were induced by heat shock at 37°C for 60 min during first and second instar larvae. klp64D mutant clones were marked by the absence of arm-lacZ expression.

For cuticle preparation, dechorionated embryos were mounted in Hoyer's medium (Gergen and Wieschaus, 1986), incubated at 60°C overnight and viewed with an AxioImager microscope (Zeiss). Wings from adult flies were dissected in isopropanol and mounted in Canada Balsam mounting medium (Gary's Magic Mount) following the protocol of Roberts (Roberts, 1986).

Wing imaginal discs were dissected from third instar larvae, fixed in 4% paraformaldehyde in PBS and incubated with primary antibodies overnight at 4°C. Embryos were prepared according to a standard method (Foe and Alberts, 1983; Mitchison and Sedat, 1983). More details are described in the supplementary Materials.

Drosophila S2R+ cells express all Wg signaling components necessary to respond to exogenously added Wg (Yanagawa et al., 1998). Details of cell transfection and TopFlash assay are described in the supplementary materials.

Immunoprecipitation from S2 cell lysates and embryos were performed by standard protocols as described in the supplementary materials.

GST pull-down of bacterially expressed proteins was performed by a standard method. Details are described in the supplementary materials.

RNAi in S2 cells was carried out with a Megraw protocol (Kao and Megraw, 2004) with minor modifications. More details are described in the supplementary materials.

To quantify the Arm protein levels in klp64D mutant embryos and the wing disc, and the number of puncta, immunostaining in the embryo and the wing pouch was measured using ImageJ software.

We are grateful to Drs Kyung-Ok Cho, Amit Singh and Sang-Chul Nam for crucial comments on the manuscript. We acknowledge the Bloomington Drosophila Stock Center, the Vienna Drosophila RNAi Centre and the Developmental Studies Hybridoma Bank for fly stocks and antibodies. We thank Drs Tadashi Uemura, Hugo Bellen and J. P. Vincent for providing Dsh antibody, Sens antibody and wisir plasmid, respectively.