Drosophila CK1-γ, gilgamesh, controls PCP-mediated morphogenesis through regulation of vesicle trafficking

Cellular morphogenesis requires polarized vesicle trafficking and cytoskeletal rearrangements to promote asymmetric shape changes (Lecuit, 2003). Diverse cell types rely on polarized protrusive activity to perform specialized functions. In yeast, polarized vesicle trafficking directs membrane deposition to discrete regions during budding and mating (Madden and Snyder, 1998). Directed cell migration and neurite outgrowth require polarized trafficking for asymmetric cytoskeletal accumulation and expansion at the leading edge (Zhou and Cohan, 2004; Heasman and Ridley, 2008; Fletcher and Rappoport, 2009). Furthermore, ciliogenesis requires a polarized cytoskeleton and vesicle trafficking (Nachury et al., 2007; Yoshimura et al., 2007). Lastly, planar polarized cytoskeletal protrusions are required for the elongation of the body axis in convergent extension (Shih and Keller, 1992; Wallingford et al., 2002).

Frizzled (Fz)/planar cell polarity (PCP) signaling promotes the orientation of cells in vertebrate and invertebrate tissues (Adler, 2002; Klein and Mlodzik, 2005; Lawrence et al., 2007; Wang and Nathans, 2007; Wu and Mlodzik, 2009). In Drosophila melanogaster, Fz/PCP signaling controls the formation of single, distally oriented actin-based hairs (trichomes) on each wing cell (Wong and Adler, 1993). The core group of Fz/PCP proteins, including membrane-associated Fz, Flamingo (Fmi), and Strabismus (Stbm)/Van Gogh and cytoplasmic Dishevelled (Dsh), Diego, and Prickle, localize asymmetrically during the formation of polarized cells (Axelrod, 2001; Das et al., 2002; Strutt et al., 2002; Jenny et al., 2003; Rawls and Wolff, 2003; Das et al., 2004).

In the wing, a bias in Fz transport along microtubules has been proposed as a mechanism for asymmetric polarization and distal trichome placement along the proximodistal (PD) axis (Shimada et al., 2006; Harumoto et al., 2010). Several PCP signaling effectors restrict trichome number downstream of the core PCP proteins, including Drosophila rho-associated kinase (Drok), inturned, fuzzy, and multiple wing hairs (mwh; Wong and Adler, 1993; Winter et al., 2001; Strutt and Warrington, 2008; Yan et al., 2008). However, the precise mechanism by which these effectors interact and coordinate single trichome formation with other cellular components is not well understood.

From a PCP modifier screen, we have identified and characterized the function of the Drosophila casein kinase 1 γ (CK1-γ) homologue gilgamesh (gish) in the regulation of trichome morphogenesis. Genetic analysis in Drosophila has revealed a role for CK1-γ/gish in glial cell migration (Hummel et al., 2002), olfactory learning (Tan et al., 2010), and sperm individualization, a process requiring membrane remodeling (Nerusheva et al., 2009). CK1-γ/gish is evolutionarily conserved, and the redundant yeast orthologues YCK1/YCK2 are essential for bud site selection, morphogenesis, and cytokinesis (Wang et al., 1992; Robinson et al., 1993). GFP-Yck2p localizes to sites of polarized bud growth (Lew and Reed, 1995; Robinson et al., 1999). CK1-γ requires C-terminal palmitoylation for membrane localization and function (Robinson et al., 1993; Vancura et al., 1994; Davidson et al., 2005). Therefore, evidence from yeast budding and Drosophila spermatogenesis suggests that CK1-γ/gish may regulate polarized membrane trafficking.

Here, we describe a function for CK1-γ/gish in the regulation of PCP-directed morphogenesis. We find that CK1-γ/gish and the PCP effector mwh cooperate in an independent manner to restrict trichome formation to a single cell site. CK1-γ/gish regulates Rab11-mediated polarized vesicle trafficking. In animal cells, Rab11 localizes to recycling endosomes and derived vesicles, where it regulates vesicle trafficking during such processes as cell migration (Mammoto et al., 1999; Jing et al., 2009) and cell polarization (Prekeris et al., 2000; Wang et al., 2000; Pelissier et al., 2003). We also find that CK1-γ/gish regulates the localization of Rab11 effectors nuclear fallout (nuf; Rab11-FIP3 homologue) and Sec15 (Riggs et al., 2003; Zhang et al., 2004; Langevin et al., 2005). nuf and Rab11 are mutually required for localization to the recycling endosome (Riggs et al., 2003), and nuf is required for microtubule-dependent trafficking of recycling endosomes (Riggs et al., 2007; Horgan et al., 2010). Sec15, a component of the exocyst complex, is associated with secretory vesicles and is required for polarized exocytosis (TerBush et al., 1996). In summary, our data support a model by which CK1-γ/gish regulates morphogenesis in development through the spatial control of polarized vesicle trafficking.

To identify genes involved in PCP establishment, we performed a genetic modifier screen using the DrosDel deficiency collection (unpublished data; Ryder et al., 2007). Overlapping deficiencies defined a region on chromosome 3R that enhanced the PCP defects (Materials and methods and unpublished data). Subsequent analysis using upstream activation sequence (UAS)-RNAi (IR) identified gish (unpublished data), the Drosophila CK1-γ homologue (Hummel et al., 2002).

Expression of gish^IR under the control of engrailed (en)-Gal4 (Fig. 1, B compare with A) or FLP-FRT–induced gish^e01759 (strong hypomorphic allele; Jia et al., 2005) mutant clones (Fig. 1 C) produced similar trichome defects. These results were confirmed with an independent RNAi sequence, gish^IR2 (also RNAi knockdown of the GFP fusion allele gish^spider-GFP was observed by en-Gal4 UAS-gish^IR; Fig. S1, A [diagram] and B–C′′; Morin et al., 2001; Frescas et al., 2006).

To corroborate these results and determine the requirement of CK1-γ/Gish kinase activity, rescue experiments were performed. en-Gal4 expression of the Myc-Gish^WT transgene exhibited no phenotype (unpublished data), whereas its coexpression with gish^IR partially rescued the trichome phenotype (Fig. S1, D, E and H, quantification). Two independent Myc-Gish^KD lines (Fig. S1 A, diagram for D187N kinase-dead mutation) failed to rescue the gish^IR phenotype (Fig. S1, F and H, quantification; and not depicted). Myc-Gish^ΔC, a C-terminal truncation lacking the conserved palmitoylation site (Fig. S1 A), was ubiquitously localized and also failed to rescue the gish^IR phenotype (Fig. S1, G and H, quantification). These data indicated that CK1-γ/Gish kinase activity and membrane association are required for trichome restriction.

Furthermore, to determine a general tissue requirement of CK1-γ/gish within the context of PCP, we analyzed whether trichome formation in the thorax results in the PCP morphogenetic process of ommatidial rotation in the eye. In the thorax, pannier (pnr)-Gal4 expression of gish^IR (Fig. 1, E compare with D) induced a phenotype with most cells projecting multiple trichomes (fz or mwh loss in the thorax and trichome orientation and morphogenesis defects; Krasnow and Adler, 1994; Olguín et al., 2011). Expression of gish^IR in photoreceptors under the control of the sevenless (sev)-Gal4 promoter or FLP-FRT gish^e01759 clones induced misrotated ommatial clusters (Fig. 1, G and H compare with F).

Just before prehair formation at 26–28 h after puparium formation (APF), much of the apical cell cortex consists of polymerized actin filaments (Eaton, 1997). At this stage, we observed no defects in actin polymerization in gish^IR clones when compared with adjacent wild-type tissue (Fig. 2, A and A′). At 30–32 h APF, trichome formation appears at distal cell vertices as single focused actin-rich domains in wild-type tissue (Eaton, 1997). These domains often displayed multiple nucleation centers in gish^IR clones (Fig. 2, B and B′, yellow arrowheads), indicating the primary defect in trichome morphogenesis occurred at this stage. These data suggested that the CK1-γ/gish loss-of-function (LOF) phenotype stemmed from a failure to restrict actin bundle nucleation to a single initiation site within the distal cell region. This was supported by later analysis (32–34 h APF), in which elongated trichomes displayed completely separate actin structures in CK1-γ/gish mutant cells (Fig. 2, C and C′, yellow arrowheads). This defect was observed in FLP-FRT gish^e01759 clones (Fig. S3, A and A′). This was also confirmed by transmission EM (TEM) on pupal wings expressing gish^IR by nubbin (nub)-Gal4, which revealed prehair separation to the apical cell surface (Fig. 2 D, red arrows).

Before prehair nucleation in wing cells, PCP proteins become asymmetrically localized through their interactions within the PD axis (Strutt and Strutt, 2009). Accordingly, mutations in core PCP genes result in symmetrical localization of the other PCP proteins (Strutt and Strutt, 2009). To determine whether CK1-γ/gish was required for PCP protein localization, we examined Fmi in gish^IR tissue and gish^e01759 mutant clones. Asymmetric Fmi localization was not affected (Fig. 2, A′′ and B′′; and Fig. S3, A and compare C and C′ with B and B′). Collectively, our data support the hypothesis that CK1-γ/gish limits nucleation to a single location at the distal cell vertex through a mechanism downstream or in parallel to the core PCP components.

Interestingly, an increase in apical cell area was also observed in gish^IR tissue (Fig. S2, A–B′, compare green FLP-out clone with adjacent tissue). Measurement of the apical area revealed a 60% increase relative to the adjacent wild-type cells (Fig. S2, B–C), whereas analysis of the basolateral area did not show a significant increase (Fig. S2, B–C). TEM of wing cells expressing gish^IR revealed an increase in apical membrane projections (Fig. S2, D and E, compare boxed regions). From these observations, we suspected that the gish^IR phenotype was associated with decreased apical contractility or an increase in apical membrane trafficking. The defect was not associated with actomyosin dynamics, as the levels and localization of Myosin II heavy chain (Fig. S3, E–E′′) and activation of Myosin regulatory light chain (MRLC; Fig. S3 F) were not affected. Moreover, no modification of the gish^IR phenotype was observed when activated MRLC (Sqh^EE; Winter et al., 2001) was coexpressed (Fig. S2 C). These data suggested that expansion of the apical compartment was caused by increased membrane deposition. Support was generated from coexpression of activated Rab11^CA (Zhang et al., 2007), which enhanced gish^IR-induced membrane expansion (Fig. S2 C). This result was consistent with a role for Rab11 in apical membrane trafficking in polarized epithelial cells (Prekeris et al., 2000).

Because polyploid or enlarged cells have been shown in some cases to correlate with trichome defects (Adler et al., 2000), we analyzed the nuclear morphology of wing cells expressing gish^IR and also tested whether larger cells induced by other means displayed multiple trichomes. Cells expressing gish^IR had no observable change in intensity or area of nuclear staining compared with wild-type cells, indicating similar DNA content (Fig. S2, F, F′ [Hoechst], and G [quantification]). To generate larger cells, we expressed an RNAi against the tumor suppressor PTEN (Li et al., 1997). Similar to the enlarged cell phenotype associated with mutations such as gigas (Ito and Rubin, 1999), the increase in cell size of PTEN^IR did not affect trichome morphogenesis (Fig. S2, H and H′) or asymmetric Fmi localization (Fig. S2, H′′). Overall, these data suggest a direct role for CK1-γ/gish in regulating trichome formation.

The CK1-γ/Gish protein was associated with the base of developing prehairs as observed in apical confocal projections of 32–34-h APF wings (Fig. 3, A and A′ and magnified in B and B′). Loss of staining in clones expressing gish^IR confirmed antibody specificity (Fig. 3, A and A′). A similar pattern was detected upon expression of Myc-Gish^WT (Fig. 3, C and C′). Myc-Gish^KD did not rescue the gish^IR phenotype (see previous section) but was enriched at the membrane (unpublished data). Lastly, Myc-Gish^ΔC, which did not rescue gish^IR (see previous section), was ubiquitously localized in cells and prehairs (Fig. 3, D and D′). Analysis of subapical confocal projections (basal to prehair) revealed puncta and membrane association for both endogenous and Myc-Gish^WT proteins (Fig. 3, E and F). In summary, CK1-γ/Gish is associated with the cell membrane and base of elongating prehairs.

Fz promotes trichome formation, and increased Fz levels result in multiple trichomes, likely through ectopic activation of the nucleation machinery (Krasnow and Adler, 1994). We reasoned that if the gish^IR phenotype resulted from Fz misregulation, removal of fz (null fz^p21 allele) would reduce the gish^IR effect. We observed no change in the gish phenotype in the fz-null background (Fig. S3 D), suggesting that gish function is Fz independent. Furthermore, as gish LOF does not include trichome orientation defects, a characteristic of core PCP phenotypes, we hypothesized that gish may be linked to an effector cascade downstream of the core PCP group. We thus examined potential genetic relationships between gish and known effectors of the core PCP factors that control trichome number.

The Drok/zipper (zip) arm of PCP effector signaling failed to interact genetically with CK1-γ/gish (unpublished data). Furthermore, Zip localization or activation was not affected in gish^IR tissue (Fig. S3, E–F). We next analyzed the FH3 domain protein mwh, which is a PCP effector required to restrict actin filament formation and trichome number at the apical cell cortex (Wong and Adler, 1993; Strutt and Warrington, 2008; Yan et al., 2008). The null allele mwh¹ dominantly enhanced the phenotype of gish^e01759 homozygous mutant clones (Fig. 4, A and B). Coexpression of mwh^IR and gish^IR also resulted in a synergistic enhancement of trichome number as compared with expression of either RNAi alone (unpublished data). We expressed gish^IR in an mwh-null mutant background, reasoning that modification of the null phenotype would support a parallel relationship. Strikingly, expression of gish^IR strongly enhanced the mwh¹-null phenotype (Fig. 4, D–F, quantification). These data support a genetic model whereby CK1-γ/gish and mwh cooperate in a parallel manner to restrict trichome formation.

As CK1-γ/gish and mwh appeared to cooperate to restrict trichome formation in parallel, we wished to determine whether overexpression of one could suppress the defect of the other. Expression of either Mwh or Myc-Gish^WT alone had no phenotype (not depicted), whereas expression of Mwh strongly suppressed the gish^IR phenotype (Fig. 4, G, H, and M, quantification). Similarly, overexpression of Myc-Gish^WT suppressed the mwh^IR phenotype (Fig. 4, I, J, and M, quantification), whereas Myc-Gish^KD failed to do so (Fig. 4, K and M, quantification). Lastly, the ubiquitously localized Myc-Gish^ΔC (see previous section) also failed to suppress mwh^IR (Fig. 4, L and M, quantification). Collectively, these data suggest that mwh and CK1-γ/gish (via kinase activity and membrane localization) cooperate in a parallel to regulate trichome formation.

The polarized localization of yeast CK1-γ to regions of membrane deposition/bud growth (Robinson et al., 1999), the morphogenesis phenotype of YCK1/YCK2 (Robinson et al., 1993), and the sperm individualization phenotype of Drosophila CK1-γ/gish (Nerusheva et al., 2009) support a role for CK1-γ in membrane dynamics (see Introduction). We hypothesized that CK1-γ/gish function in trichome morphogenesis was related to polarized membrane trafficking.

To test this hypothesis, we analyzed well-characterized markers, including the recycling endosome-associated Rab11 GTPase (Ullrich et al., 1996; Dollar et al., 2002; Pelissier et al., 2003). During early prehair formation (30–32 h APF), Rab11 was detected in puncta in the subapical region (Fig. 5 A′′ and Fig. S4 B, diagram) and was detected in strongly stained foci at the base of the initiating prehairs in the apical region (Fig. 5, A and A′, yellow arrowheads; and Fig. S4 B). Strikingly, the apical Rab11 foci were reduced from the prehair base of gish^IR-expressing cells, leaving mostly diffuse apical puncta (Fig. 5, A and A′). Abnormal Rab11 localization was also apparent in gish^IR clone tissue in 32–34-h APF wings; Rab11 was present uniformly in subapical cell regions in wild-type and gish^IR cells (Fig. 5, B′′ and D–D′′, z sections), but puncta detected along the length of the developing prehairs in wild-type cells (Fig. 5, B and B′, insets) were mostly absent in gish^IR cells (Fig. 5, B, B′, and E, quantification). Again, diffuse apical Rab11 staining remained in the gish^IR tissue (Fig. 4 B′). In support of these data, the Rab11 binding partner dRip11 (Drosophila class 1 Fip; Prekeris et al., 2000; Li et al., 2007) was enriched to the prehairs but appeared mislocalized in gish^IR tissue (Fig. S4, G and G′). The Rab11-interacting protein MyoV (Hales et al., 2002) did not localize to the prehairs (Fig. S4, F and F′), suggesting a context-specific association of these proteins. Furthermore, Rab11 localization is likely dependant on Gish kinase activity and membrane enrichment, as we detected a partial rescue of Rab11 localization only upon coexpression of the Gish^WT transgene in gish^IR tissue (Fig. S1, I–L′). These data suggested that CK1-γ/gish was required to enrich Rab11 endosomal structures to the developing prehairs.

As CK1-γ/gish and mwh interacted in trichome restriction, we tested whether mwh regulated Rab11 localization in a manner similar to CK1-γ/gish. mwh^IR-expressing cells exhibited normal Rab11 localization to elongating prehairs (Fig. 5, C, C′, and E, quantification). These data and the genetic analysis led us to conclude that the functions of CK1-γ/gish and mwh converge to regulate trichome morphogenesis but use distinct cellular mechanisms (see Discussion).

To confirm Rab11 localization and to visualize recycling endosome and vesicle dynamics near the prehairs, we performed in vivo imaging of YFP-Rab11^WT trafficking in the dorsal thorax (notum). At 36 h APF, we observed dynamic YFP-Rab11^WT foci enriched to the base and within the developing prehairs, as revealed by the membrane-associated CD8-RFP (Fig. 5 F and Video 1). In contrast, coexpression of gish^IR exhibited a diffuse YFP-Rab11 pattern (Fig. 5 G and Video 2), similar to endogenous Rab11. At 38 h APF, we observed trafficking of YFP-Rab11^WT within the elongating prehair (Fig. 5, H–J, yellow arrowheads; and Video 3). This activity was markedly reduced in the presence of a gish^IR trichome phenotype (Fig. 5, K–M, magenta arrowhead indicates remaining trafficking; and Video 4). These data support a requirement for gish in Rab11 dynamics associated with developing prehairs.

We extended the endosomal localization experiments to other markers of vesicle trafficking. In gish^IR-expressing cells, we found no defects in the level or localization of Rab5, which is required for early endocytic trafficking (Fig. S4, A–A′′; Bucci et al., 1992), the endosomal marker Hrs (Fig. S4, C–C′′; Lloyd et al., 2002), or the Golgi marker Lava Lamp (Fig. S4, D–D′′; Sisson et al., 2000). As studies in yeast have supported a role for CK1-γ in endocytosis (Panek et al., 1997; Marchal et al., 2000), we analyzed dextran uptake in third-instar wing disc FLP-out clones expressing gish^IR but observed no effects (Fig. S4, E and E′). These results specifically implicate CK1-γ/gish in the regulation of Rab11/recycling endosomal (or vesicle) localization.

The failure to focus Rab11 in gish^IR cells supported the presence of a spatial bias in polarized trafficking within the distal cell vertex, which may be required for single trichome formation. To test this hypothesis, we reasoned that overexpression of a constitutively activated Rab11 (Rab11^CA; Zhang et al., 2007) might overwhelm the restriction and result in ectopic trichomes. Indeed, expression of YFP-Rab11^CA (29°C) in the posterior wing resulted in multiple trichomes (Fig. 6 A). We also tested for a genetic interaction between Rab11 and gish. Although lower expression of YFP-Rab11^CA (25°C) had no phenotype (Fig. 6 B), multiple trichomes were induced in a gish^e01759 heterozygous background (Fig. 6, C and F, quantification). Consistently, coexpression of YFP-Rab11^CA and gish^IR exhibited a strong synergistic enhancement (Fig. 6, E and F, quantification) compared with either one alone (Fig. 6, B, D, and F, quantification). In contrast to Rab11, overexpression of Rab5^CA resulted only rarely in multiple trichomes, and Rab7^CA and Rab4^CA had no effect in our assay (unpublished data). In summary, our data indicate that CK1-γ/gish and Rab11 interact specifically during trichome formation.

To further dissect the gish and Rab11^CA interaction, we analyzed YFP-Rab11^CA localization in 32–34-h APF wing cells. Expression of gish^IR resulted in diffuse YFP-Rab11^CA localization (Fig. S5, B and B′), in contrast to focused Rab11^CA associated with control prehairs (Fig. S5, A and A′). These data indicated that the distribution of endogenous Rab11, YFP-Rab11^WT, and constitutively active YFP-Rab11^CA is altered in gish^IR cells. The enhancement of the Rab11^CA effect by gish^IR correlates with the diffuse localization of Rab11^CA away from the trichome. Furthermore, to determine the localization of Rab11 on an ultrastructural level, we analyzed the distribution of YFP-Rab11 in pupal wings by immuno-EM. In the pupal wing, YFP-Rab11 was observed in trichomes near the membrane (Fig. S5, E–E′′, black arrows), and abundant YFP-Rab11–positive areas were identified in the apical cell region (Fig. S5 F).

The model that gish enriches Rab11 vesicles to promote single trichomes suggests vesicle recycling is required to deliver trichome nucleation activity. Therefore, we analyzed the effect of decreasing Rab11 function. Expression of a dominant-negative Rab11, YFP-Rab11^DN (Zhang et al., 2007), or Rab11^IR (Vienna Drosophila RNAi Center; Satoh et al., 2005) in FLP-out clones resulted in cells with short/missing and malformed trichomes in pupal wings (Fig. 7, A–A′′ and D; Fig. S5, D and D′; and not depicted). We also confirmed Rab11 knockdown and antibody specificity in Rab11^IR FLP-out clones (Fig. S5, C and C′). Interfering with trafficking from the cell membrane to the early endosome via Rab5^DN resulted in trichome orientation and number phenotypes (unpublished data), consistent with the function of the Rab5 effector rabenosyn-5 in PCP establishment (Mottola et al., 2010). Furthermore, interfering with late endocytic trafficking via Rab7^DN (Chavrier et al., 1990) or the fast endocytic recycling route through Rab4^DN (Van Der Sluijs et al., 1991) had no effect in our assay (unpublished data).

To further characterize the defects associated with recycling endosomal trafficking, we explored additional Rab11-associated proteins. The Rab11 effector nuf can bind Rab11 and link recycling endosome trafficking along microtubules via association with dynein (Riggs et al., 2007; Horgan et al., 2010). Nuf enrichment to the prehair base was dissociated in the apical domain of gish^IR tissue (Fig. 7, B and B′). We next explored the morphological effect of reducing Nuf by analyzing nuf^IR in adult wings. nuf^IR expression resulted in shortened or malformed trichomes (Fig. 7, C [arrowheads and asterisks] and D [quantification]). We also observed patches of missing trichomes (unpublished data). In addition, the exocyst component Sec15 is a downstream effector of Rab11 required for polarized membrane delivery (Langevin et al., 2005; Wu et al., 2005; Oztan et al., 2007). The Rab11–Sec15 complex is required to initiate the formation of the exocyst and promote tethering of recycling endosome–derived vesicles to the plasma membrane for subsequent membrane fusion (Wu et al., 2008). We therefore analyzed the effect of reducing Sec15 function on trichome development. To circumvent early lethality, we initiated expression of Sec15^IR in third instar larval development. Sec15^IR induced defects in trichome development, similarly to Rab11 and Nuf, and, strikingly, caused an accumulation of Rab11 (see also Jafar-Nejad et al., 2005; Langevin et al., 2005) away from the base of the trichome across the cell (Fig. 7, compare E with F–F′′). Accumulated Rab11 and defective trichome formation in the context of Sec15 knockdown, in conjunction with the Rab11 and Nuf phenotypes displaying missing and malformed prehairs, reveal (a) a requirement for Rab11–Nuf–Sec15 trafficking in trichome formation and (b) a requirement for the Sec15–exocyst complex in localized trafficking of nucleation activity through the Rab11 recycling endosome to nucleation sites.

As sec15^IR exhibited effects on Rab11 localization and trichome formation, we analyzed the effect of gish^IR on endogenous Sec15. In 32–34-h APF wings, Sec15 was observed in a punctate distribution in the prehair, whereas apical Sec15 was mislocalized in gish^IR tissue (Fig. 8, A–A′′). Furthermore, large apical patches of aggregated vesicles (dependent on exocyst function) have been reported in yeast, Drosophila epithelia, and mammalian cells upon Sec15GFP expression (Salminen and Novick, 1989; Zhang et al., 2004; Guichard et al., 2010). Thus, as an assay for Sec15GFP localization/function, we analyzed the formation of these large puncta in pupal wing cells (Fig. 8, B, B′, and E, quantification). Strikingly, in gish^IR cells, we observed an overall dissociation of these large puncta (Fig. 8, C, C′, and E; and Fig. S4, H–I′), whereas mwh-null wings resembled the control (Fig. 8, D–E). Rab11 colocalized with Sec15GFP as previously reported (Zhang et al., 2004; Jafar-Nejad et al., 2005; Guichard et al., 2010), supporting that these structures were recycling vesicles trafficking to the plasma membrane (Fig. 8, F and F′). Sec15GFP and Rab11 were dissociated by gish^IR (Fig. 8, G and G′). Collectively, our data suggest that CK1-γ/gish specifically regulates Rab11–Nuf–Sec15 vesicle localization and polarized trafficking between the recycling endosome and the plasma membrane.

Membrane trafficking is a key mechanism during morphogenesis and cell polarization (Lecuit, 2003; Mellman and Nelson, 2008). Studies in Drosophila have established a requirement for trafficking in core PCP protein localization, trichome orientation, and morphogenesis (Shimada et al., 2006; Strutt and Strutt, 2008; Fricke et al., 2009; Mottola et al., 2010; Pataki et al., 2010; Purvanov et al., 2010). Membrane trafficking has also been associated with PCP establishment in cell packing (Classen et al., 2005) and in vertebrate morphogenesis during cilia formation (Park et al., 1994; Gray et al., 2009). In the Drosophila wing, the core PCP proteins are required for the formation of a single trichome at the distal cell vertex, but these proteins are found along the entire distal (Fz/Dsh) and proximal (Stbm–Van Gogh/Prickle) cell sides (Adler, 2002). It is unclear how trichomes are restricted to one position within a broader PCP domain. It is likely that proteins, such as Fz, broadly define prehair formation and, within that domain, recruit the refinement proteins necessary to restrict the nucleation machinery to a central subdomain (Adler, 2002). The function of CK1-γ/gish in polarized vesicle trafficking provides a mechanism for refining a trichome to a single position.

Gish is localized to the cell periphery and the prehair base. This localization is consistent with a requirement for CK1-γ/gish in targeting nucleation activity to that region, as we have identified ectopic trichome nucleation as the primary defect. Moreover, our data implicate vesicle trafficking, as genetic alteration of Rab11 or effectors (nuf and Sec15) resulted in lost and/or malformed trichomes. Collectively, these data support a model that targeted membrane recycling through Rab11, Nuf, and Sec15 is required to build a trichome (Fig. 9, compare A and B), and spatial regulation of this trafficking by CK1-γ/gish is required to counteract lateral membrane localization of nucleation activity and ectopic trichome formation (Fig. 9, A and C). A similar model exists for polarized growth in yeast (e.g., bud growth), in which uptake and recycling is a mechanism to balance cdc42 lateral diffusion (Marco et al., 2007; Slaughter et al., 2009).

Our data provide evidence that Rab11–Nuf–Sec15 trafficking promotes trichome formation. Importantly, this study supports that CK1-γ/Gish regulates localized actin nucleation by directing Rab11–Nuf–Sec15-mediated vesicle trafficking between the recycling endosome and a distinct region of the membrane. Evidence for polarized recycling includes that (a) Sec15^IR tissue displays trichome malformation in the presence of accumulated Rab11 vesicles away from the trichome base (reduced vesicle tethering and fusion can result in vesicle accumulation in the cell) and (b) dissociation of Rab11–Nuf–Sec15 from the prehair in gish^IR cells is associated with multiple trichomes. These results suggest that polarized and focused membrane delivery of a nucleation factors may be disrupted. Interestingly, knockdown of the Arp2/3 nucleation machinery or regulators, such as Wasp, resulted in multiple trichomes (Fig. S5, G–I; Fricke et al., 2009) and can genetically enhance the gish^IR phenotype (Fig. S5, J–M, quantification). CK1-γ/Gish may promote delivery of Rab11–Nuf–Sec15 recycling vesicles carrying Arp2/3 activity to the prehair. Further experiments are needed to explore the relationship between localized Rab11 trafficking and the branched nucleation machinery.

As Gish is associated with the base of developing trichomes, it is possible that CK1-γ/gish, in analogy to the yeast homologue Yck2p, is localized to the proximal prehair membrane in an area of membrane deposition. Such localization suggests Gish could regulate localized trafficking by tethering Rab11 vesicles to the plasma membrane through the Sec15–exocyst complex. As in other contexts, Rab11 and Sec15 are functionally linked on vesicles en route from the recycling endosome to the plasma membrane. As gish reduction mislocalized endogenous Rab11–Nuf–Sec15 and dissociated the aggregated Sec15GFP and Rab11 colocalization pattern (but had no effect on alternative trafficking compartments, such as the Rab5 early endosome), this suggests CK1-γ/gish is required at a late vesicle recycling step. Although coimmunoprecipitation experiments with Myc-Gish failed to detect an association with Rab11, Nuf, or Sec15 (unpublished data), the potential phosphorylation targets may be binding partners or associate through weak/transient interactions with Gish.

Alternatively, CK1-γ/gish may regulate polarized vesicle trafficking through an effect on the cytoskeleton, associated motor proteins, or adaptors. As mentioned, the Rab11 effector Nuf can bind the motor protein dynein and link Rab11 endosomal structures to microtubules (Riggs et al., 2007; Horgan et al., 2010). A recent study showed that Nuf phosphorylation by IKK-ε regulates trafficking of Rab11 vesicles along developing bristles (Otani et al., 2011). Thus, one possibility is that CK1-γ/gish may act through Nuf to affect Rab11 trafficking. We have observed a correlation between Rab11 localization and microtubule networks within the prehair (unpublished data). Thus, Rab11 vesicles may require microtubules for localization, as previously reported in other contexts (Riggs et al., 2007). Consistent with this notion, microtubule depolymerization with nocodazole dissociated Sec15 vesicle aggregates in mammalian cells (Zhang et al., 2004). Intriguingly, disrupting microtubules resulted in multiple trichomes in Drosophila (Turner and Adler, 1998). Further analysis is necessary to address the role of gish on microtubule-based vesicle trafficking in the context of trichome formation.

PCP studies have identified effectors required to restrict trichomes, such as fuzzy, inturned, frtz, and mwh (Adler, 2002). Based on genetic epistasis and localization experiments, mwh is downstream as a modulator of the cytoskeleton (Wong and Adler, 1993; Strutt and Warrington, 2008; Yan et al., 2008). Mwh is enriched proximally, and a model has been proposed whereby prehair initiation is restricted to the distal cell region by the proximal repression of Mwh (and promoted by Fz in the distal domain). Taking these observations together with our data regarding the relationship of mwh and CK1-γ/gish, the restriction of prehair initiation requires two parallel mechanisms in orthogonal axes: (1) prehair initiation is restricted by the gradient of Mwh along the PD axis (e.g., wing) of the cell and (2) the tight restriction of Fz-directed trichome nucleation along a second axis of refinement by CK1-γ/gish (Fig. 9). This model is supported by both genetic and cell biological data: (a) analyses of CK1-γ/gish and mwh LOF reveal phenotypic differences—the initial prehair phenotype of CK1-γ/gish is restricted to the distal cell region and results in multiple, distally oriented trichomes, whereas the mwh phenotype is first observed as excess actin filaments over the entire apical cell cortex and results in multiple, randomly oriented trichomes (Strutt and Warrington, 2008), (b) gish^IR strongly enhances the mwh-null phenotype, indicative of two genetically independent pathways, and (c) CK1-γ/gish is required for Rab11 localization in the proximity of the developing prehair, whereas mwh has no effect on Rab11 (or Sec15GFP). Collectively, these data imply that these proteins perform two independent functions in the cell that converge to restrict trichome formation to a single site. In support of this model, our data indicate that excess of either of these proteins can partially suppress the defects associated with the loss of the other. Thus, we suspect a failure to focus trichome nucleation activity to the distal cell vertex by gish LOF can be corrected by increasing Mwh levels to repress actin nucleation/polymerization by an independent mechanism.

In summary, we define a novel mechanism by which trichome formation is restricted to a single domain in epithelial cells. These data support the model that CK1-γ kinases can regulate cellular morphogenesis through controlling the localization of Rab11 vesicle recycling. Our study has identified parallels between Drosophila and yeast CK1-γ in cellular morphogenesis, supporting a conserved mechanism of action. Collectively, our findings reveal that an independent mechanism of CK1-γ–regulated vesicle trafficking converges to refine Fz/PCP-directed morphogenesis.

Drosophila experiments were performed at 25°C unless otherwise indicated. Deficiency collection stocks used in the modifier screen were obtained from Exelixis, Szeged, and Bloomington Stock centers. Phenotypic analysis was performed on genes isolated in the modifier screen through transgenic UAS-RNAi flies obtained from the Vienna Drosophila RNAi Center (Dietzl et al., 2007). The following alleles were used in this study and obtained from the Bloomington Stock Center: Drok², zip¹, gish^e01759, gish^spider-GFP, mwh¹, and scar^Δ37 (described in Flybase). wsp³ was a gift from E.D. Schejter (Weizmann Institute of Science, Rehovot, Israel; Ben-Yaacov et al., 2001). gish^e01759 is a transposon insertion allele and has been characterized as a strong hypomorphic mutation (Jia et al., 2005). gish^e01759 was recombined onto an FRT82B chromosome, and mitotic clones of FRT82B-gish^e01759 were generated via the FLP/FRT system (Xu and Rubin, 1993). Clones were unmarked or marked with the forked mutation in adults or marked by the absence of β-galactosidase in pupal tissue. gish^spider-GFP is a protein trap insertion within the gish locus, generating a Gish-GFP fusion protein (Morin et al., 2001; Frescas et al., 2006). Overexpression of cDNA transgenes or RNAi (IR) was performed using the Gal4/UAS system (Brand and Perrimon, 1993). The Gal4 expression drivers used were as follows: sev-Gal4, en-Gal4, nub-Gal4, and dpp-Gal4. FLP-out expression clones of the indicated genotypes were performed using hs-FLP; actin>y>Gal4, UAS-GFP (Struhl and Basler, 1993). UAS-gish^IR (v26003), UAS-mwh^IR (v41514), UAS-Rab11^IR (v22198), UAS-Sec15^IR (v105126), and UAS-PTEN^IR (v35731) were obtained from the Vienna Drosophila RNAi Center collection. Where indicated, UAS-dicer2 was included with UAS-IR expression to increase RNAi efficiency (Bloomington Stock Center). The following transgenes were also used: UAS-sqh^E20E21 (Winter et al., 2001), UAS-mwh (Yan et al., 2008), UAS-Sec15GFP (Jafar-Nejad et al., 2005), UAS-Rab11^IR (Satoh et al., 2005), UAS-YFP-RabGTPases (including wild type, dominant negative, and constitutively active; Zhang et al., 2007), arm-fz-GFP (Strutt, 2001), actin-stbm-YFP (Strutt et al., 2002), and UAS-kette^Myr (Bogdan and Klämbt, 2003). Temperature shifting experiments were performed via an 18–29°C shift during the third-instar larval stage using the tub-Gal80^ts transgene (Bloomington Stock Center).

The following transgenes were generated from the Berkeley Drosophila Geneome Project gish cDNA clone LD04357: UAS-gish^IR2 was created by PCR amplifying a cDNA sequence (independent of the Vienna Drosophila RNAi Center probe sequence; Fig. S1 A) into a modified pWizDir vector (Jenny and Mlodzik, 2008). The PCR primers used to generate gish^IR2 were (including restriction sites) forward, 5′-ATCCCTAGGC­TCCACCGGATC­GAATATGTT-3′, and reverse, 5′-GCTGGATCCC­ACCGATTTTTT­GATATCTCT-3′. UAS-myc-gish^WT and UAS-myc-gish^ΔC were PCR isolated and correspond to transcripts gish-RB and gish-RF, respectively (described in Flybase). The PCR primers used to generate these transcripts were (including restriction sites and linkers) forward primer (common for gish-RB and gish-RF), 5′-ATCGAATTCG­GCGGCATGCAG­CGACGAGAACG­GCAA-3′, reverse (gish-RB only), 5′-GCTCTCGAGT­CATTTTTGGCG­CGTCGATTT-3′, and reverse (gish-RF only), 5′-GCTCTCGAGC­TATGTCTCCAT­TGTCTTCCC-3′. These PCR products were cloned into pCS2-Myc, and myc-gish was subcloned into the pUAST vector to generate transgenic flies. myc-gish^KD (D187N substitution) was generated by site-directed mutagenesis (Agilent Technologies) of pCS2-myc-gish^WT followed by subcloning of myc-gish^KD into pUAST. The PCR primers used to generate the D187N substitution were forward, 5′-GGCACTTAATATATAGGAATGTGAAACCAGAGAAC-3′, and complementary reverse, 5′-GTTCTCTGGTTTCACATTCCTATATATTAAGTGCC-3′. Substituted nucleotides are in bold. The gish kinase-dead D187N mutation is based on the comparable Xenopus laevis CK1-γ mutant on a conserved residue in the ATPase domain (Davidson et al., 2005).

For analysis of wing trichomes, adult wings were removed, incubated in wash buffer (PBS and 0.1% Triton X-100), and mounted on a slide in 80% glycerol in PBS. To analyze trichomes in adult nota (dorsal thorax), flies were partially dissected, incubated at 95°C in 10% KOH for 10 min to clear fat tissue, washed (PBS and 0.1% Triton X-100), and then placed in 80% glycerol in PBS. Nota were then fully dissected and mounted on a slide in 80% glycerol in PBS. Adult eye section analyses were performed as previously described (Gaengel and Mlodzik, 2008). In brief, fly heads were fixed in 2% glutaraldehyde in PBS and treated with 2% osmium tetroxide. Samples were dehydrated in a graded ethanol series (50, 70, 90, and 100%) and propylene oxide and embedded in Durcupan resin. Tangential sections were made and mounted on a slide using DPX mounting medium. Adult wings, nota, and eye sections were imaged at room temperature on a microscope (Axioplan; Carl Zeiss). Images were acquired with a camera (Zeiss AxioCam Color type 412–312; Carl Zeiss) and AxioCam software. For analysis of pupal wings, white pupae were collected (0 h APF) and aged at 25°C. Dissections were performed as follows: in brief, pupae were immobilized on double-sided tape, removed from the pupal case, and placed into PBS, in which pupae were partially dissected to remove fat tissue and then fixed in 4% formaldehyde in PBS and washed three times (PBS and 0.1% Triton X-100). Wing membranes were removed, and tissue was incubated in wash buffer containing 10% normal goat serum overnight for primary antibody (4°C), washed three times with PBS, and then incubated with secondary for 1 h (25°C). Wings were washed three times with PBS and mounted in glycerol/PBS supplemented with 1% N-propyl gallate. Pupal wing images were acquired at room temperature using a confocal microscope (63×, 1.4 NA; LSM 510 Meta; Carl Zeiss) using LSM software (Carl Zeiss). Images were processed with ImageJ (National Institutes of Health) and Photoshop (CS4; Adobe).

Rabbit anti–β-galactosidase (1:1,000; European Molecular Biology Laboratory), rat anti-Rab11 (Dollar et al., 2002), rabbit anti-Rab5 (1:1,000; ab31261; Abcam), guinea pig anti-Hrs (Lloyd et al., 2002), guinea pig anti-Sec15 (1:500; Mehta et al., 2005), rabbit anti–Lava lamp (Sisson et al., 2000), mouse anti-Fmi (1:10; Developmental Studies Hybridoma Bank), rabbit anti-MyoV (1:1,000; Li et al., 2007), rabbit anti-dRip11 (1:1,000; Li et al., 2007), rabbit anti-Sec5 (Murthy et al., 2003), rabbit anti-Nuf (1:1,000; Riggs et al., 2003), mouse anti-Myc (1:250; 9E10; Santa Cruz Biotechnology, Inc.), mouse anti-En (1:2; Developmental Studies Hybridoma Bank), rabbit anti-Zip (Liu et al., 2008), and rabbit anti-GFP (1:1,000; Invitrogen). Rhodamine-phalloidin was obtained from Invitrogen. All fluorophore-conjugated secondary antibodies, including Cy5, TRITC, and FITC, were used at 1:200 and were obtained from Jackson ImmunoResearch Laboratories, Inc. Donkey anti–rabbit 10-nm gold was used at 1:200 (ab27235; Abcam).

Cell area measurements were performed using ImageJ. The apical cell periphery was identified by Fmi staining (overlapping apical adherens junctions), and basolateral area measurements were identified by rhodamine-phalloidin staining of cortical actin basal to nuclear level (Hoechst 33342; Thermo Fisher Scientific). Individual cell membranes were traced using the polygon drawing tool, and measurements were determined. Measurements were performed on ≥30 cells for control and adjacent mutant tissue (32–34 h APF). The percent increase of the area of mutant tissue was calculated relative to adjacent control tissue in the same wing. Means and SDs were calculated from at least three wings from individual animals. Sec15GFP puncta number per cell was determined using the particle analysis function of ImageJ. Three independent wings were analyzed, and the number of Sec15GFP-positive cells in each field was used to determine the mean number of puncta per cell.

White pupae (0 h APF) of the indicated genotypes were collected into separate vials, aged to 36 or 38 h APF at 25°C, and mounted for imaging as previously described (Bellaïche et al., 2001). In brief, aged pupae were fixed on slides with double-sided tape in between stacks of four coverslips. Pupal cases were partially removed to expose head and notum. A drop of halocarbon oil (Sigma-Aldrich) was placed onto the bottom of a cover slide and gently applied to the notum, supported by the adjacent stacks of coverslips. Images were acquired at room temperature using a confocal microscope (63× oil immersion, 1.4 NA; SP5 DMI; Leica) with LAS AF (Leica) software. A single confocal plane (1 µm) was taken at 6-s intervals for a total of 2 min.

The dextran uptake assay was performed as previously described (Entchev et al., 2000) with some modifications: third-instar larvae were partially dissected in S2 medium supplemented with 10% FBS to expose wing discs. The medium was replaced with one containing 5 mg/ml Texas red dextran (lysine fixable; 3,000 molecular weight; Invitrogen) and pulsed for 10 min at 25°C. The samples were washed three times with ice-cold S2 medium (with 10% FBS) and then incubated at 25°C for 20 min (chase) to visualize uptake and the early endosomal compartment. The samples were fixed for 20 min in 4% formaldehyde in PBS and washed three times in PBS with 0.1% Triton X-100. Mounting was in 30% glycerol (Mowiol and 2.5% Dabco). Images were acquired at room temperature using a confocal microscope (63×, 1.4 NA; LSM 510 Meta) using LSM software. Images were processed with ImageJ and Photoshop CS4.

White pupae (0 h APF) of the indicated genotypes were isolated and staged to 32 h APF, and pupal wings were dissected as described in Immunohistochemistry and histology. Wings were fixed in 3% glutaraldehyde and then in 1% osmium tetroxide. Samples were dehydrated in a graded ethanol series (50, 70, 90, and 100%) and propylene oxide and embedded in Epon resin. Ultrathin sections were made along oblique angles to the apical epithelial surface of pupal wing epithelia. Immuno-EM sample preparation was as previously described (Shimada et al., 2006). In brief, white pupae (0 h APF) of the indicated genotype was isolated, staged to 32 h APF, dissected in PBS, and fixed for 1 h in periodate-lysine-paraformaldehyde fix. Pupae were washed three times in PBS. Blocking was performed by incubation for 1 h in glycine-PBS supplemented with 4% normal donkey serum, 0.1% saponin, and 0.05% Triton X-100. The primary antibody was diluted in blocking solution and incubated overnight at 4°C. Pupae were washed five times in glycine-PBS and then incubated for 3 h with secondary antibody and washed five times. All subsequent steps were as described in the previous paragraph for standard TEM. The electron microscope used in all cases was an H-7650 (Hitachi) with MaxIm DL software (Diffraction Ltd.).

Third-instar wing imaginal discs (50 discs/genotype) were dissected and placed directly into SDS sample buffer to dissolve the tissue. These samples were boiled at 95°C for 10 min and then centrifuged at 14,000 rpm for 10 min. The supernatant was run on a 10% SDS-PAGE gel and transferred to polyvinylidene fluoride membrane (Millipore). The membranes were probed with rabbit antiphospho-MRLC (Ser19; Cell Signaling Technology). ECL Plus was used for detection (GE Healthcare).

293T cells were cultured in DME supplemented with 10% FBS and maintained in 5% CO₂ at 37°C. Cells were transfected with siRNA against human CSNK1G1, CSNK1G2, and CSNK1G3 or negative control siRNA (Silencer Select siRNA and negative control #1; Invitrogen). Transfection was performed using siPORT NeoFX (Invitrogen). After 48 h, cells were transfected with pEGFP-Sec15GFP (0.4 µg per well; 8-well Laboratory-Tek glass chamber slide; a gift from C.A. Mitchell, Monash University, Clayton Victoria, Australia; Zhang et al., 2004) with the transfection reagent (FuGENE HD; Promega). After 24 h, cells were fixed with 4% paraformaldehyde, washed in PBS (0.1% Triton X-100), and mounted with Vectashield (including DAPI; Vector Laboratories). Images were acquired using a confocal microscope (63×, 1.4 NA; LSM 510 Meta) using LSM software. Images were processed with ImageJ and Photoshop CS4.

Fig. S1 shows in vivo RNAi strategy used for gish knockdown and the requirement for CK1-γ/Gish kinase activity and membrane association in trichome formation and Rab11 localization. Fig. S2 displays data characterizing the apical membrane expansion phenotype of gish^IR. Fig. S3 displays data supporting a PCP-independent role for CK1-γ/gish in trichome morphogenesis. Fig. S4 shows control stainings of additional trafficking compartments and Rab11-associated proteins. Fig. S5 shows the effect of gish on YFP-Rab11^CA localization, YFP-Rab11^WT localization in the context of trichome formation by immuno-EM, Rab11^IR knockdown of endogenous Rab11, and genetic data implicating Arp2/3 nucleators in gish regulation of trichome morphogenesis. Video 1 shows YFP-Rab11^WT trafficking within initiating prehair in a 36-h APF notum. Video 2 shows diffuse localization of YFP-Rab11^WT with gish^IR coexpression in a 36-h APF notum. Video 3 shows YFP-Rab11^WT trafficking within elongating prehair in a 38-h APF notum. Video 4 shows diffuse YFP-Rab11^WT trafficking with gish^IR coexpression during prehair elongating in a 38-h APF notum.

We are grateful to Paul Adler, Hugo Bellen, William Sullivan, Ken Prehoda, Jennifer Zallen, Thomas Schwarz, Christina Mitchell, Robert Cohen, Ronald Davis, Donald Ready, Nicholas Baker, and the Bloomington, Exelixis, Szeged, and Vienna Drosophila RNAi Center stock centers for fly stocks and reagents. We thank Jeffery Bush, Robert Krauss, and Alvaro Glavic for advice and critical reading of the manuscript and members of the Mlodzik laboratory for critical comments and support. We also thank Suzanna Franks for diagrams and Aurore Dussert, Andrea Blitzer, and Sarah McGowan for technical support.

This work was supported by National Institutes of Health grant GM62917 to M. Mlodzik. W.J. Gault was also supported by award number T32CA078207 from the National Cancer Institute.