Rab8a and Rab8b are essential for several apical transport pathways but insufficient for ciliogenesis

The small GTP-binding protein Rab8 is reported to be involved in basolateral transport in epithelial cells (Huber et al., 1993a; Ang et al., 2003; Henry and Sheff, 2008) and in dendritic transport in neurons (Huber et al., 1993b). Rab8 is also reported to be involved in trafficking of neurotransmitter receptors in postsynaptic terminals (Gerges et al., 2004) and the formation of outer photoreceptor segments in the retina (Moritz et al., 2001; Bachmann-Gagescu et al., 2011). We have shown that Rab8a, an isoform of Rab8, is essential for localizing apical proteins in the small intestine by analysing Rab8a-knockout (AKO) mice (Sato et al., 2007). In addition, AKO mice display a phenotype that is strikingly similar to that shown in the human disease microvillus atrophy. One of the genes responsible for this hereditary disease is myosin Vb (Müller et al., 2008), which binds Rab8 (Roland et al., 2007). Thus, Rab8 is closely associated with the pathogenesis of this disease. Recently, the human disease Bardet-Biedl syndrome (BBS) was determined to result from mutations of a group of genes within the BBSome complex (a complex of seven BBS proteins and BBIP10), which is associated with the formation of cilia (Nachury et al., 2007). Interestingly, Rabin8, a guanine nucleotide exchange protein that activates Rab8, binds this complex for ciliogenesis (Nachury et al., 2007; Westlake et al., 2011). Using different experimental systems, other groups independently reported that Rab8 is involved in the elongation of ciliary membranes (Yoshimura et al., 2007; Omori et al., 2008). Thus, Rab8 is currently regarded as being involved in the formation of cilia and the pathogenesis of BBS.

However, it has been difficult to determine the function of Rab8 for several reasons. Rab8 consists of the two closely related proteins Rab8a and Rab8b, which are encoded by different genes (Armstrong et al., 1996). Both proteins are highly similar in their sequence and are expressed at similar levels. Therefore, knocking out the Rab8a gene is insufficient to uncover the role of Rab8 because Rab8b might compensate for Rab8a. In addition, regarding Rab8, an apparent contradiction was observed in previous findings that describe its role in basolateral transport, and our findings that indicate it has a role in apical transport. Thus, to determine the function of Rab8, an examination of the function of Rab8b is crucial. In addition, Rab8 has been proposed as being essential for ciliogenesis and for the pathogenesis of ciliopathies, including BBS. Therefore, determining the functional role of Rab8 has both scientific and clinical importance. To determine its role in polarised transport in epithelial cells, neurons and in ciliogenesis, we generated Rab8b knockout (BKO) mice and Rab8a and Rab8b double-knockout (DKO) mice.

To determine the function of Rab8 in intracellular transport and ciliogenesis, we generated BKO mice (Fig. 1A,B) and confirmed that the Rab8b transcript was, indeed, absent in the BKO mice (Fig. 1C). As we were unable to generate a Rab8b-specific antibody, despite a number of trials, we used an antibody that recognised both Rab8a and Rab8b for western blot analysis of the control, AKO, BKO and DKO mice (Fig. 2A). Because we were unable to detect any Rab8 protein in the intestines of the DKO mice, the BKO mice were assumed to be deficient in the Rab8b protein. The BKO mice neither displayed any overt abnormalities (Fig. 1D) nor showed any apparent histological abnormalities in the various tissues analysed (Fig. 1E).

To determine whether any redundancy exists between Rab8a and Rab8b, we generated DKO mice. When we intercrossed Rab8a^+/−; Rab8b^−/− males and females, the DKO mice were born at a Mendelian ratio, but gradually died 2 weeks after birth. Almost all DKO mice had perished by postnatal week 3 (Fig. 2B), approximately 1 week earlier than Rab8a single-knockout mice (Sato et al., 2007). To detect abnormalities in the tissues of the DKO mice, we examined their livers and kidneys. Immunofluorescence microscopy of the tissues failed to detect any abnormalities until postnatal week 2 (Fig. 2C,D). Previous results have indicated the involvement of Rab8 in the dendritic transport process in neurons (Huber et al., 1993b). We, therefore, used the cerebellum to test this result in vivo. Because Purkinje cells are the only calbindin-positive cells in the cerebellum, the axons of Purkinje cells in the deep cerebellar nuclei (DCN) region can be easily identified. Previously, dentritic markers were found to be mislocalised to the axons of Purkinje cells of mice in which the adaptor protein AP4 had been knocked out (Matsuda et al., 2008). However, we did not observe a mislocalisation of the dendritic marker GluA1 in our DKO mice until postnatal week 2 (Fig. 2F). In addition, the shape of the Purkinje cell dendrites was similar in control and DKO mice (Fig. 2E).

Having previously identified the mislocalisation of apical markers in AKO mice at postnatal day 17 (P17) (Sato et al., 2007), we next examined the small intestine of the DKO mice at different postnatal (P) days. When we examined the localisation of apical markers, such as aminopeptidase N (APN), dipeptidyl peptidase IV (DPPIV) and sodium-dependent glucose transporter (SGLT), their localisation was similar in control and DKO mice small intestine at P7 (Fig. 3A). However, we found intracellular accumulation of these markers in DKO enterocytes from P10, i.e. 1 week earlier than those found in the AKO enterocytes. Interestingly, the pattern of intracellular accumulation differed depending on the apical markers. APN (Fig. 3A, arrows) and DPPIV (supplementary material Figs S1, S2, arrows) displayed punctate accumulation, whereas SGLT displayed a diffuse staining pattern (Fig. 3B). Interestingly, APN and SGLT presented mutually exclusive localisation (Fig. 3C, top). A similar mutually exclusive staining pattern was also observed in the AKO mice at later postnatal ages (Fig. 3C, bottom). Because SGLT was also stained in mutually exclusive manner to Lamp2, SGLT was thought to localise to a structure distinct from the lysosomes (Fig. 3D).

Another apical marker, sucrase-isomaltase (SI), did not – unlike APN or SGLT – accumulate subapically in the DKO mice at P10, but localised to the apical plasma membrane at P14 (Fig. 4A,B). In the DKO, SI began to localise near the nucleus at P7 (Fig. 4B, indicated by an arrow in P7 DKO) and at the apical plasma membrane at P10 (Fig. 4B, indicated by an arrow in P10 DKO). At the later stages, SI mislocalised to the basolateral plasma membrane at P12 (Fig. 4A,B, indicated by arrowheads in P12 DKO and in Fig. 4D) and, finally, to the lysosomes at P14 (Fig. 4B, indicated by an arrowhead in P14 DKO and in Fig. 4E). Perinuclear SI staining at P7 in DKO enterocytes, and at P10 and P12 in control enterocytes (Fig. 4B, arrows in P10 and P12 Ctrl) corresponded to the Golgi complex (Fig. 4F,G; arrows). Interestingly, Rab8a also localised to the Golgi (Fig. 4H, arrows). Colocalisation of Rab8a and SI at these ages suggest that Rab8a interacts with SI at the Golgi to retard the transport of SI. This assumption is supported by the fact that the loss of Rab8 speeds up apical localisation of SI. These observations indicate that there are at least three pathways that lead to the apical plasma membrane.

In contrast to apical markers, the basolateral markers did not accumulate intracellularly, as observed in the AKO mice (supplementary material Fig. S2B,C). Thus, both Rab8a and Rab8b are involved in apical transport rather than basolateral transport. The microvillus atrophy-like phenotype of the DKO mice was very similar to that of the AKO mice, although the onset of microvillus atrophy in DKO is about 1 week earlier than the one in AKO mice. When using electron microscopy, we observed a reduction of microvilli (Fig. 5A,B) and the appearance of microvillus inclusion bodies (Fig. 5C) in DKO mice at P14, the time point when AKO mice appeared normal. These phenotypes indicate a synergistic interaction between Rab8a and Rab8b.

Recently, a number of papers have reported that Rab8 is essential for providing ciliary membranes and, thus, for the elongation of cilia (Nachury et al., 2007; Yoshimura et al., 2007; Omori et al., 2008). In recent years it was established that BBS is caused by the mutation of genes, whose products form the BBSome complex, which is essential for ciliogenesis. BBSomes have been shown to interact with Rabin8, a guanine exchange factor for Rab8, indicating the involvement of Rab8 in ciliogenesis (Nachury et al., 2007). Indeed, when Rab8 is knocked down, the lengths of the cilia are reduced (Yoshimura et al., 2007). Thus, to determine the function of Rab8 in ciliogenesis in vivo, we observed primary and motile cilia, outer disc membranes and cilium-derived structures in the photoreceptor neurons (Deretic et al., 1995; Moritz et al., 2001).

First, the morphology of the cells and the lengths of the cilia were not significantly different in DKO and control mouse embryonic fibroblasts (MEFs) (Fig. 6A,B). In addition, the percentage of ciliated cells is similar in DKO and control MEFs (Fig. 7A,B). Second, the structures of the retina and the outer segments of the photoreceptors are similar in DKO and control MEFs (Fig. 6C,D). Third, when we examined the olfactory epithelium, we observed a layer of cilia clearly stained by acetylated tubulin (supplementary material Fig. S3), which is in contrast to the observed marked reduction of cilial staining in BBS1- or BBS4-deficient mice (Kulaga et al., 2004). As a representative of a motile cilium, we examined the ciliated epithelium in the trachea using electron microscopy. The overall structures of the ciliated cells were similar in control and DKO MEFs (supplementary material Fig. S3). The ‘9+2’ microtubule structure in the motile cilia was observed in the DKO as well as the control mice (supplementary material Fig. S3, inset). The structure and length of the motile cilia appeared similar in DKO and the control mice (Fig. 6E,F).

From these observations, and the absence of cysts in the kidney (Fig. 1E) and retinal degeneration (Fig. 6A,B), both of which are hallmarks of some ciliopathies, we concluded that Rab8a and Rab8b are not sufficient for the formation of cilia, at least during the lifespan of the DKO mice.

Because there are several species of Rabs (Rab10 and Rab13) in the Rab8 family, we knocked down Rab10 and Rab13 together in Rab8 DKO and control MEFs to determine the functional redundancy within the Rab8 family (supplementary material Fig. S4). When we knocked down these Rabs in DKO MEFs, the percentage of MEFs with primary cilia was greatly reduced (Fig. 7A,B). The percentage was almost equal to that of ciliated MEFs deficient in intraflagellar transport 20 and 88 (Ift20 and Ift88 respectively) by double knockdown (Fig. 7B), which is the most potent knockdown treatment on ciliogenesis (Clement et al., 2009). To understand which Rab is more important, we knocked down either Rab10 or Rab13 in Rab8 DKO MEFs (Fig. 7C). We found that Rab10 knockdown is as effective as a double knockdown of Rab10 and Rab13 in DKO MEFs, but Rab13 knockdown had only a subtle effect on ciliogenesis (Fig. 7C). To confirm which combination of Rab8 family proteins does affect ciliogenesis, we used three knockdown oligonucleotides in wild-type MEFs and measured the rate of ciliated cells. Knocking down Rab8a, Rab8b and Rab10 in wild-type MEFs had the most effect on ciliogenesis, whereas other combinations of knocked out Rab proteins had almost no effect (Fig. 7D). Expressing knockdown-resistant Rab10 cDNA into the Rab10 knockdown DKO MEFs rescued the ciliary phenotype (Fig. 7E). Collectively, these results suggested that Rab8a, Rab8b and Rab10 work synergistically for ciliogenesis, whereas Rab13 is not required.

Rab8 has been thought to be involved in basolateral transport in epithelial cells and dendritic transport in neurons (Huber et al., 1993a; Huber et al., 1993b). In addition, several groups have reported that Rab8 is involved in ciliogenesis (Nachury et al., 2007; Yoshimura et al., 2007). Previously, we have shown that Rab8a is essential for apical transport (Sato et al., 2007). However, Rab8a has a closely related isoform, Rab8b. Thus, it is possible that Rab8b functions in basolateral transport. Moreover, the function of Rab8a and Rab8b in neurons and ciliated epithelium was not examined in our previous study. To clarify their roles, we generated Rab8a–Rab8b double-knockout mice and found that both proteins function synergistically in apical transport but do not have a function in basolateral and dendritic transport.

Knocking out both the Rab8 isoforms showed that there are at least three pathways to the apical plasma membrane. The first pathway transports DPPIV and APN. In the absence of Rab8, these cargos accumulate in the lysosomes. The second pathway transports SGLT and in the absence of Rab8, SGLT accumulates in an unknown structure that is different from the lysosomes. The third pathway transports SI. In the DKO intestine, SI is transported to the apical plasma membrane earlier than it is in the control intestine (Fig. 4A,B). As the localisation of SI resembles that of Rab8a in the Golgi, Rab8a might interact with SI in the Golgi, thereby regulating its transport.

Previously, many researchers suggested that Rab8 is essential for ciliogenesis. However, in our current study, when observing Rab8a and Rab8b DKO mice, the absence of both Rab8 isoforms had no effect on the generation of primary cilia (olfactory epithelium, retina outer segment and MEF) or motile cilia (trachea). As the DKO mice died within 3 weeks after birth, Rab8a and Rab8b might function after the establishment of cilia. Previously, either Rab8 isoform was shown to function by itself in ciliogenesis (Nachury et al., 2007; Yoshimura et al., 2007). However, our results indicate that Rab8a, Rab8b and Rab10 function simultaneously in ciliogenesis rather than alone, whereas Rab13 does not seem to be necessary for the process. The fact that Rab10 is involved in generating cilia in the apical membrane is surprising because it was thought to be involved in basolateral transport (Chen et al., 2006; Schuck et al., 2007). A previous report describes the localisation of Rab10 on the cilia and the interaction of Rab10 with the exocyst complex, but did not present evidence that Rab10 is involved in ciliogenesis (Babbey et al., 2010). The involvement of Rab10 in ciliogenesis in the MEFs might be explained by the fact that MEFs are not polarised. Several papers have described the specialised membranous area that segregates the cilia from the surrounding apical membrane (Nachury et al., 2010). Thus, it is also possible that transport to the basolateral plasma membrane and ciliary membrane use common molecular machinery. For further elucidation of the mechanism of apical transport and ciliogenesis, more studies – including Rab10-knockout studies – are necessary.

Rab10 is also reported to be necessary for axonal membrane trafficking and axonal elongation (Wang et al., 2011b; Liu et al., 2013). Rab13, a Rab closely related to Rab8 and Rab10, is reported to be necessary for neurite elongation (Sakane et al., 2010). Additionally, Rab17 and Rab22, which belong to different subfamilies of Rab8, have been reported to be necessary for dendrite (Mori et al., 2012) and neurite elongation (Wang et al., 2011a), respectively. Thus, no obvious abnormality in the nervous system in Rab8a and Rab8b DKO mice could be explained by compensation through Rab17 and Rab22. Recently, Rab10 has been shown to be involved in various kinds of intracellular trafficking events. Beside any aforementioned role in neurite elongation and ciliogenesis, Rab10 has recently proposed to be involved in ER dynamics (English and Voeltz, 2013) and the fusion of the Glut4 storage compartment to the plasma membrane (Chen et al., 2012). To assess these roles of Rab10 and the fact that Rab10 might compensate the lack of other Rabs in vivo, we believe that Rab10-knockout studies in any model organism (e.g. mouse, worm, fly) are necessary, as well as studies that use organisms in which Rab10 and other Rab family proteins are knocked out simultaneously.

All animal procedures were performed within the guidelines of the Animal Care and Experimentation Committee of Gunma and Osaka University, and all animals were bred at the Institute of Animal Experience Research of Gunma and Osaka University.

Generation of Rab8b knockout (BKO) mice was performed largely according to a previous report (Harada et al., 1994). Rab8b genomic clones were isolated from a mouse genomic BAC library from the 129Sv/J mouse strain (RPCI-22: Children's Hospital Oakland Research Institute, Oakland, CA), using a fragment of the mouse Rab8b second intron (indicated as probe 2 in Fig. 1A) as a probe. The targeting vector consisted of a 1-kb 5′ homologous region, a SA-IRES-βgeo-polyA cassette flanked with two FRT sites and a downstream loxP site (Sato et al., 2007), a 1-kb genomic sequence region including exon 2 (from 350 bp upstream of exon 2 to 400 bp downstream of exon 2), a single loxP site, and a 6.7-kb 3′ homologous region. Two targeted clones (B19, C21) were identified by Southern blot analysis using probes 1 and 2 (Fig. 1B,C). To generate the null mice, we crossed Rab8b βgeo/+ mice with Act-Flp-e transgenic mice and then with CMV-cre transgenic mice (Jackson Laboratory, Bar Harbor, ME).

Mice at various postnatal days were anesthetised, injected intracardially with 3% paraformaldehyde (PFA) in 0.1 M phosphate buffer (pH 7.2) and kept for another 2 hours in that same fixative. Fixed tissues were soaked in 4, 10, 15 and 20% sucrose in 0.1 M phosphate buffer (pH 7.2) at 4°C for more than 30 minutes each step for cryoprotection. The tissues were then frozen in isopentane chilled in liquid nitrogen and stored in liquid nitrogen until cryosectioning. The tissues were cut at 5–10 µm, except for the cerebellum (shown in the next section).

Antibodies, dye and lectin used: APN (1∶100; BMA Biomedicals, Augst, Switzerland); DPPIV (1∶100; R&D Systems, Minneapolis, MN); Lotus-lectin (1∶100; Honen, Tokyo, Japan); DAPI (1∶1000; 4′,6-Diamidine-2′-phenylindole dihydrochloride; Roche, Basel, Switzerland); Rab8 (1∶1000 for WB; BD, San Jose, CA); SI (1∶100; a gift from S. Matsumoto; Umesaki et al., 1982); Rab8a (1∶100; Sato et al., 2007); SGLT (1∶100; Santa Cruz, Dallas, TX); Lamp2 (1∶100; clone Abl-93; Developmental Studies Hybridoma Bank, IA); E-cadherin (1∶50; TAKARA, Otsu, Shiga, Japan); and Na⁺-K⁺ ATPase (1∶100; a gift from H. Homareda; Homareda et al., 1993); acetylated tubulin (1∶100; Sigma). We used our antibody for Rab8a for immunofluorescence (Sato et al., 2007) because an antibody that recognises both Rab8a and Rab8b (BD, San Jose, CA) did not work for immunofluorescence in our hands. For secondary antibodies, Alexa-Fluor-488- or Alexa-Fluor-546-labeled species-specific secondary antibodies (1∶400; Life Technologies, Carlsbad, CA) were used.

Histological assays for mouse cerebellar slices were performed as previously described (Hirai et al., 2005). Briefly, micro-slicer sections were incubated overnight with the following primary antibodies: mouse anti-calbindin (1∶1000; Swant, Marly, Switzerland) and rabbit anti-GluA1 (1∶200; Millipore, Billerica, MA). Sections were incubated for 1 hour with Alexa-Fluor-488- or Alexa-Fluor-546-labeled species-specific secondary antibodies (1∶1000; Life Technologies, Carlsbad, CA). The stained slices were analysed by using confocal microscopy (Olympus, Tokyo, JAPAN) as described previously (Hirai et al., 2005).

For immunofluorescence microscopy, DKO and control mice were anaesthetised and perfused with phosphate-buffered saline (PBS), followed by 4% PFA in PBS. The nose, containing the olfactory epithelium, was removed, post-fixed for 3 hours in the same fixative, de-calcified in 0.5 M EDTA for 2 days, and cryoprotected by immersion in 20% sucrose (Koshimoto et al., 1992). The frontal sections (20 µm thickness) of the nasal epithelium were obtained using a cryostat and mounted on MAS-coated slide glass (Matsunami, Osaka, Japan). The sections were treated with 0.2% Triton X-100 in PBS containing 5% normal horse serum for 60 minutes and then incubated overnight at room temperature with a rabbit anti-golgin 97 antibody (1∶500; a gift from N. Nakamura; Yoshimura et al., 2004) and mouse anti-acetylated tubulin antibody (1∶1000; Sigma). After washing in PBS, the sections were incubated with 3.75 µg/ml Cy3-conjugated donkey anti-rabbit IgG (Jackson ImmunoResearch, West Grove, PA) and 5 µg/ml Alexa-Fluor-488-conjugated donkey anti-mouse IgG (Invitrogen, Carlsbad, CA) for 60 minutes at room temperature and washed in PBS. The mounted specimens were then examined using confocal laser scanning microscopy (Olympus FV1000, Olympus, Tokyo, Japan) or fluorescence microscopy (Zeiss Axio Observer Z1, Carl Zeiss Japan, Tokyo, Japan).

Eight- to nine-week-old mice were used for histology and immunofluorescence microscopy. Samples were fixed using perfusion with 4% (w/v) PFA in 0.1 M phosphate buffer (pH 7.4) and processed as previously described (Harada et al., 1990, Sato et al., 2011). We performed hematoxylin-eosin staining using standard histological procedures. Western blot analyses were performed as previously described (Sato et al., 2011) and 10 µg of protein was loaded per lane.

Mice were perfused with 2% PFA and 2.5% glutaraldehyde in 0.1 M cacodylate buffer (pH 7.4). Tissues were dissected and further fixed for 2 hours at RT and treated with 1% OsO4 in 0.1 M cacodylate buffer followed by 0.5% uranyl acetate in water. The samples were dehydrated and embedded in Epon, and thin sections were post-stained with uranyl acetate and lead citrate as described previously (Harada et al., 1990). Sections were then studied under an electron microscope (model 1010; JEOL, Tokyo, Japan) at 80 kV.

For scanning EM, fresh tissues were dissected out and fixed for 4 hours in 2% PFA and 2.5% glutaraldehyde in 0.1 M cacodylate buffer (pH 7.4). Tissues were further fixed in 1% tannic acid in 0.1 M cacodylate buffer for 2 hours and 1% OsO₄ in 0.1 M cacodylate buffer for 1 hours. Tissues were then transferred to 50% DMSO in water, frozen and fractured in liquid nitrogen. The samples were substituted in t-butyl alcohol, freeze-dried and sputter-coated with Pt-Pd (E-1010; Hitachi High-Technologies Corporation, Tokyo, Japan). Samples were viewed at 15 kV using scanning electron microscopy (S-4100; Hitachi High-Technologies Corporation).

Control and DKO mice, generated by intercrossing Rab8a^+/−; Rab8b^−/− mice, were dissected on embryonic day 13.5, and MEFs isolated by trypsinisation of the embryos as described previously (Uemura et al., 2009). MEFs were cultured in Dulbecco's modified minimal Eagle's medium (DMEM) supplemented with 10% foetal calf serum (FCS) at 37°C in a humidified 5% CO₂ 95% air atmosphere. To stain cilia, MEFs were fixed in 3% PFA in PBS for 15 minutes at RT, permeabilised in 0.05% saponin and then stained with an anti-acetylated tubulin antibody.

To analyze cilia, 2.5×10⁵ MEFs were cultured on a 12-mm round coverslip in normal culture medium for 2 days and in DMEM containing 0.5% FCS for an additional 2 days. After starvation, cells were fixed with 4% PFA in PBS for 20 minutes at RT. Cells were treated with methanol∶acetone (1∶1) solution at −20°C for 15 minutes, followed by permeabilisation with 0.5% Triton X-100 in PBS. The permeabilised cells were incubated with the blocking solution (5% normal donkey serum and 0.01% Triton X-100 in PBS) for 1 hour at RT, then with the blocking solution containing mouse anti-acetylated tubulin monoclonal antibody (1∶4000; T6793, Sigma), overnight at 4°C. Cells were washed with 0.01% Trion X-100 in PBS and incubated with Alexa-Fluor-488-conjugated anti-mouse IgG (1∶200; Invitrogen, Carlsbad, CA) and DAPI (1∶1000) for 45 minutes. After washing with PBS, the coverslips were mounted and viewed on a microscope. For knockdown experiments, 100 pmol small interfering RNAs (siRNAs) targeting Rab8a, Rab8b, Rab10, Rab13, Ift20 and Ift88 (Thermo Scientific Dharmacon, Lafayette, CO, siGENOME set of four; catalogue numbers MQ-040860-00-0002, MQ-055301-00-0002, MQ-040862-01-0002, MQ-045749-01-0002, MQ-050410-01-0002 and MQ-050417-00-0002) was electroporated into 5×10⁵ MEF cells. One 100 V pulse of 10 mseconds and subsequent five square driving pulses (20 V) of 50 mseconds, at 50-msecond intervals, were applied using a pulse generator CUY21 edit II and a cuvette (SE-202P, BEX, Tokyo, Japan). In the rescue experiment for Rab10 knockdown, 8 µg of FLAG-tagged human Rab10 subcloned into pcDNA5 plasmid (Life Technologies) and 2 µg of pEGFP-N2 plasmid (Clontech) were electroporated into MEF cells together with the relevant siRNA. In the human Rab10 gene, we designed silent mutations for the resistance against mouse Rab10 siRNAs. The percentage of ciliated cells was calculated in electroporated cells that express EGFP.

Images were processed using Adobe Photoshop® (Adobe Systems, Inc., San Jose, CA) version 7.0.

We thank R. Hirai, M. Takano, T. Horie, Y. Okada, H. Togawa and A. Goto for technical assistance. We thank H. Homareda, S. Matsumoto and N. Nakamura for providing us with antibodies.