ERK7 regulates ciliogenesis by phosphorylating the actin regulator CapZIP in cooperation with Dishevelled

Cilia are essential for embryogenesis and maintenance of homeostasis, but little is known about the signalling pathways that regulate ciliogenesis. Here, we identify ERK7, an atypical mitogen-activated protein kinase, as a key regulator of ciliogenesis. ERK7 is strongly expressed in ciliated tissues of Xenopus embryos. ERK7 knockdown markedly diminishes both the number and the length of cilia in multiciliated cells, and it inhibits the apical migration of basal bodies. Moreover, ERK7 knockdown results in a loss of the apical actin meshwork, which is required for the proper migration of basal bodies. We find that the actin regulator CapZIP, which has been shown to regulate ciliogenesis in a phosphorylation-dependent manner, is an ERK7 substrate, and that Dishevelled, which has also been shown to regulate ciliogenesis, facilitates ERK7 phosphorylation of CapZIP through binding to both ERK7 and CapZIP. Collectively, these results identify an ERK7/Dishevelled/CapZIP axis that regulates ciliogenesis. The actin regulator CapZIP has been shown to regulate ciliogenesis in a phosphorylation-dependent manner. Miyatake et al.show that the atypical MAP kinase ERK7 promotes ciliogenesis by phophorylating CapZIP, and that interactions between both proteins and Dishevelled are required for CapZIP phosphorylation.

T he mitogen-activated protein (MAP) kinase family plays an essential role in various processes such as cell proliferation, responses to stress and differentiation 1 . In vertebrates, there are 14 members of the MAP kinase family 2 . Among these, ERK7 (also known as ERK8 or MAPK15) is the most recently identified member, and is evolutionarily conserved throughout the animal kingdom. ERK7 undergoes autophosphorylation and is constitutively active 3 . Previous studies in cultured cells have shown that ERK7 is involved in the protection of genomic integrity, response to amino-acid starvation and autophagy [4][5][6] . However, its function in vivo remains largely unknown.
Cilia, which are microtubule-based organelles, exert various important functions, such as the generation of fluid flow, sensing of extracellular stimuli and Hedgehog signal transduction, and they are essential for embryogenesis and the maintenance of homeostasis 7,8 . Defects in cilia cause a wide range of disorders, known as ciliopathy 9 . Ciliogenesis is a multistep process that includes basal body migration 7 . Although structural components of cilia and transcription factors that are essential for ciliogenesis have been extensively investigated, little is known about signalling pathways that regulate ciliogenesis.
In this study, we aimed to identify the physiological function of ERK7 during embryogenesis using Xenopus embryos. Our results show that ERK7 is required for ciliogenesis. They further demonstrate that ERK7 is required for basal body migration and apical actin organization in multiciliated cells (MCCs). Mechanistically, ERK7 phosphorylates the actin regulator CapZIP that regulates ciliogenesis in a phosphorylation-dependent manner 10 . Moreover, Dishevelled (Dvl), which also regulates ciliogenesis 11 , facilitates ERK7 phosphorylation of CapZIP by binding to both ERK7 and CapZIP. These results identify an ERK7/Dvl/CapZIP axis that regulates ciliogenesis.

Results
Expression patterns of ERK7 during Xenopus development. Whole-mount in situ hybridization showed maternal expression of ERK7 in the animal hemisphere ( Supplementary Fig. 1a). Interestingly, zygotic ERK7 messenger RNA (mRNA) was highly expressed in ciliated tissues. At stage 12 (late gastrula), ERK7 expression was detected in the dorsal involuting marginal zone (DIMZ; Fig. 1a). Costaining of ERK7 with FoxJ1, which is expressed in the DIMZ 12 , showed that the region expressing ERK7 overlapped with the region expressing FoxJ1 ( Supplementary Fig. 1b), confirming that ERK7 is expressed in the DIMZ. After involution through the blastopore, DIMZ forms the gastrocoel roof plate (GRP), on which motile monociliated cells are located 13 . ERK7 was expressed in the GRP at stage 16 (neurula stage; Fig. 1b). At stage 25 (early tailbud stage) and stage 33/34 (late tailbud stage), ERK7 was expressed in the floor plate, cloaca and ear vesicle ( Fig. 1a and Supplementary Fig. 1c), all of which are ciliated tissues. At stage 33/34 (late tailbud stage), ERK7 expression was also detected in nephrostomes, which are ciliated funnels. In addition, a salt-and-pepper expression pattern of ERK7 was observed in the epidermis (Fig. 1a). This expression pattern was observed at stage 13 and thereafter (Fig. 1c). In the Xenopus epidermis, four types of cells, alpha-ionocytes, betaionocytes, small secretory cells and MCCs, are distributed in a salt-and-pepper pattern [14][15][16][17][18][19] . MCCs bear many cilia on their surface, and therefore, they can be detected by immunostaining with an antibody against acetylated a-tubulin (a marker of cilia). To determine whether ERK7 is expressed in MCCs on the epidermis, we combined ERK7 in situ hybridization with acetylated a-tubulin immunostaining. The in situ staining of ERK7 on the epidermis completely coincided with acetylated atubulin immunostaining (Fig. 1d), indicating that ERK7 is specifically expressed in MCCs on the epidermis. No detectable signal was observed with the sense probe ( Supplementary  Fig. 1d). These results indicate that ERK7 is specifically expressed in ciliated tissues during Xenopus development.
ERK7 is required for ciliogenesis. To examine ERK7 function, we performed knockdown experiments using an antisense morpholino oligonucleotide (MO), which was designed to block the translation of ERK7. We also generated an antibody against Xenopus ERK7. Immunoblotting analysis showed that ERK7 MO inhibited the translation of exogenous ERK7 (Fig. 1e), and significantly reduced endogenous ERK7 protein levels (Fig. 1f). Embryos that had been dorsally injected with ERK7 MO displayed reduced head structures, close-set eyes and dorsal curvature (Fig. 1g). On the other hand, embryos that had been ventrally injected with ERK7 MO displayed loss of fin (Fig. 1h). A higher dose of ERK7 MO resulted in the inhibition of antero-posterior axis elongation and breakdown of the epidermis, which resulted in exposure of the inner tissues to the outer environment (Fig. 1h). In fact, the higher dose of ERK7 MO caused a loss of epidermal cells such as small secretory cells ( Supplementary  Fig. 2). ERK7 knockdown with a second non-overlapping ERK7 MO caused defects similar to those caused by the first MO in both dorsal and ventral injections ( Supplementary Fig. 3a,b).
We noticed that ERK7 knockdown in the epidermis abrogated the hovering movement of embryos, which is driven by MCCgenerated fluid flow (Supplementary Movies 1 and 2). To directly monitor the fluid flow, we added fluorescent beads to the culture media and monitored the bead movement. While beads moved fast in a cranio-caudal direction in control embryos as previously reported 20 , beads hardly moved in ERK7 morphants (Fig. 2a,b and Supplementary Movies 3 and 4). These results indicate that ERK7 knockdown results in the loss of MCC-generated fluid flow.
Next, we performed immunostaining using an acetylated atubulin antibody to visualize the cilia. Although the staining intensity of each spot on the epidermis was drastically reduced in ERK7 morphants, the number of spots was comparable to that in control embryos (Fig. 2c). As each spot represents MCC, this result indicates that ERK7 knockdown does not affect the number of MCCs. Higher magnification images revealed that both the number of cilia and the length of the formed cilia were strikingly diminished in ERK7 morphants (Fig. 2d). Moreover, ciliary beating was lost in ERK7 morphants (Supplementary Movies 5 and 6). Injection of the second non-overlapping ERK7 MO resulted in the same defects in ciliogenesis ( Supplementary  Fig. 3c). In addition, the defect in cilia formation in ERK7 morphants was significantly rescued by the injection of MOresistant ERK7 mRNA ( Supplementary Fig. 4). These results indicate that ERK7 is required for the process of ciliogenesis but not for the specification of ciliated cells.
We next examined whether ERK7 is required for ciliogenesis in the GRP. Motile monocilia are known to be located on the GRP 13 , which is the functional equivalent of the node in the mouse and Kupffer's vesicle in zebrafish [21][22][23] . The GRP cilia were shorter in ERK7 morphants than in control embryos ( Fig. 2e and Supplementary Fig. 5a), indicating that ERK7 regulates ciliogenesis in many ciliated tissues. The GRP cilia are known to be posteriorly polarized 13 . We then analysed the polarization of the GRP cilia. ERK7 knockdown disturbed the polarization of the GRP cilia ( Supplementary Fig. 5b-d), indicating that ERK7 is important for the polarization of GRP cilia.
The GRP cilia are important for left-right axis specification 13 . Thus, we analysed the expression of Pitx2c, which is asymmetrically expressed in the left lateral plate mesoderm 24 . ERK7 knockdown resulted in a loss of asymmetric Pitx2c expression in the left lateral plate mesoderm ( Supplementary  Fig. 6), indicating that ERK7 is important for left-right axis specification.
ERK7 expression is under the control of FoxJ1. The above results indicate that ERK7 is required for ciliogenesis and is specifically expressed in ciliated tissues. Thus, we examined the regulation of ERK7 expression. FoxJ1, one of the forkhead family genes, is a master regulator of the motile ciliogenic programme 12,25 , and its spatial expression pattern 12,26 is very similar to that of ERK7. In addition, we noticed that in the deposited microarray data 12 , ERK7 was among the genes that were upregulated by the injection of FoxJ1 under the condition of ectopic expression of a Notch intracellular domain (NICD), which caused a loss of MCCs in the skin 14 . Thus, we examined whether ERK7 expression is regulated by FoxJ1. We injected FoxJ1 mRNA into four-cell-stage embryos and harvested the embryos at stage 13 to evaluate ERK7 expression. Quantitative reverse transcription (RT)-PCR analysis showed that overexpression of FoxJ1 strongly induced ERK7 expression (Fig. 3a). Next, we examined the effect of NICD on ERK7   ARTICLE expression. Animal caps were prepared at stage 10 and then harvested at the equivalent of stage 19-20 to analyse ERK7 expression. The analysis showed that overexpression of NICD, which acts upstream of FoxJ1 and suppresses FoxJ1 expression 27,28 , decreased ERK7 expression (Fig. 3b). In agreement with the deposited microarray data 12 , the overexpression of FoxJ1 increased ERK7 expression even in the presence of NICD (Fig. 3b). Furthermore, we examined the effect of FoxJ1 knockdown on ERK7 expression. FoxJ1 MO resulted in a drastic decrease in ERK7 expression levels (Fig. 3c).
These results indicate that ERK7 expression is under the control of FoxJ1, the master regulator of ciliogenesis. Next, we examined whether FoxJ1 expression is affected by ERK7. The result showed that neither overexpression of ERK7 nor ERK7 MO affected the expression level of FoxJ1 ( Supplementary Fig. 7a,b). Thus, the FoxJ1 expression level is not affected by ERK7.
ERK7 is required for basal body migration in MCCs. We determined the subcellular localization of ERK7. In MCCs, an ERK7-EGFP (enhanced green fluorescent protein) fusion protein was observed in a punctate pattern, suggesting the localization of ERK7 in the base of the cilia (Fig. 4a). In addition, ERK7 localized to cell-cell junctions (Fig. 4a) and weakly to cilia ( Supplementary  Fig. 8a). EGFP did not specifically localize to these regions ( Supplementary Fig. 8b). Coexpression of ERK7-EGFP with Centrin2-mCherry (a basal body marker) showed that ERK7 localization overlapped with Centrin2 localization and was adjacent to a basal body (Fig. 4b). Coexpression of ERK7-EGFP with CLAMP-mCherry (a marker for striated ciliary rootlet) showed that ERK7 also localized to the striated ciliary rootlet (Fig. 4c). Ciliogenesis is a multistep process 7 . During ciliogenesis, basal bodies are formed either from pre-existing centrioles or de novo. After formation, basal bodies migrate apically and nucleate  axonemal microtubules. To investigate the behaviour of basal bodies, we performed immunostaining with a g-tubulin antibody.
In control embryos, basal bodies of MCCs were located apically and docked at the apical membrane (Fig. 5a), as previously reported 11 . However, in ERK7 morphants, most of basal bodies failed to migrate apically and remained located deep within the cytoplasm (Fig. 5a). This mislocalization of basal bodies in ERK7 morphants was more clearly visualized in transverse sections ( Fig. 5b). These results indicate that ERK7 is required for basal body migration in MCCs.
In Xenopus MCCs, two distinct pools of actin, apical actin and subapical actin, form actin meshworks at the apical surface 29 . It has previously been shown that the apical actin meshwork is important for basal body migration 11,30,31 . Phalloidin staining in MCCs showed that knockdown of ERK7 caused a drastic loss of both the apical and subapical actin meshwork (Fig. 5c). These results strongly suggest that ERK7 regulates basal body migration by organizing the actin meshwork.
ERK7 phosphorylates CapZIP. To examine the molecular mechanisms by which ERK7 regulates ciliogenesis and actin organization, we searched for a direct substrate of ERK7 in ciliogenesis. CapZIP emerged as a potential substrate of ERK7 based on the following previous observations: (1) CapZIP has been shown to regulate ciliogenesis in a phosphorylationdependent manner 10 ; (2) the phosphorylation sites essential for CapZIP function in ciliogenesis coincide with MAP kinase consensus sequences; and (3) CapZIP regulates actin formation 10 . CapZIP was originally identified as a substrate for MAPKAP-K2, p38g and p38d in skeletal muscle 32 , but it remains unknown which kinase phosphorylates CapZIP during ciliogenesis. Therefore, we hypothesized that ERK7 would regulate ciliogenesis by phosphorylating CapZIP.
To investigate whether ERK7 phosphorylates CapZIP, we first performed coexpression experiments. Coexpression of CapZIP with ERK7, but not with a kinase-dead ERK7 mutant (ERK7 KD), caused a mobility shift of CapZIP, which suggested that ERK7 phosphorylates CapZIP (Fig. 6a). Xenopus CapZIP has nine MAP kinase consensus sequences (serine/threonine followed by proline) (Supplementary Fig. 9). Of the nine residues, six are well conserved among mouse, human and Xenopus. Of the six residues, two serine residues (S81 and S108 in Xenopus CapZIP) have been reported to be important for the function of CapZIP during ciliogenesis 10   reported residues were replaced with alanine, and in CapZIP 6A, the six conserved residues were replaced with alanine. When coexpressed with ERK7, the mobility of the shifted bands of CapZIP 2A was much faster than that of CapZIP WT (Fig. 6a), and CapZIP 6A produced no mobility shifted bands (Fig. 6a). Thus, it is likely that ERK7 phosphorylates CapZIP at several of these six residues, including the two residues. CapZIP has been shown to be essential for ciliogenesis in zebrafish, and phosphorylation of CapZIP on the two serine residues is important for its function 10 . We next examined the expression patterns of CapZIP during early Xenopus development by quantitative RT-PCR and whole-mount in situ hybridization. Quantitative RT-PCR experiments showed that CapZIP expression was detected at stage 12 (late gastrula stage) and thereafter, and then gradually increased until stage 25 (early tailbud stage; Supplementary Fig. 10a). Whole-mount in situ hybridization experiments showed that CapZIP was expressed in the DIMZ at stage 14 ( Supplementary Fig. 10b) and in the epidermis at stages 14, 17 and 23 ( Supplementary Fig. 10b-d), indicating that CapZIP and ERK7 are expressed in overlapping regions. CapZIP expression was also detected in the somites and heart field ( Supplementary  Fig. 10d,e). Next, we examined whether the roles of CapZIP and its phosphorylation in ciliogenesis are conserved in Xenopus. Injection of CapZIP 2A mRNA, but not CapZIP WT mRNA, resulted in a significant reduction in the length of cilia in the GRP (Fig. 6b and Supplementary Fig. 10f). CapZIP 6A expression had the same ARTICLE effect on the length of cilia as CapZIP 2A (Fig. 6b and Supplementary Fig. 10f). These results indicate that CapZIP also regulates ciliogenesis in a manner that is dependent on its phosphorylation at S81 and S108 in Xenopus.
To investigate whether ERK7 directly phosphorylates CapZIP, we performed an in vitro kinase assay using purified FLAGtagged ERK7 and FLAG-tagged CapZIP. The incubation of these two proteins in the presence of ATP produced mobility shifted bands of CapZIP (Fig. 6c), indicating that ERK7 directly phosphorylates CapZIP.
To provide evidence for the functional significance, we next examined whether the expression of a phosphomimetic mutant of CapZIP rescues the ciliogenesis defect in the ERK7 morphants. To achieve this goal, we made a CapZIP 6E mutant, in which the six conserved residues were replaced with glutamic acid. The injection of CapZIP 6E mRNA significantly rescued the defect in cilia formation in ERK7 morphants (Fig. 6d), indicating that ERK7-mediated phosphorylation of CapZIP is important for ciliogenesis in vivo.
Dvl facilitates ERK7 phosphorylation of CapZIP. CapZIP has been shown to be a non-canonical Wnt signalling mediator and to function downstream of Dvl 10 . Remarkably, the knockdown phenotypes and the subcellular localization pattern of Dvl in MCCs 11 are similar to those of ERK7. Knockdown of Dvl leads to a failure of basal body migration and loss of the apical actin meshwork in Xenopus MCCs 11 , and the subcellular localization of Dvl2, which localized near basal bodies in MCCs, is important for ciliogenesis 11 . In addition, we found that overexpression of ERK7 resulted in a severely shortened anterior-posterior axis ( Supplementary Fig. 11a), and the phenotype is related to the function of Dvl 33,34 . These observations prompted us to investigate the relationship between ERK7 and Dvl. First, to examine whether ERK7 co-localizes with Dvl2 in MCCs, we coinjected Dvl2-EGFP and ERK7-mCherry mRNAs. As previously reported 11 , Dvl2-EGFP localized both at smaller foci (near the basal bodies) and at larger aggregates in MCCs (Fig. 7a and Supplementary Fig. 11b). ERK7-mCherry co-localized with Dvl2-EGFP at smaller foci and partly at larger aggregates. These results suggest that ERK7 and Dvl2 would form a complex. To test this idea, we performed coimmunoprecipitation assays in cultured cells and found that Dvl2 coimmunoprecipitated with ERK7 (Fig. 7b) and vice versa (Supplementary Fig. 11c). Our results also showed coimmunoprecipitation of mouse ERK7 with mouse Dvl2 (Supplementary Fig. 11d).
Given that Dvl functions as a scaffold by providing a signalling platform in diverse contexts 35 , we hypothesized that Dvl should facilitate ERK7 phosphorylation of CapZIP by serving as a scaffold. To examine this possibility, we first investigated whether Dvl binds to CapZIP. Coimmunoprecipitation experiments showed that Dvl2 binds to CapZIP (Fig. 7c). Next, by using a series of Dvl2 deletion mutants, we determined an ERK7-binding site and a CapZIP-binding site in Dvl2. ERK7 coimmunoprecipitated with Dvl2 WT, Dvl2 DDIX (a deletion mutant of the DIX domain) and Dvl2 DPDZ (a deletion mutant of the PDZ domain), but not with Dvl2 DC (a deletion mutant of the C-terminal region containing the DEP domain; Fig. 7d). CapZIP coimmunoprecipitated with Dvl2 WT and Dvl2 DDIX, but not with Dvl2 DPDZ or Dvl2 DC. These results indicate that ERK7 and CapZIP bind to Dvl2 at the C-terminal region and at the region spanning both the PDZ domain and the C-terminal region, respectively (Fig. 7f). Finally, to examine whether Dvl2 could enhance ERK7 phosphorylation of CapZIP, we examined the effect of increasing amounts of Dvl2 on ERK7-mediated phosphorylation of CapZIP. The data show that the phosphorylation of CapZIP was enhanced by increasing the concentration of Dvl2, but the extent of phosphorylation tended to decrease when the concentration of Dvl2 became too high (Fig. 7e). This biphasic nature of the effect of its increasing concentrations is a well-known characteristic of the action of a scaffold protein. Thus, these results taken together indicate that Dvl2 functions as a scaffold in ERK7-mediated phosphorylation of CapZIP (Fig. 7f).
Mouse ERK7 is involved in ciliogenesis. We finally investigated whether the role of ERK7 in ciliogenesis is conserved in mammals. To achieve this goal, we used mouse tracheal epithelial cells (mTECs), in which ciliogenesis is triggered by exposure to an airliquid interface (ALI) 36 . It has been previously established that MCCs start to be induced at B3-4 days after exposure of confluent mTECs to ALI 37 . First, we examined the changes in gene expression during ciliogenesis in mTECs. ERK7 mRNA exhibited peak expression at B5 days after ALI exposure, whereas the expression of FoxJ1 was initiated earlier and peaked at B2 days after ALI exposure (Fig. 8a). This result is consistent with our hypothesis that ERK7 expression is under the control of FoxJ1. Dnah9, Tekt2, Spag6 and Rsph1 (also known as radial spoke protein 44) are present in axonemal structures exclusive to motile cilia and are known to be upregulated by FoxJ1 (refs 12,25). These four genes also exhibited peak expression of B5 days after ALI exposure, similarly to ERK7 expression (Fig. 8a). We next designed mouse ERK7 short hairpin RNA (shRNA) constructs (nos. 1 and 2) and confirmed that these shRNA sequences effectively suppressed ERK7 expression in mIMCD3 cells (Fig. 8b). To investigate the function of ERK7 in the ciliogenesis of mTECs, we introduced our ERK7 shRNA constructs using lentivirus. While B17% of mTECs expressing control shRNA formed mature multicilia, the proportion of mTECs with mature multicilia was reduced to B8% by ERK7 shRNA (Fig. 8c,d). These results show that ERK7 plays a role in ciliogenesis in mammals.

Discussion
In this study, we identified ERK7 as a key regulator of ciliogenesis. In addition, we showed that ERK7 expression is under the control of FoxJ1. Because ERK7 is constitutively active 3 , the regulation at the expression level should be crucial for ERK7 function. It has been previously shown that knockdown of FoxJ1, similarly to knockdown of ERK7, results in a failure in basal body migration and a loss of the actin meshwork in MCCs 12 . Thus, our results strongly suggest that FoxJ1 regulates actin organization and basal body migration at least in part through the induction of ERK7.
Knockdown of ERK7 resulted in reduced head structures. Because Wnt signalling is known to be important for head formation 38 , there may be a link between ERK7 and Wnt signalling. Interestingly, FoxJ1 expression has been shown to be regulated by canonical Wnt signalling 39,40 . Thus, a canonical Wnt signalling pathway may regulate ERK7 expression levels through FoxJ1.
Dvl regulates various biological processes, including ciliogenesis 11 , and it has various modes of action that are achieved with different interaction partners 35 . Our findings demonstrate that Dvl regulates the efficiency of ERK7-mediated phosphorylation of CapZIP, which has been shown to regulate ciliogenesis in a phosphorylation-dependent manner 10 , by binding to both ERK7 and CapZIP. In addition, knockdown of ERK7 disturbed the polarization of GRP cilia. Interestingly, non-canonical Wnt signalling regulates the polarization of cilia in Xenopus GRP, mouse node and zebrafish Kuppfer's vesicle [41][42][43][44] . CapZIP has been shown to be a non-canonical Wnt signalling mediator that functions downstream of Dvl 10 , and thus, ERK7 should play a crucial role in linking the FoxJ1-induced ciliogenesis programme to non-canonical Wnt signalling.  45 . The in vitro synthesis of capped mRNA was performed using mMESSAGE mMACHINE (Ambion) according to the manufacturer's instructions. Antisense MOs were obtained from Gene Tools, Inc. The MO sequences were as follows: ERK7 MO, 5 0 -ATCTTCCACCTCTGGGCCACTCATG-3 0 ; second non-overlapping ERK7 MO, 5 0 -ATGTCACACAACAACTGGCCCCTCC-3 0 ; FoxJ1 MO (described as XFoxJ1-MOATG in previous study 12 ), 5 0 -GCAGGTCAAACA TTAATAAAGCCCT-3 0 and a standard control oligo (control MO), 5 0 -CCTCTTA CCTCAGTTACAATTTATA-3 0 . To assess cilium formation in the GRP, dorsal explants were dissected at stage 16, and placed underneath a glass coverslip with silicone vacuum grease. To monitor the hovering movement of embryos, we anaesthetized embryos with MS-222 (Sigma) on agar plates. First, the embryo was placed on the uninjected left side, and then, it was flipped to the injected right side as described in a previous study 46 . To assess the fluid flow, we immobilized anaesthetized embryos at stage 33/34 in 1% Steinberg agar plates and added fluorescent beads (F8844, Invitrogen) to the culture media. To measure the flow velocity of the beads, 10 beads were chosen from the beads above an embryo at random and the flow speed was measured using ImageJ. Images and movies were acquired using the stereomicroscope SZX12 (Olympus) equipped with the digital camera DP50 (Olympus), SZX16 (Olympus) equipped with the digital camera DP73 (Olympus) or the Delta Vision optical sectioning system (Applied Precision) equipped with the inverted microscope IX71 (Olympus), the charge-coupled device camera CoolSNAP HQ Monochrome (Photometrics) and SoftWoRx software (Applied Precision). Image deconvolution was performed using SoftWoRx software. To display the trajectories of the fluorescent beads, movies of the bead movement were processed as follows, which was modified from a previous study 13 : (1) each frame of the movie was converted to a binary image and then applied to Watershed segmentation using ImageJ; and (2) each of the frames was assigned a pseudocolour ranging from yellow (frame 1, time 0 s) to red (frame 33, time 3.3 s), and all of the frames were merged using Adobe Photoshop CS5 (Adobe Systems).
Whole-mount in situ hybridization and immunostaining. Whole-mount in situ hybridization was performed according to the standard protocol 45 using a robot (InsituPro, Intavis). Briefly, embryos were fixed in MEMFA (100 mM MOPS (pH 7.4), 2 mM EGTA, 1 mM MgSO 4 and 3.7% formaldehyde) and dehydrated in methanol. After rehydration, embryos were treated with 5 mg ml À 1 proteinase K in PBS containing 0.1% Tween20, incubated in 100 mM triethanolamine (pH 8.0), incubated in 100 mM triethanolamine containing 0.25% acetic anhydride and then refixed with 4% paraformaldehyde in PBS. After washing, embryos were hybridized with RNA probes and then subjected to anti-digoxigenin antibody staining (no. 11093274910, Roche; 1:2,000). The colour reaction was performed using BM purple. For double in situ hybridization, the first colour reaction was performed using BM purple. To inactivate remaining alkaline phosphatase derived from the first reaction, stained embryos were incubated with the maleic acid buffer (100 mM maleic acid (pH 7.5) and 150 mM NaCl) containing 10 mM EDTA at 65°C for 15 min, and dehydrated in methanol. After rehydration, embryos were refixed with 4% paraformaldehyde in PBS, and then subjected to anti-fluorescein antibody staining (no. 11426338910, Roche; 1:4,000). The second colour reaction was performed using 5-bromo-4-chloro-3-indolyl phosphate (BCIP). The following digoxigenin-labelled RNA probes were used: X. laevis ERK7 (the entire coding region of NM_001095966), X. laevis tph1 (the entire coding region of NM_001087454), X. laevis Pitx2c (the entire region of AJ243596) and X. laevis CapZIP (bp 601-978 of AB920341). The fluorescein-labelled RNA probe of FoxJ1 was synthesized from the cDNA corresponding to the coding region of X. laevis FoxJ1 (AJ609390). For in situ hybridization followed by immunostaining, we omitted the proteinase K step. Immunostaining procedures for Xenopus embryos were modified from previously described methods 48 . Briefly, the embryos were fixed in 4% paraformaldehyde in PBS for 2 h at room temperature or overnight at 4°C. For g-tubulin immunostaining, the embryos were dehydrated in Dent's fixative. Fixed embryos were washed with PBS containing 0.1% Triton X-100 (PBT) and blocked with 10% fetal bovine serum (FBS) and 5% dimethylsulphoxide in PBT for 1 h. Embryos were incubated with primary antibodies at 4°C overnight. After washing five times with PBT for 1 h, the embryos were incubated with the appropriate secondary antibodies at 4°C overnight followed by washing with PBT. The following antibodies were used at the indicated dilutions. The primary antibodies were mouse anti-acetylated a-tubulin (T7451, Sigma; 1:1,000), rabbit anti-GFP (A-11122, Invitrogen; 1:1,000) and mouse anti-g-tubulin (T5236, Sigma; 1:500). The secondary antibodies were Alexa Fluor 350-conjugated anti-mouse IgG (A-21049, Invitrogen; 1:500), Alexa Fluor 488-conjugated anti-mouse IgG (A-11029, Invitrogen; 1:500), Alexa Fluor 488-conjugated anti-rabbit IgG (A-11034, Invitrogen; 1:500) and Alexa Fluor 568-conjugated anti-mouse IgG (A-11031, Invitrogen; 1:500). To visualize actin, we used Alexa Fluor 488-conjugated phalloidin (Invitrogen). The length of the cilia was measured with SoftWoRx software. For cryosectioning, embryos were washed with PBS, and embedded in 7.5% gelatin and 15% sucrose in PBS. The embedded embryos were frozen in liquid N 2 and sectioned at 10 mm using a Microm HM500 OM microtome cryostat (CarlZeiss). For immunostaining of mTECs, cells were fixed in 4% paraformaldehyde at room temperature for 10 min and then submerged in cold methanol for 5 min at À 20°C. After blocking with 3% bovine serum albumin in PBS at room temperature for 30 min, cells were incubated with 3% bovine serum albumin in PBS containing the primary antibody at 4°C overnight, washed with PBS and then incubated with the appropriate secondary antibodies for 30 min at room temperature. Finally, the cells were mounted in Mowiol.
Cell culture and transfection. HEK293T and mIMCD3 cells were obtained from the American Type Culture Collection. HEK293T cells were cultured in Dulbecco's modified Eagle's medium (DMEM) containing 10% FBS. mIMCD3 cells were cultured in DMEM/F12 containing 10% FBS. HEK293T cells were cultured on collagen-coated plates. Cells were transfected using Lipofectamine 2000 (Invitrogen) or Fugene HD (Promega) according to the manufacturer's protocol.
Antibody production and affinity purification. The anti-ERK7 polyclonal antibody was produced in rabbits by immunizing them with a synthetic peptide corresponding to residues 414-430 of Xenopus ERK7 (QKDSTQQSPKIKAPSSN) plus an additional N-terminal cysteine for conjugation. Rabbit antisera were affinity purified using the synthetic peptide. The antibody was dialysed and concentrated. The peptide synthesis, immunizations and affinity purification were performed by MBL Co.
Kinase assays. For protein preparations, FLAG-tagged ERK7 and FLAG-tagged CapZIP were expressed separately in HEK293T cells, precipitated by anti-FLAG M2 affinity gel (Sigma), and eluted using 3 Â FLAG peptide. For the kinase assays, FLAG-tagged ERK7 protein was mixed with FLAG-tagged CapZIP protein in a kinase reaction buffer (50 mM Tris-Cl (pH 7.5), 0.1 mM EGTA, 10 mM magnesium acetate, 0.1 mM sodium orthovanadate and 1 mM ATP), which was modified from a previous study 49 , and incubated for 30 min at 30°C. The reactions were stopped by addition of Laemmli's sample buffer and boiling.