Ndel1 suppresses ciliogenesis in proliferating cells by regulating the trichoplein–Aurora A pathway

The primary cilium projects from the cell surface and is considered to function as a chemo- and/or mechanosensor (Singla and Reiter, 2006; Anderson et al., 2008; Gerdes et al., 2009; Nigg and Raff, 2009; Goetz and Anderson, 2010; Seeley and Nachury, 2010; Ishikawa and Marshall, 2011). Upon cell cycle exit, the mother centriole frequently gives rise to a basal body to nucleate a nonmotile and microtubule-rich protrusion ensheathed by the plasma membrane. Dysfunction of a primary cilium is associated with a broad spectrum of diseases such as polydactyly, craniofacial abnormalities, brain malformation, congenital heart diseases, situs inversus (defects of left–right patterning), obesity, diabetes, and polycystic kidney disease (Gerdes et al., 2009; Nigg and Raff, 2009; Li et al., 2015).

With the exception of some cells possessing primary cilia during cell proliferation, most cells begin to retract their primary cilia at the cell cycle reentry (Quarmby and Parker, 2005; Kim and Tsiokas, 2011; Goto et al., 2013). Forced induction of primary cilia can affect cell cycle progression (Kim et al., 2011; Li et al., 2011; Inoko et al., 2012), suggesting a possible checkpoint role for primary cilia in cell cycle progression. Recent studies have highlighted a mitotic kinase Aurora A as a negative regulator of primary cilia (Pugacheva et al., 2007; Kinzel et al., 2010; Inoko et al., 2012; Plotnikova et al., 2012). Several proteins were identified as Aurora A activators to disassemble primary cilia at cell cycle reentry (the G0/G1 transition; Pugacheva et al., 2007; Kinzel et al., 2010; Plotnikova et al., 2012) or inhibit their regeneration during cell proliferation (Inoko et al., 2012). Among them, trichoplein, a protein originally identified as a keratin intermediate filament scaffold protein (Nishizawa et al., 2005), localizes at mother and daughter centrioles in proliferating cells (Ibi et al., 2011). Trichoplein binds and activates Aurora A especially in G1 phase, which suppresses unscheduled primary cilia formation during cell proliferation (Inoko et al., 2012). As cells exit the proliferation cycle, trichoplein is polyubiquitinated at the mother centriole by Cul3-RING E3 ligase (CRL3)–KCTD17 complex (Kasahara et al., 2014). This CRL3^KCTD17-mediated trichoplein degradation enables quiescent cells to assemble primary cilia by limiting Aurora A activity (Kasahara et al., 2014).

Nuclear distribution element (NDE)-like 1 (Ndel1; also known as Nudel; Yamada et al., 2010; Chansard et al., 2011a; Bradshaw et al., 2013) was originally identified as a binding partner of Lis1, a dynein regulatory protein, from two-hybrid screening (Niethammer et al., 2000). Because Ndel1 also interacts with dynein and modifies its activity, Ndel1 is considered to regulate microtubule (MT) dynamics and MT-based transport (Sasaki et al., 2000; Liang et al., 2004; Vergnolle and Taylor, 2007; Yamada et al., 2008; Zyłkiewicz et al., 2011). Numerous proteins have been identified as Ndel1-binding partners including kinases, ATPases and GTPases, some activities and functions of which are modulated by the interaction with Ndel1 (Kim et al., 2009; Mori et al., 2009; Bradshaw et al., 2011; Chansard et al., 2011b). Therefore, Ndel1 is known as a scaffold protein involved in numerous cellular processes such as mitosis, neuronal development, and neuronal migration (Yamada et al., 2010; Chansard et al., 2011a; Bradshaw et al., 2013). Here, we have unexpectedly identified Ndel1 as a suppressor of primary cilia assembly likely through the stabilization of trichoplein at the mother centriole.

By searching a public database (Human Gene and Protein Database, http://www.HGPD.jp), we found that 77 proteins including trichoplein possess putative trichohyalin and plectin homology domain (TPHD; Nishizawa et al., 2005; Table S1). A comprehensive siRNA screen for TPHD-containing proteins revealed that four proteins might show ciliary phenotypes similar to trichoplein (Inoko et al., 2012; unpublished data; Fig. 1, B and C). Here, we focused on Ndel1 (Yamada et al., 2010; Chansard et al., 2011a; Bradshaw et al., 2013; Fig. S1 A). To examine Ndel1 localization, we stained HeLa (human cervical carcinoma) and RPE1 (h-TERT–immortalized retinal pigment epithelia) cells with appropriate markers, together with anti-Ndel1 (Fig. 1 A and Fig. S1, B and C). Ndel1 was diffusely localized in the cytoplasm but concentrated in the centrosome (Fig. S1 B and Fig. 1 A, magnified images). Ndel1 localized at proximal sites of CEP164, a marker of distal appendage on the mother centriole (Graser et al., 2007; also see the bottom of merged images with cartoons in Fig. 1 A). Anti-ninein signals were detected as four dots (Figs. 1 A and S1 C), supporting previous studies that ninein was localized not only at subdistal appendages of the mother centriole but also at the proximal ends of both centrioles (Mogensen et al., 2000; Ou et al., 2002; Delgehyr et al., 2005). Anti-Ndel1 signals closely overlapped with two dots of anti-ninein signals: Ndel1 appeared to exist inside subdistal appendage-associated ninein. As shown in Fig. S1 C, anti-Ndel1 signals partially overlapped with Odf2 (a protein of distal/subdistal appendages on the mother centriole; Lange and Gull, 1995; Nakagawa et al., 2001; Ishikawa et al., 2005) or centrin 2 (a protein at the distal ends of centrioles; Paoletti et al., 1996; Laoukili et al., 2000). These results suggested that Ndel1 localizes at the subdistal appendage of mother centriole, which is consistent with a previous study (Guo et al., 2006).

Because Odf2 (Lange and Gull, 1995; Nakagawa et al., 2001; Ishikawa et al., 2005) or ninein (Mogensen et al., 2000; Ou et al., 2002; Delgehyr et al., 2005) was reportedly localized at the subdistal appendage, we analyzed the relationship between Ndel1 and Odf2/ninein. Ndel1 localization was impaired by Odf2 depletion in HeLa cells, whereas Ndel1 depletion did not affect Odf2 localization (Fig. S2, A and B). Because Odf2 is essential for distal and subdistal appendage formation on the mother centriole (Tateishi et al., 2013), the presence of a subdistal appendage may be required for Ndel1 targeting to centrosome. However, neither ninein nor Ndel1 knockdown affected their mutual localization (Fig. S2, C and D). To address functional consequences of their depletion by RNAi, we performed a MT-regrowth assay after cold depolymerization (Dammermann and Merdes, 2002; Delgehyr et al., 2005; Ibi et al., 2011; Fig. S2 E). Similar to control cells, Ndel1-depleted cells assembled an extensive array of cytoplasmic MTs that were centered at the centrosome at 5 min after warming. In contrast, this radial, centrosome-based organization was severely impaired by ninein knockdown, supporting the importance of ninein in MT anchoring as reported previously (Dammermann and Merdes, 2002; Delgehyr et al., 2005; Ibi et al., 2011). Therefore, Ndel1 may function independently of ninein at the subdistal appendage of mother centriole.

To confirm our siRNA data (unpublished data), we stained transfected RPE1 cells with anti–acetylated tubulin and anti-Ndel1 48–72 h later. Ndel1 knockdown led to increased number of cells with a linear object that stained with anti–acetylated tubulin and protruded from the cell surface (Fig. 1, B and C). Transmission electron microscopy (TEM) analysis verified this structure as a primary cilium, characterized by the presence of membranous sheaths surrounding the axonemal MTs and a clear structural transition between the basal body and the cilium (Fig. 1 D). Using electron microscopy, we also detected electron-dense, appendage-like structures in Ndel1-depleted, ciliated cells (Fig. 1 D). In addition, Ndel1 depletion did not affect the localization of appendage proteins even in the ciliated cells (Fig. 1 E). These results suggested that Ndel1 depletion induced unscheduled primary ciliary assembly without affecting appendage structures.

The human NDEL1 gene gives rise to two splice variants (isoforms A and B; Fig. S3 A). Deletion of amino acids 256 to 291 (Δ256–291; Fig. S3 A) from the isoform B splice variant inhibited its localization at the centrosome (Liang et al., 2004). We established three types of RPE1 cell lines allowing the expression of each Ndel1 not only tagged with a C-terminal myc-epitope but also possessing silent mutations to resist each siNdel1-mediated knockdown in a doxycycline (Dox)-dependent manner. Unscheduled cilia formation caused by siNdel1 RNAs (target sequence 2 or 3) was reverted by Dox-induced expression of Ndel1 isoform A or B (Fig. S3, B–D). Therefore, the ciliary phenotype was caused by the reduction of Ndel1 protein level regardless of the isoform. Because Dox-induced expression of Δ256–291 did not revert the ciliary phenotype induced by Ndel1 knockdown (Fig. S3, B–D), it was hypothesized that centrosomal targeting of Ndel1 is required for the suppression of unscheduled cilia formation in proliferating cells. Based on these results (Fig. S3), we focused on Tet-On RPE1 cells expressing the myc-tagged Ndel1 isoform B (called Ndel1-Myc) in subsequent experiments. Because Ndel1-Myc induction improved unscheduled cilia formation by the transfection with each siNdel1 (target sequence 1, 2, or 3; Fig. 1, F–H; and Fig. S3, B–D), we used either of three siRNAs for the following Ndel1 knockdown in RPE1 cells. The results suggested that Ndel1 exerts its inhibitory effect on ciliogenesis in proliferating cells, a function similar to trichoplein (Inoko et al., 2012).

NDE1 (also known as NudE) is highly similar to Ndel1 at the amino acid sequence level (Bradshaw et al., 2013). These paralogous proteins share some common functions (Bradshaw et al., 2013). Like Ndel1, NDE1 localized to the subdistal appendages, whereas neither NDE1 nor Ndel1 knockdown affected their mutual localization (Fig. 1 C). One NDE1-specific siRNA (target sequence 2) slightly increased the proportion of ciliated cells, whereas the other siRNA (target sequence 1) displayed no significant increase (Fig. 1, B and C). Thus, with regard to the suppression of ciliogenesis in proliferating cells, NDE1 participation was only marginal at best. In addition, unscheduled cilia formation caused by Ndel1 depletion was not reverted by Dox-induced expression of NDE1 (Fig. 1, F–H), suggesting that NDE1 expression may not compensate for Ndel1 loss.

Because NDE1 was reportedly required for deciliation at cell cycle reentry (Kim et al., 2011), we examined whether Ndel1 exerts the same function. The cells were treated with each siRNA for 24 h, followed by serum-free medium for 48 h and finally in serum-containing medium for the indicated time (Fig. 2). As reported previously (Kim et al., 2011), NDE1 depletion delayed deciliation process (Fig. 2, A and B). The delay was also observed in Ndel1-depleted cells (Fig. 2, A and B). Thus, both proteins participate in the deciliation processes when cells reenter the cell cycle.

Because NDE1 is also known to regulate ciliary length (Kim et al., 2011), we measured the length, using RPE1 cells cultured without serum for 48 h (Fig. 2 C; also see magnified micrographs at 0 h in Fig. 2 B). NDE1-depleted cells had longer cilia than control cells. In contrast, we observed no significant differences between control and Ndel1-depleted cells.

Because Ndel1 interacts with Lis1 (Niethammer et al., 2000), we examined the effects of Lis1 depletion. Lis1 knockdown not only induced unscheduled ciliary assembly (Fig. 3, A and B) but also increased the distance between nucleus and centrosome (Fig. 3 C); the latter phenotype was previously reported in LIS1^+/− neurons (Tanaka et al., 2004). However, no apparent phenotypes of nucleus–centrosome coupling were observed in Ndel1-depleted cells (Fig. 3 C). Lis1 depletion delayed deciliation process at cell cycle reentry, like Ndel1 depletion (Fig. 3 D).

We next examined the effects of cytoplasmic dynein 1 heavy chain 1 (DYNC1H1) and intermediate chain 2 (DYNC1I2), because Ndel1, NDE1, and Lis1 interact with dynein and modulate its activity (Yamada et al., 2010; Chansard et al., 2011a; Bradshaw et al., 2013). The knockdown of DYNC1H1 (Fig. 3, E and F) or DYNC1I2 (Fig. 3, H and I) might increase the proportion of ciliated cells under cultivation with serum, but the effects were if at all only marginal. In contrast, deciliation process after serum addition was significantly delayed by the treatment with each DYNC1H1 (Fig. 3 G) or DYNC1I2 siRNA (Fig. 3 J), except for one DYNC1I2 siRNA (target sequence #1; Fig. 3 J). Cytoplasmic dynein 2 heavy chain 1 depletion exhibited similar tendency in the two assays (unpublished data; Fig. 3,E–J).

Because dynein light-chain Tctex-type 1 (Tctex-1) was reported to participate in the deciliation process (Li et al., 2011) and control ciliary length (Palmer et al., 2011), we further examined the effects of Tctex-1 knockdown on ciliary dynamics. Tctex-1 depletion elevated the proportion of ciliated cells under cultivation with serum, but the effect was much weaker compared with Ndel1 depletion (Fig. 3, K and L). Meanwhile, Tctex-1 depletion delayed the deciliation process after serum addition (Fig. 3 M) as reported previously (Li et al., 2011).

Collectively, Ndel1, NDE1, Lis1, and dynein complexes participate in ciliary absorption at cell cycle reentry (Fig. 2 and Fig. 3, D, G, J, and M; Kim et al., 2011; Li et al., 2011). NDE1 (Fig. 2 C; Kim et al., 2011), Lis1 (unpublished data), and dynein complexes (Palmer et al., 2011) negatively regulate ciliary length in quiescent RPE1 cells, whereas Ndel1 is unlikely to share this function (Fig. 2 C). In addition, Ndel1 and Lis1 suppress unscheduled cilia formation in proliferating cells, whereas NDE1 and dynein complexes only marginally contributed this process.

We examined the effect of Ndel1 depletion on cell cycle progression in RPE1 cells in the presence of serum. Staining for cyclin A as a marker for the transition of G1/S to pro(meta)phase (Murray, 2004) revealed that Ndel1 depletion reduced both the proportion of cyclin A–positive cells (Fig. 4, A and B) and the protein level of cyclin A (Fig. 4 C). Pulse-labeling experiments using BrdU also demonstrated that Ndel1 knockdown reduced the number of proliferating cells (Fig. 4 D). FACS analysis showed that Ndel1 depletion increased the percentage of cells with 2N DNA content (Fig. 4, E and F). These results suggested that Ndel1 depletion induces cell cycle arrest at the G0/G1 phase.

Because IFT20 is essential for ciliogenesis (Follit et al., 2006), we analyzed the effect of IFT20 coknockdown. With regard to ciliary formation, the effect of Ndel1 knockdown was alleviated by IFT20 coknockdown (Fig. 4, A and B). This restored cell proliferation to almost normal levels, based upon the analysis of several cell markers (Fig. 4). In contrast, single knockdown of IFT20 had little impact on cell cycle progression (Fig. 4). Therefore, in Ndel1-depleted cells, the cell cycle arrest is mediated by the formation of an unscheduled primary cilium, similar to trichoplein-depleted cells (Inoko et al., 2012).

Using RPE1 cells, we examined the behavior of Ndel1 after serum starvation. Ndel1 protein levels were reduced 4 to 24 h after serum depletion (Fig. 5, A and B). This was accompanied by decreased Ndel1 staining at the centrosome for ∼24 h after the starvation (Fig. 5, C and D). Treatment with protease inhibitors restored Ndel1 protein levels almost completely (Fig. 5 E), indicating that serum depletion caused the reduction in Ndel1 protein levels by protein degradation. Interestingly, the time course of degradation appears to differ between trichoplein and Ndel1. Trichoplein was decreased at a slower rate and remained at low levels after serum starvation (Fig. 5, A and B). As reported previously (Kasahara et al., 2014), serum depletion results in the disappearance of trichoplein at the mother centrioles (Fig. 5 D, arrowheads). In contrast, Ndel1 protein levels were initially reduced but recovered ∼48 h after serum starvation (Fig. 5, A and B). Immunocytochemical analyses also revealed that centrosome-associated Ndel1 was decreased after serum depletion but recovered after cilia formation (Fig. 5, C and D). Because Ndel1-Myc induction by Dox decreased the proportion of ciliated cells in the absence of serum (Fig. 6 A), transient Ndel1 degradation is required for ciliary assembly, at least in the absence of serum.

We examined the relationship between Ndel1 and trichoplein. Ndel1-Myc induction by Dox recovered not only trichoplein protein levels (Fig. 6 B) but also its association with the mother centriole (Fig. 6, C and D) in serum-starved cells. In contrast, Dox-induced expression of maltose-binding protein (MBP)-trichoplein-Flag (Inoko et al., 2012) had little impact on transient Ndel1 degradation in response to serum depletion (Fig. 7 A). Thus, Ndel1 degradation likely has to precede trichoplein degradation during the course of ciliogenesis.

Conversely, Dox-mediated induction of MBP-trichoplein-Flag (Ibi et al., 2011) rescued the unscheduled primary cilia formation induced by Ndel1 depletion in the presence of serum (Fig. 7 B). Meanwhile, Ndel1-Myc induction did not compensate for trichoplein depletion (Fig. 7 C). Because trichoplein binds and activates Aurora A (Inoko et al., 2012), we examined the effect of Ndel1 knockdown on the level of Aurora A phosphorylated at Thr288 (Aurora A–pT288), which implies an active form of Aurora A. As shown in Fig. 7 D, Ndel1 depletion reduced Aurora A–pT288 level, like trichoplein depletion (Inoko et al., 2012). Because trichoplein is polyubiquitinated by CRL3^KCTD17 and then degraded specifically at the mother centriole (Kasahara et al., 2014), we performed a KCTD17 knockdown. KCTD17 codepletion alleviates both Aurora A–pT288 reduction and unscheduled ciliation by the knockdown of Ndel1, but not by the knockdown of trichoplein (Fig. 7 D). Ndel1 knockdown also reduced mother-centriole–associated trichoplein, and this effect was also reversed by KCTD17 coknockdown (Fig. 7 E). Because the KCTD17 knockdown itself had little impact on both Ndel1 protein levels (Fig. 7 D) and centrosome-associated Ndel1 (Fig. 7 E), it is unlikely that Ndel1 degradation is directly regulated by CRL3^KCTD17. It rather suggests that Ndel1 likely protects mother-centriole–associated trichoplein from CRL3^KCTD17-mediated degradation, resulting in Aurora A activation.

Ciliogenesis is a multistep process (Sorokin, 1968; Molla-Herman et al., 2010). Cytoplasmic vesicles dock to the distal appendage of mother centriole/basal body (step 1a) and then fuse with each other to form ciliary vesicle (CV; step 1b); electron-dense materials, termed the ciliary bud (CB), accumulate at the distal ends of a basal body (just beneath the CV; step 2); MTs start to elongate from a basal body and then form the axonemal shaft and the ciliary pocket (step 3); the CV finally fuses with plasma membrane before or after step 3 and the cilium emerges into the extracellular environment (see the cartoon [Kasahara et al., 2014] in Fig. S4 A). In serum-starved cells expressing Ndel1-Myc, CVs docked to distal appendages on mother centrioles in most cells (9 out of 11), some of which (5 out of 11) also formed the CB (Figs. 8 A and S4). However, the induction of Ndel1-Myc by Dox prevented the majority of serum-starved cells from proceeding to the step 3 in ciliogenesis (Figs. 8 A and S4). In the aforementioned experimental condition, exogenous Ndel1-Myc properly localized to the centrosome (i.e., subdistal appendage) and cytoplasm (Fig. 8 B) as an endogenous protein (Figs. 1 A and S1 B). Similar disturbance of ciliation steps was induced by exogenous trichoplein expression (Fig. 8 A) or KCTD17 knockdown (Kasahara et al., 2014). Therefore, Ndel1 negatively regulates axonemal extension during ciliogenesis by stabilizing trichoplein at the mother centriole (Fig. 8 C).

Because NDEL1 knockout mice were reported to be embryonic lethal (Sasaki et al., 2005), we analyzed mice with low levels of Ndel1 expression (NDEL1^cko/cko; Fig. 9 A; Sasaki et al., 2005) to elucidate Ndel1 function in vivo. NDEL1^cko/cko mice exhibited abnormalities in telencephalon development and hippocampus pyramidal cell morphology (Sasaki et al., 2005). We also found obesity in NDEL1^cko/cko male mice (Fig. 9, B and C). Because obesity is frequently observed in cilia-related diseases (Gerdes et al., 2009; Nigg and Raff, 2009; Li et al., 2015), we analyzed newborn kidney. To that end, histological sections of proximal and distal tubules and collecting ducts in the kidney of control newborn mice (NDEL1^WT/WT; Fig. 9 D) were compared with NDEL1^cko/cko at postnatal day 0. This revealed three major abnormalities in the latter. First, NDEL1^cko/cko increased the proportion of ciliated cells in proximal tubules compared with NDEL1^WT/WT (Fig. 9, E and F). Second, cilia were longer in proximal and distal tubules and collecting ducts of NDEL1^cko/cko (Fig. 9, E and G). Third, judged by anti–Ki-67 staining, proliferating cells were decreased in tubular epithelia of NDEL1^cko/cko kidney (Fig. 9, H and I). Thus, the increase of ciliated and quiescent cells in the kidney of newborn NDEL1^cko/cko confirmed the observations made with RPE1 cells based on RNAi against Ndel1 (Figs. 1 and 4).

Using serum-starved RPE1 cells, we observed only marginal differences in ciliary length between control and Ndel1 knockdown (Fig. 2, B and C). Less than 10% of proliferating RPE1 cells had primary cilia (e.g., Fig. 1 B). However, in newborn kidney, ∼70–90% cells had cilia (Fig. 9, E and F), whereas ∼30% cells were proliferating (Fig. 9, H and I). Thus, not a few proliferating cells were assumed to possess primary cilia. Thus, we used another cell line, Swiss 3T3 (mouse fibroblast), because ∼50% of Swiss 3T3 cells possessed primary cilia under the cultivation with serum (Fig. 10 A). For mouse Ndel1 knockdown, we use a siRNA specific to mouse Ndel1 (target sequence m1) in addition to siNdel1 3 (the target sequence of which is completely conserved between human and mouse Ndel1). In Swiss 3T3 cells, Ndel1 depletion induced only marginal changes in the percentage of ciliated cells under cultivation with serum (Fig. 10 A). This observation was in contrast to results obtained in RPE1 cells (e.g., Fig. 1 B), which was likely caused by the difference in ciliary dynamics during cell proliferation (i.e., Swiss 3T3 cells form cilia under the cultivation with serum). Next, we analyzed the effect on quiescent Swiss 3T3 cells. As shown in Fig. 10 B, Ndel1-Myc induction by Dox inhibited ciliation in Swiss 3T3 cells, like RPE1 cells (Fig. 6 A), suggesting that Ndel1 also functions as a negative regulator of cilia formation in quiescent Swiss 3T3 cells. Meanwhile, Ndel1 knockdown lengthened the primary cilia during quiescent state (Fig. 10, C–E). Therefore, Ndel1 also limits ciliary length during quiescence in Swiss 3T3 cells, further sustaining the observation in the kidney of newborn NDEL1^cko/cko (Fig. 9, E and G).

In the present study, we have discovered a novel function of Ndel1 in promoting cell cycle progression through the inhibition of ciliogenesis in proliferating cells (Figs. 1 and 4). This function likely requires Ndel1 associated with the subdistal appendage of the mother centriole (Fig. S3). During the course of ciliogenesis, Ndel1 is transiently decreased at the centrosome just before trichoplein degradation starts (Fig. 5). This transient Ndel1 degradation is required for axonemal MT extension during ciliogenesis (Figs. 6 and 8).

How does Ndel1 exert these functions? In the presence of serum, Ndel1 knockdown not only induces unscheduled primary cilia reassembly (Fig. 1) but also decreases mother-centriole–associated trichoplein (Fig. 7 E). The unscheduled cilia formation is restored by exogenous trichoplein induction (Fig. 7 B) or KCTD17 codepletion (a treatment suppressing trichoplein degradation at the mother centriole; Fig. 7, D and E). In serum-starved cells, Ndel1 induction increases mother-centriole–associated trichoplein and blocks primary cilia assembly (Fig. 6). Upon serum depletion, we observed only marginal differences in the morphology of mother centrioles and basal bodies among exogenous Ndel1 and trichoplein induction (Fig. 8 A) and KCTD17 knockdown (Kasahara et al., 2014). These observations led us to suggest the following model (Fig. 8 C). Ndel1 protects mother-centriole–associated trichoplein from CRL3^KCTD17-mediated degradation. The stabilized trichoplein binds Aurora A, which stimulates its autophosphorylation (Inoko et al., 2012). This elevation of the kinase activity destabilizes axonemal MTs, resulting in the deciliation at the cell cycle reentry (Pugacheva et al., 2007) or the inhibition of unscheduled ciliary assembly (Inoko et al., 2012; Kasahara et al., 2014).

Judged by our pull-down (Fig. S5, A and B) and in vitro binding (Fig. S5, C and D) assays, Ndel1 is unlikely to directly bind trichoplein or the CRL3^KCTD17 complex. Our in vitro analyses also revealed that Ndel1 might not directly inhibit CRL3^KCTD17-mediated trichoplein ubiquitination (Fig. S5 E). These observations raise the question of how Ndel1 inhibits CRL3^KCTD17-mediated trichoplein degradation. We consider the following two possibilities. (1) Because Ndel1 functions as a scaffold protein (Yamada et al., 2010; Chansard et al., 2011a; Bradshaw et al., 2013), its binding partners may directly suppress CRL3^KCTD17-mediated trichoplein degradation through the interaction with trichoplein or CRL3^KCTD17 complex (also see next chapter). (2) Ndel1 may function as a gate equipped in the mother centriole. When Ndel1 exists in subdistal appendages, this gate is closed and the CRL3^KCTD17 complex can therefore scarcely enter mother centriole. When Ndel1 is removed from the appendages (e.g., by Ndel1 depletion or in response to serum depletion), the gate is open and then CRL3^KCTD17 complex is able to attack mother-centriole–associated trichoplein.

Accumulating evidence has revealed that Ndel1 and NDE1 share some common functions but also exert unique functions (Bradshaw et al., 2013). Two proteins are also known to modulate dynein activity (Yamada et al., 2010; Chansard et al., 2011a; Bradshaw et al., 2013). These observations raise the question of whether Ndel1 controls ciliary dynamics coordinately with dynein and dynein-associated proteins. With regard to ciliary absorption at cell cycle reentry, knockdown of NDE1 (Fig. 2, A and B; Kim et al., 2011) or dynein (Fig. 3, G, J, and M) produces phenotypes similar to Ndel1 knockdown. However, NDE1 (Fig. 2 C; Kim et al., 2011) or dynein (Palmer et al., 2011) also limits ciliary length in quiescent RPE1 cells, which raises the possibility that abnormally longer cilia may affect ciliary resorption processes at cell cycle reentry. Meanwhile, Ndel1 depletion has no effect on ciliary length in quiescent RPE1 cells (Fig. 2 C). Furthermore, unlike Ndel1, NDE1 (Fig. 1) or dynein (Fig. 3, E, F, H, I, K, and L) only marginally contribute the process to suppress ciliogenesis in proliferating RPE1 cells. Together with the in vitro binding and ubiquitination analyses (Fig. S5), these observations increase the possibility that NDE1 and/or dynein may control ciliary dynamics in a way different from Ndel1. In contrast, ciliary phenotypes of Lis1 depletion resemble those of Ndel1 depletion (Fig. 3, A–D). Thus, Lis1 may be one of likely candidates to transduce Ndel1 signaling. However, we cannot rule out the possibility that Lis1 and Ndel1 may regulate ciliary dynamics in different ways, because these are a few apparent differences in their knockdown phenotypes (e.g., nucleus–centrosome coupling or ciliary length in quiescent RPE1 cells).

Both Ndel1 protein level and centrosome-associated Ndel1 are transiently decreased but finally recovered after serum starvation (Fig. 5). Because primary cilia are lengthened by reduced expression of Ndel1 in newborn mouse kidney (Fig. 9, E and G) or serum-starved Swiss 3T3 cells (Fig. 10, C–E), Ndel1 also limits ciliary length during quiescent state. Thus, after the construction of primary cilia, Ndel1 returns to the basal body and may adjust proper ciliary length, a process required for the maintenance of primary cilia. These observations also raise a new question of why Ndel1 knockdown did not affect ciliary length in quiescent RPE1 cells (Fig. 2 C). One possible explanation is the existence of NDE1, which reportedly limited ciliary length during ciliogenesis (Kim et al., 2011). With regard to ciliary length control, NDE1 may compensate for reduced Ndel1 expression in some types of cells, such as RPE1 cells. Because the recovery of Ndel1 at the centrosome unlikely affects the behavior of trichoplein after serum starvation (Fig. 5, A–D), Ndel1 may control other targets to limit ciliary length. In this regard, NDE1 negatively regulates ciliary length via the interaction with a dynein light-chain protein, LC8 (Kim et al., 2011). In addition, dynein complexes share similar function in RPE1 cells (Fig. 3, G, J, and M; Palmer et al., 2011). Therefore, like NDE1, Ndel1 may also limit ciliary length by modulating dynein activity.

The present study documents novel functions of Ndel1 as a negative regulator of ciliary assembly and paves the way for future studies evaluating the importance of ciliary dynamics in development and tissue homeostasis.

Commercial antibodies and their dilution were as follows: mouse anti–acetylated α-tubulin diluted at 1:200 (clone 6-11B-1; Sigma-Aldrich), anti–α-tubulin diluted at 1:100 (clone DM 1A; Sigma-Aldrich), anti–cyclin A antibody diluted at 1:100 for immunofluorescence and at 1:1,000 for immunoblotting (cyclin A clone 25; BD Biosciences), anti–centrin 2 diluted at 1:100 (clone 20H5; Merck Millipore), anti-dynein intermediate chain diluted at 1: 1,000 (clone 74.1; Merck Millipore), anti-Flag diluted at 1:10,000 (clone M2, Sigma-Aldrich), anti-Myc diluted at 1:1,000 (clone 9B11; Cell Signaling Technology; or clone 4A6; Merck Millipore), rabbit anti–phospho–Aurora A (Thr288) antibody diluted at 1:1,000 (clone C39D8; Cell Signaling Technology), anti-CEP164 antibody diluted at 1:100 (Sigma-Aldrich), anti-DYNC1H1 antibody diluted at 1:1,000 (Proteintech Group), anti-DYNC1I2 antibody diluted at 1:1,000 (Proteintech Group), anti-DYNLT1 (Tctex-1) diluted at 1:1,000 (Proteintech Group), anti-IFT20 antibody diluted at 1:1,000 (Proteintech Group), anti-Lis1 antibody diluted at 1:1,000 (Santa Cruz Biotechnology), anti-NDE1 antibody diluted at 1:3,000 (Proteintech Group), anti-Odf2 antibody diluted at 1: 100 for immunofluorescence and at 1:500 for immunoblotting (Sigma-Aldrich), anti-ninein diluted at 1:100 for immunofluorescence (BioLegend) and at 1:1,000 for immunoblotting (a gift from A. Kikuchi, Osaka University, Osaka, Japan; Fumoto et al., 2006), and goat anti-actin antibody diluted at 1:1,000 (Santa Cruz Biotechnology). The following rabbit polyclonal antibodies were produced as described previously (Sasaki et al., 2000; Nishizawa et al., 2005; Ibi et al., 2011; Kasahara et al., 2014): anti-KCTD17 diluted at 1:20,000, anti-Ndel1 diluted at 1:3,000–1:5,000, and anti-trichoplein diluted at 1:200 for immunofluorescence and at 1: 2,000 for immunoblotting. Fluorescent secondary antibodies and Zenon antibody labeling kits were obtained from Molecular Probes and Life Technologies.

hTERT-immortalized human retinal pigment epithelial (RPE1) cells were cultured in DMEM and F12 nutrient mix (1:1) supplemented with 10% FBS. HeLa and Swiss 3T3 cells were cultured in DMEM supplemented with 10% FBS. MG132, Lactacystin, and ALLN were purchased from Merck. Epoxomicin was obtained from the Peptide Institute.

Tet-On RPE1 or Swiss 3T3 cell lines that expressed GFP, MBP-trichoplein-Flag, NDE1-Myc, or RNAi-resistant Ndel1-Myc (isoform A, B, and Δ256–291) were established as described previously (Ibi et al., 2011; Li et al., 2011). For induction, Tet-On RPE1 or Swiss 3T3 cells were treated with 3–300 ng/ml Dox (Sigma-Aldrich).

Transfection of siRNA duplexes was performed with Lipofectamine RNAiMAX reagent according to the manufacturer’s protocol (Life Technologies); each duplex for Odf2, IFT20, or others was used at final concentration of 20, 50, or 10 nM, respectively. The double-stranded RNAs were purchased from Qiagen or Ambion and Life Technologies. Target nucleotide sequences are shown in Table S2. Negative control 2 siRNA (Silencer Select; Life Technologies) was used as a negative control.

Immunofluorescence microscopy was performed using confocal microscopy (LSM510 META; Carl Zeiss) equipped with a microscope (Axiovert 200M; Carl Zeiss), a plan Apochromat 100×/1.4 NA oil-immersion lens, and LSM image browser software (Carl Zeiss) or a DeltaVision system (Applied Precision) equipped with a microscope (IX70; Olympus), a Plan Apochromat 100×/1.4 NA oil-immersion lens, a cooled charge-coupled device camera, and softWoRx software (Applied Precision) as described previously (Inoko et al., 2012). RPE1, HeLa, and Swiss 3T3 cells were grown on coverslips (Iwaki Glass) and fixed with methanol or formaldehyde as described previously (Sugimoto et al., 2008). For immunostaining with anti-trichoplein, cells were fixed with −20°C methanol for 10 min after 0.01% saponin in PBS at room temperature for 15 min. For detection of primary cilia, cells were placed on ice for 30 min before fixation and stained with anti–acetylated tubulin antibody (Inoko et al., 2012). BrdU incorporation was evaluated using the DNA Replication Assay kit (Merck Millipore). Quantification of fluorescence intensity was performed using ImageJ 1.47v Software (National Institutes of Health). Ciliary length of cultured cells was measured using LSM image browser software. Immunofluorescence images of mouse kidneys were acquired using confocal microscopy (LSM700; Carl Zeiss) equipped with a plan Apochromat 100×/1.4 NA oil-immersion lens and Zen software (Carl Zeiss). NDEL1^WT/WT and NDEL1^cko/cko newborn mouse kidneys were fixed with 4% paraformaldehyde overnight and embedded in Optimal Cutting Temperature compound (Sakura Finetek) after 30% sucrose infiltration and processed for immunofluorescence. Kidney sections were permeabilized with 0.3% Triton X-100. Reconstructed images were used to determine cilium length and rates of cilia formation as follows. To calculate cilia length, points corresponding to the ciliary base and tip form a right triangle with a third point corresponding to a projection of tip on the x, y plane. In this triangle, the hypotenuse is the cilia length. For rates of cilia formation, objects (cells and cilia) were defined in the 3D images and numbers of object obtained for each. All images were taken at room temperature. Figures were generated using Photoshop Elements 11 and Illustrator CS6 (Adobe).

Cul3-3xFlag and Rbx1-GFP were coexpressed in HEK293T cells using Lipofectamine 2000. 1 d after transfection, cells were lysed in cell lysis buffer (20 mM Tris-HCl, pH 7.5, 150 mM NaCl, 2 mM β-glycerophosphate, 50 mM NaF, 1 mM Na₃VO₄, 2.5 mM sodium pyrophosphate, 2 mM EDTA, 1 mM EGTA, and 1% Triton X-100) containing protease inhibitor cocktail (Nacalai Tesque). Cul3-3xFlag immunocomplexes immobilized on anti–DYKDDDDK-tag antibody beads (Wako Laboratory Chemicals) were washed three times with cell lysis buffer and twice with TBS (50 mM Tris-HCl, pH 7.4, 150 mM NaCl) and then eluted with 100 µg/ml 3xFlag peptide (Sigma-Aldrich) dissolved in PBS containing 0.1% Tween 20 (PBST). For protein purification from bacteria, Ndel1 or NDE1 with an N-terminal His-tag and C-terminal Myc-tag was expressed in BL21-CodonPlus (DE3)-RIPL (Agilent Technologies) and then purified through nickel-affinity chromatography according to manufacturer’s protocol (Qiagen). GST-tagged KCTD17 and MBP-trichoplein were expressed in DH5α strain (Invitrogen) and BL21 CodonPlus RP stain (Agilent Technologies), respectively. Each protein was purified through the affinity chromatography with glutathione-Sepharose 4B (GE Healthcare) or amylose resin (New England Biolabs).

Bacterially purified GST-KCTD17 (20 µg) or His-Ndel1/NDE1-myc (10 µg) was incubated with RPE1 cell lysates for 3 h at 4°C and then affinity purified with glutathione-sepharose 4B beads (GE Healthcare) or anti–c-Myc (MC045) agarose conjugate beads (Nacalai Tesque), respectively. After washing with cell lysis buffer three times, the beads were subjected to immunoblotting.

MBP-trichoplein (1.0 μg), Cul3-3xFlag immunocomplex (0.3 μg), and GST-KCTD17 (1.0 μg) were incubated in cell lysis buffer in the presence of His-Ndel1-myc (1.0 μg) or His-NDE1-myc (1.0 μg) for 3 h at 4°C. The mixtures were subjected to immunoprecipitation using anti-GST (Cell Signaling Technology) or anti-Myc (MC045).

His-ubiquitin (12 μg; LifeSensors), MBP-trichoplein (1.0 μg), Cul3-3xFlag immunocomplex (0.1 μg), GST-KCTD17 (1.0 μg), His-Ube1 (0.4 μg; ENZO Life Sciences), and His-UbcH5a (0.5 μg; Enzo Life Sciences) were incubated in 30-μl reaction mixture (50 mM Tris-HCl, pH 7.5, 5 mM MgCl₂, 2 mM NaF, 2 mM ATP, and 0.6 mM DTT) in the presence or absence of His-Ndel1/NDE1-Myc (1.0 μg) at 37°C for 1 h. The reaction was subjected to immunoprecipitation using anti-trichoplein before immunoblotting with anti-ubiquitin.

Immunoblotting (Kasahara et al., 2014), FACS analysis (Inoko et al., 2012), TEM observation (Inoko et al., 2012), and the MT regrowth assay (Ibi et al., 2011) were performed as described previously.

Fig. S1 shows localization of Ndel1 at the centrosome. Fig. S2 demonstrates relationship between Ndel1 and Odf2 or ninein. Fig. S3 shows that centrosomal targeting domain of Ndel1 is necessary to inhibit unscheduled primary cilia assembly in proliferating cells. Fig. S4 demonstrates electron microscopic analysis of a basal body/mother centriole in serum-starved, Ndel1-overexpressing cells. Fig. S5 shows that Ndel1 neither interacts with trichoplein or Cul3-KCTD17 nor affects ubiqutination of trichoplein by Cul3-KCTD17 in vitro. Table S1 lists THPD-containing proteins for siRNA screening. Table S2 lists siRNA sequences used in this study. Additional data are available in the JCB DataViewer at http://dx.doi.org/10.1083/jcb.201507046.dv.

We thank Y. Hayashi, N. Tanigawa, K. Kobori, and E. Kawamoto for technical assistance; H. Tanaka for the FACS analysis; M. Aoki for helpful discussion, and A. Kikuchi for the gift of rabbit polyclonal anti-ninein antibody. We are grateful to R.A. Quinlan and T. Magin for helpful discussion and critical comments on the manuscript.

This work was supported in part by Grants-in-Aid for Scientific Research from the Japan Society for the Promotion of Science and from the Ministry of Education, Culture, Sports, Science, and Technology of Japan; by research grants from the Takeda Science Foundation, the Astellas Foundation for Research on Metabolic Disorders, and the Princess Takamatsu Cancer Research Fund; and by the Uehara Memorial Foundation.

The authors declare no competing financial interests.