Cullin-3–KCTD10-mediated CEP97 degradation promotes primary cilium formation

Primary cilia are non-motile cilia protruding from the plasma membrane of the majority of vertebrate cells. They function as cellular antennae that sense and transmit chemical and mechanical signals from the extracellular milieu and, thereby, play essential roles in tissue development and homeostasis (Basten and Giles, 2013; Gerdes et al., 2009; Goetz and Anderson, 2010). Deficits in the formation of primary cilia cause various human disorders that are collectively termed ciliopathies, including polycystic kidney disease, retinal degeneration, visceral inversion, polydactyly and neurodegeneration (Fliegauf et al., 2007).

Primary cilia are generated by the extension of a microtubule-based axoneme from the basal body, which is derived from the mother centriole, the older of the two centrioles in the centrosome (Basten and Giles, 2013; Gerdes et al., 2009; Goetz and Anderson, 2010). Whereas extensive studies have identified a large number of molecules that participate in primary cilium formation, the molecular mechanisms of ciliogenesis still remain elusive. In cultured mammalian cells, ciliogenesis is triggered by serum starvation, indicating that ciliogenesis is tightly correlated with exit from the cell cycle (Izawa et al., 2015; Nigg and Raff, 2009; Sánchez and Dynlacht, 2016). The centrosomal protein CP110, which localizes on the distal ends of both mother and daughter centrioles, forms a cap to block the elongation of axonemal microtubules and, thereby, prevents inappropriate ciliogenesis in proliferating cells (Bettencourt-Dias and Carvalho-Santos, 2008; Kobayashi and Dynlacht, 2011; Sánchez and Dynlacht, 2016; Tsang and Dynlacht, 2013). In response to cell cycle exit signals, CP110 is removed selectively from the mother centriole, which is an essential step for the quiescence-induced ciliogenesis (Bettencourt-Dias and Carvalho-Santos, 2008; Kobayashi and Dynlacht, 2011; Sánchez and Dynlacht, 2016; Tsang and Dynlacht, 2013). Previous studies demonstrated that TTBK2, MARK4 and Centrin2 are required for this step (Goetz et al., 2012; Kuhns et al., 2013; Prosser and Morrison, 2015), and some miRNAs affect ciliogenesis by controlling expression levels of CP110 (Cao et al., 2012; Song et al., 2014). However, how these molecules regulate the removal of CP110 from the mother centriole at the early step of ciliogenesis is largely unknown.

CEP97 interacts and cooperates with CP110 to suppress primary cilium formation (Spektor et al., 2007). Depletion of CEP97 decreases CP110 protein level and suppresses CP110 localization to the centrioles, suggesting that CEP97 is crucial for regulating the stability and centriolar recruitment of CP110 (Spektor et al., 2007). Similar to CP110, CEP97 is specifically removed from the mother centriole in quiescent cells. However, the mechanisms of CEP97 removal from the mother centriole and its roles in CP110 removal and ciliogenesis have remained unknown.

The ubiquitin-proteasome system is a protein degradation system that regulates many cellular and physiological processes (Hershko and Ciechanover, 1998). Previous studies have shown that E3 ubiquitin ligases, such as pVHL, CRL3-KCTD17 and MIB1, promote or suppress ciliogenesis through the ubiquitylation and degradation of negative or positive regulators of ciliogenesis (Kasahara et al., 2014; Thoma et al., 2007; Villumsen et al., 2013; Wang et al., 2016). CP110 was reported to be ubiquitylated and degraded by the E3 ligase SCF^cyclinF during G2 and M phases (D'Angiolella et al., 2010). Recent studies also showed that NEURL-4 and EDD-DYRK2-DDB1^VprBP, the components of E3 ligases, are involved in ciliogenesis through CP110 ubiquitylation (Hossain et al., 2017; Loukil et al., 2017). However, the precise mechanism of the removal of CP110 and CEP97 during the early stage of ciliogenesis is still unknown.

In this study, we explore the mechanism of CEP97 degradation and its roles in CP110 removal and ciliogenesis induced by serum-starvation. We show that the ubiquitin-proteasome system controls serum-starvation-induced CEP97 degradation. We provide evidence that the cullin-3 (CUL3)–RBX1–KCTD10 complex acts as the E3 ligase that mediates serum-starvation-induced CEP97 degradation, leading to CP110 removal and ciliogenesis.

CEP97 and CP110 cooperatively suppress cilium formation, and the removal of these proteins from the mother centriole is required for serum-starvation-induced ciliogenesis (Spektor et al., 2007). We examined the effects of CEP97 knockdown on the protein level and centriolar localization of CP110 in human telomerase-immortalized retinal pigmented epithelial 1 (RPE1) cells. As reported (Spektor et al., 2007), knockdown of CEP97 decreased CP110 protein level (Fig. 1A) and suppressed centriolar localization of CP110 (Fig. S1A), suggesting that CEP97 plays a crucial role in supporting the stability and centriolar recruitment of CP110. In contrast, knockdown of CP110 did not affect CEP97 protein level (Fig. 1A).

A previous study has shown that the CEP97 protein level was reduced in serum-starved cells (Spektor et al., 2007). We also observed that the Cep97 protein level decreased after serum starvation in RPE1 cells (Fig. 1B). CEP97 did not completely disappear in serum-starved cells, which probably reflects its retention at the daughter centriole even after serum starvation (Fig. 1C, left). To examine the mechanism of the decrease in CEP97 protein level, RPE1 cells were serum-starved for 24 h and treated with the proteasome inhibitor MG-132. This treatment blocked the serum-starvation-induced decrease in CEP97 protein level (Fig. 1B), suggesting that CEP97 is degraded upon serum starvation by the ubiquitin-proteasome system.

In serum-starved cells, CEP97 is removed from the mother centriole and remains present only on the daughter centriole (Spektor et al., 2007). We next examined the effect of MG-132 treatment on CEP97 localization in serum-starved RPE1 cells. Immunostaining with anti-CEP97 antibody revealed that CEP97 was localized to only one centriole in serum-starved control cells, whereas MG-132 treatment caused a marked accumulation of CEP97 on two centrioles and several dots near the centrosome (Fig. 1C,D). Treatment with small interfering RNA (siRNA) targeting  CEP97 abolished CEP97 immunofluorescence signals in MG-132-treated cells (Fig. S1B), indicating that the Cep97 immunostaining signal was specific. MG-132 treatment did not increase the number of cells with >2 centrioles (as measured by centrin and γ-tubulin immunostaining) in serum-starved RPE1 cells (Fig. S1C,D), indicating that MG-132-induced CEP97 accumulation near the centrosome did not result from centriole amplification. In addition, we examined the effect of MG-132 on serum-starvation-induced cell quiescence by immunostaining for Ki-67, a commonly used marker for proliferating cells. In the control experiments, the number of Ki-67-positive cells was decreased upon serum starvation and increased after serum had been added again (Fig. S1E,F). MG-132 treatment did not increase the number of Ki-67-positive cells under serum-starved conditions (Fig. S1G,H), indicating that MG-132 treatment had no effect on serum starvation-induced cell quiescence under our experimental conditions.

To examine whether CEP97 is ubiquitylated in cultured cells, HEK293T cells were co-transfected with YFP-tagged CEP97 and HA-ubiquitin, and cultured in the presence or absence of MG-132. The cell lysates were then analyzed by immunoprecipitation with an anti-GFP nanobody, followed by immunoblotting with an anti-HA antibody. Polyubiquitylation of YFP-CEP97 was observed predominantly after the treatment with MG-132 (Fig. 1E). We also demonstrated that endogenous CEP97 was more prominently polyubiquitylated in MG-132-treated RPE1 cells under serum-starved quiescent conditions than in the same cells during S phase arrest (Fig. 1F). These results suggest that CEP97 was preferentially degraded upon serum starvation by the ubiquitin-proteasome system.

To investigate the role of CEP97 ubiquitylation in ciliogenesis, we constructed a non-ubiquitylatable mutant of CEP97 and analyzed the effect of its overexpression on ciliogenesis. A proteomic database of post-translational modifications predicted K582 and K618 as the ubiquitylated residues of CEP97 (Mertins et al., 2013). We, therefore, constructed the CEP97(3KR) mutant, in which K582, K618 and K617 were replaced by arginine residues (Fig. 2A). Ubiquitylation assays revealed that the level of ubiquitylation of CEP97(3KR) was reduced in MG-132-treated cultured cells, compared with that of wild-type (WT) CEP97 (Fig. 2B). It has been reported that overexpression of CEP97(WT) suppresses serum-starvation-induced ciliogenesis, but its suppressive effect is limited (Spektor et al., 2007), possibly because CEP97(WT) is unstable due to ubiquitylation and proteasomal degradation under serum-starved conditions. To test this assumption, we compared the effect of overexpression of non-ubiquitylatable CEP97(3KR) with that of CEP97(WT) on serum-starvation-induced ciliogenesis. RPE1 cells transfected with YFP-tagged CEP97(WT) or CEP97(3KR) were serum-starved and then immunostained with antibodies against Arl13b (as a marker of ciliary membranes) and FGFR1 oncogene partner (FOP; as a marker of mother and daughter centrioles; Lee and Stearns, 2013) (Fig. 2C). CEP97(WT)-YFP and CEP97(3KR)-YFP were enriched on two centrioles in non-ciliated cells (Fig. 2C). Whereas primary cilia were formed in 63% of control YFP-expressing cells, overexpression of CEP97(WT) only modestly reduced the fraction of ciliated cells to 44%. Notably, overexpression of CEP97(3KR) reduced the fraction of ciliated cells more severely to 27% (Fig. 2D). These results suggest that CEP97 ubiquitylation is crucial for serum-starvation-induced ciliogenesis.

Both CEP97 and CP110 are removed from the mother centriole upon serum starvation but remain localized to the daughter centriole in ciliated cells (Spektor et al., 2007). We next examined the effects of overexpressing CEP97(WT) or CEP97(3KR) on CP110 removal from the mother centriole after serum starvation. RPE1 cells were transfected with YFP-tagged CEP97(WT or 3KR), serum-starved, and stained with antibodies against CP110, FOP and CEP170 (as a marker of the mother centriole; Guarguaglini et al., 2005) (Fig. 2E). In most control YFP-expressing cells, CP110 was localized to a CEP170-negative single centriole, indicating that CP110 had been removed from the mother centriole. Overexpression of CEP97(WT) or CEP97(3KR) decreased the population of cells with CP110 localized to a CEP170-negative single centriole and inversely increased the population of cells with CP110 localized to two centrioles (Fig. 2E). Quantitative analyses showed that the population of cells with CP110 localized to one centriole was 66% in control YFP-expressing cells, but it was decreased to 45% and 30% by overexpression of CEP97(WT) and CEP97(3KR), respectively (Fig. 2F). These results suggest that the non-ubiquitylatable CEP97 variant more severely suppresses the removal of CP110 from the mother centriole and that CEP97 ubiquitylation plays a crucial role in serum-starvation-induced CP110 removal from the mother centriole.

To explore the mechanism of serum-starvation-induced CEP97 ubiquitylation, we next analyzed the effects of inhibitors of E3 ubiquitin ligases on serum-starvation-induced decrease in CEP97 protein level. We used inhibitors of E3 ligases that had been shown to be involved in ciliogenesis, i.e. MLN4924, an inhibitor of cullin-RING type E3 ligases (CRLs) (Kasahara et al., 2014; Pan et al., 2004; Soucy et al., 2009); Heclin, an inhibitor of HECT-type E3 ligases (Loukil et al., 2017; Mund et al., 2014); and VH298, an inhibitor of pVHL (Frost et al., 2016; Thoma et al., 2007). Treatment with MLN4924 effectively blocked the decrease in CEP97 after serum starvation but treatments with Heclin or VH298 had no apparent effect on CEP97 protein levels (Fig. 3A and Fig. S2A). A previous study has shown that MG-132 treatment blocks ciliogenesis (Kasahara et al., 2014). MLN4924 also inhibited serum-starvation-induced ciliogenesis (Fig. S2B). These results suggest that CRLs are involved in CEP97 degradation. Immunoblot analyses showed multiple anti-CEP97 immunoreactive bands. Treatment with λ-phosphatase abolished the upwards mobility shift (Fig. S2C), indicating that the upper bands are the phosphorylated forms of CEP97. Intensity levels of the upper band frequently increased after treatment with MG-132 but not with MLN4924. Thus, the relationship between CEP97 phosphorylation and degradation remains unknown.

To identify which member of the cullin family is involved in CEP97 degradation, we examined the effects of siRNA-based knockdown of each member of the cullin family (CUL1, CUL2, CUL3, CUL4A, CUL4B, CUL5 and CUL7) (Petroski and Deshaies, 2005) on the amount and localization of CEP97 to the centrosome. Transfection of cullin-targeting siRNAs effectively decreased the expression levels of each target protein in RPE1 cells (Fig. S2D). RPE1 cells were transfected with siRNAs, serum-starved, and stained with antibodies against CEP97, FOP and CEP170 (Fig. 3B). Cells were categorized into three classes according to the phenotype of CEP97 localization to the centrosome; class 1, cells with CEP97 localization to a single centriole (presumably a CEP170-negative daughter centriole); class 2, cells with CEP97 localization to two centrioles; and class 3, cells with marked accumulation of CEP97 on and near the centrosome (Fig. 3B). Quantitative analyses of the population of cells in each class revealed that most of the cells treated with control siRNAs were scored as class 1 (Fig. 3C). In contrast, treatments with two independent CUL3-targeting siRNAs significantly decreased the population of cells categorized into class 1, and inversely increased the population of cells categorized into class 2 and 3 (Fig. 3C). The siRNAs targeting other cullin members had only minor effects on CEP97 localization to the centrosome, compared with the effects of CUL3-targeting siRNAs. These results indicate that knockdown of CUL3 suppresses CEP97 degradation and removal from the mother centriole and causes its retention on the mother centriole in serum-starved cells. Additionally, immunoblot analyses showed that the level of CEP97 protein decreased upon serum starvation but CUL3 knockdown partially recovered the decrease in CEP97 protein in serum-starved cells (Fig. S3A). These results suggest that CUL3 is involved in serum-starvation-induced CEP97 degradation and its removal from the mother centriole.

We also examined the effects of knockdown of each of the cullin members on serum-starvation-induced ciliogenesis. RPE1 cells transfected with siRNAs were serum-starved and stained for CEP97, Ac-tubulin and γ-tubulin (Fig. 3D). We observed that treatments with two independent CUL3-targeting siRNAs significantly suppressed ciliogenesis (Fig. 3D,E). Several siRNAs targeting other cullin members also suppressed ciliogenesis but only one of each of the two siRNAs targeting CUL1, CUL4A, CUL5 and CUL7 suppressed ciliogenesis, probably due to their off-target effects on the gene products related to ciliogenesis, and the inhibitory effect of one of the CUL4B-targeting siRNAs was weaker than that of either of CUL3-targeting siRNAs (Fig. S3D). Together with the results shown in Fig. 3C, CUL3 is the most likely E3 ligase candidate that is involved in serum-starvation-induced CEP97 degradation and ciliogenesis.

We next examined the effect of CUL3 knockdown on the removal of CP110 from the mother centriole. RPE1 cells transfected with CUL3 siRNAs were serum-starved and stained with antibodies against CP110, FOP and CEP170 (Fig. 4A). Whereas most of cells transfected with control siRNA exhibited CP110 localization to one centriole (presumably a CEP170-negative daughter centriole), treatment with CUL3 siRNAs significantly decreased the population of cells with CP110 localization to one centriole and increased the cells with CP110 localization to two centrioles (Fig. 4A,B), indicating that CUL3 is involved in the removal of CP110 from the mother centriole in serum-starved cells.

To examine whether CUL3 knockdown affects serum-starvation-induced cell quiescence, RPE1 cells treated with CUL3 siRNAs were serum-starved and stained for Ki-67. Knockdown of CUL3 did not increase the population of Ki-67-positive cells (Fig. S4A,B), indicating that CUL3 knockdown had no significant effect on serum-starvation-induced cell quiescence.

As an E3 ligase, CUL3 functions by forming a complex with broad complex/tramtrack/bric-a-brac (BTB) domain-containing proteins (adaptors that link CUL3 to substrate proteins) at the N-terminal region, and with RBX1 (a RING motif-containing protein that cooperatively functions with CUL1–4 for ubiquitylation of substrate proteins) at the C-terminal region (Furukawa et al., 2003; Genschik et al., 2013). To further define the role of CUL3 in CEP97 degradation, we analyzed the effect of expression of the CUL3 N-terminal fragment, CUL3(N418), which binds to the BTB proteins but not to RBX1 and, thereby, acts as a dominant-negative form of CUL3 (Cullinan et al., 2004; Furukawa et al., 2003). Overexpression of CUL3(N418) caused centrosomal accumulation of CEP97, whereas overexpression of CUL1(N452), a dominant-negative form of CUL1 (Wu et al., 2000), did not (Fig. S4G,H). Fluorescence image analyses showed that YFP-CUL3(N418), but not CUL1(N452), was enriched on the centrosome (Fig. S4G). These results further support the notion that CUL3 plays a crucial role in CEP97 degradation.

Because RBX1 cooperatively functions with CUL3 for ubiquitylation of substrate proteins, we next examined the effect of RBX1 knockdown on the amount and localization of centrosomal CEP97 and ciliogenesis under serum-starved conditions. RBX1-targeting siRNAs suppressed RBX1 protein expression in RPE1 cells (Fig. S2D). Similar to the effect of CUL3 knockdown, the treatment with two independent RBX1-targeting siRNAs significantly decreased the population of class 1 cells with CEP97 localization to a single centriole, and increased the population of class 2 cells with CEP97 localization to two centrioles and class 3 cells with CEP97 accumulation on the centrosome (Fig. 5A,B). Immunoblot analyses also showed that knockdown of RBX1 partially recovered the level of CEP97 protein in serum-starved cells (Fig. S3B). These results suggest that RBX1 is involved in serum-starvation-induced CEP97 degradation and removal from the mother centriole. Similar to CUL3 knockdown, RBX1 knockdown significantly decreased the number of cells with CP110 localization to one centriole (Fig. 5C,D) and the number of ciliated cells (Fig. 5E,F) under serum-starved conditions. Knockdown of RBX1 did not increase the number of Ki-67-positive cells (Fig. S4C,D). Taken together, these results suggest that the CUL3–RBX1 complex is involved in CEP97 degradation, removal of CEP97 and CP110 from the mother centriole, and ciliogenesis.

CUL3-mediated ubiquitylation requires BTB domain-containing proteins that serve as adaptors to link CUL3 to substrate proteins (Genschik et al., 2013). In the human genome, there are ∼180 genes encoding BTB-containing proteins (Genschik et al., 2013; Stogios et al., 2005). To identify the BTB protein(s) involved in CUL3-mediated CEP97 degradation, we first screened for the BTB proteins that have the potential to bind to CEP97 by using the visible immunoprecipitation (VIP) assay, a recently developed method to easily visualize protein-protein interactions (Fig. S5A; Katoh et al., 2015). HEK293T cells were co-transfected with CEP97-mCherry and each of Venus-tagged BTB-containing proteins. The cDNAs coding for BTB-containing proteins were obtained from a human proteome expression resource (HuPEX) cDNA library (Maruyama et al., 2012) (Table S1). We also used YFP and YFP-CP110 as negative and positive controls for CEP97-binding ability, respectively. Cell lysates were immunoprecipitated with GST-tagged anti-mCherry-nanobody and glutathione-Sepharose beads. The fluorescence of co-precipitated Venus (or YFP) on the beads bearing CEP97-mCherry was directly observed by using fluorescence microscopy. As expected, YFP fluorescence was detected on the beads for lysates of cells co-transfected with YFP-CP110, but not of those transfected with control YFP (Fig. S5B). When Venus fluorescence intensity on the beads was higher than or equal to that of YFP-CP110, we judged that this Venus-BTB protein has the potential to bind to CEP97 (Fig. S5B and Table S2). Of the 150 Venus-tagged BTB proteins examined, 24 proteins exhibited Venus fluorescence on the beads bearing CEP97-mCherry (Fig. S5B and Table S2). Reproducible results were obtained in three independent experiments.

In the second screening, we tested the effects of knockdown of individual candidates of BTB proteins on serum-starvation-induced ciliogenesis. Of the siRNAs targeting the 24 BTB proteins, both independent siRNAs targeting each of six BTB proteins significantly suppressed serum-starvation-induced ciliogenesis (Fig. S6A). As an example, the effects of KCTD10 knockdown on ciliogenesis are shown in Fig. 6A,B. Thus, we selected these six BTB proteins as the candidates that could be involved in CEP97 degradation and ciliogenesis. We further analyzed the effect of knockdown of each of these six BTB proteins on the centrosomal localization of CEP97. Notably, among the siRNAs targeting the genes encoding those six BTB proteins, only the two independent siRNAs targeting KCTD10 caused accumulation of CEP97 on the centrosome (Fig. 6C, Fig. S6B). Although one of the two siRNAs targeting BTBD11 and KCTD1 modestly increased the number of cells with CEP97 accumulation on the centrosome, their effects were weaker than those of KCTD10-targeting siRNAs (Fig. S6B). Knockdown efficiency of siRNAs targeting KCTD10, KCTD1 and BTBD11 was shown in Figs S2D and S6C,D. Moreover, KCTD10 knockdown decreased the number of class 1 cells with CEP97 localization to a single centriole and increased the number of class 2 cells with CEP97 localization to two centrioles, in addition to the number of class 3 cells with CEP97 accumulation on the centrosome (Fig. 6C,D). Immunoblot analyses also showed that KCTD10 knockdown partially recovered the level of CEP97 protein in serum-starved cells (Fig. S3C). Thus, we selected KCTD10 as the BTB protein most likely to be involved in serum-starvation-induced CEP97 degradation and ciliogenesis. Similar to knockdowns of CUL3 and RBX1, knockdown of KCTD10 significantly decreased the percentage of cells with CP110 localization to one centriole in serum-starved cells (Fig. 6E,F), indicating a crucial role of KCTD10 in the removal of CP110 from the mother centriole. Additionally, KCTD10 knockdown did not increase the number of Ki-67-positive cells (Fig. S4E,F). These results suggest that KCTD10 is involved in degradation of CEP97, removal of CEP97 and CP110 from the mother centriole, and in ciliogenesis. The effect of KCTD10 knockdown on CEP97 dysregulation was weaker than those of knocking down CUL3 or RBX1, suggesting that other BTB components of the CUL3 E3 ligase complex are partially involved in CEP97 degradation.

To confirm the results of the VIP assay, we analyzed the interaction between CEP97 and KCTD10 by coimmunoprecipitation assays. mCherry-tagged CEP97 was coexpressed with YFP-tagged KCTD10 or control YFP in HEK293T cells and immunoprecipitated with an anti-GFP nanobody. CEP97-mCherry was effectively co-precipitated with YFP-KCTD10 but barely detectable with control YFP (Fig. 6G), indicating that KCTD10 associates with CEP97. Taken together, these results suggest that the CUL3–RBX1–KCTD10 complex is the strong candidate E3 ligase for CEP97 degradation.

BTB-containing adaptor proteins directly bind to CUL3 via the N-terminal BTB domains (Furukawa et al., 2003; Pinkas et al., 2017; Xu et al., 2003). To investigate the importance of the CUL3–KCTD10 interaction for CEP97 degradation and ciliogenesis, we constructed three KCTD10 mutants: a C-terminal fragment (C142), a BTB-domain mutant (3A) in which Ile-83, Leu-84 and Ile-85 were replaced by alanine residues that are known to impair CUL3-binding ability (Chen et al., 2009), and an N-terminal fragment (N145) (Fig. 7A). Co-precipitation assays revealed that CUL3 effectively co-precipitated with full-length (FL) KCTD10 and KCTD10(N145) but it barely with KCTD10(C142) or KCTD10(3A) (Fig. 7B). These results suggest that KCTD10 binds to CUL3 principally through the N-terminal BTB domain. In contrast, CEP97 effectively co-precipitated with KCTD10(C142) and KCTD10(3A) but not with KCTD10(N145) (Fig. 7C), suggesting that KCTD10 binds to CEP97 via its C-terminal region.

We then examined whether overexpression of these KCTD10 mutants affects the amount and localization of centrosomal CEP97 and ciliogenesis under serum-starved conditions. Overexpression of YFP-tagged KCTD10(C142) or (3A), but not (N145), significantly decreased the number of class 1 cells with CEP97 localization to a single centriole, and increased the number of class 2 cells with CEP97 localization to two centrioles and class 3 cells with CEP97 accumulation on the centrosome (Fig. 7D,E). Furthermore, overexpression of YFP-KCTD10(C142) or (3A), but not (N145), significantly suppressed ciliogenesis under serum-starved conditions (Fig. 7F). Fluorescence image analyses showed that YFP-KCTD10(C142) and (3A), but not (N145), were enriched on the centrosome (Fig. 7D). These results suggest that KCTD10 mutants (C142 and 3A) have dominant-negative effects on serum-starvation-induced CEP97 degradation and ciliogenesis, and that the interaction between KCTD10 and CUL3 is crucial for CEP97 degradation and ciliogenesis.

By using YFP-KCTD10(WT) and (3A) constructs, we further carried out rescue experiments with KCTD10 knockdown cells. We generated an siRNA-resistant (sr) silent mutation into YFP-KCTD10 constructs and confirmed the expression of these proteins even in the KCTD10 siRNA-treated cells (Fig. S7A). Immunostaining for CEP97 showed that expression of sr-KCTD10(WT) partially rescued the reduction of class 1 cells with CEP97 localization to one centriole, whereas expression of sr-KCTD10(3A) had no such effect (Fig. S7B,C). These results further suggest that the KCTD10 BTB domain is required for CEP97 degradation and its removal from the mother centriole.

Upon serum starvation, CEP97 and CP110 are removed from the mother centriole but not the daughter centriole. The specific removal of these proteins from the mother centriole might be regulated by the localization of KCTD10 to the mother centriole. To examine this possibility, we analyzed the localization of KCTD10 in RPE1 cells by immunostaining with an anti-KCTD10 antibody. Intriguingly, KCTD10 was localized to one of the two γ-tubulin-positive centrioles in serum-fed cells and at the base of the primary cilium in serum-starved cells (Fig. 8A). KCTD10 was also detected as punctate foci in the nucleus. To examine whether KCTD10 is localized to the mother centriole, we also analyzed KCTD10 localization by co-staining with CEP170, a protein that is localized to the subdistal appendage of the mother centriole and in the transition zone of the primary cilium (Guarguaglini et al., 2005). KCTD10 was located in close proximity to CEP170, indicating that KCTD10 was localized to the mother centriole in serum-fed cells and in the basal region of the cilium in serum-starved cells (Fig. 8B and Fig. S7D). Moreover, the anti-KCTD10 immunostaining signals were diminished by treatment with KCTD10 siRNAs (Fig. S7D). These results suggest that KCTD10 is specifically localized to the mother centriole where it promotes CEP97 degradation and removal from the mother centriole.

CP110 and CEP97 cooperatively function to block unscheduled ciliogenesis in proliferating cells. When cells enter the quiescent phase, these proteins are removed from the mother centriole, which triggers axoneme extension to form primary cilia (Spektor et al., 2007). In this study, we examined the mechanism of quiescence-induced removal of CEP97, and its role in CP110 removal and ciliogenesis. We showed that treatment with MG-132 suppressed serum-starvation-induced degradation of CEP97 and caused aberrant accumulation of CEP97 on the centrosome, indicating that upon serum starvation, centrosomal CEP97 is degraded by the ubiquitin-proteasome system. In fact, CEP97 was preferentially polyubiquitylated in serum-starved cells. Overexpression of a non-ubiquitylatable CEP97(3KR) blocked CP110 removal from the mother centriole and ciliogenesis more severely than did overexpression of CEP97(WT), indicating the crucial role of CEP97 ubiquitylation in those processes. Several steps of screening, by using siRNA-based knockdown experiments and VIP protein-binding assays, revealed that the CUL3–RBX1–KCTD10 complex is the most potent candidate E3 ubiquitin ligase responsible for CEP97 degradation. Depletion of each component of this E3 ligase complex or overexpression of KCTD10 mutants caused dysregulation of CEP97 at the centrosome, suppressed CEP97 removal from the mother centriole and blocked ciliogenesis. Moreover, KCTD10 was selectively localized to the mother centriole. In addition, knockdown of CUL3, RBX1 or KCTD10 significantly decreased the population of cells with CP110 localized to one centriole under serum-starved conditions. Taken together, these results strongly suggest that the CUL3–RBX1–KCTD10 complex acts as the E3 ligase for CEP97, and plays a crucial role in the removal of CEP97 and CP110 from the mother centriole and in ciliogenesis induced by serum starvation (Fig. 8C).

Knockdown of CUL3 increased the population of class 2 cells (with CEP97 localization to two centrioles) and class 3 cells (with CEP97 accumulation on the centrosome) and inversely decreased the population of class 1 cells (with CEP97 localization to one centriole, presumably a daughter centriole) (Fig. 3C). The sum of the increase in the percentages of class 2 and class 3 cells (or the decrease in the percentage of class 1 cells) (∼50%) was comparable to the drop in the percentage of ciliated cells in CUL3 knockdown cells (Fig. 3E). Thus, these results suggest that knockdown of CUL3 causes the defects in ciliogenesis principally by stabilizing centrosomal CEP97 and impairing CEP97 removal from the mother centriole. On the other hand, whereas the percentage of cells with CEP97 localized to one centriole (class 1 cells) was ∼40% (Fig. 3C), the percentage of ciliated cells was ∼10% (Fig. 3E). Thus, the percentage of cells with CEP97 removed from mother centriole but still no cilium formation is ∼30%. This percentage is comparable to that for control siRNA cells, in which the percentages of class 1 cells and ciliated cells were 80–90% and 60%, respectively. Thus, 20–30% of either control cells or CUL3-knockdown cells failed to form cilia even when they lost CEP97 from the mother centriole. These results suggest that CEP97 removal alone is not sufficient for ciliogenesis, and other mechanisms that properly control the following processes, such as axonemal microtubule extension and intraflagellar transport, are required to complete ciliogenesis.

In accordance with previous reports (Spektor et al., 2007), we showed that CEP97 depletion reduced the level of CP110 protein and caused CP110 disappearance from the centrioles, indicating that CEP97 plays a crucial role in controlling protein stability and centriolar recruitment of CP110. However, CP110 depletion did not affect CEP97 protein levels but abrogated CEP97 centriolar localization (Spektor et al., 2007), indicating that CP110 is required for CEP97 recruitment to the centrioles but not for its stability. These results suggest that, whereas CEP97 and CP110 are mutually required for their co-recruitment to the centrioles, CEP97 appears to play a dominant role in controlling the stability of the CEP97–CP110 complex on the centrioles. Thus, it is conceivable that CEP97 ubiquitylation and degradation play a key role in the removal of the CEP97–CP110 complex from the mother centriole at the early stage of ciliogenesis.

Previous studies have demonstrated that CP110 is degraded during G2 and M phases, with SCF^CyclinF and USP33 mediating ubiquitylation and deubiquitylation of CP110, respectively (D'Angiolella et al., 2010; Li et al., 2013). However, the roles of SCF^CyclinF and USP33 in ciliogenesis in quiescent cells have not yet been elucidated. Recent studies demonstrated that EDD-DYRK2-DDB1^VprBP and NEURL-4, the E3 ligase components, regulate ubiquitylation and stability of CP110, and that inhibition or depletion of these proteins suppresses ciliogenesis (Hossain et al., 2017; Loukil et al., 2017). In this study, we demonstrated for the first time that CEP97 is degraded in serum-starved cells through the ubiquitin-proteasome system. We also showed that expression of the non-ubiquitylated CEP97(3KR) mutant severely suppressed serum-starvation-induced CP110 removal and ciliogenesis, which strongly suggests that CEP97 ubiquitylation is a prerequisite for CP110 removal and ciliogenesis. Because CEP97 knockdown causes CP110 disappearance, CEP97 degradation is likely to trigger CP110 degradation and removal from the mother centriole. However, further studies are needed to elucidate the precise mechanism of how CEP97 regulates CP110 ubiquitiylation and stability, and the suppressive function of the CEP97–CP110 complex in ciliogenesis.

Knockdown of CUL3, RBX1 or KCTD10 did not completely suppress the removal of CEP97 and CP110 from the mother centriole but effectively blocked ciliogenesis, suggesting that this E3 complex can have other targets for ciliogenesis. As for CUL3, a previous study has shown that CUL3 promotes ubiquitylation and degradation of trichoplein, a negative regulator of ciliogenesis, through binding to the alternative BTB domain-containing protein KCTD17 (Kasahara et al., 2014). Here, we showed that the CUL3–KCTD10 complex is involved in ciliogenesis by promoting CEP97 degradation. Therefore, CUL3 has, at least, dual roles in ciliogenesis through ubiquitylation of trichoplein and CEP97, mediated by KCTD17 and KCTD10, respectively.

KCTD10 has previously been shown to be a binding partner of B9-domain-containing protein-1 (B9D1), which is localized to the transition zone of the primary cilium, and mutations of which cause ciliopathies (Chih et al., 2012). In fact, we showed here that KCTD10 is specifically localized to the mother centriole in serum-fed cells and to the base of the primary cilium in ciliated cells cultured under serum-starved conditions. The selective localization of KCTD10 to the mother centriole further supports the notion that the CUL3–KCTD10 complex plays a key role in the selective removal of CEP97 from the mother centriole. However, KCTD10 is localized to the mother centriole, irrespective of the presence or absence of serum in culture medium. Thus, the serum-starvation-induced degradation and removal of CEP97 from the mother centriole are not regulated by the translocation of KCTD10. Additional mechanisms are required to activate KCTD10-mediated CEP97 degradation upon serum starvation. Further investigations on the mechanism controlling CUL3–RBX1–KCTD10-mediated CEP97 degradation and removal from the mother centriole during early ciliogenesis will provide a better understanding of quiescence-induced ciliogenesis and pathogenesis of ciliopathies.

MG-132 (C2211; MilliporeSigma, St Louis, MO), 4′,6-diamidino-2-phenylindole (DAPI) (09224; Polysciences, Warrington, PA), MLN4924 (I-502; Boston Biochem, Cambridge, MA), Heclin (5433/10; R&D Systems, Minneapolis, MN), VH298 (6156/10; R&D Systems) and horseradish peroxidase-conjugated streptavidin (SA-5004; Vector Laboratories, Burlingame, CA) were purchased from indicated suppliers. Antibodies were purchased as follows: rabbit polyclonal antibodies against CEP97 (Novus Biologicals, Littleton, CO), CP110 (Bethyl Laboratories, Montgomery, TX), GFP (Thermo Fisher Scientific, Waltham, MA), KCTD10 (Novus Biologicals), Arl13b, CUL1, CUL2, CUL4A, CUL4B and CUL7 (Proteintech, Rosemont, IL), CUL3 (Cell Signaling, Danvers, MA), Centrin (MilliporeSigma), RBX1 and mCherry (Genetex, Irvine, CA); mouse monoclonal antibodies against CEP170 (Thermo Fisher Scientific), β-actin, α-tubulin, γ-tubulin and acetylated (Ac)-tubulin (MilliporeSigma), Myc-tag (Medical & Biological Laboratories, Nagoya, Japan), CUL5 (Santa Cruz Biotechnology, Dallas, TX) and FOP (Abnova, Taipei, Taiwan); Alexa488-conjugated mouse monoclonal antibody against Ki-67 (BioLegend, San Diego, CA); rat monoclonal antibody against HA (MilliporeSigma); HRP-conjugated sheep anti-mouse IgG, donkey anti-rabbit IgG or goat anti-rat IgG (GE Healthcare, Chalfont St Giles, UK), Alexa350, Alexa488, Alexa568, Alexa633 or Alexa647-conjugated goat anti-mouse IgG, goat anti-mouse IgG1, goat anti-mouse IgG2b or goat anti-rabbit IgG (Thermo Fisher Scientific). Detailed information about antibodies used in this study is shown in Table S3.

Human cDNAs for CEP97 and CP110 were provided by Tetsuo Kobayashi (Nara Institute of Science and Technology, Ikoma, Japan) and Brian D. Dynlacht (New York University, New York, NY). Plasmids encoding HA-ubiquitin (#18712), HA-streptag-CUL3 (#36970) and HA-RBX1 (#19897) were purchased from Addgene. The human cDNA for CUL1 was cloned by PCR amplification. To screen for CEP97 E3 ligase, we selected 179 genes encoding BTB domain-containing proteins by searching Swiss-Prot database and a previous report (Stogios et al., 2005). Of these 179 genes, we used 150 genes in this study; and of those, 29 genes had two variant clones in a HuPEX library (Maruyama et al., 2012) (Table S1). DNA of 101 clones was each inserted into the pDESTMN-Venus206K mammalian expression vector. DNA of 78 clones was each inserted into pENTR vector, and the resulting 78 cDNAs were transferred into the pDESTMN-Venus206K vector in an LR reaction using a GATEWAY LR clonase mix (#11791-043; Thermo Fisher Scientific). The cDNA plasmids for mutants of CEP97 and KCTD10 were constructed using a site-directed mutagenesis kit (Agilent, Santa Clara, CA). The cDNA plasmid for siRNA-resistant (sr)-KCTD10 was constructed by introducing three silent mutations into a target sequence of KCTD10 siRNA (5′-GAAGCACTTTGGCACCATC-3′, mutated nucleotides are underlined) by using a site-directed mutagenesis kit (Agilent).

The list of siRNAs used in this study is shown in Table S4. Silencer negative control siRNA (4390843; Thermo Fisher Scientific) and ON-TARGETplus non-targeting siRNA (D-001810-01-05; GE Dharmacon, Lafayette, CO) were used as control siRNA #1 and #2, respectively. In knockdown experiments, we quantified the effects of knockdown based on the premise that siRNAs were effectively transfected into almost all cells, since we observed that CUL3 siRNAs grossly reduced the fluorescence intensity of anti-CUL3 immunostaining in almost all cells on the plate (Fig. S8).

Human telomerase reverse transcriptase-immortalized retinal pigmented epithelial (RPE)-1 cells were provided by Hiroyuki Nakanishi (Kumamoto University, Kumamoto, Japan). Cells had recently been authenticated and tested for contamination. RPE1 cells were cultured in Dulbecco's modified Eagle's medium (DMEM)/Ham's F-12 (042-30555; Wako Pure Chemical, Osaka, Japan) supplemented with 10% fetal calf serum (FCS; Biosera, Kansas City, MO). HEK293T cells obtained from ATCC (No. CRL-3216; Manassas, VA) were cultured in DMEM (044-29765; Wako Pure Chemical) supplemented with 10% FCS. Both cell lines were tested for mycoplasma contamination and found to be negative. Transfection with plasmids was performed using Lipofectamine LTX with PLUS reagent (Thermo Fisher Scientific) in RPE1 cells, and FuGENE HD (Promega, Madison, WI) or JetPEI (PolyPlus, New York, NY) in HEK293T cells. Transfection with siRNAs was performed using Lipofectamine RNAiMAX (Thermo Fisher Scientific). For knockdown experiments, RPE1 cells were plated on coverslips in 6-well culture plates at an average density of 1.0×10⁴ cells/cm². At 24 h after transfection, cells were serum-starved for 48 h before fixation. For rescue experiments, RPE1 cells were plated on coverslips in 6-well culture plates at an average density of 1.0×10⁴ cells/cm², transfected with plasmids and cultured for 4 h. Then, cells were transfected with siRNAs, cultured for 24 h and serum-starved for 48 h before fixation. To examine the effects of MG-132 or E3 ligase inhibitors, RPE1 cells were plated on coverslips in 6-well culture plates at high density of 3.5×10⁴ cells/ cm² and serum starved for 24 h. Exposure of 0.3 µM MG-132, 0.5 µM MLN4924, 100 µM Heclin or 100 µM VH298 was carried out for 24 h before fixation or cell harvesting.

Immunoblotting was carried out as described previously (Nagai et al., 2013; Nagai and Mizuno, 2017). To examine the levels of CEP97 protein, cells were lysed with lysis buffer (150 mM NaCl, 50 mM HEPES-KOH pH 7.2, 1 mM MgCl₂, 1% Triton X-100, 1 mM PMSF, 10 µg/ml leupeptin, 3 µg/ml pepstatin, 10 µM E-64). To examine CEP97 phosphorylation, MG-132-treated cell lysates were incubated with 0.25 U/µl of λ-phosphatase (New England BioLabs, Ipswich, MA). Proteins were transferred to Immobilon-P (MilliporeSigma) polyvinylidene difluoride membrane using semi-dry transfer apparatus (BE-300. BIO CRAFT, Tokyo, Japan) according to the manufacturer's instructions. Protein-transferred membranes were subjected to immunoblot analyses using a standard protocol.

Total RNA was isolated from RPE1 cells using an Isogen II kit (Nippon Gene, Tokyo, Japan). One µg total RNA was subject to reverse transcription to yield single-stranded cDNAs using a PrimeScript RT Reagent Kit (TaKaRa, Shiga, Japan). For quantitative PCR, cDNA fragments were amplified using SYBR Premix Ex Taq II (TaKaRa) and the products was detected by Applied Biosystems 7300 (Thermo Fisher Scientific). All reactions were done according to the manufacturer's instruction. Sequences of primers for quantitative PCR are shown in Table S5. Analyses of the data were done using 7300 System Software (Thermo Fisher Scientific) and Microsoft Excel (Microsoft, Redmond, WA).

Immunostaining was carried out as described previously (Oda et al., 2014). Cells were fixed with 4% paraformaldehyde in phosphate-buffered saline (PBS) followed by methanol fixation or with methanol alone. Cells were stained with primary antibody diluted with Can-Get-Signal Immunostain (Toyobo, Osaka, Japan) and secondary antibody diluted with 2% FCS in PBS. Fluorescence images were obtained using a DMI6000B fluorescence microscope (Leica Microsystems, Wetzlar, Germany) equipped with a PL Apo 63× and 100× oil objective lens and CCD camera (Cool SNAP HQ, Roper Scientific, Vianen, The Netherlands). Image acquisition was performed using LAS AF software (Leica Microsystems). Image processing was performed using ImageJ (NIH, Bethesda, MD).

Immunoprecipitation was performed as described previously (Toshima et al., 2001). Briefly, cells were lysed with lysis buffer (150 mM NaCl, 50 mM HEPES-KOH pH 7.2, 1 mM MgCl₂, 1% Triton X-100, 1 mM PMSF, 10 µg/ml leupeptin, 3 µg/ml pepstatin, 10 µM E-64) and centrifuged. Lysates were pre-cleared with Protein A-Sepharose (GE Healthcare) for 1 h at 4°C, and the supernatants were incubated with antibodies and Protein A-Sepharose or GST-anti-GFP-nanobody pre-bound to glutathione-Sepharose beads (GE Healthcare) for 4 h at 4°C. After centrifugation, immunoprecipitates were washed three times and used for the immunoblot analysis. Detection of protein polyubiquitylation was performed as described previously (Yatsu et al., 2015). Briefly, cells were treated with 10 µM MG-132 before harvesting and lysed with lysis buffer (150 mM NaCl, 50 mM HEPES-KOH pH 7.2, 1 mM MgCl₂, 1% Triton X-100, 1 mM PMSF, 10 µg/ml leupeptin, 3 µg/ml pepstatin, 10 µM E-64, 10 mM N-ethylmaleimide) and centrifuged. Lysates were pre-cleared with Protein A-Sepharose for 1 h at 4°C and the supernatants were incubated with antibodies and Protein A-Sepharose or GST-anti-mCherry-nanobody prebound to glutathione-Sepharose beads (GE Healthcare) for 2 h at 4°C. After centrifugation, immunoprecipitates were washed twice and used for the immunoblot analysis.

VIP assay was performed as described previously (Katoh et al., 2015). Briefly, HEK293T cells were co-transfected with plasmids encoding CEP97-mCherry and Venus-tagged BTB domain-containing proteins and cultured for 2 days. Cells were lysed with HNTG buffer (150 mM NaCl, 20 mM HEPES-KOH pH 7.4, 0.1% Triton X-100, 10% glycerol, 1 mM PMSF, 10 µg/ml leupeptin, 3 µg/ml pepstatin, 10 µM E-64) and incubated on ice for 15 min. After centrifugation, supernatants were incubated with GST-anti-mCherry nanobody–prebound glutathione-Sepharose 4B (GE Healthcare) for 1 h at 4°C. Precipitates were washed with HNTG buffer three times and then transferred into a 96-well plate. Fluorescence on the beads was observed using a DMI6000B fluorescence microscope (Leica Microsystems) equipped with a C PLAN L20× dry objective lens and CCD camera (Cool SNAP HQ, Roper Scientific). Image acquisition was performed using LAS AF software (Leica Microsystems). Image processing was performed using ImageJ (NIH). A protein was considered as a candidate if its Venus fluorescence on the beads was more intense than that of YFP-CP110 beads.

All statistical analyses were carried out by Prism 6 (GraphPad Software, La Jolla, CA). Data are expressed as the mean±standard error of the mean (±s.e.m.) of at least three independent experiments. Sample size and number of repeated experiments are described in figure legends. P-values were calculated using the Student's t-test for two-group comparisons and one-way ANOVA followed by the Tukey's test for multiple data set comparisons. All statistical analyses were conducted with a significance level of α=0.05 (P<0.05).

We thank Kazuki Irie, Shinya Umeda, Yo Hasegawa and Ryo Tamura (Tohoku University, Sendai, Japan) for technical assistance. We also thank Dr Tadashi Nakagawa and Dr Keiko Nakayama (Tohoku University) for helpful advice, Dr Yohei Kato and Dr Kazuhisa Nakayama (Kyoto University, Kyoto, Japan) for helpful advice and for providing the cDNA plasmids for GST-tagged anti-GFP and anti-mCherry-nanobodies, Dr Tetsuo Kobayashi (Nara Institute of Science and Technology, Ikoma, Japan) and Dr Brian D. Dynlacht (New York University, New York, NY) for providing cDNAs for CEP97 and CP110, and Dr Hiroyuki Nakanishi (Kumamoto University, Kumamoto, Japan) for providing RPE1 cells.