Roles of TOG and jelly-roll domains of centrosomal protein CEP104 in its functions in cilium elongation and Hedgehog signaling

Primary cilia are generated through the extension of the micro-tubule-based axoneme. Centrosomal protein 104 (CEP104) localizes to the tip of the elongating axoneme, and CEP104 mutations are linked to a ciliopathy, Joubert syndrome. Thus, CEP104 has been implicated in ciliogenesis. However, the mechanism by which CEP104 regulates ciliogenesis remains elusive. We report here that CEP104 is critical for cilium elongation but not for initiating ciliogenesis. We also demonstrated that the tumor-overexpressed gene (TOG) domain of CEP104 exhibits microtu-bule-polymerizing activity and that this activity is essential for the cilium-elongating activity of CEP104. Knockdown/rescue experiments showed that the N-terminal jelly-roll (JR) fold partially contributes to cilium-elongating activity of CEP104, but neither the zinc-finger region nor the S X IP motif is required for this activity. CEP104 binds to a centriole-capping protein, CP110, through the zinc-finger region and to a microtubule plus-end – binding protein, EB1, through the S X IP motif, indicating that the binding of CP110 and EB1 is dispensable for the cilium-elongating activity of CEP104. Moreover, CEP104 depletion does not affect CP110 removal from the mother centriole, which suggests that CEP104 functions after the removal of CP110. Last, we also showed that CEP104 is required for the ciliary entry of Smoothened and export of GPR161 upon Hedgehog signal activation and that the TOG domain plays a critical role in this activity. Our results define the roles of the individual domains of CEP104 in its functions in cilium elongation and Hedgehog signaling and should enhance our understanding of the mechanism underlying CEP104 mutation – associated ciliopathies.

Primary cilia are antenna-like protrusions that emanate from the surface of most vertebrate cells, and these structures consist of a microtubule (MT)-based axoneme that extends from a basal body and a ciliary membrane that envelopes the axoneme (1,2). Primary cilia are generally formed when a cell exits from the cell cycle and enters the quiescent phase (1,2). At the onset of ciliogenesis, the mother centriole is translocated to the plasma membrane and converted into the basal body, from which axonemal MTs extend. Two centrosomal proteins, CP110 and CEP97, localize at the distal ends of both mother and daughter centrioles and block axonemal MT assembly, and thus the removal of CP110 and CEP97 from the mother centriole is essential for initiating axoneme extension and ciliogenesis (3)(4)(5)(6)(7). Subsequently, the axoneme extends and ciliary components are transported into the cilia by the intraflagellar transport (IFT) system (8,9).
The IFT system recruits various types of ion channels and receptors to the ciliary membrane (10)(11)(12), and this allows primary cilia to serve as cellular antennae that sense and transmit diverse chemical signals, including Hedgehog (Hh) and Wnt, and mechanical signals (13)(14)(15). Thus, primary cilia play crucial roles in the development and homeostasis of several tissues, and defects in the formation of primary cilia lead to various types of human disorders, collectively termed ciliopathies, that present a complex set of symptoms, including neuronal and retinal degeneration, polycystic kidney, situs inversus, and polydactyly (16,17). Genetic studies of ciliopathies and knockdown-based screening have identified many genes involved in ciliogenesis, but the molecular mechanisms underlying ciliogenesis and the etiology of ciliopathies remain largely unknown (16,17).
Centrosomal protein 104 (CEP104), an evolutionarily conserved protein, is encoded by CEP104, which was recently reported to be a causative gene for JBTS (27,28). CEP104 binds to CP110 and CEP97 and colocalizes with these proteins at the distal ends of both mother and daughter centrioles in nonciliated cells (29,30). When cells are growth-arrested, CEP104 on the mother centriole is dissociated from the CP110-CEP97 complex and transferred to the tip of the elongating axoneme (30,31), and depletion of CEP104 causes the loss or shortening of the cilia (30). These findings strongly suggest that CEP104 plays a critical role in ciliogenesis. However, the mechanism by which CEP104 regulates ciliogenesis remains elusive.
CEP104 contains a jelly-roll fold (JR fold; also termed IFT25like fold) at the N terminus, a tumor-overexpressed gene (TOG) domain in the central region, and four tandem zinc-finger (ZF) repeats and an SXIP motif in the C-terminal region (32,33). CEP104 was recently reported to bind to CEP97 and CSPP1 (another JBTS-associated protein, named for centrosome and spindle pole-associated protein 1) through the JR domain, tubulins through the TOG domain, CP110 and NIMArelated kinase-1 (NEK1) through the ZF region, and end-binding protein-1 (EB1) through the SXIP motif (29,(32)(33)(34). These results suggest that by interacting with the aforementioned proteins, CEP104 can potentially regulate multiple steps in ciliogenesis, such as the conversion of the mother centriole into the basal body through its interaction with CP110 and CEP97, and the extension of the axoneme through the interaction with tubulins and EB1. However, the precise function of CEP104 at the distinct stages of ciliogenesis and the roles of the individual domains in CEP104 localization and function are unknown.
Here, we show that CEP104 is essential for the elongation of the ciliary axoneme but is not involved in the initiation of cilium formation. We provide evidence that the TOG domain of CEP104 exhibits MT-polymerizing activity and that this activity is essential for cilium elongation. We also show that the Nterminal JR fold partially contributes to the cilium-elongating activity of CEP104, but neither the C-terminal ZF region nor the SXIP motif is necessary for this activity, which indicates that the cilium-elongating activity of CEP104 does not require its binding to CP110, NEK1, or EB1. A recent study showed that the CEP104 mutation suppresses ciliary import of Smoothened (Smo) upon Hh signal activation (34). We here demonstrate that CEP104 is required for ciliary import of Smo and export of G-protein-coupled receptor 161 (GPR161) in Hh signal transduction and that the TOG domain of CEP104 plays a critical role in this activity.

Knockdown of CEP104 shortens cilia but does not affect frequency of cilium formation
To investigate the role of CEP104 in ciliogenesis, we first examined the effects of CEP104 knockdown on the formation and length of primary cilia. Human telomerase reverse transcriptase-immortalized retinal pigment epithelial-1 (RPE1) cells were transfected with CEP104-targeting siRNAs and cultured under serum-starved conditions. Immunoblotting of the cell extracts revealed that two independent siRNAs effectively suppressed CEP104 expression (Fig. 1A). To examine the population of ciliated cells and the length of cilia, cells were immunostained with an anti-acetyl (Ac)-tubulin antibody (Fig. 1B). Quantification of the results showed that whereas CEP104 knockdown exerted no apparent effect on the population of ciliated cells (Fig. 1C), the knockdown significantly shortened primary cilia (Fig. 1, B and D). The average 6 S.D. lengths of primary cilia were 2.79 6 0.65 mm in control cells and 1.66 6 0.52 and 1.80 6 0.49 mm in CEP104-knockdown cells (Fig. 1D). These results indicate that CEP104 functions in cilium elongation but is not involved in triggering cilium formation. In accordance with our results, a recent study showed that CEP104 silencing reduces the cilium length in Kupffer's vesicle in zebrafish (34).

CEP104 TOG domain promotes MT polymerization
To examine the mechanism by which CEP104 promotes the cilium elongation, we analyzed the function of the TOG domain in the central region of CEP104 ( Fig. 2A). TOG domains are recognized to bind to tubulins and regulate MT dynamics (35), and the CEP104 TOG domain was recently shown to exhibit tubulin-binding ability (32,33). However, whether the TOG domain of CEP104 can promote MT polymerization has remained unknown. To address this question, recombinant CEP104 TOG domain was expressed in Escherichia coli and purified (Fig. 2, A and B) and used in in vitro assays in which MT polymerization was measured by monitoring tubulin solution turbidity; the analysis revealed that the CEP104 TOG domain effectively promotes MT polymerization (Fig. 2C). For this analysis, we also constructed a mutant form of the CEP104 TOG domain, W448A/V493D/R626A (WA/VD/RA), in which three amino acid residues predicted to participate in tubulin binding (32,33) were point-mutated (Fig. 2, A and B), and analyzed its MT-polymerizing activity. In the in vitro assay, the WA/VD/RA mutant of the CEP104 TOG domain showed decreased MT-polymerizing activity (Fig. 2C). Thus, the TOG domain of CEP104 has the ability to promote MT polymerization, and this depends on the domain's tubulin-binding activity.

MT-polymerizing activity of TOG domain is required for cilium-elongating activity of CEP104
We next investigated the role of the MT-polymerizing activity of the TOG domain in the cilium-elongating activity of CEP104; we constructed plasmids encoding yellow fluorescent protein (YFP)-tagged, siRNA-resistant (sr) WT CEP104 and its TOG domain mutant (WA/VD/RA) and used these in knockdown/rescue experiments. Immunoblot analysis confirmed the expression of CEP104(WT)-YFP and its WA/VD/RA mutant (Fig. 2D). In control YFP-expressing cells, treatment with CEP104 siRNA caused cilium shortening, but the siRNA effect was rescued in cells expressing CEP104(WT) (Fig. 2, E and F), indicating that the cilium shortening occurred due to CEP104 depletion and not any off-target effect of CEP104 siRNA. By contrast, expression of the WA/VD/RA mutant did not rescue the cilium shortening induced by CEP104 siRNA (Fig. 2, E and F), indicating that the MT-polymerizing activity of the TOG domain is essential for the cilium-elongating activity of CEP104. Fluorescence image analyses showed that CEP104(WT)-YFP and its WA/VD/RA mutant localize to the ciliary base and axoneme (Fig. S1), suggesting that the tubulin binding is not required for ciliary localization of CEP104.
EB1 binding is not required for ciliary tip localization and cilium-elongating activity of CEP104 CEP104 was previously shown to bind to EB1 through an SXIP motif (SKIP, amino acids 905-908) located near the C terminus of CEP104 (29). In Chlamydomonas and human epithelial cells, EB1 and EB3 are reported to localize to the basal body and the tip of cilia (36,37), and CEP104 also localizes to the tip of cilia (31); this raised the possibility that EB1 participates in CEP104 localization to the ciliary tip. To test this, we treated RPE1 cells with an EB1-targeting siRNA, which effectively suppressed EB1 protein expression (Fig. 3A). In cells transfected with a control siRNA, immunostaining with an anti-CEP104 antibody revealed that CEP104 localizes to the two (daughter and mother) centrioles in nonciliated cells and to the ciliary tip and a single (daughter) centriole in ciliated cells. Unexpectedly, EB1 knockdown exerted no marked effect on CEP104 localization in either nonciliated or ciliated cells (Fig.  3B), which suggests that EB1 is not involved in the localization of CEP104 to the ciliary tip.
To examine whether EB1 binding contributes to the ciliumelongating function of CEP104, we constructed a CEP104 mutant, in which the SKIP sequence (amino acids 905-908) was replaced with SKNN ( Fig. 3C). Co-precipitation assays revealed that EB1 binds to CEP104(WT) but not the SKNN mutant (Fig.  3D), which confirmed that EB1 binds to CEP104 through the SKIP sequence. We next analyzed whether expression of YFPtagged CEP104(WT) or its SKNN mutant affects ciliary length in CEP104-knockdown cells. Immunoblotting confirmed the expression of CEP104(WT)-YFP and its SKNN mutant (Fig.   3E). The results of knockdown/rescue experiments showed that the cilium shortening induced by CEP104 knockdown was rescued to a similar level by the expression of the WT and the SKNN mutant of CEP104 (Fig. 3, F and G). Thus, the binding of EB1 is not essential for the cilium-elongating activity. Fluorescence image analysis showed that CEP104(SKNN)-YFP localizes to the ciliary base and axoneme (Fig. S1), further suggesting that the EB1 binding is not required for ciliary localization of CEP104.
CEP104 is not involved in CP110 removal from mother centriole CP110 localizes on the distal ends of both mother and daughter centrioles and suppresses inadequate ciliogenesis by blocking ciliary axoneme extension in proliferating cells (4). Upon serum starvation, CP110 is removed from the mother centriole, which results in axoneme extension and cilium formation (4). Intriguingly, CEP104 binds to and colocalizes with CP110 at the mother and daughter centrioles in serum-fed cells, and upon serum starvation, CP110 is removed from the mother Figure 1. CEP104 knockdown shortens cilia but exerts no effect on frequency of cilium formation. A, immunoblot analysis of the knockdown efficiency of CEP104 siRNAs. RPE1 cells were transfected with control or CEP104-targeting siRNAs and cultured for 48 h, and then cell lysates were immunoblotted with the indicated antibodies. B, effects of CEP104 knockdown on cilium formation and length. RPE1 cells were transfected with control or CEP104 siRNAs, serumstarved for 48 h, fixed, and immunostained with anti-Ac-tubulin (red) and anti-centrin (green) antibodies. DNA was stained with DAPI (blue). Scale bar, 10 mm. Magnified images of boxed regions are shown on the right. C, quantification of the percentage of ciliated cells, based on staining of Ac-tubulin. Lines and error bars show means and S.D. from three independent experiments. In each experiment, more than 50 cells were analyzed. D, quantification of cilium length in ciliated cells, based on staining of Ac-tubulin. Dot plots indicate cilium length of individual cells. Lines and error bars show means and S.D. (n = total number of cells used for measuring cilium length). p values were calculated by ordinary one-way ANOVA followed by Tukey's test.
centriole, and the CEP104 localized on the mother centriole is dissociated from CP110 and translocated to the tip of the elongating cilium (30,38). In accord with these findings, the results of co-precipitation analyses showed that CEP104 was associated with CP110 under serum-fed conditions but dissociated from CP110 under serum-starved conditions (Fig. 4A).
Because CEP104 interacts with CP110 at centrioles, CEP104 could potentially function in CP110 removal from the mother centriole. To test this possibility, we analyzed the effect of CEP104 knockdown on serum starvation-induced CP110 removal from the mother centriole. RPE1 cells were treated with CEP104 siRNAs, serum-starved, and then stained with antibodies against CP110, Ac-tubulin, and g-tubulin. In 67% of control cells, CP110 was localized to a single centriole (Fig. 4, B and C), which indicated CP110 removal from the mother centriole in these cells, and the percentage of cells with CP110 localized to a single centriole was unaffected by CEP104 knockdown (Fig. 4, B and C). This result suggests that CEP104 is not involved in serum starvation-induced CP110 removal from the mother centriole.

Effects of double knockdown of CEP104 and CP110 on cilium formation and length
To further investigate the functional relationship between CEP104 and CP110, we examined the effects of double knockdown of CEP104 and CP110 on cilium formation and length in serum-fed RPE1 cells by Ac-tubulin staining, and we compared these effects with those in CEP104 or CP110 single-knockdown cells. Treatment with CP110 siRNA effectively suppressed CP110 expression (Fig. 4D), and, as reported (4), CP110 single knockdown significantly increased the population of ciliated cells under the serumsupplemented conditions (Fig. 4, E and F). Ac-tubulin-positive structures formed by CP110 knockdown were costained with Arl13b and IFT88 (Fig. S2), indicating that they are the bona fide cilia but not the elongated centrioles, which were previously reported in nonciliated U2OS and extraembryonic endoderm stem (XEN) cells (39)(40)(41)(42). Notably, CEP104 knockdown did not affect the number of ciliated cells in the case of either control or CP110-knockdown cells (Fig. 4, E and F), which indicates that CEP104 neither promotes nor suppresses the inhibitory function of CP110 in ciliogenesis. Furthermore, whereas CEP104 single knockdown led to significant ciliary shortening, concurrent knockdown of CP110 did not additionally affect ciliary length in CEP104-knockdown cells (Fig. 4, E and G); this indicates that CP110 is not involved in the function of CEP104 in cilium elongation. Thus, CP110 and CEP104 function independently in cilium formation and elongation, Figure 2. CEP104 TOG domain exhibits MT-polymerizing activity, and this activity is required for cilium-elongating activity of CEP104. A, schematic representation of CEP104 and its mutants. Numbers indicate amino acid residues. B, purification of CEP104 TOG domain and its mutant. Recombinant GST-CEP104 TOG proteins expressed in E. coli were bound to GSH-Sepharose, and GST was removed using PreScission Protease. Purified proteins were analyzed using SDS-PAGE and CBB staining. C, in vitro MT polymerization assay. Purified tubulin (12.5 mM) was incubated at 37°C in the absence (buffer) or presence of each of 2 mM TOG domain protein of CEP104 (WT) or WA/VD/RA mutant. Kinetics of tubulin polymerization was measured based on turbidity at 350 nm. AU, arbitrary unit. Lines and error bars show means and S.D. from three independent experiments. Dot plots indicate the turbidity in each experiment. D, expression of CEP104-YFP and its mutant in CEP104-depleted cells. RPE1 cells were co-transfected with CEP104 siRNA and CEP104-YFP or its WA/VD/RA mutant and cultured for 48 h. Cell lysates were immunoblotted with anti-CEP104, anti-GFP, and anti-b-actin antibodies. E, CEP104(WT), but not CEP104(WA/VD/RA), rescues cilium shortening in CEP104-depleted cells. RPE1 cells were co-transfected with CEP104 siRNA and CEP104-YFP or its WA/VD/RA mutant, serum-starved for 48 h, fixed, and then immunostained with anti-Ac-tubulin (magenta) and anti-Arl13b (green) antibodies. DNA was stained with DAPI (blue). Cells were also imaged by YFP fluorescence (red). Scale bar, 10 mm. Magnified images of boxed regions are shown on the right. F, quantification of cilium length in ciliated cells, based on Ac-tubulin staining. Dot plots indicate the cilium length of individual cells. Lines and error bars show means and S.D. (n = total number of cells used for measuring cilium length). p values were calculated by ordinary one-way ANOVA followed by Tukey's test.
respectively, and CEP104 functions in promoting cilium elongation after CP110 removal from the mother centriole.
JR fold, but not ZF region, is involved in cilium-elongating activity of CEP104 CEP104 was shown to bind to CEP97 and CSPP1 through its N-terminal JR fold (amino acids 1-163) and to CP110 and NEK1 through the C-terminal ZF region (amino acids 730-887) (32)(33)(34). To examine the roles of the JR fold and ZF region in the cilium-elongating activity of CEP104, we constructed and tested two CEP104 deletion mutants: DJR (D1-164), lacking the JR fold, and DZF (D731-886), lacking the ZF region (Fig. 5A). The results of co-precipitation assays confirmed that CEP97 binds to CEP104-WT and DZF but not DJR (Fig. 5B) Figure 3. EB1 binding is not required for ciliary tip localization and cilium-elongating activity of CEP104. A, immunoblot analysis of knockdown efficiency of EB1 siRNA. RPE1 cells were transfected with control or EB1-targeting siRNA and cultured for 48 h, and cell lysates were immunoblotted with anti-EB1 and anti-a-tubulin antibodies. B, EB1 knockdown does not affect CEP104 localization at ciliary tip. RPE1 cells were transfected with control or EB1 siRNA, serum-starved for 48 h, fixed, and immunostained with anti-CEP104 (green), anti-Arl13b, and anti-pericentrin (red) antibodies. DNA was stained with DAPI (blue). Scale bar, 20 mM. Magnified images of boxed regions are shown on the right. C, schematic representation of CEP104 and its SKNN mutant. Numbers indicate amino acid residues. D, CEP104(SKNN) does not bind to EB1. HEK293T cells were co-transfected with CEP104-(Myc1His) and EB1-CFP (cyan fluorescent protein). Cell lysates were immunoprecipitated with GST-tagged anti-GFP-nanobody, prebound to GSH-Sepharose beads. Precipitates and lysates were immunoblotted with anti-GFP and anti-Myc antibodies. E, expression of CEP104-YFP or its SKNN mutant in CEP104-depleted cells. RPE1 cells were co-transfected with CEP104 siRNA and CEP104(WT) or its SKNN mutant and cultured for 48 h. Cell lysates were immunoblotted with anti-CEP104, anti-GFP, and anti-b-actin antibodies. F, CEP104(WT) and its SKNN mutant rescue cilium shortening in CEP104-depleted cells. RPE1 cells were co-transfected with CEP104 siRNA and CEP104-YFP or its SKNN mutant, serum-starved for 48 h, fixed, and then stained with anti-Ac-tubulin (magenta) and anti-Arl13b (green) antibodies. DNA was stained with DAPI (blue). Cells were also imaged by YFP fluorescence (red). Scale bar, 10 mm. Magnified images of boxed regions are shown on the right. G, quantification of cilium length in ciliated cells. Dot plots indicate the cilium length of individual cells. Lines and error bars show means and S.D. (n = total number of cells used for measuring cilium length). p values were calculated by ordinary one-way ANOVA followed by Tukey's test. and that CP110 and NEK1 bind to CEP104-WT and DJR but not DZF (Fig. 5, C and D). Next, we analyzed the rescue effects of the expression of YFP-tagged CEP104(WT) or its deletion mutants on cilium length in CEP104-knockdown cells. Immunoblotting results confirmed the expression of CEP104(WT)-YFP and its DJR and DZF mutants (Fig. 5E). Expression of CEP104(WT) or its DZF mutant almost fully rescued the cilium-shortening effect produced by CEP104 siRNA (Fig. 5, F and G). On the other hand, the rescue effect of CEP104(DJR) was modest and significantly weaker than that of CEP104(WT) (Fig. 5, F and G). Similarly, it was reported that the cells bearing the N-terminally truncated CEP104 mutant exhibited the shortened cilium (34). These results suggest that the N-terminal JR fold contributes, at least in part, to the cilium-elongating activity of CEP104, whereas the C-terminal ZF region and the binding of CP110 and NEK1 are dispensable for the function of CEP104 in cilium elongation. Fluorescence image analyses showed that CEP104(DJR) and CEP104(DZF) localize to the ciliary base and axoneme (Fig. S1), suggesting that neither the JR fold nor the ZF region is required for ciliary localization of CEP104.

CEP104 binding is not required for ciliogenesis-suppressing activity of CP110
Our aforementioned results showed that CP110 is not involved in the cilium-elongating function of CEP104; we examined here whether CEP104 is involved in the CP110 function of suppressing ciliogenesis. We constructed a deletion mutant, CP110(D907-937), which lacks CEP104-binding ability (33) (Fig. 6A), and by performing co-precipitation assays, we confirmed that this mutant had lost the CEP104-binding activity (Fig. 6B). We next tested the effect of overexpression of CP110(WT) or CP110 (D907-937) on serum starvation-induced ciliogenesis. Under serum starvation, 70% of control YFP-expressing cells were ciliated, but in cells overexpressing either CP110(WT) or CP110(D907-937), the serum starvation-induced ciliogenesis was significantly suppressed (with only 13 and 12% of the cells being ciliated, respectively) (Fig. 6, C and D). This result suggests that CP110 suppresses ciliogenesis independently of CEP104 binding.  (Fig. 7, A  and B), the relative intensity of ciliary Smo was markedly lower in SAG-treated CEP104-knockdown cells than in SAG-treated control siRNA-treated cells (Fig. 7, A and B). These results suggest that CEP104 is required for the Hh signal-induced ciliary entry of Smo.
To further examine the role of CEP104 in the ciliary trafficking of Hh-related proteins, we tested the effect of CEP104 knockdown on SAG-induced ciliary exit of GPR161, a receptor that is reported to localize to the ciliary membrane in the absence of Hh signal but to be removed from cilia upon Hh-signal activation (44,45). Accordingly, SAG treatment caused a reduction in the fluorescence intensity of ciliary GPR161 in control siRNA-transfected cells (Fig. 7, C and D). Notably, after CEP104 knockdown, the intensity of ciliary GPR161 was significantly increased in both untreated and SAG-treated cells (Fig.  7, C and D). These results suggest that CEP104 plays a role in the export of GPR161 from cilia in both unstimulated and Hh signal-stimulated cells.
To investigate the function of CEP104 in regulating Hh signaling, we analyzed the effect of CEP104 knockdown on the expression of the Hh-target gene GLI1. Quantitative PCR analyses of SAG-treated and -untreated cells revealed that GLI1 mRNA level was significantly increased after SAG treatment in control cells but not in CEP104-knockdown cells (Fig. 7E) and Figure 5. CEP104 JR fold, but not ZF region, is required for cilium-elongating activity of CEP104. A, schematic representation of CEP104 and its DJR and DZF mutants. Numbers indicate amino acid residues. B, CEP97 does not bind to CEP104(DJR). HEK293T cells were co-transfected with CEP97-FLAG and CEP104-YFP or its DJR or DZF mutant. Cell lysates were immunoprecipitated with anti-GFP antibody, and precipitates and lysates were immunoblotted with anti-FLAG and anti-GFP antibodies. C, CP110 does not bind to CEP104(DZF). HEK293T cells were co-transfected with FLAG-CP110 and CEP104-YFP or its mutants. Cell lysates were immunoprecipitated with anti-GFP antibody, and precipitates and lysates were immunoblotted with anti-FLAG and anti-GFP antibodies. D, NEK1 does not bind to CEP104(DZF). HEK293T cells were co-transfected with Halo-NEK1 and CEP104-YFP or its mutants. Cell lysates were immunoprecipitated with GST-tagged anti-GFP-nanobody, prebound to GSH-Sepharose beads, and precipitates and lysates were immunoblotted with anti-Halo and anti-GFP antibodies. E, expression of CEP104-YFP and its mutants in CEP104-depleted cells. RPE1 cells were co-transfected with CEP104 siRNA and CEP104-YFP or its DZF or DJR mutant and cultured for 48 h. Cell lysates were immunoblotted with anti-CEP104, anti-GFP, and anti-b-actin antibodies. F, CEP104(WT) and (DZF), but not (DJR), rescue cilium shortening in CEP104-depleted cells. RPE1 cells were co-transfected with CEP104 siRNA and CEP104-YFP or its mutants, serum-starved for 48 h, fixed, and then stained with anti-Ac-tubulin (magenta) and anti-Arl13b (green) antibodies. DNA was stained with DAPI (blue). Cells were also imaged by YFP fluorescence (red). Scale bar, 10 mm. Magnified images of boxed regions are shown on the right. G, quantification of cilium length in ciliated cells, based on Ac-tubulin staining. Dot plots indicate cilium length of individual cells. Lines and error bars show means and S.D. (n = total number of cells used for measuring cilium length). p values were calculated by ordinary one-way ANOVA followed by Tukey's test.
further that SAG-induced GLI1 gene expression was significantly lower in CEP104-knockdown cells than in control cells (Fig. 7E). These results indicate that CEP104 plays a critical role in the cilium-mediated Hh signal transduction.
Last, we investigated the roles of the TOG and JR domains of CEP104 in Hh signaling by knockdown/rescue experiments. Expression of CEP104(WT) and CEP104(DJR), but not CEP104 (WA/VD/RA), rescued the defects in SAG-induced ciliary entry of Smo and exit of GPR161 in CEP104-knockdown cells (Fig. 7, F and G), indicating that the TOG domain is critical, but the JR fold is dispensable, for the function of CEP104 in Hh signal-induced ciliary trafficking of Smo and GPR161.

Discussion
CEP104 has been previously shown to bind to and colocalize with two centriolar capping proteins, CP110 and CEP97, on the distal ends of both mother and daughter centrioles in nonciliated cells (29,30). In response to cell-cycle exit signals, CP110 and CEP97 are removed from the mother centriole, and CEP104 on the mother centriole is dissociated from these two proteins and translocated to the tip of the elongating cilium (29,30,38). The removal of CP110 and CEP97 is a key step required to initiate cilium extension, and CEP104 binds to these proteins at the distal ends of centrioles; therefore, CEP104 was proposed to play a role in ciliogenesis by affecting the centriolar-capping activity of CP110 and CEP97 or the removal of the two proteins from the mother centriole (29,30). However, we showed here that CEP104 depletion exerts no effect on CP110 removal from the mother centriole and further that a CP110 mutant (D907-937) lacking CEP104-binding ability suppresses ciliogenesis to a similar extent as WT CP110. Thus, CEP104 is involved in neither the centriolar removal of CP110 nor the centriolar-capping activity of CP110. Our results also showed that CEP104 depletion shortens cilia but does not significantly alter the frequency of ciliated cells. A recent study also showed that CEP104 mutation significantly reduces axoneme length (34). Together, these findings suggest that CEP104 does not participate in the initiation of cilium formation but preferentially functions at the stage of axoneme elongation after the CP110-CEP97 complex is removed from the mother centriole.
We investigated the roles of the individual domains of CEP104 in its function in cilium elongation by knockdown/rescue experiments. CEP104(DZF), which does not bind to either CP110 or NEK1, and CEP104(SKNN), which does not bind to EB1, retain the ability to promote cilium elongation; these results suggest that the interactions with CP110, NEK1, and EB1 are not involved in the cilium-elongating activity of CEP104. By contrast, the rescue effect of CEP104(DJR) was modest and significantly weaker than that of CEP104(WT), which suggests that the N-terminal JR fold plays some role in this activity. CSPP1 and CEP97 interact with CEP104 through the N-terminal JR fold-containing region (33,34). CSPP1 is a JBTS-associated gene product that colocalizes with CEP104 at the tip of cilia, and CSPP1 depletion induces cilium shortening (34), which suggests the possibility that the interaction with CSPP1 through the N-terminal JR fold-containing region is involved in the cilium-elongating activity of CEP104.
TOG domains are generally recognized to bind to tubulins and regulate MT dynamics (35). Recently, the crystal structure of the TOG domain of CEP104 was solved and its tubulin-binding activity was demonstrated (32,33). Here, we first showed that the purified TOG domain of CEP104 exhibits MT-polymerizing activity in cell-free assays, and then that mutation of the TOG domain at three key residues involved in tubulin binding impairs the domain's MT-polymerizing activity in cell-free assays and abolishes the cilium-elongating activity of CEP104 in cells. These results indicate that the tubulin-binding and MT-polymerizing activities of the TOG domain are essential for the cilium-elongating activity of CEP104.Whereas most TOG domain-harboring proteins, such as XMAP215, contain tandem repeats of TOG domains and regulate MT dynamics by means of multiple TOG domains (35), CEP104 harbors only one TOG domain. Intriguingly, a single TOG domain (TOG2) of CLASP was recently shown to hold the potential to suppress MT catastrophes (46). Thus, similar to TOG2 of CLASP, the single TOG domain of CEP104 might mediate CEP104-promoted MT polymerization. CEP104 also binds to EB1-which can potentially recruit binding proteins to MT plus-endsthrough the SXIP motif located near the C terminus, but our results showed that EB1 binding is dispensable for CEP104 localization and function. Moreover, a recent study showed that CSPP1 is not required for CEP104 localization to the ciliary tip (34). We also showed that all mutations introduced in this study have no apparent effect on ciliary localization of CEP104. Thus, further studies are required to enhance our understanding of the molecular mechanisms underlying the MT-polymerizing activity of the TOG domain of CEP104 and the localization of CEP104 at the ciliary tip.
CEP104 was previously shown to be a causative gene for JBTS (27,28), and mutations of other JBTS-associated genes, such as ARL13B, cause defects in Hh signaling (47). We have provided evidence here that CEP104 is critical for the control of cilium-dependent Hh signaling. Our results agree with the recently reported findings from a CRISPR-based screening of genes involved in Hh signaling (48,49) and a functional analysis of CEP104 in cultured cells and zebrafish (34). We showed that CEP104 knockdown suppressed the ciliary entry of Smo and the exit of GPR161; this suggests that CEP104 depletion causes defects in Hh signaling through not only the shortening of cilia, but also the disruption of normal ciliary trafficking.
Knockdown/rescue experiments revealed that expression of CEP104(WT), but not its TOG domain mutant, rescues the defects in SAG-induced ciliary Smo/GPR161 translocation in CEP104 knockdown cells, indicating that the TOG domain plays a critical role in the function of CEP104 in ciliary Smo/ GPR161 trafficking upon Hh signal activation. In contrast, expression of CEP104(DJR) almost fully restored the defects in ciliary Smo/GPR161 translocation in CEP104-knockdown cells, indicating that the JR fold is dispensable for the function of CEP104 in ciliary Smo/GPR161 trafficking. However, considering that our data are based on the result of overexpression of CEP104(DJR), we cannot exclude the possibility that the JR fold has some role in the function of CEP104 in Hh signaling and overexpression of CEP104(DJR) compensates for its functional attenuation. In this respect, a recent study showed that the Nterminally deleted mutation of CEP104 strongly suppresses SAG-induced ciliary entry of Smo (34). Inconsistent results may reflect the differences in their "Experimental Procedures" and the length of deleted region of CEP104 mutants used in these two studies; our study used the overexpression of CEP104 (DJR) (lacking the N-terminal 164 amino acids) in CEP104knockdown cells, whereas the previous study used the CRISPR-Cas9 system to edit endogenous CEP104 gene to its mutant gene (lacking the N-terminal 203 amino acids) (34). Taken together, these results suggest that the N-terminal JR fold-containing region of CEP104 plays some role in Smo trafficking. Further studies are required to define the roles of the JR fold and the following region (165-203 amino acids) for the function of CEP104 in Smo/GPR161 trafficking and Hh signaling.
A bioinformatics study predicted that the JR fold of CEP104 is structurally related to IFT25 (32). IFT25 and its binding partner IFT27 form a subcomplex of IFT-B and are essential for ciliary Hh signaling (50,51), and, interestingly, recombinant CEP104 was recently shown to interact with IFT27 (34). Thus, CEP104 could function in the trafficking of Hh signal-related ciliary membrane proteins by regulating the activity of the IFT machinery.
In conclusion, we showed that CEP104 plays a crucial role in cilium elongation and that this requires the MT-polymerizing activity of the TOG domain. We also showed that CEP104 plays an essential role in Hh signaling by regulating the ciliary trafficking of Smo and GPR161. Further investigation into CEP104 functions and regulation mechanisms will contribute to enhanced understanding of the pathogenesis of ciliopathies.
RPE1 cells were transfected with plasmids and siRNAs by using Lipofectamine LTX with PLUS reagent (Thermo Fisher Scientific #15338100) and Lipofectamine RNAi MAX (Thermo Fisher Scientific #13778150), respectively. For EB1/CEP104/ CP110 knockdown experiments, RPE1 cells were plated on coverslips in 6-well culture plates and transfected with the corresponding siRNAs, cultured for 24 h, and then serum-starved for 48 h before fixation. For knockdown/rescue experiments, RPE1 cells were plated on coverslips in 6-well culture plates, transfected with plasmids, cultured for 4 h, transfected with CEP104 siRNAs, cultured for 20 h, and then serum-starved for 48 h before fixation.
Immunofluorescence and fluorescence microscopy RPE1 cells were washed with PBS and then fixed/permeabilized by treating with 4% paraformaldehyde for 5 min at 37°C followed by methanol for 15 min at 220°C, with 4% paraformaldehyde for 30 min at 37°C followed by 0.1% Triton X-100 for 5 min at room temperature, with 4% paraformaldehyde for 30 min at 37°C followed by 1% SDS for 5 min and 0.1% Triton X-100 for 5 min at room temperature or with methanol alone. Cells were blocked with 2% FCS in PBS (30 min) and incubated with primary antibodies diluted in Can Get Signal immunostain solution A (Toyobo #NKB-501) (1 h at room temperature or overnight at 4°C) and then with secondary antibodies diluted in 2% FCS/PBS (1 h at room temperature). DNA was stained with 0.5 mg/ml DAPI (5 min) and mounted in embedding medium (100 mM Tris-HCl, pH 7.4, 1 mg/ml p-phenylenediamine, and 90% glycerol). Fluorescence images were obtained using a DMI 6000B fluorescence microscope (Leica Microsystems) equipped with a PL Apo 363 oil objective lens (numerical aperture 1.3) and CCD camera (Cool SNAP HQ, Roper Scientific) and an A1 confocal microscope (Nikon) with a 360 objective lens (numerical aperture 1.27). ImageJ software (National Institutes of Health) was used for Z-stack projection and measurement of cilium length.

In vitro MT polymerization assay
In vitro MT polymerization was performed as described previously (53). Briefly, 12.5 mM tubulin and 2 mM recombinant CEP104 TOG domain protein were incubated for 15 min on ice in assembly buffer (50 mM MES, pH 6.6, 3.4 M glycerol, 5 mM DTT, 1 mM EGTA, 5 mM MgSO 4 , and 1 mM GTP), and then 100ml aliquots of samples were pipetted into cold 96-well plates. Next, the plates were moved to a Styrofoam water box and floated on water at 25°C for 1 min and then immediately placed into a Benchmark Plus microplate reader (Bio-Rad), which was prewarmed at 37°C. Scattering was monitored at 350 nm in 10-s intervals over 1500 s. Minimum A 350 was used for baseline adjustment.

Quantitative RT-PCR
Total RNA was isolated from RPE1 cells by using an Isogen II kit (Nippon Gene #311-07361), and then 1 mg of total RNA was reverse-transcribed to obtain single-stranded cDNA by using a PrimeScript RT reagent kit (TaKaRa #RR037A). For quantitative PCR, cDNA fragments were amplified using SYBR Premix Ex Taq II (TaKaRa #RR081A), and the products were detected using an Applied Biosystems 7300 system (Thermo Fisher Sci-entific). All reactions were performed as per the manufacturer's instructions. The sequences of the primers for quantitative PCR were as follows: 59-CAGGGAGTGCAGCCAATACAG-39 (Gli1 forward primer), 59-GAGCGGCGGCTACAGTATA-39 (Gli1 reverse primer), 59-ATTTGGTCGCGGTTCTTG-39 (UBC forward primer), and 59-TGCCTTGACATTCTC-GATGGT-39 (UBC reverse primer). Data were analyzed using 7300 System Software (Thermo Fisher Scientific) and Microsoft Excel (Microsoft).

Statistical analysis
Data are expressed as scatter dot plots from more than three independent experiments. All statistical analyses were performed using Prism 8 (GraphPad Software), and p values were calculated using ordinary one-way ANOVA followed by Tukey's test for multiple-data set comparisons.

Data availability
All of the data described in this study are contained within the article.