HsSAS-6-dependent cartwheel assembly ensures stabilization of centriole intermediates

The centrosome is the main microtubule-organizing center of animal cells. It comprises a pair of centrioles surrounded by pericentriolar material (PCM). Centriole formation is indispensable for centrosome duplication and must be strictly regulated during cell cycle progression to ensure robust formation of bipolar spindles and correct chromosome segregation (Brito et al., 2012; Fu et al., 2015; Godinho and Pellman, 2014; Nigg and Raff, 2009).

Centriole formation starts with the assembly of a cartwheel structure at the proximal wall of the pre-existing centriole, which dictates the 9-fold radial symmetrical arrangement of the centriole (Hilbert et al., 2016; Kilburn et al., 2007; Kitagawa et al., 2011; Nakazawa et al., 2007; van Breugel et al., 2011). Following the initial assembly of cartwheel layers, the peripheral triplet microtubules are attached and distally capped, allowing further centriole elongation and maturation (Gönczy, 2012). In mammalian cells, the cartwheel is detectable throughout procentriole formation, but is absent in the mature mother centrioles in the subsequent G1 phase (Guichard et al., 2010).

Evolutionarily conserved proteins of the SAS-6 family (Dammermann et al., 2004; Leidel et al., 2005) are recognized as a core component of the cartwheel structure (Culver et al., 2009; Guichard et al., 2012; Kitagawa et al., 2011; Nakazawa et al., 2007; Stevens et al., 2010; van Breugel et al., 2011). SAS-6 proteins form rod-shaped homodimers with their coiled-coil domains, and these homodimers are thought to self-assemble with their globular heads oriented towards the central part of the cartwheel (Guichard et al., 2012; Hilbert et al., 2016; Kitagawa et al., 2011; Nakazawa et al., 2007; van Breugel et al., 2011, 2014). In most cells, a lack of SAS-6 proteins results in failure of centriole formation. In Chlamydomonas and Drosophila, defective centrioles with an aberrant number of triplet microtubules are detectable in the mutants lacking SAS-6 proteins (Nakazawa et al., 2007; Rodrigues-Martins et al., 2007). In fact, 90% of cells mutant for bld12, the SAS-6 homolog in Chlamydomonas, do not have centrioles, and aberrant centriole symmetries were only seen in a small fraction of the remaining 10% of cells (Nakazawa et al., 2007). Meanwhile, in human cells, a lack of the human SAS-6 (herein denoted HsSAS-6, also known as SASS6) results in the complete failure of centriole duplication (Leidel et al., 2005; Strnad et al., 2007).

Recent reports have further revealed the unique molecular properties of SAS-6 proteins and their essential role in the initiation of centriole duplication (Fong et al., 2014; Keller et al., 2014; Qiao et al., 2012). However, it is still poorly understood whether the self-assembly ability of SAS-6 proteins is conserved in higher eukaryotes, including humans, and how exactly additional factors facilitate their assembly in vivo. Furthermore, after the early stages of procentriole assembly, whether the cartwheel is dispensable or otherwise necessary for the maintenance of growing procentrioles remains to be clarified. These questions stem from technical limitations in the use of conventional knockout and RNAi-based knockdown approaches in cultured human cells. Here, we show the exact functions of HsSAS-6 in centriole duplication, via transient removal of endogenous HsSAS-6 in human cells achieved by using the auxin-inducible degron (AID)-based protein degradation system. Although HsSAS-6 was known to be essential for the initiation of cartwheel assembly, we find that it is also required for the maintenance of centriole intermediates. In addition, the presence of HsSAS-6 at procentrioles limits centriolar loading of Plk4 and thus ensures formation of a single procentriole per parental centriole.

Previous studies have demonstrated, through in vitro reconstitution, that SAS-6 protein fragments from Chlamydomonas and zebrafish form the central hub of the cartwheel structure (Kitagawa et al., 2011; van Breugel et al., 2011). However, we found that HsSAS-6 failed to self-assemble to form the central part of the cartwheel structure under the same conditions (data not shown). Therefore, we next attempted to examine whether HsSAS-6 is also the core component of the cartwheel structure in humans. Considering that the N-terminal globular domain of SAS-6 family proteins, which is required for self-oligomerization, is highly conserved among species (the K_d values of N–N self-interaction of SAS-6 family proteins are 50, 90, 60 and 110 µM for human, zebrafish, Chlamydomonas and C. elegans, respectively) (Kitagawa et al., 2011; van Breugel et al., 2011), we speculated that HsSAS-6 self-assembly could be induced by promoting protein–protein interactions under concentrated conditions. Lipid monolayers have been used not only for the formation of two-dimensional crystals of soluble proteins, but also for reconstitution of the early stages of endocytosis (Chiu, 1997; Ford et al., 2001, 2002; Lévy et al., 1999). We therefore employed the lipid monolayer system to test whether we could promote HsSAS-6 self-assembly in vitro (Fig. S1). Purified HsSAS-6 NL fragments [amino acids (aa) 1–487], which contain the N-terminal globular head and long coiled-coil domain were observed after performing the lipid monolayer assay and negative staining electron microscopy (Fig. 1A–F; Fig. S1). Intriguingly, we found that a ring-like structure had locally assembled from the HsSAS-6 NL fragments on the layer (Fig. 1C). In contrast, we could not find any such structures with the HsSAS-6 F131E mutant proteins, which are defective in the self-assembly through their N-terminal heads (data not shown) (Keller et al., 2014; Kitagawa et al., 2011). Furthermore, the mean diameter of the rings (21.7±1.8 nm) and the mean angle between the spokes (41.9±9.6°) were in good agreement with those previously reported for 9-fold symmetric cartwheel structures in vitro (Fig. 1D–F) (Kitagawa et al., 2011; van Breugel et al., 2011). Notably, this is the first evidence that HsSAS-6 proteins can also self-assemble to form the hub of the cartwheel structure in vitro. Thus, our data strongly support previous studies indicating that self-oligomerization of SAS-6 is a conserved molecular property that is necessary for cartwheel assembly.

The oligomerization ability of HsSAS-6 is essential for centriole formation (Kitagawa et al., 2011), but additional factors are thought to be required to facilitate cartwheel assembly in vivo. It has previously been suggested that procentriole formation starts with the interaction between Plk4 and STIL and the associated phosphorylation of STIL by Plk4, which is followed by HsSAS-6 loading to the assembly site of procentrioles (Arquint et al., 2015; Dzhindzhev et al., 2014; Kratz et al., 2015; Moyer et al., 2015; Ohta et al., 2014). These proteins are present at the proximal end of procentrioles throughout the growth of the centrioles, until they are removed during mitosis (Arquint and Nigg, 2014; Strnad et al., 2007). To understand the detailed process of the cartwheel assembly in cells, with a focus on centriolar HsSAS-6, we mapped the precise localization of Plk4, STIL and HsSAS-6 at the super-resolution level by using gated stimulated emission depletion (STED) microscopy. As expected, we observed similar localization patterns for these three proteins at the centrioles, relative to the reference markers Cep192 and polyglutamylated tubulin (Fig. 2A). Notably, we also found that Plk4, STIL and HsSAS-6 exhibited unique behavior during procentriole formation: the three proteins started to localize at the procentriole assembly site on the mother centriole wall (denoted stage II in Fig. 2), elongated (stage III) and then finally separated into two parts (stage IV) (Fig. 2A,B). This observation raises the possibility that these proteins exist in two different states, one between mother and procentrioles as a part of ‘connecting stalk’ (Guichard et al., 2010), and the other at the base of the procentriole as a part of the ‘cartwheel’ (Guichard et al., 2010). Interestingly, we also found, under super-resolution, that before STIL and HsSAS-6 loading, Plk4 was distributed as discrete dots, as opposed to a continuous ring, around the mother centriole (Fig. 2A,B). Overall, our data suggest that Plk4, STIL and HsSAS-6 assemble and continue to associate with each other closely throughout centriole duplication.

Although the cartwheel structure is known to be essential for the initiation of procentriole formation, whether it has an additional role in the subsequent processes of procentriole formation remained to be elucidated, presumably because of technical limitations. Therefore, to investigate the possible functions of the cartwheel structure in each step of procentriole formation, we generated an HsSAS-6-AID human cell line, which enabled swifter removal of endogenous HsSAS-6 proteins than the conventional reverse-genetic approach of RNAi. We applied a combination of genome editing by CRISPR/Cas9 and the AID-based protein degradation systems to human HCT116 cells (Natsume et al., 2016; Nishimura et al., 2009; Ran et al., 2013). In an HCT116 cell line constitutively expressing Oryza sativa (os)TIR1 (Natsume et al., 2016), which is an essential component of the SCF/E3 ligase required for AID-based degradation, both endogenous HsSAS-6 alleles were AID-tagged at their C-terminal ends (Fig. S2A,B). We note that even in the HsSAS-6-AID cell line even without auxin treatment, there are lower protein expression levels of HsSAS-6-AID compared to the endogenous HsSAS-6 expression levels in the unmodified HCT116 cell line (Fig. 3D; Fig. S2E). This may be because constitutively expressed TIR induces a leaky degradation of HsSAS-6-AID, and/or because the AID tag itself affects the stability of the protein. Nevertheless, the expression levels of HsSAS-6 mRNA and centrosome duplication in the HsSAS-6-AID cells were not significantly affected, although the cell cycle profile of the HsSAS-6-AID cells shows a slight accumulation in G1 compared to WT cells (Fig. S2C,F–H). We cannot rule out the possibility that the slight cell cycle delay might stem from the reduced amount of total HsSAS-6, or a compromised function for HsSAS-6 resulting from the AID tag. We next confirmed that HsSAS-6-AID protein localized at centrioles, and that the intensity of centriolar HsSAS-6 significantly decreased after 6 h of treatment with indole-3-acetic acid (IAA; natural auxin) (89% decreased) (Fig. 3A,B; Fig. S2D).

We further tested the degradation properties of HsSAS-6-AID proteins in these cells. We found that after addition of IAA, the level of cytoplasmic HsSAS-6-AID proteins was quickly reduced (within 1 h), whereas it took 6 h to completely remove HsSAS-6-AID proteins from the centrioles (Fig. 3C,D). This observation implies that HsSAS-6 proteins that are incorporated into the centriole structure might be more difficult to access and target for protein degradation. Next, we examined whether centriole duplication occurs when HsSAS-6-AID is degraded before centriole formation starts in the late G1 phase. HsSAS-6-AID cells were synchronized in the G2/M phase by treatment with nocodazole, released and then simultaneously treated with IAA until the S phase or the next G1 phase (Fig. 3E–I). The recruitment of HsSAS-6-AID to the centriole was inhibited as a result of the addition of IAA (−IAA 80%, +IAA 0%; Fig. 3F,G), and centriole duplication was effectively blocked when assessed by monitoring recruitment of centriole markers, specifically CP110 (also known as CCP110) in S phase (−IAA 79%, +IAA 14%; Fig. 3F,H), and centrin (centrin 1) and Cep192 in the next G1 phase (−IAA 100%, +IAA 32%; Fig. 3F,I), respectively. These results indicate that HsSAS-6-AID degradation induced by IAA before the initiation of centriole formation causes centriole duplication failure, as is the case with HsSAS-6 depletion in RNAi experiments (Leidel et al., 2005; Strnad et al., 2007). Overall, our results confirm that HsSAS-6 is required for the initiation of centriole formation. Additionally, we established the HsSAS-6-AID cell line, in which endogenous HsSAS-6-AID can be rapidly knocked down at the protein level.

Next, we sought to analyze whether HsSAS-6 is dispensable after the early stages of procentriole assembly, or alternatively whether it is essential for further elongation (Fig. 4B). To examine procentrioles, which are partially formed as intermediate structures, we synchronized cells in the early S phase by using a double thymidine block (DTB) (Fig. S3A). We first confirmed that HsSAS-6 was recruited to most of the centrioles, and that the capping protein CP110 was largely incorporated into the procentrioles (HsSAS-6, 97%; CP110, 79%; Fig. S3B). The lengths of the procentrioles in the early S phase were significantly shorter than those of the fully elongated procentrioles in the G2/M phase, when tested by measuring the distance between CP110 in the mother and procentrioles (0.52 µm at early S, and 0.75 µm at G2/M; Fig. 4A). On the other hand, human (h)Poc5, which only appears in fully elongated, mature procentrioles (Azimzadeh et al., 2009), had not yet been recruited to the procentrioles in the early S phase (0%, Fig. S3B).

Given that it normally takes 6 h to remove HsSAS-6-AID proteins from the centrioles in asynchronous cells (Fig. 3C), we arrested cells in the early S phase, and then treated them with 0.5 mM IAA to remove HsSAS-6-AID from centrioles for the next 6 h prior to analysis (Fig. S3A). Contrary to our expectation, we found that the centrioles remained very weakly stained with HsSAS-6 in 65% of the cells, even in the S-phase arrested cells after 6 h of IAA treatment (Fig. S3C–E). This is presumably because of continuous loading of HsSAS-6 at centrioles during the extended S phase. Hence, HsSAS-6 that is stably incorporated in the centriole structure may be difficult to efficiently target for protein degradation. Centriolar loading of HsSAS-6 is known to be dependent on Plk4 kinase activity (Dzhindzhev et al., 2014; Ohta et al., 2014). Therefore, to completely remove HsSAS-6 from the centrioles, we combined IAA treatment with simultaneous transient Plk4 inactivation via the addition of centrinone (Wong et al., 2015), a Plk4-specific inhibitor (Fig. 4C; Fig. S3F). We found that the addition of 0.5 mM IAA and 100 nM centrinone for 3 h in the early S phase or for 4 h in mid-S phase, was sufficient to remove HsSAS-6 from the centrioles (Fig. 4D–F; Fig. S3G).

After release from DTB in the early S phase, centriole duplication was complete in control cells as shown by the number of Cep192, γ-tubulin and NEDD-1 foci in the next G1 phase (94% duplicated; Fig. 4G; Fig. S3H). We found that in the cells treated with IAA and transiently with centrinone, the number of cells with duplicated centrioles was dramatically decreased in the next G1 phase (24% duplicated; Fig. 4G; Fig. S3H). By contrast, in the cells only treated transiently with centrinone, 81% of the procentrioles became mature centrioles in the next G1 phase, indicating that transient Plk4 inactivation itself affected procentriole formation less (Fig. 4G).

We then sought to investigate the effect of HsSAS-6 removal in the mid S phase, by releasing cells from DTB for 3 h and arresting them again through the addition of thymidine, IAA and centrinone for 4 h to remove HsSAS-6 (Fig. 4C; Fig. S3F). In this experiment, we found that the number of duplicated centrioles in the next G1 phase was dramatically increased compared with what was seen in the case of HsSAS-6 removal in early S phase (83% duplicated; Fig. 4H; Fig. S3H). Taken together, these data suggest that HsSAS-6, and probably the cartwheel structure, are necessary for the integrity of procentriole intermediates at least up to the early S phase, but are thereafter dispensable for subsequent centriole elongation and maturation (Fig. 4I,J).

The intermediate structures harboring cartwheel structures that were observed in DTB-arrested cells were reminiscent of the proximal half of procentrioles formed in Cep295-depleted cells (Chang et al., 2016; Tsuchiya et al., 2016). Cep295 is a conserved protein that is not required for the initiation of cartwheel assembly, but is necessary for the proper elongation and maturation of centrioles through recruitment of Cep135, Cep192, Cep152 and other PCM proteins (Chang et al., 2016; Fu et al., 2016; Saurya et al., 2016; Tsuchiya et al., 2016). Among these proteins, Cep135, also known as Bld10 in Chlamydomonas (Matsuura et al., 2004), binds to the C-terminus of HsSAS-6, presumably to promote procentriole formation in human cells (Lin et al., 2013). However, the precise time of the centriolar loading of Cep135 and its functions are still controversial.

Considering the above, we examined the localization patterns of these proteins using super-resolution microscopy (Fig. S4A–D,F,G). In the early S phase, Cep295 was already detectable at the procentriole assembly site in most centrosomes, and acetylation of centriolar microtubules also could also be seen; however, Cep135 and Cep192 had not yet been loaded (Fig. S4A,C,D). As centriole formation proceeded during the S phase, the border of Cep295 signals between the mother centriole and procentriole became clearer, and interestingly, loading of Cep135 to procentriole assembly sites was frequently observed (Fig. S4A,B). The recruitment of Cep192 at the walls of the procentrioles was frequently detectable only during the late S and G2 phases (Fig. S4A).

A recent study using cell-free cartwheel reconstitution with C. reinhardtii (Cr)SAS-6 revealed that Cep135/Bld10 is required for well-organized stacking of cartwheels (Guichard et al., 2017). Our data support the idea that that Cep135 loading may be a key event for stabilizing the procentriole structure (Fig. S4E). Overall, our results experimentally demonstrated for the first time the importance of HsSAS-6 until the procentriole structure is stabilized for subsequent maturation.

We next focused on the localization of Plk4 and STIL at the centrioles upon HsSAS-6 removal. The phosphorylation of STIL by Plk4 has been shown to facilitate STIL–HsSAS-6 interaction and centriolar loading of HsSAS-6, which drives negative regulation to ensure that only a single duplication site is initiated for each mother centriole (Ohta et al., 2014, 2018). Interestingly, a previous study showed that after AID-mediated Plk4 destruction or Plk4 inactivation with an analog-sensitive mutant (Plk4^AS), STIL rapidly disappeared from the centriole whereas HsSAS-6 was retained (Kim et al., 2016; Lambrus et al., 2015). This suggests that once the cartwheel structure is assembled, the interaction between HsSAS-6 and Plk4–STIL might be dispensable for the maintenance of the cartwheel structure.

Hence, we tested the behavior of Plk4 and STIL after HsSAS-6 removal from the centriole in the HsSAS-6-AID cell line. Interestingly, after 6 h of IAA treatment during the S and G2 phases, the intensities of HsSAS-6 and STIL at the centriole were similarly decreased, with only a slight difference [83% decreased (Fig. 5A,B,D,G) and 59% decreased (Fig. 5A,B,E,G) for HsSAS-6 and STIL, respectively]. By contrast, the intensity of Plk4 was increased more than twofold and returned to the ring-like pattern that is normally seen around the mother centriole wall during G1 and prior to procentriole formation (Fig. 5C,F,H). These results suggest that HsSAS-6 is required for restricting the presence of the Plk4–STIL module only at the procentriole assembly site, presumably to suppress the formation of additional centrioles after the intermediate structure is assembled (Fig. 5I).

In this study, we demonstrated human SAS-6 self-assembly in vitro and established that the ability of SAS-6 to assemble in the central part of the cartwheel structure is conserved. We also characterized the function of the cartwheel in centriole formation by developing a system for inducible HsSAS-6 degradation in human cells. We demonstrated that HsSAS-6 is essential not only for the initiation of procentriole formation, but also for the stabilization of the proximal half of the procentriole (Fig. 5J). However, it seems that the cartwheel structure is dispensable for the subsequent steps of full elongation and maturation of centrioles. Such requirement of the cartwheel structure implies a stepwise procentriole assembly in the following way. First, several layers of cartwheel structures are stacked on the mother centriole wall to establish the overall geometry of a procentriole. At this stage, the procentriole assembly is already limited to one site, while the centriole intermediate still requires the cartwheel structures for its maintenance. Thereafter, some critical components, such as Cep295 and Cep135, are incorporated into the intermediate (Fig. S4A–C,E), presumably in a cartwheel-dependent manner, and these would be expected to facilitate arrangement of centriolar triplet microtubules and further procentriole elongation. This step could ensure the stability of the centriole intermediate, thus serving as a critical transition point for the necessity of the cartwheel structure, as a scaffold (Fig. 5J). In Chlamydomonas, Bld10p/Cep135 is known to be an essential protein for cartwheel assembly by interacting with SAS-6 at its C-terminus and connecting the central hub with triplet microtubules. However, in humans, how Cep135 is involved in the procentriole assembly is not fully understood. In our previous study, we showed that Cep295 is required for the recruitment of Cep135 to the daughter centriole. Together with the findings in this study, it is possible that the acquisition of Cep135 leads to the stabilization of the centriole intermediate, and it would allow further elongation and maturation of the procentriole.

Additionally, our data suggest a novel function of HsSAS-6 at procentrioles. Given the observation that the depletion of HsSAS-6 after procentriole formation led to a significant increase in Plk4 around the mother centriole wall, the existence of an intact cartwheel structure is presumably necessary to maintain the system and suppress the formation of extra procentrioles (Fig. 5I,J). It would be interesting to clarify how the structural integrity of the cartwheel is monitored, and how such information is transmitted, so as to block centriolar accumulation of Plk4 and the formation of extra procentrioles. Our approach of using transient degradation of HsSAS-6 might provide insight into this as-yet-unexplored mechanism.

U2OS cells were obtained from the European Collection of Cell Culture (ECACC). Cells were cultured in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum at 37°C in 5% CO₂. HCT116 cells were cultured in McCoy's 5A medium supplemented with 10% fetal bovine serum and 2 mM L-glutamine. All cell lines were tested and shown to be without mycoplasma contamination.

To synchronize HCT116 cells at prometaphase, cells were treated with 100 ng/ml nocodazole for 12 h, washed three times with PBS and released into fresh medium. For cell synchronization in the early S phase, cells were treated with 2 mM thymidine for 12 h, released 8 h and treated with thymidine for 12 h again (double thymidine block; DTB).

To induce degradation of HsSAS-6-AID, 0.5 mM indole-3-acetic acid (IAA) was added to the culture medium. For HsSAS-6-AID removal in the S phase-arrested cells, they were treated with 0.5 mM IAA and/or 100 nM centrinone.

For flow cytometry analysis, cells were harvested and fixed in 70% ethanol. Fixed cells were washed and resuspended in PBS containing 1% BSA, 50 µg/ml RNaseA and 40 µg/ml propidium iodide. Then cells were incubated at 37°C for 30 min, filtered through a nylon mesh filter and analyzed by the Accuri C6 FACS machine (BD) equipped with FCS4 Express Cytometry software (De Novo Software).

The HsSAS-6-AID cell line was generated in the HCT116 CMV-OsTIR1 cell line (RIKEN BRC, RCB4662) by using the CRISPR/Cas9 system, as previously described (Natsume et al., 2016; Ran et al., 2013). GuideRNA oligonucleotides were hybridized and cloned into the BbsI-digested pX330 (Addgene #42230). To generate the homology-directed repair (HDR) targeting plasmids, a 500 bp synthetic target genomic DNA in modified pUC57 (GENEWIZ) was digested by BamHI, and an AID tag with a hygromycin or neomycin resistance cassette was cloned into the site. Genome information was obtained from NCBI RefSeq (NG_051914.1). Oligonucleotides and synthetic DNA sequences used here are listed in Table S1. Plasmids were generated using PrimeSTAR MAX DNA polymerase and/or the In-Fusion cloning system (TaKaRa). Transfection and colony selection were performed as previously described (Natsume et al., 2016).

To prepare genomic DNA, cells were lysed in 50 mM NaOH solution, boiled and neutralized with 100 mM Tris-HCl pH 8.0. The lysate was briefly spun down and the supernatant was served as a template. Genomic PCR was carried out using Phire Hot Start ll DNA polymerase (Thermo Fisher Scientific), according to the manufacturer's instructions. The following primers were used in this study: a, 5′-ATCTTCCAAACCCACAGCGC-3′; b, 5′-TTCTTGACGAGTTCTTCTGA-3′; c, 5′-GCCAAGTTCTAATTCCATCAG-3′; and d, 5′-TTTACCAATCCCCAAGTACA-3′.

For the detection of mRNA, total RNA was prepared with TRIzol Reagent (Life Technologies). Reverse transcription and following DNA amplification were performed by using a One Step PrimeScript RT-PCR Kit (TaKaRa). For each sample, 10 ng of total RNA was served as a template for semi-quantitative RT-PCR. The following primers were used in this study: Hprt_fw, 5′-GCAGACTTTGCTTTCCTTGG-3′ and Hprt_rv, 5′-CTCCAGATGTTTCCAAACTCAA-3′; HsSAS-6_ex14_fw, 5′-GTTGATGGTAGACTGACTTA-3′ and HsSAS-6_ex15_rv, 5′-TTTTCCTTATCAGTTGAGCA-3′. The following siRNAs were used (Life Technologies): Silencer Select siRNA against HsSAS-6 ORF (s46487), negative control #1 (4390843).

The following primary antibodies were used in this study: mouse monoclonal antibodies against Plk4 [Merck Millipore, 6H5, MABC544, for immunofluorescence (IF) 1:250], HsSAS-6 (Santa Cruz Biotechnology, sc-81431, IF 1:250), centrin-1 (Millipore, 20H5, IF 1:500), centrobin (Abcam, ab70448, IF 1:1000), γ-tubulin (GTU88) (Sigma, T5192, IF 1:1000), polyglutamylation modification (GT335, mAb) (AdipoGen, AG-20B-0020-C100, IF 1:1000), mAID (MBL, IF 1:500), NEDD1 (Abcam, ab57336, IF 1:1000) and Cyclin A (Abcam, 6E6, ab16726, IF 1:1000); and rabbit polyclonal antibodies against HsSAS-6 [Santa Cruz Biotechnology, sc-98506, for western blotting (WB) 1:2000], Cep295/KIAA1731 (Sigma-Aldrich, HPA038596, IF 1:1000), Cep192 (a gift from Laurence Pelletier, Lunenfeld-Tanenbaum Research Institute, Department of Molecular Genetics, University of Toronto, Canada, IF 1:1000), Cep192 (Bethyl, A302-324A, IF 1:1000), Cep152 (Bethyl, A302-479A and A302-480A, IF 1:1000), CP110 (proteintech, 12780-1-AP, IF 1:500), hPoc5 (Bethyl, A303-341A, IF 1:1000), STIL (Abcam, ab89314, IF 1:250, WB 1:2000) and CENP-F (Bethyl A301-617A, IF 1:1000). Secondary antibodies used were conjugated to Alexa Fluor 488, 568, 555, 594 and 647 (Molecular Probes, IF 1:1000). Alexa Fluor 488 or 647-labeled Cep192 and Cep152 were generated with Alexa Fluor labeling kits (Life Technologies) and used for triple staining (IF 1:500). Hoechest 33258 (Dojindo, IF 2 mg/ml) was used for nuclear staining. Goat polyclonal antibodies against mouse IgG (Promega, W402B, 1:5 000) and rabbit IgG (Promega, W401B, 1:5000) conjugated to horseradish peroxidase were used for western blotting.

For immunofluorescence analysis, cells were cultured on the coverslips, fixed with ice-cold methanol for 5 min at 20°C and washed with PBS. Then cells were permealized with 0.05% Triton X-100 in PBS (PBST), and incubated with primary antibodies in 3% BSA in PBST overnight at 4°C. After cells were washed with PBST twice, they were incubated with secondary antibodies for 1 h at room temperature. For triple staining with Alexa Fluor-labeled antibodies, cells were treated after secondary antibodies were washed out. Then cells were stained with Hoechst 33258 for 5 min, washed with PBST again and mounted onto the glass slides for observation.

For preparation of total cell lysates, cells were washed with PBS, lysed in lysis buffer [20 mM Tris-HCl pH 7.5, 150 mM NaCl, 0.5% Triton X-100, 1 mM DTT, 2 mM MgCl₂ and 1:1000 protease inhibitor cocktail (Nacalai Tesque)] and insoluble material was removed by centrifugation (15,000 g for 20 min). SDS-PAGE was performed using 6–12% polyacrylamide gels, followed by transfer on Immobilon-P membrane (Millipore Corporation). The membrane was probed with the primary antibodies, washed with 0.02% Tween 20 in PBS and incubated with HRP-conjugated secondary antibodies. The signals were detected with ECL Prime/Select reagents (GE healthcare) via the Chemi Doc XRSþ system (Bio-Rad).

The number of fluorescence signals were observed and counted from images acquired with an Axioplan2 fluorescence microscope (Carl Zeiss) with a 100×/1.4 NA plan-APOCHROMAT objective lens. Image capture and the signal quantification were performed using a DeltaVision deconvolution microscope (GE Healthcare) with softWoRx software (Applied Precision). For quantification of centriole signals, the values of the intensities were obtained from ‘Data Inspector’ in Tools in softWoRx. The values of cytoplasmic signals were subtracted from those of centriole signals. The reminder after the subtraction was taken as the value of centriole signal, and analyzed for statistical significance.

To obtain high-resolution images, the Leica TCS SP8 Hyvolution system with Huygens essential software (Scientific Volume Imaging; SVI) was used. The parameters of the settings were as previously described (Tsuchiya et al., 2016).

STED mages were taken with a Leica TCS SP8 STED system with a Leica HC PL APO 100×/1.40 oil STED WHITE objective, and 660 nm gated STED. The scan speed was set to 100 Hz in combination with five fold line average in a 512×80 format (pixel size 15 or 20 nm). The images were collected at 180 nm z steps. The STED images were processed by deconvolution with Huygens professional software (SVI). For the projection of Z images, max projection was applied by ImageJ. In this study, we visualized proteins of interest via Alexa Fluor 555-conjugated antibodies whereas the marker proteins were visualized by using Alexa Fluor 488-conjugated antibodies. Each pattern was tested at least in five cells from two independent experiments.

The GST–HsSAS-6 NL fragment was expressed in E. coli and purified by using glutathione–Sepharose beads (GE 17075601). The GST tag was cleaved by treatment with prescission protease (GE 27084301) and removed. The cleaved sample was gel-filtrated with Superdex 200 10/300 GL column (GE Healthcare), and assessed by SDS-PAGE and Simply Blue Safe staining. For in vitro reconstitution on lipid monolayer, used a modified version of the ‘Lipid Monolayer Assay Protocol’ from the Harvey McMahon laboratory website (http://www2.mrc-lmb.cam.ac.uk/groups/hmm/techniqs/mono.html). A lipid mixture containing 10% cholesterol, 40% phosphatidylehanolamine (PE), 40% phosphatidylcholine (PC) and 10% phosphatidylserine (PS) to a final concentration of 0.1 mg/ml in chloroform was loaded on the chamber filled with HKM buffer (25 mM Hepes pH7.4, 125 mM potassium acetate, 5 mM magnesium acetate, 1 mM dithiothreitol), and incubated for 1 h at room temperature to form a lipid monolayer. Then an electron microscopy (EM) grid was set on the layer, and protein solution was applied and incubated for 3 h at room temperature for the assembly of protein complex on the lipid monolayer. The sample was stained with 2% uranyl acetate and observed with a transmission electron microscope JEM-1010 (JEOL). Images were acquired by a VELETA CCD camera and analyzed with iTEM software (SEIKA Corporation).

Statistical analyses were performed using GraphPad Prism software (version 7.0) and Microsoft Excel. Comparisons between two groups were performed with a two-tailed t-test and comparisons among groups were performed by one-way ANOVA. Statistical significance was assigned for P<0.05, unless otherwise mentioned in the text.

We are grateful to Rie Hayashi at Leica Microsystems for technical support and access to STED microscopy facilities. We thank Andrew Shiau and Karen Oegema at the Ludwig Institute for Cancer Research for providing the PLK4 inhibitor centrinone, Harvey McMahon at MRC Laboratory of Molecular Biology for kindly giving us a Teflon block for the lipid monolayer assay, and Kazuko Ohishi for technical assistance. We also thank Kitagawa lab members for critical reading of the manuscript and advice on the experiments.