Deubiquitylase USP9X maintains centriolar satellite integrity by stabilizing pericentriolar material 1 protein

The centriolar satellite, centrosome and primary cilium network plays an important role in development by mediating signal transduction (Chavali et al., 2014; Conkar et al., 2017; Hori and Toda, 2017; Kurtulmus et al., 2016; Tollenaere et al., 2015). Centriolar satellites are small cytoplasmic granules that cluster in the vicinity of centrosomes (Bärenz et al., 2011; Hori et al., 2015; Hori and Toda, 2017; Tollenaere et al., 2015). Centriolar satellites contain more than 100 proteins (Chavali et al., 2014; Hori and Toda, 2017). One major proposed function of centriolar satellites is to transport proteins from the cytoplasm into centrosomes or primary cilia to regulate their assembly (Hori et al., 2015; Hori and Toda, 2017). In addition, centriolar satellites could be involved in regulating stress response and autophagy (Joachim et al., 2017; Shearer et al., 2016). Mutations in genes encoding centriolar satellite proteins, including CEP290, OFD1 and BBS4, cause ciliopathies that are genetic disorders causing kidney diseases, blindness, developmental retardation and neurological problems (Coene et al., 2009; Ferrante et al., 2009; Frank et al., 2008; Helou et al., 2007; Katsanis et al., 2001; Lopes et al., 2011; Perrault et al., 2007; Valente et al., 2006). Currently, mutations in a large body of genes in the centriolar satellite, centrosome and primary cilium network have been identified to cause ciliopathies.

Pericentriolar material 1 (PCM1) is a structural scaffold protein and a cellular marker of centriolar satellites (Hori and Toda, 2017; Tollenaere et al., 2015). PCM1 recruits proteins, including CEP290, OFD1, BBS4 and CEP131, to form the structural platform for assembly of centriolar satellites. Depletion of PCM1 disrupts centriolar satellites and causes mislocalization of centrosomal and satellite proteins (Dammermann and Merdes, 2002; Kim et al., 2008; Wang et al., 2016). Post-translational modification of PCM1 was found to regulate centriolar satellite integrity. Polo-like kinase 4 (PLK4) binds and phosphorylates PCM1 at serine 372, which mediates PCM1 dimerization and interaction with other satellite proteins such as BBS4 and CEP290. Depletion of PLK4 leads to dispersion of centriolar satellite particles (Hori et al., 2016; Hori and Toda, 2017). PCM1 can also be phosphorylated by other cell cycle-related kinases, including CDK1 and PLK1 (Hori and Toda, 2017; Olsen et al., 2010; Santamaria et al., 2011). It is speculated that phosphorylation of PCM1 by different kinases may regulate, at least in part, the cell cycle-dependent centriolar satellite remodeling (Hori et al., 2016; Hori and Toda, 2017; Zhang et al., 2017). The E3 ubiquitin ligase mindbomb 1 (MIB1) localizes to centriolar satellites by binding to PCM1 (Wang et al., 2016). MIB1 catalyzes PCM1 polyubiquitylation to promote proteasomal degradation (Wang et al., 2016), thereby functioning as a destabilizer of centriolar satellites. Another study found that MIB1 mediates monoubiquitylation of PCM1, CEP131 and CEP290, which was proposed to maintain centriolar satellite structures under non-stressed conditions, although the underlying mechanism is not known (Villumsen et al., 2013).

USP9X is a large deubiquitylase (∼290 kDa) that deubiquitylates substrates conjugated with monoubiquitin and/or polyubiquitin chains (Al-Hakim et al., 2008; Dupont et al., 2009; Schwickart et al., 2010). Loss-of-function mutations in USP9X cause female-specific syndromes, including intellectual disability and defects in neural development, typical phenotypes seen in ciliopathies (Homan et al., 2014; Paemka et al., 2015; Reijnders et al., 2016). In this regard, USP9X was found to localize along the ciliary axoneme in fibroblasts, but knockdown of USP9X has no effect on ciliogenesis (Reijnders et al., 2016). In another study, IQCB1 was found to recruit USP9X into centrosomes, where USP9X protects IQCB1 from ubiquitylation and degradation, which promotes ciliogenesis in human retinal pigment epithelium (RPE) cells (Das et al., 2017). In addition, two recent studies have found that USP9X regulates centrosome duplication (Li et al., 2017; Wang et al., 2017). Wang et al. (2017) showed that USP9X colocalizes with PCM1 and CEP55 in centrosomes. USP9X controls the protein abundances of PCM1 and CEP55, which could contribute to the requirement of USP9X in centrosome duplication. Li et al. (2017) found that USP9X colocalizes with CEP131 in centrosomes. USP9X binds and deubiquitylates CEP131 to antagonize proteasomal degradation, which could also contribute to the requirement of USP9X in centrosome duplication. Intriguingly, both PCM1 and CEP131 are also key centriolar satellite proteins. Whether USP9X is a centriolar satellite protein and its role in regulating centriolar satellite functions have not been investigated. In this study, our results reveal that USP9X deubiquitylates PCM1 to protect it from proteasomal degradation, by which USP9X stabilizes PCM1 and is required for maintaining centriolar satellite integrity.

In a previous study, we identified survival motor neuron (SMN) protein as a substrate of USP9X-mediated deubiquitylation. USP9X stabilizes the SMN complex and plays an important role in regulating Cajal body formation in the nucleus (Han et al., 2012). In that study, we performed a proteomics study to identify USP9X-interacting proteins; several proteins in the centriolar satellite, centrosome and primary cilium network, including CEP290, IQCB1, ATXN10 and CEP170, were identified with trypsinization-derived peptides (Han et al., 2012) (Fig. S1 and data not shown). We initiated our current study by investigating the interaction between USP9X and CEP290, because CEP290 is an important protein in the centriolar satellite, centrosome and primary cilium network. First, we found that endogenous USP9X interacted with CEP290 in 293T cells in a co-immunoprecipitation assay (Fig. 1A). Second, immunostaining showed that CEP290 existed as cytoplasmic foci, and USP9X primarily localizes in the cytoplasm of HeLa cells (Fig. 1B), 293T and HCT116 cells (data not shown). Remarkably, USP9X colocalized with CEP290 in foci in these cell lines. Lastly, using FLAG-tagged USP9X deletion mutants expressing USP9X(1–966), USP9X(967–1537), USP9X(1531–1971) or USP9X(1971–2554), immunoprecipitation assays revealed that the N-terminal USP9X fragment, USP9X(1–966), interacted with endogenous CEP290 (Fig. 1C,D). Collectively, these results demonstrate that USP9X and CEP290 form a protein complex in the cell, requiring the N-terminal region of USP9X.

CEP290 resides in centriolar satellites, centrosomes and primary cilia (Coppieters et al., 2010; Drivas and Bennett, 2014; Kim et al., 2008). To identify subcellular structures in which USP9X colocalizes with CEP290, we colocalized USP9X with a centrosomal marker, γ-tubulin, and a centriolar satellite marker, PCM1. USP9X primarily localized within the cytoplasm (Fig. 1B), but a small population of USP9X formed foci partially colocalizing with and largely surrounding γ-tubulin foci (Fig. 1E,F). Thus, the majority of foci-like USP9X surrounds centrosomes. In contrast, foci-like USP9X primarily colocalized with the centriolar satellite component PCM1 (Fig. 1E,F). PCM1 can dynamically traffic to centrosomes (Dammermann and Merdes, 2002; Kim et al., 2008; Lopes et al., 2011). Co-staining of PCM1 with γ-tubulin showed that PCM1 foci largely surrounded γ-tubulin with partial colocalization (Fig. 1E,F). These results indicate that the majority of PCM1 foci in HeLa and HCT116 are centriolar satellites, in which USP9X colocalizes.

The clustering of centriolar satellites around centrosomes depends upon microtubule-based transport (Dammermann and Merdes, 2002; Hori and Toda, 2017). Disrupting the microtubule network impairs centriolar satellites, but not centrosomes (Hori and Toda, 2017; Kim et al., 2008). We therefore examined whether PCM1-positive large foci are affected by microtubule depolymerization. Indeed, nocodazole-mediated microtubule depolymerization caused dispersion of PCM1 (Fig. S2), further supporting that the PCM1-positive foci in which USP9X colocalizes are centriolar satellites.

USP9X was found to bind PCM1 in co-immunoprecipitation assays (Li et al., 2017; Wang et al., 2017). Consistent with previous reports, endogenous PCM1 and USP9X in HeLa cells formed a protein complex in a co-immunoprecipitation assay (Fig. 2A). Knockdown of CEP290 had no effect on the protein levels of USP9X and PCM1, as well as their interaction (Fig. 2B), indicating that CEP290 does not mediate the PCM1 and USP9X interaction. Next, we generated deletion mutants of the PCM1 isoform 14 (1859 amino acids) (Fig. 2C). Expressed PCM1(1–1300) and PCM1(851–1859), but not PCM1(1–400), pulled down endogenous USP9X in co-immunoprecipitation assays. PCM1(1–850) and PCM1(1301–1859) weakly interacted with USP9X (Fig. 2D). These results indicate that the middle region of PCM1, mainly between 851 and 1300 amino acids, interacts with USP9X. We then generated a PCM1 fragment expressing Myc-tagged PCM1(851–1300). Co-immunoprecipitation assays revealed that Myc-PCM1(851–1300) interacted with FLAG-USP9X(1971–2554) when co-expressed in 293T cells (Fig. 2E). The USP9X(967–1537) fragment was expressed at low levels. We were unable to obtain high-purity recombinant GST-PCM1 or GST-PCM1(851–1300), which restrained us from assessing whether PCM1 directly binds USP9X using GST pulldown assays. Nevertheless, these results indicate that USP9X forms a protein complex with PCM1 in the cell, requiring the middle region of PCM1 and the C-terminal region of USP9X.

siRNA-mediated depletion of USP9X reduces PCM1 protein levels (Li et al., 2017; Wang et al., 2017). We confirmed that stable knockdown of USP9X using two distinct shRNA constructs in HeLa or HCT116 cells caused significant reduction of PCM1 protein levels (Fig. 3A). Interestingly, CEP290 protein levels were reduced in HCT116 cells, but not in HeLa cells, upon depletion of USP9X, indicating a cell type-specific effect (Fig. 3A). In contrast, knockdown of another deubiquitylase, USP7, had no effect on protein levels of PCM1 or CEP290 (Fig. 3B). The proteasome inhibitor MG132, but not the lysosome inhibitor bafilomycin A1, blunted the PCM1 protein level difference between control and USP9X knockdown cells (Fig. 3C). Lower levels of PCM1 protein in MG132-treated cells than those in dimethyl sulfoxide (DMSO)- or bafilomycin-treated cells were likely due to cell death, as judged by low α-tubulin (a loading control) protein levels in these samples. Thus, USP9X knockdown may promote proteasomal degradation of PCM1.

Next, we asked whether PCM1 is a substrate of USP9X-mediated deubiquitylation. To examine this, endogenous PCM1 in USP9X or control knockdown cell lysates were subjected to denaturing immunoprecipitation, which can avoid non-specific binding. Immunoblotting of ubiquitin showed that depletion of USP9X resulted in a substantial elevation of PCM1 ubiquitylation, despite less PCM1 being immunoprecipitated in USP9X knockdown cells than that in GFP knockdown cells (Fig. 3D). Lastly, we examined the protein half-life of endogenous PCM1 in USP9X or control knockdown cells using the cycloheximide-chase assay. Depletion of USP9X using either of the shRNA constructs greatly shortened the protein half-life of PCM1 compared with that of control knockdown cells, when analyzed by immunoblotting assays (Fig. 3E) and densitometric quantification of immunoblots (Fig. 3F). Taken together, these results demonstrate that USP9X stabilizes PCM1 by deubiquitylating PCM1 to protect it from proteasomal degradation.

PCM1 is a core structural protein of centriolar satellites (Hori and Toda, 2017; Tollenaere et al., 2015). Because USP9X stabilizes PCM1, we predicted USP9X to be a key regulator of centriolar satellite integrity. To assess this prediction, we mixed GFP knockdown cells with USP9X knockdown cells to create two cell populations that express either high or low USP9X protein levels. The mixed cells were co-immunostained with anti-USP9X and anti-PCM1 antibodies to visualize their colocalization. In ∼80% of HeLa cells expressing high USP9X protein levels (marked with a white arrowhead in Fig. 4A and quantification results in B), PCM1 formed large foci that colocalized with USP9X. In contrast, PCM1-positive large foci were only detected in ∼10% of cells expressing low USP9X protein levels (marked with a white arrow in Fig. 4A and quantification results in B). Similar results were also obtained when the same assay was performed by mixing GFP knockdown or USP9X knockdown HCT116 cells (Fig. 4C,D). These results indicate that depletion of USP9X disrupts centriolar satellites.

Centriolar satellites are crucial for the proper localization of some proteins in the centriolar satellite, centrosome and primary cilia network (Kim et al., 2008). CEP290 is found in all three subcellular structures of the network. We next examined whether the localization of CEP290 is affected by USP9X knockdown. In GFP knockdown HeLa cells, CEP290 stained as a few foci that overlapped with PCM1 (Fig. 5A), and surrounded γ-tubulin with partial overlap (Fig. 5B), indicating that the majority of CEP290 resides in centriolar satellites. In striking contrast, in USP9X knockdown HeLa cells, CEP290 stained as two foci that barely co-stained with dispersed PCM1 dots (Fig. 5C), but largely overlapped with γ-tubulin (Fig. 5D). Thus, CEP290 localizes in centrosomes when USP9X is depleted. In quantification, ∼5% GFP knockdown HeLa cells had two CEP290 foci, whereas ∼70% USP9X knockdown cells showed two CEP290 foci (centrosome localization) (Fig. 5E). Overall, these results indicate that loss of USP9X causes localization of CEP290 to centrosomes.

MIB1 is a ubiquitin ligase that ubiquitylates PCM1 and promotes its proteasomal degradation (Wang et al., 2016). Consistent with the previous study, stable knockdown of MIB1 using two distinct shRNA constructs resulted in a significant accumulation of PCM1 in HeLa cells, despite only ∼50% MIB1 being depleted (Fig. 6A). We therefore expected depletion of MIB1 in USP9X knockdown cells to increase PCM1 protein levels, which, in turn, could rescue defects caused by the loss of USP9X. Indeed, PCM1 protein levels increased from ∼10% in USP9X knockdown cells to ∼50% in USP9X/MIB1 double knockdown cells when referenced to endogenous PCM1 protein levels (comparing lanes 3 and 4 with lane 1 in Fig. 6B). Centriolar satellites re-appeared in USP9X knockdown cells upon MIB1 depletion (marked with a white arrowhead in Fig. 6C). A few USP9X knockdown cells still had high USP9X expression that showed colocalization of USP9X and PCM1 (an example marked with a white arrow in Fig. 6C). Quantification analyses showed that cells with large PCM1 foci increased from ∼6% in USP9X knockdown cells to ∼35% in USP9X and MIB1 double knockdown cells (Fig. 6D). In experiments in which CEP290 and PCM1 were co-immunostained, MIB1 depletion caused re-localization of CEP290 in centriolar satellites (Fig. 6E). Cells with CEP290 and PCM1 colocalization increased from ∼3% in USP9X knockdown cells to ∼36% in USP9X and MIB1 double knockdown cells (Fig. 6F). Collectively, these results indicate that increasing PCM1 protein levels in USP9X knockdown cells corrects defects caused by USP9X depletion. Thus, the role of USP9X in maintaining centriolar satellite integrity is likely due to its ability to stabilize PCM1.

In this study, we found that USP9X is required to maintain centriolar satellite integrity. USP9X interacts with and deubiquitylates the centriolar satellite scaffold protein PCM1, and this protects PCM1 from proteasomal degradation. Knockdown of USP9X decreases PCM1 protein levels, disrupts centriolar satellite particles and localizes CEP290 in centrosomes. Strikingly, depletion of MIB1, a ubiquitin ligase of PCM1, in USP9X knockdown cells increases PCM1 protein levels and corrects defects caused by the loss of USP9X. Thus, USP9X stabilizes PCM1 that is required for maintaining centriolar satellite integrity (Fig. 6G). Recent studies have also found that USP9X regulates centrosome duplication and primary cilium formation (Das et al., 2017; Li et al., 2017; Wang et al., 2017). Altogether, USP9X is an important regulator of the centriolar satellite, centrosome and primary cilium network.

USP9X is a highly conserved and multifunctional deubiquitylase (Murtaza et al., 2015). USP9X primarily localizes in the cytoplasm, but it was also found in cytoplasmic vesicles, Golgi complex, mitochondria and the nucleus (Murray et al., 2004; Murtaza et al., 2015; Schwickart et al., 2010). Recent studies revealed that USP9X localizes in centrosomes by interacting with centrosomal proteins, including IQCB1 (Das et al., 2017), CEP131 (Li et al., 2017) and PCM1 (Li et al., 2017; Wang et al., 2017), through which USP9X regulates centrosome duplication and primary cilia formation. In our study, we observed that USP9X colocalizes with the centrosomal marker γ-tubulin in 293T cells, but sparsely overlaps with γ-tubulin staining in HeLa and HCT116 cells. In contrast, in all examined cell types, USP9X colocalizes with the centriolar satellite marker PCM1, indicating that USP9X is a centriolar satellite protein. A recent study has found that an overexpressed N-terminal region of USP9X can pull down endogenous PCM1 (Wang et al., 2017). In our assay, the N-terminal USP9X(1–966) was expressed well, but could not immunoprecipitate any examined PCM1 fragments. Instead, we identified that USP9X(1971–2554), a C-terminal fragment, interacts with the middle region of PCM1, PCM1(850–1300). Formation of a protein complex between USP9X and PCM1 (via a direct or an indirect interaction), and the requirement of USP9X in maintaining centriolar satellite integrity, suggest that USP9X is a bona fide constituent of centriolar satellites.

A few centriolar satellite and centrosomal proteins are substrates for ubiquitylation, including PCM1, CEP290, CEP131 and KIAA0586 (Li et al., 2017; Wang et al., 2016). Knockdown of USP9X could reduce levels of these proteins in cells (Wang et al., 2017). For PCM1, we provided further evidence that USP9X deubiquitylates PCM1, and that depletion of USP9X shortens the protein half-life of PCM1. Thus, USP9X stabilizes PCM1 at the post-translational level by deubiquitylation of PCM1 to prevent proteasomal degradation. Functionally, defects caused by the loss of USP9X are similar to those that have been observed through the loss of PCM1 (Wang et al., 2016), including disruption of centriolar satellite particles and localization of satellite proteins in centrosomes. Thus, USP9X plays a role in maintaining centriolar satellite integrity, as important as that of the core satellite scaffold protein PCM1.

The ubiquitin ligase MIB1 binds PCM1 and catalyzes PCM1 polyubiquitylation to promote proteasomal degradation (Wang et al., 2016). Interestingly, MIB1 is also a USP9X-interacting protein (Choe et al., 2007; Mertz et al., 2015). We found that knockdown of MIB1 in USP9X-depleted HeLa cells largely rescues PCM1 protein levels, restores centriolar satellite particles, and restores localization of CEP290 in satellites. The incomplete rescue effect could be due to the low MIB1 knockdown efficiency (∼50%), and/or the involvement of other unknown ubiquitin ligases that mediate PCM1 ubiquitylation and degradation. In addition to increasing PCM1 protein levels, knockdown of MIB1 could increase the levels of PLK4 protein, which in turn phosphorylates PCM1 to promote its oligomerization and satellite formation (Hori et al., 2016). Centriolar satellites were proposed as reservoirs of MIB1, which prevent inappropriate localization of MIB1 into centrioles. Otherwise, MIB1 mediates degradation of KIAA0586 and CEP131, which inhibits ciliogenesis (Wang et al., 2016). Our current study found that USP9X counteracts MIB1-mediated ubiquitylation of PCM1 to protect it from proteasomal degradation, which maintains centriolar satellites and thus, constrains MIB1, CEP290 and other proteins in centriolar satellites.

The centriolar satellite, centrosome and primary cilium network is important for human health. Mutations in the growing number of genes encoding proteins of this network are implicated in ciliopathies, including Bardet–Biedl syndrome (BBS4), Joubert syndrome (OFD1), Meckel Gruber syndrome (MKS1, TMEM216 and CEP290), primary microcephaly (MCPH; WDR62, STIL and CEP135) and oral-facial-digital syndrome (OFD1) (Faheem et al., 2015; Ko, 2012; Lopes et al., 2011; Pan, 2008; Quinlan et al., 2008; Tollenaere et al., 2015). Primary cilia are signaling hubs that coordinate Hedgehog and GPCR pathways to mediate vertebrate development (Pala et al., 2017; Pedersen et al., 2016; Wheway et al., 2018). Two recent studies identified mutations in USP9X that cause female-specific ciliopathy-like syndromes (Homan et al., 2014; Reijnders et al., 2016). Our study adds USP9X to the expanding list of centriolar satellite proteins for which mutations in the encoding genes cause ciliopathies or similar syndromes. It is unknown how USP9X mutations cause ciliopathy-like syndromes; we speculate that the loss-of-function mutations of USP9X impair its role in the centriolar satellite, centrosome and primary cilium network, which in turn causes developmental defects and the development of ciliopathy-like syndromes.

The following antibodies were purchased: anti-CEP290 (A301-659A) and anti-USP9X (A301-351A) (Bethyl Laboratories); anti-CEP290 (SC-390462), anti-USP9X/Y (SC-365353), anti-PCM1 (SC-67204, SC-398365), anti-MIB1 (SC-393811) and anti-ubiquitin (SC-8017) (Santa Cruz Biotechnology); anti-γ-tubulin (T35559) and anti-FLAG (F1804) (Sigma-Aldrich); and anti-CEP131 (25735-I-AP), anti-KIAA0586 (24421-I-AP) and anti-α-tubulin (66031-I-Ig) (Proteintech).

Expression plasmids of USP9X, USP9X deletion mutants, and PCM1 and PCM1 deletion mutants were constructed by PCR, followed by subcloning into the pRK7 vector containing an N-terminal FLAG or Myc tag. All plasmids were validated by DNA sequencing.

Human embryonic kidney (HEK) 293T, HeLa and HCT116 cells were originally purchased from the American Type Culture Collection, and grown in Dulbecco's modified Eagle medium supplemented with 10% fetal bovine serum (VWR, 89510-186) and 100 μg/ml of penicillin and streptomycin at 37°C with 5% CO₂. Cells were also periodically authenticated and tested for contamination by the Protein Production Core at the University of Colorado Anschutz Medical Campus. The 293T cells were grown to 50–60% confluence and transfected using the standard calcium phosphate precipitation method (Jordan et al., 1996). Typically, 10 μg of each plasmid was used for transfection of cells in a 100-mm dish. Usually, cells were harvested after 48 h transfection. For experiments where indicated, 10 µM MG132 (UBPBio, F1101) was supplemented in medium for 8–16 h before harvesting.

A lentivirus-based method was used to generate stable cell lines as described previously (Han et al., 2012). Two different shRNA constructs in pLKO.1 vector targeting human USP9X (TRCN0000007361 and TRCN0000007364), human MIB1 (TRCN0000352693 and TRCN0000004554) or human CEP290 (TRCN000142313 and TRCN000145444), and a control EGFP shRNA (SHC005), were purchased from the Functional Genomics Core Facility at the University of Colorado. To stably knock down MIB1 in USP9X knockdown cells, we swapped the puromycin-resistant gene in the USP9X shRNA construct (TRCN0000007361) with a neomycin-resistant gene. HeLa cells were infected with USP9X shRNA-containing virus and selected with 2 mg/ml neomycin for 2 weeks. The cells were further infected with MIB1 shRNA-containing virus, and then selected with 2 µg/ml puromycin for at least 5 days, or until all cells in a negative control infection were killed.

For non-denaturing immunoprecipitation, cells in a 100-mm dish (90% confluence) were harvested and washed with 1× PBS, then lysed with 1.0 ml cold cell lysis buffer (20 mM Tris-HCl, pH 7.6, 150 mM NaCl, 2 mM EDTA, 0.5% Triton X-100, 10% glycerol) with protease inhibitor cocktails (Roche). After clearing lysates by centrifugation, supernatants were incubated with 1 μg of an appropriate antibody or control IgG for 4 h at 4°C, then supplemented with 20 μl protein A beads that were preincubated with 2 mg/ml bovine serum albumin (BSA) to reduce non-specific binding. After overnight rocking, protein A beads were pelleted by centrifugation and washed three times with the cell lysis buffer plus 0.5 M NaCl, unless otherwise specified. Bound proteins were eluted in 50 μl 1× SDS sample buffer. For denaturing immunoprecipitation, cells in a 100-mm dish were lysed in 1 ml cell lysis buffer plus 1% SDS. Cell lysates were collected and then heated at 95°C for 15 min. After centrifugation, 0.3 ml supernatants were diluted with 1.2 ml cell lysis buffer to reduce SDS concentration to 0.2%. The immunoprecipitation assay was performed as described above, except that 5 μg anti-FLAG M2 antibody or mouse IgG was used in each reaction. Eluates (20 μl) were resolved in SDS-PAGE and transferred to nitrocellulose membranes for immunoblotting assays. Immunoblotting images were captured using a ChemiDoc MP Imaging system (Bio-Rad).

Immunofluorescence staining and imaging were performed as described previously (Han et al., 2016). Briefly, cells cultured on glass coverslips were fixed with 100% methanol at −20°C for 10 min and then blocked with blocking buffer (3% BSA in TTBS) for 1 h. Cells were then incubated with an appropriate primary antibody (1:200–1:1000 dilution) in blocking buffer overnight at 4°C, followed by incubation with an anti-mouse or -rabbit IgG (H+L) (Alexa Fluor 488 or 555 conjugated) (Cell Signaling Technology, 4412 and 4409, respectively) for 1 h at room temperature. The coverslip was mounted with the ProLong^® Gold Antifade Kit (Thermo Fisher Scientific, P7481) containing the blue fluorescent nuclear counterstain DAPI. Images were captured using a Leica DM 500B fluorescence microscope with a 100× oil immersion objective lens.

To quantify PCM1 foci and their colocalization with USP9X or CEP290, at least 100 cells in each of the three different sections of a slide were counted. Data represent the mean±s.d. Statistical analysis was performed using two-tailed unpaired Student's t-tests, from which s.d. and P-values were calculated. A P-value <0.05 was considered statistically significant.

We thank Dr Bruce Appel for insightful discussion.