WDR8 is a centriolar satellite and centriole-associated protein that promotes ciliary vesicle docking during ciliogenesis

The centrosome is the main microtubule-organising centre (MTOC) of most animal cells. Centrosomes generally consist of two orthogonally arranged centrioles surrounded by layers of pericentriolar material (PCM). Centrioles bear a defined structure of nine triplet microtubules of defined length. In addition, one of the two centrioles of the pair acquires appendage proteins that assemble at the subdistal and distal ends of the centriole. The centriole associated with appendage proteins is called the mother centriole. The daughter centriole lacks appendage proteins. When cells exit mitosis and withdraw from the cell cycle, for example, upon differentiation, the mother centriole turns into the basal body that functions as a platform for the formation of a single non-motile cilium (primary cilium). Primary cilia serve as mechano- and chemo-sensing organelles. They consist of a bundle of axonemal microtubules surrounded by the ciliary membrane. The latter accumulates a number of membrane receptors, which detect environmental cues with high sensitivity (Singla and Reiter, 2006; Ishikawa and Marshall, 2011; Kim and Dynlacht, 2013). A rising number of mutations in genes coding for ciliary proteins cause a variety of genetic disorders that collectively are called ‘ciliopathies’. Ciliopathies are often a consequence of impaired signalling functions of primary cilia. Hence, understanding the molecular basis of primary cilia formation is of the highest medical relevance (Badano et al., 2006; Fliegauf et al., 2007; Baker and Beales, 2009).

Assembly of the primary cilium starts at the mother centriole with the docking of ciliary vesicles, most likely derived from the Golgi (Das and Guo, 2011; Qin, 2012). Distal appendage proteins, including Cep164 and Cep123 (also known as Cep89), have been reported to be indispensable for this initial step of ciliary membrane biogenesis (Schmidt et al., 2012; Tanos et al., 2013; Joo et al., 2013; Sillibourne et al., 2013; Burke et al., 2014; Ye et al., 2014). Moreover, a cascade of small GTPases regulates vesicle docking and fusion at the mother centriole. The GTPase Rab11 targets Rabin8 (also known as RAB3IP), a guanine nucleotide exchange factor (GEF) for the GTPase Rab8, to centrioles. Rabin8 activates Rab8 at the distal end of centrioles leading to vesicle extension (Knodler et al., 2010; Westlake et al., 2011). Recently, the membrane-shaping proteins EHD1 and EHD3, together with the SNARE SNAP29, have been shown to drive vesicle fusion and tubulation of ciliary vesicles (Lu et al., 2015). Following the assembly of the ciliary membrane, the CP110–Cep97 complex (which inhibits axonemal extension) is removed from the distal end of the centrioles (Spektor et al., 2007; Tsang et al., 2008). The microtubule triplets of the basal body then serve as seeds for the elongation of the axonemal microtubules.

Assembly of primary cilia, as well as their maintenance, requires constant exchange between cytoplasmic and centriolar components. Accumulating evidence indicates that macromolecular assemblies of centriolar proteins (known as centriolar satellites) actively support the communication of centrioles with the cytoplasm. Satellites assist centrosome rearrangements and diverse centriolar functions throughout the cell cycle and are involved in centriole duplication, mitotic spindle and primary cilia formation (Bärenz et al., 2011; Tollenaere et al., 2015). A number of proteins with known functions in ciliogenesis localise, at least in part, to centriolar satellites (Tollenaere et al., 2015). SSX2IP, PCM1 and Cep290 are satellite proteins that function in ciliogenesis (Jin et al., 2010; Klinger et al., 2014; Craige et al., 2010; Garcia-Gonzalo et al., 2011; Hori et al., 2014). How satellites are assembled and how they contribute to ciliogenesis is only rudimentarily understood.

Here, we have identified human WDR8 (also known as WRAP73) as an interacting partner of the satellite components SSX2IP and PCM1 (Shen and Osmani, 2013; Yukawa et al., 2015). We show that, in contrast to SSX2IP and PCM1, which only associate with cytoplasmic satellite complexes, a large fraction of WDR8 localises to centrioles and basal bodies even when satellites are dispersed. We identified the centriole proximal end protein Cep135 as a WDR8 interactor that recruits WDR8 to centrioles. Strikingly, knockdown of WDR8 blocked the formation of ciliary vesicles at mother centrioles without affecting the localisation of components relevant to ciliation at the distal end of the centriole. WDR8, therefore, represents a new satellite and centriolar protein with ciliary vesicle docking function that is, however, not associated with the site of vesicle docking.

To identify new satellite proteins with functions in ciliogenesis, we searched for interaction partners of SSX2IP by an affinity purification and mass spectrometry approach. We used a murine NIH/3T3 cell line stably expressing human SSX2IP fused to the localisation and purification (LAP) tag (LAP–SSX2IP), under the control of the inducible tetracycline operator (Klinger et al., 2014). LAP–SSX2IP, induced by addition of doxycycline (dox), is known to associate with centriolar satellites in the vicinity of centrosomes in both cycling and ciliated serum-starved NIH/3T3 cells (Klinger et al., 2014). We used LAP–SSX2IP protein complexes under native conditions followed by quantitative mass spectrometry analysis (Boersema et al., 2009; Klinger et al., 2014). We identified endogenous SSX2IP, the known SSX2IP-interacting protein PCM1 (Bärenz et al., 2013; Klinger et al., 2014) and a previously uncharacterised protein named WD repeat containing protein 8 (WDR8), also known as WRAP73 (Fig. 1A; Tables S1 and S2). The interaction between endogenous murine WDR8 and human SSX2IP was confirmed by co-immunoprecipitation assays followed by immunoblotting (Fig. 1B). Likewise, GFP pulldowns from NIH/3T3 cells expressing LAP-tagged human WDR8 (LAP–hWDR8) revealed co-purifying endogenous murine SSX2IP (Fig. 1C). Interestingly, we also found that the Xenopus laevis ortholog of human WDR8 (xlWDR8) co-immunoprecipitated with xlSSX2IP (O.G., unpublished data). Thus, WDR8 and SSX2IP are components of conserved protein complexes.

To determine where WDR8, SSX2IP and PCM1 interact, we analysed the subcellular localisation of LAP–hWDR8 in NIH/3T3 cells (Fig. 1D, upper images). LAP–hWDR8 partially colocalised with PCM1 and SSX2IP, yet the satellite colocalisation of WDR8 with SSX2IP was not as prominent as between SSX2IP and PCM1. This suggests that a satellite and a non-satellite fraction of hWDR8 exist. Given that it is known that microtubules organise centriolar satellites (Kubo et al., 1999), we asked whether microtubule depolymerisation with the drug nocodazole disturbs WDR8 localisation. Indeed, colocalisation of WDR8 with the centriolar satellite markers PCM1 and SSX2IP was lost after treatment of NIH/3T3 cells with nocodazole (Fig. 1D, lower images). In addition, we observed two prominent, nocodazole-resistant LAP–hWDR8 signals that overlapped with γ-tubulin (Fig. 1D; Fig. S1). This implies that a pool of WDR8 associates with centrosomes independently of centriolar satellites.

To investigate the localisation of endogenous protein, we raised antibodies against WDR8 in guinea pigs. The antibodies specifically detected exogenous WDR8 in total cell lysates (Fig. S1B,C). This reactivity was largely inhibited by pre-adsorption with the blocking antigen (Fig. S1B,C), indicating specificity to WDR8. In immunofluorescence experiments, the same antibodies recognised endogenous WDR8 at centrosomes and satellites in human retina epithelial (RPE-1) cells (Fig. 1E). The staining was specific to WDR8 given that it markedly decreased upon knockdown of WDR8 using two synthetic small interfering RNAs (siRNAs; Fig. 1F,G, siWDR8_1; siWDR8_2). The localisation of WDR8 was further assessed relative to PCM1 and SSX2IP as well as poly-glutamylated tubulin, which decorates centrioles (Fig. 1E). Similar to NIH/3T3 cells, WDR8 in RPE-1 cells was readily detected at poly-glutamylated-marked centrioles (Fig. 1E). In addition, WDR8 colocalised with a fraction of SSX2IP- and PCM1-stained centriolar satellites (Fig. 1E). Taken together, our data indicate that WDR8 is a centriolar protein that is also present in centriolar satellites.

Colocalisation of WDR8 with SSX2IP and PCM1 at satellites suggests a functional interplay between these proteins at this location. We therefore investigated how depletion of WDR8 affected the localisation of a subset of satellite proteins (Fig. 2). The accumulation of PCM1 (Dammermann and Merdes, 2002), SSX2IP (Bärenz et al., 2013; Hori et al., 2014), Cep290 (Kim et al., 2008) and Cep90 (Kim and Rhee, 2011; Kim et al., 2012) in pericentriolar satellites was apparent in control-depleted cells, but it was considerably diminished upon WDR8 depletion (Fig. 2A–D). Quantification revealed a significant reduction of PCM1, SSX2IP, Cep290 and Cep90 in the vicinity of centrioles in response to WDR8 depletion (Fig. 2E). The steady-state levels of Cep290, PCM1 and SSX2IP were, however, not affected in WDR8-depleted cells (Fig. 2F). Thus, WDR8 influences Cep290, PCM1 or SSX2IP localisation rather than the stability of these proteins. Conversely, downregulation of SSX2IP or PCM1 led to a reduction in the amount of satellite-localised WDR8, implying interdependency in satellite localisation between PCM1, SSX2IP and WDR8 (Fig. S2).

Ciliary proteins as well as satellite proteins including SSX2IP, PCM1 and Cep290 are important for proper assembly and function of primary cilia (Craige et al., 2010; Jin et al., 2010; Garcia-Gonzalo et al., 2011; Hori et al., 2014; Klinger et al., 2014). We therefore asked whether WDR8 has a similar role in primary cilia formation. We first analysed the localisation of WDR8 in RPE-1 cells that were grown in low-serum conditions to induce ciliogenesis (Fig. 3A). Cilia were stained with antibodies that recognised poly-glutamylated tubulin, which, together with acetylation, is a tubulin modification enriched at the cilium (Bre et al., 1994) (Fig. 3A). In serum-starved cells, WDR8 colocalised with SSX2IP and PCM1 at centriolar satellites that accumulated around the basal body (i.e. the mother centriole that gave rise to the cilium) (Fig. 3A, ‘merged high’). Additionally, WDR8, but not SSX2IP or PCM1, associated with the basal body and the adjacent centriole (Fig. 3A, ‘merged low’; see colocalisation with poly-glutamylated tubulin). This indicates that during ciliogenesis WDR8 is a satellite- and basal-body-associated protein.

To investigate the function of WDR8 in cilia formation, we depleted WDR8 in RPE-1 cells prior to serum withdrawal. Both, in control- and WDR8-siRNA-treated cells, the percentage of cells arrested in the G1/G0 phase of the cell cycle upon serum starvation was comparable, as demonstrated by FACS analysis (Fig. S3A). However, whereas the majority of control-depleted cells formed a cilium at the mother centriole (stained for Cep164, Fig. 3B), the vast majority of WDR8-depleted cells failed to form a cilium (Fig. 3B,D). The diminished cilia signal upon WDR8 depletion was not an indirect reflection of reduced tubulin glutamylation or acetylation, because the lack of cilia was confirmed with antibodies against the axoneme-associated markers IFT88 and BBS4 (Fig. 3B–D). However, although cilia were absent in WDR8-depleted cells, IFT88 and BBS4 were anchored at the Cep164-stained mother centriole (Fig. 3B,C, arrowheads). This indicates that centrosomal recruitment of IFT88 and BBS4 in response to serum starvation does not require WDR8 function, whereas cilia formation does. Importantly, transient expression of a FLAG-tagged siRNA-resistant version of WDR8 restored primary cilia formation in cells after depletion of endogenous WDR8 (Fig. 3E,F). Collectively, our data show that WDR8 is essential for ciliogenesis.

Assembly of primary cilia requires removal of the CP110–Cep97 complex from the distal end of the mother centriole to allow growth of axonemal microtubules (Spektor et al., 2007; Tsang et al., 2008). To investigate whether WDR8 is involved in this process, we determined the centriolar association of CP110 and Cep97 in serum-starved control- and WDR8-depleted cells (Fig. 4). As previously reported (Spektor et al., 2007), CP110 and Cep97 localise at the daughter centriole but not at the basal body of ciliated cells (Fig. 4A,B, control). The majority of non-ciliated WDR8-depleted cells maintained CP110 and Cep97 at the basal body and the daughter centriole (Fig. 4A–C). This indicates that WDR8 function is required for CP110–Cep97 removal from the mother centriole during the early steps of ciliogenesis.

We next asked whether the persistence of CP110–Cep97 binding to the basal body is the only cause for the cilia assembly defect in WDR8-depleted RPE-1 cells. To test this hypothesis, we ectopically downregulated CP110 in cells lacking WDR8 (Fig. 4D). The combined treatment of cells with WDR8 and CP110 siRNAs led to a substantial reduction in CP110 at centrioles without affecting centriole binding of IFT88 (Fig. 4D). However, the percentage of cells proficient in primary cilia formation was indistinguishable between WDR8-knockdown and WDR8 and CP110 double-knockdown cells (Fig. 4E). This indicates that WDR8 has at least one more function in ciliogenesis than the displacement of CP110 from the distal end of the mother centriole (basal body).

We next asked whether WDR8 could be involved in the establishment of the ciliary vesicle at the mother centriole. This is one of the first steps in ciliogenesis that occurs prior to CP110–Cep97 removal (Das and Guo, 2011; Qin, 2012). To investigate ciliary vesicle formation at the mother centriole, we followed the localisation of the GTPases Arl13b and Rab8a, which are required for ciliary membrane extension at the mother centriole. Upon mock depletion, Arl13b docked to the basal body of cells that started to assemble a cilium (Fig. 5A, top panels) and then decorated the ciliary membrane during axoneme extension (Fig. 5A, middle panel). In contrast, WDR8 depletion almost completely abolished the recruitment of Arl13b to the basal body after serum starvation (Fig. 5A,B). Similar results were obtained with RPE-1 cells stably expressing GFP–Rab8a (Fig. 5C). WDR8 knockdown abolished basal body localisation of Rab8a in 93% of the cells (Fig. 5C,D). We suggest that WDR8 function promotes the docking of Rab8a and Arl13b vesicles to the basal body during ciliogenesis.

It is possible that WDR8 might regulate activation of the GTPase Rab8a. If this hypothesis was correct, one would expect that Rab8a centrosomal localisation and ciliogenesis in WDR8-depleted cells would be rescued by overexpression of a constitutively active GTP-bound form of Rab8a. This hypothesis was tested using RPE-1 cells that stably expressed the hydrolysis-deficient active Rab8a-Q67L mutant protein (Fig. 5E–G). Previously, we have shown that Rab8a-Q67L rescues the ciliogenesis defects of Cep164-depleted cells (Schmidt et al., 2012). However, in WDR8-depleted cells GFP–Rab8a-Q67L failed to localise to basal bodies and did not promote cilia formation (Fig. 5E,F). Therefore, our data indicate that Rab8a activation is not the sole cause of cilia deficiency in cells lacking WDR8.

Next, we used transmission electron microscopy (TEM) to analyse the ultrastructure of the basal body and/or cilia in control and WDR8-knockdown cells. In all control RPE-1 cells (n=12), the distal end of the basal bodies was in contact with the ciliary membrane (Fig. 6A,C). In marked contrast, vesicle docking was absent in nine out of ten basal bodies of WDR8-knockdown cells (Fig. 6B,C). This indicates that vesicle docking to the mother centriole does not take place in the absence of WDR8.

Vesicle docking at the mother centriole requires the presence of distal appendage components, including the proteins Cep164 and Cep123 (Schmidt et al., 2012; Sillibourne et al., 2013; Tanos et al., 2013). As assessed by TEM, distal appendages were present in WDR8-depleted cells (Fig. 6B), indicating that WDR8 is not needed for distal appendage formation. This notion was confirmed by quantitative immunofluorescence microscopy, which indicated that the levels of Cep123 or Cep164 were not reduced in WDR8-depleted cells (Fig. S3B,C). We conclude that WDR8 promotes ciliary vesicle docking at the mother centriole by a mechanism unrelated to formation of distal appendages.

To analyse how WDR8 associates with centrioles, we used murine cell lines stably expressing LAP–hWDR8 to search for WDR8-associated components. In addition to SSX2IP, we found the highly conserved centrosomal proximal end protein Cep135 (Ohta et al., 2002) to be a WDR8 interaction partner in quiescent and cycling cells (Fig. 7A; Tables S3 and S4). Interaction between WDR8 and Cep135 was confirmed by immunoprecipitation experiments using anti-Cep135 antibodies (Fig. 7B). Strikingly, when we compared the centriolar localisation of WDR8 with Cep135 in stimulated emission depletion (STED) and 3D-structured illumination (3D-SIM) microscopy experiments, we observed that LAP–hWDR8 largely colocalised with Cep135 (Fig. 7C, DOX+). The LAP–hWDR8 and Cep135 signals were clearly distinct from the ring-like structure formed by the distal end protein Cep164 (Fig. 7C). Our data thus indicate that Cep135 and WDR8 form a complex at the proximal end of the mother centriole.

To explore the functional relationship between Cep135 and WDR8, we investigated whether Cep135 was required for centrosomal association and/or the function of WDR8 in ciliogenesis. To circumvent the previously described centriole duplication defect that arises from prolonged treatment of cells with Cep135 siRNA (Inanc et al., 2013), we analysed cells in which Cep135 was depleted with siRNA (siCep135_1) for only a short period of time (48 h). Under these conditions, the Cep135 knockdown efficiently reduced Cep135 centrosomal levels without compromising centriole duplication, as inferred by the two side-by-side centrin signals that labelled the two centrioles in G1 cells (Fig. 7D). In cells lacking Cep135, FLAG–WDR8 was absent from centrosomes in the large majority (>99%) of the cells (Fig. 7D). The levels of endogenous WDR8 at centrosomes were significantly decreased upon Cep135 depletion (Fig. 7E,F, EGFP control; Fig. S2D). The expression of a siRNA-resistant rodent (Chinese hamster Cricetulus griseus) Cep135 cDNA (CgCep135) (Ohta et al., 2002) in Cep135-depleted cells partly rescued WDR8 localisation at centrioles and, to the same extent, to pericentriolar satellites (Fig. 7E,F). Our data thus indicate that WDR8 is recruited to centrioles in a Cep135-dependent manner.

Considering that the composition of centriolar satellites was strongly dependent on WDR8, we tested whether a similar phenotype would be observed in cells lacking the WDR8 interaction partner Cep135. Short-term knockdown of Cep135 significantly reduced the centrosomal accumulation of Cep290, PCM1 and SSX2IP without compromising centrosome duplication, as judged by the presence of two pericentrin (PCNT)-stained foci (Fig. 8A,B; Fig. S2E). This indicates that Cep135, like WDR8, is required for the targeting of satellite proteins to the centrosome area.

We then investigated whether Cep135 plays a role in ciliogenesis. Under conditions of short-term Cep135 knockdown and serum starvation, primary cilia formation was strongly decreased, as indicated by the lack of staining for the axonemal microtubule markers poly-glutamylated tubulin or IFT88 (Fig. 8C,D). The diminished ciliogenesis in Cep135-depleted cells was rescued by expression of siRNA-resistant CgCep135 (Fig. 8E,F; Fig. S2D). Despite the lack of cilia, IFT88, as well as the appendage proteins Cep164 and Cep123, were still recruited to centrioles in the absence of Cep135 (Fig. 8C; Fig. S4), resembling WDR8-depleted cells (Fig. 3C). Cep135 depletion affected primary cilia formation almost as strong as WDR8 knockdown (compare Fig. 8D and Fig. 3F). To determine whether Cep135 influenced ciliogenesis in a way similar to WDR8, we analysed the ciliary membrane markers Rab8a and Arl13b. Active GFP–Rab8a (Rab8a-Q67L) and Arl13b accumulated at centrosomes in serum-starved control cells but failed to do so in the majority of Cep135-depleted cells (Fig. 8G,H). Taken together, our data indicate that, similar to WDR8, Cep135 is required for the initial steps of ciliogenesis.

The establishment of the ciliary membrane at the mother centriole is a remarkable process that requires the concerted action of mother centriolar appendage proteins and Golgi-derived vesicular components. Here, we established WDR8 as a centriolar and satellite-associated protein that is indispensable for ciliary vesicle docking at the mother centriole during the early steps of ciliogenesis.

We initially identified WDR8 as a new interaction partner of the previously characterised centriolar satellite protein SSX2IP. WDR8 has also been found to be in complex with SSX2IP in high-throughput mass spectrometry analysis in human cells (Hutchins et al., 2010). Several lines of evidence indicate that WDR8 is a bona fide centriolar satellite component in human and murine cells. Firstly, WDR8 interacts with the satellite components SSX2IP and PCM1. Secondly, endogenous or ectopically expressed fluorescently tagged WDR8 colocalises with SSX2IP and PCM1 in both human and murine cells in the vicinity of centrosomes. Thirdly, WDR8 localisation in satellites, as seen with other satellite protein markers (Kubo et al., 1999), requires an intact microtubule cytoskeleton for centrosome targeting. Finally, targeting of SSX2IP, PCM1 and Cep290 to centrosomes requires WDR8, and vice-versa, indicating interdependency in satellite organisation. Similarly, SSX2IP depletion also affects satellite localisation of PCM1 and Cep290 (Klinger et al., 2014). Thus, the satellite protein network is highly interdependent.

How WDR8 maintains satellite integrity is not clear. Each WD domain of WDR8 consists of 40 amino acids that are regularly involved in mediating protein–protein interactions (Zhang and Zhang, 2015). Six WD repeats are predicted in the human WDR8 primary structure, and, interestingly, we observed that only full-length and not truncated versions of WDR8 were able to co-immunoprecipitate with SSX2IP (our unpublished data). It is therefore possible that the WD repeats of WDR8 have a scaffold and crosslinking function in promoting satellite complex formation and stabilisation.

Importantly, our data indicate that, in contrast to SSX2IP and PCM1, a pool of WDR8 associates with centrioles even when satellites are dispersed. This data is consistent with previous systematic proteomic studies in which WDR8 has been found in human centrosome fractions (Andersen et al., 2003; Jakobsen et al., 2011). All these findings imply that a pool of WDR8 must tightly associate with centrosomes. The dual localisation at centrioles and satellites is, however, not unique to WDR8, as it has been reported for other satellite proteins such as OFD1 and Cep290, which also localise to centrioles independently of satellites (Tollenaere et al., 2015).

Using WDR8 as a bait, we identified the centrosomal protein Cep135 as a WDR8-binding partner. Short-term depletion of Cep135 almost completely abolished WDR8 centriolar localisation and led to the dispersal of centriolar satellite components, without causing the severe centriole duplication defect and loss of γ-tubulin localisation that has been previously reported for long-term depletion of Cep135 (Inanc et al., 2013). This indicates that Cep135 targets WDR8 to the proximal end of centrioles. Cep135 might also affect the protein stability of WDR8, although difficulty in detecting endogenous WDR8 by immunoblotting prevented us from analysing this possibility in detail. The remarkably similar phenotypes of WDR8 and Cep135 knockdown in ciliogenesis suggest that both proteins share a common function in this process. As Cep135 functions at the proximal end of centrioles (Sonnen et al., 2012; Lin et al., 2013), we suggest that the interacting WDR8 is also associated with the proximal end of the centriole. Our super-resolution data confirmed this hypothesis, as both proteins largely colocalised distantly from the distal end protein Cep164. Interestingly, we sporadically observed a weak Cep135 and LAP–Wdr8 signal closer to the Cep164 ring-like structure. This suggests that low levels of the protein complex could also be present at the distal end of the centriole. Future detailed analysis of endogenous Wdr8 will, however, be necessary to substantiate this conclusion.

Our work establishes that WDR8 plays an essential role in primary cilia biogenesis. The penetrance of cilia loss exceeds the knockdown efficiency indicating that a critical threshold of WDR8 is needed to maintain competence for ciliogenesis. The loss of primary cilia formation upon WDR8 knockdown is considerably more severe than the defect observed upon depletion of SSX2IP (Klinger et al., 2014). SSX2IP-depleted cells are still able to assemble a primary cilium; however, that cilium shows defects in ciliary membrane protein targeting and length (Hori et al., 2014; Klinger et al., 2014). This might suggest that the centriolar WDR8–Cep135 complex plays a more important role in the initiation of cilia biogenesis than the pericentriolar WDR8–SSX2IP complex.

Using fluorescence and electron microscopy, we demonstrate that WDR8 is required for ciliary membrane docking at the mother centriole and displacement of the ciliary inhibitory complex composed of Cep97 and CP110 (which inhibits axoneme extension) (Spektor et al., 2007). Ectopic removal of CP110 did not rescue the cilia biogenesis defect observed upon WDR8 depletion. We thus postulate that the axoneme extension phenotype in WDR8 knockdown cells is most likely a consequence of impaired vesicle docking rather than persistence of the inhibitor Cep97–CP110 protein at the mother centriole per se.

Failure in vesicle docking to distal ends of centrioles has been previously reported for components of distal appendages, such as Cep164 and Cep123, that serve as adaptors for the binding of vesicles to the distal ends (Schmidt et al., 2012; Sillibourne et al., 2013; Tanos et al., 2013; Burke et al., 2014; Joo et al., 2013; Ye et al., 2014). Our data show that WDR8 or Cep135 knockdown does not decrease the centriolar levels of Cep164 or Cep123. This argues against a model in which WDR8 or Cep135 works in ciliary vesicle docking by promoting distal appendage formation, at least not through Cep123 or Cep164.

We propose a model in which Cep135 targets WDR8 to the mother centriole proximal end and this process is required for the initiation of cilia membrane biogenesis that takes place at the distal region of centrioles. In Drosophila, depletion of the Cep135 ortholog Bld10 was shown to abolish assembly of the distal central pair of axonemal microtubules in motile cilia (Mottier-Pavie and Megraw, 2009; Carvalho-Santos et al., 2012). Although primary cilia lack the central pair of microtubules, this exemplifies that Cep135 is also required for events that take place at the distal end, i.e. at the site of microtubule polymerisation. It is thus possible that Cep135 and Bld10 might share important functions in cilia formation.

It is currently unclear how loss of WDR8 function causes a block of ciliogenesis at the early stage of vesicle docking. Possibly, loss of WDR8 might abolish targeting of factors required for the function of distal ends in vesicle docking. Such a factor could comprise a yet uncharacterised component of the distal appendage or components that cooperate with the Rab8–Rab11 GTPase cascade in ciliary membrane establishment, such as the recently identified membrane shaping proteins EHD1 and EHD3, and the SNARE component SNAP29 (Lu et al., 2015). Alternatively, the mode of action of WDR8 might involve the targeting of an enzymatic activity, for example, in tubulin modification, to centrioles. The modifying enzyme might localise to the proximal end but generate a diffusible signal that would ‘reach’ out to the distal ends. Candidates would be kinases, phosphatases or tubulin-modifying enzymes, which could modulate the properties of centriolar microtubules in general.

Interestingly, orthologues of SSX2IP and WDR8 have been found in conserved complexes not only in human cells, as shown here, but previously in Aspergillus nidulans (Shen and Osmani, 2013), in Schizosaccharomyces pombe (Toya et al., 2007; Yukawa et al., 2015) and X. laevis (O.G., unpublished observation). This indicates that the SSX2IP–WDR8 protein module must be highly conserved throughout eukaryotic evolution. In S. pombe, the SSX2IP–WDR8 complex functions in interphase and mitosis to anchor microtubules at their minus end in cooperation with the γ-tubulin ring complex (Toya et al., 2007; Yukawa et al., 2015). Consistent with this, SSX2IP has been previously shown to function in primary cilia formation and also in mitosis (Bärenz et al., 2013; Klinger et al., 2014), and we are currently investigating a possible function of WDR8 in mitotic cells. Finally, yeast cells do not rearrange MTOCs for primary cilia formation as in metazoan organisms. The WDR8–SSX2IP module must therefore have acquired an additional function in cells that withdraw from the cell cycle and initiate ciliogenesis. It remains an intriguing question for the future whether a single conserved primary function of the SSX2IP–WDR8 module explains all the roles of the two individual proteins that have been described throughout the cell cycle.

Antibodies against human SSX2IP, PCM1 and γ-tubulin were as described previously (Bärenz et al., 2013). Anti-human WDR8 antibodies (1:10) were produced against a recombinant fusion of the N- and C-terminal residues of WDR8 (codons 1–45 and 410–460) purified from E. coli. Anti-IFT88 antibodies were raised in rabbit (1:200) and guinea pig (1:250) (against codons 365–824 of mouse IFT88) (Pazour et al., 2002; Gazea et al., 2016). Mouse anti-poly-glutamylated tubulin GT335 (1:1000) was a gift of Carsten Janke (Institute Curie, Paris, France) (Wolff et al., 1992). Rabbit anti-BBS4 antibody (1:250) was a gift from Andrew Fry (University of Leicester, England, UK) (Lopes et al., 2011). Anti-Cep135 (1:100) and anti-pericentrin (1:800) antibodies were gifts from Elmar Schiebel (ZMBH, University of Heidelberg, Germany). Rabbit anti-Cep123 (1:1000) antibody was a gift from Michel Bornens (Institute Curie, Paris, France) (Sillibourne et al., 2013). Rabbit anti-Cep90 (1:200) was a gift of Kunsoo Rhee (Seoul National University, Korea) (Kim and Rhee, 2011). Anti-Cep164 antibodies were generated in rabbits (1:800) and guinea pigs (1:2000) as described previously (Schmidt et al., 2012). For the immunoprecipitation assays, anti-GFP beads were generated using GFP-binding protein (GBP beads) as previously described (Rothbauer et al., 2008; Kuhns et al., 2013).

Antibodies from commercial sources were monoclonal antibodies against GFP (1:800; G1546, Sigma-Aldrich); mouse anti-acetylated tubulin (1:100; sc-23950, Santa Cruz Biotechnology); mouse anti-γ-tubulin (1:500; T6557, Sigma-Aldrich); rabbit anti-γ-tubulin (1:500; T5192; Sigma-Aldrich); rabbit anti-Cep97 and rabbit anti-CP110 (both 1:300; A301-947A and A301-343, Bethyl Laboratories); rabbit anti-Arl13b (1:500; 17711-1-AP, Proteintech); rabbit anti-FLAG antibody (1:500; F7425, Sigma-Aldrich); and rabbit anti-Cep290 (1:1000; A301-659A, Bethyl Laboratories).

Full-length human WDR8 cDNA (accession number BC086311) was obtained from Source BioScience (Nottingham, UK). Expression constructs comprising N-terminally-tagged YFP, FLAG or LAP were generated in pEYFP-C3 (Takara Bio Inc.), pCMV-3Tag-1A (Agilent Technologies) or pIC113 [LAP construct; a gift from Ian Cheeseman, Whitehead Institute, USA (Cheeseman and Desai, 2005)], respectively. A plasmid carrying Chinese hamster Cep135 cDNA (CgCep135) was a gift of Ryoko Kuriyama (University of Minnesota, USA) (Ohta et al., 2002). Point mutations in WDR8 were generated by site-directed PCR mutagenesis. Expression of TET-on-inducible constructs was induced by the addition of doxycycline (Sigma-Aldrich) at a concentration of 100–1000 ng/ml during serum starvation. For disruption of microtubules, cells were treated with nocodazole (Sigma-Aldrich) at a final concentration of 20 μM for 2 h at 37°C.

hTERT-immortalised retinal pigment epithelial (RPE-1, ATCC, CRL-4000) cells were grown in DMEM/F12 (Life Technologies) supplemented with 10% fetal calf serum (FCS, Life Technologies), 2 mM L-glutamine and 0.348% sodium bicarbonate. G418 (800 µg/ml, Life Technologies) were added to RPE-1 cells stably expressing GFP–Rab8a. NIH/3T3 cells (ATCC, CRL-1658) were grown in DMEM supplemented with 10% newborn calf serum (NCS, PAN-Biotech). All cell lines were grown at 37°C under 5% CO₂. Transient transfection in NIH/3T3 cells was performed using Lipofectamine 2000 (Life Technologies) according to the manufacturer's protocol. RPE-1 cells were transiently transfected with plasmid DNA by electroporation using the Neon^® Transfection System (Life Technologies) according to the manufacturer's protocol. Stable NIH/3T3 cells were generated as previously described (Kuhns et al., 2013). NIH/3T3 and RPE-1 cells were incubated in serum-free medium for 24–48 h to induce ciliogenesis. At least 100 cells were counted for the quantification of ciliated cells for each experimental condition. HEK293T (ATCC CRL-1573) cells were cultured in DMEM supplemented with 10% FCS and transfected with the calcium phosphate precipitation method. All cell lines were authenticated by Multiplex Cell line Authentication (MCA) and were tested for contamination by Multiplex cell Contamination Testing (McCT) v.2 service provided by Multiplexion (Heidelberg, Germany).

Transfections of the siRNAs were performed using Lipofectamine RNAiMAX reagent (Life Technologies). The following siRNAs (20 nM, silencer select, Life Technologies) were used: WDR8_1: 5′-GGAAUUCCAUCUGCGGAUA-3′, WDR8_2: 5′-CCAAGAUAGUGGUGUAUAA-3′; SSX2IP_1: 5′-GGAAGGUUGCUAUAGUGGA-3′, SSX2IP_2: 5′-GCAUGUCUAAACUUACUAA-3′; Cep135_1: 5′-GAUCUAGAGAAAUCGCUUU-3′, CEP135_2: 5′-GCAAAUUGAUGAACCGGUU-3′ CP110: 5′-AAGCAGCATGAGTATGCCAGT-3′ (Spektor et al., 2007). Unless otherwise indicated, after 24 h of knockdown, RPE-1 cells were incubated in serum-free medium for 48 h to induce ciliogenesis.

For preparation of cell lysates, cells were washed in pre-warmed PBS and re-suspended in 0.8 ml of PBS containing 0.15 ml of 1.85 M NaOH. Cells were kept on ice for 5 min before addition of 0.15 ml of 55% trichloroacetic acid (TCA; w/w). After an incubation of 10 min on ice, samples were centrifuged at 20,800 g for 15 min at 4°C. The pellet was resuspended in HU-DTT buffer (8 M urea, 0.2 M Tris-HCl pH 6.8, 5% SDS, Bromophenol Blue and 0.1 M DTT), heated for 15 min at 65°C and loaded onto SDS-PAGE gels. Proteins were transferred onto nitrocellulose membranes using a semi-dry blotter. Immunoblotting was carried out using anti-WDR8, anti-FLAG and anti-GFP antibodies. Affinity purified anti-WDR8 antibodies, pre-adsorbed with purified WDR8 antigen, were used in the experiments shown in Fig. S1B,C.

Co-immunoprecipitaion in NIH/3T3 LAP–hSSX2IP cells was carried out as described previously (Cheeseman and Desai, 2005; Klinger et al., 2014) with immobilised GBP beads. In brief, NIH/3T3 LAP–hSSX2IP cells were induced with 1 µg/ml doxycycline for 24 h. Cells were either incubated in serum-rich medium (cycling cultures) or serum starved for a total of 48 h to induce ciliogenesis. Cells were lysed and incubated for 3 h at 4°C with GBP beads and processed as described previously (Kuhns et al., 2013). Bound proteins were analysed by SDS-PAGE and immunoblotting. For quantitative mass spectrometry analysis, comparison of negative control and LAP–hSSX2IP pulldown samples was performed after dimethyl labelling using stable isotopes (Boersema et al., 2009). Proteins, separated by SDS-PAGE and stained with Coomassie Blue, were digested with trypsin as described previously (Bärenz et al., 2013) except that triethylammoniumbicarbonate buffer (TEAB) was used instead of ammoniumbicarbonate buffer. Following digestion, tryptic peptides were extracted from the gel pieces with 50% acetonitrile and 0.1% trifluoroacetic acid (TFA) and concentrated nearly to dryness in a speedVac vacuum centrifuge. As previously described (Boersema et al., 2009), dimethyl labelling was performed by incubating samples with 20 µl 50 mM phosphate buffer pH 7.5, 4 µl 150 mM NaBH₃CN and 4 µl 1% formaldehyde (‘light’ or ‘medium’) for 1 h at 25°C. The ‘light’ and ‘medium’ labelled samples were loaded separately on a C18 trapping column with a flow rate of 10 µl/min of 0.1% TFA and thereby mixed. Peptides were subsequently separated on an analytical column (75 µm×150 mm) with a flow rate of 300 nl/min coupled to a ESI LTQ Orbitrap XL MS (Thermo Fisher, USA). One survey scan (resolution: 60,000) was followed by five information-dependent product ion scans in the LTQ. Only double- and triple-charged ions were selected for fragmentation. Data were analysed using Proteome Discoverer 1.4 and Mascot (Matrix Science; version 2.4). Mascot was set up to search the SwissProt database, taxonomy Mus (16,700 entries) using trypsin as protease, a fragment ion mass tolerance of 0.50 Da and a parent mass tolerance of 4.0 ppm. Iodoacetamide derivative of cysteine was specified in Mascot as a fixed modification. Deamidation of asparagine and oxidation of methionine were specified in Mascot as variable modifications. The target decoy PSM validator was set to a target FDR (strict) of 0.01.

Cells were grown on coverslips and fixed in ice-cold methanol at −20°C for 5 min. Cells were blocked with PBS containing 10% FCS and 0.2% Triton X-100 (blocking buffer) for 60 min and incubated with primary antibodies at room temperature for 1 h. The cells were incubated with Alexa-Fluor-488-, 568- or 633-conjugated secondary antibodies (Life Technologies) for 1 h at room temperature. All antibodies were diluted in blocking solution. Coverslips were mounted on glass slides in Mowiol (EMD Millipore) containing 4′,6-diamidino-2-phenylindole (DAPI). Images were acquired as z-stacks using the Zeiss LSM780 (Carl Zeiss, Oberkochen, Germany) confocal system and software (ZEN2) or using a Zeiss Axio Observer inverted microscope with a Plan-APOCHROMAT 63×1.4 NA oil objective.

For 3D-SIM, NIH/3T3 cells stably expressing LAP–WDR8 were treated with 1 μg/ml doxycycline for 18 h and with 5 μg/ml of nocodazole for 2 h to disperse the satellite pool of WDR8. Cells were processed for immunofluorescence microscopy as described above, except that secondary antibodies were coupled to Alexa Fluor 488, 555 or 647 (Molecular Probes). Cells were mounted in Prolong Gold antifade reagent (Molecular Probes). Samples were analysed with 3D-SIM Nikon Ti inverted microscope equipped with three lasers (488, 561 and 640 nm), a Nikon Apo TIRF 100×1.49 NA oil immersion objective and an Andor iXon3 DU-897E single-photon detection EMCCD camera. After image capturing, raw images were reconstructed in the NIS-Elements program of the microscope to obtain 3D-SIM images.

STED microscopy was performed on a Leica TCS SP8 STED 3X confocal laser scanning microscope using a HC PL APO 100×1.40 NA oil STED WHITE objective. Images of single confocal sections were taken and processed using the Leica LAS X Core software with STED and deconvolution algorithms.

Images were processed in ImageJ, Photoshop CS3 (Adobe) and Illustrator CS3 (Adobe). No manipulations were performed other than brightness, contrast and colour balance adjustments. Rolling ball background subtractions (ImageJ) were performed in Figs 1E,F, 2A–D, 7E and 8A,E.

Quantification of fluorescence intensity was performed using original confocal images and the ImageJ 1.47T software (National Institutes of Health, Bethesda, MA). Measurement of integrated density was performed using the summed projections of acquired z-stacks in the areas indicated in the figures. γ-tubulin was used as a marker of the centrosome. The centrosome area was defined by a circle of 2-µm diameter and the centriolar satellite region by a circle of 5-µm diameter around γ-tubulin. The data were plotted as box plots using KaleidaGraph 4.0 (Synergy Software). Boxes showing the lower and upper quartiles represent the data value located halfway between the median and the smallest and largest data values, respectively. Outliers are the points whose value is either greater than the upper quartile value plus 1.5 times the interquartile difference or less than lower quartile value minus 1.5 times interquartile difference. Statistical analyses of fluorescence intensity measurements and ciliation assays were performed as specified in figure legends.

For rescue experiments, the plasmid containing FLAG-tagged human WDR8 was changed into an siRNA-resistant version using site-directed mutagenesis. The rescue construct (WDR8 siRNA_2 resistant or EGFP–CgCep135) or control vector (FLAG or EGFP) was transiently transfected into RPE-1 cells 2 h prior to the transfection of the cells with control siRNA (siControl) or siWDR8_2. After 24 h, cells were serum starved for 24 h, fixed and processed for immunofluorescence analysis. Approx. 100 cells were counted for each experimental condition.

Electron microscopy was performed as previously described (Kuhns et al., 2013; Schmidt et al., 2012). In brief, RPE-1 cells growing on coverslips were fixed with 2.5% glutaraldehyde, washed with 50 mM cacodylate buffer and postfixed in 2% OsO₄ at 4°C. Cells were incubated in 0.5% aqueous uranyl acetate overnight and dehydrated at 4°C using ascending ethanol concentrations and propylene oxide. Cells were incubated in a 1:1 mixture of Epon and propylene oxide, infiltrated with Epon and polymerised at 60°C. Serial ultrathin sections (70 nm) were produced and poststained with uranyl acetate and lead citrate. The stained samples were examined under a transmission electron microscope (EM 910; Carl Zeiss, Oberkochen, Germany). The brightness and contrast of the images were adjusted in Photoshop CS3 (Adobe).

We would like to thank Carsten Janke, Michel Bornens, Elmar Schiebel, Ian Cheeseman, Andrew Fry, Ryoko Kuriyama and Kunsoo Rhee for sharing reagents; Laurence Pelletier for exchange of data prior to publication; the mass spectrometry core facilities of the ZMBH for protein identification; the electron microscopy facility of the DKFZ for help with the imaging of electron microscopy samples and the microscopy facilities of the DKFZ, ZMBH and Heidelberg-University Nikon Imaging Center for providing access to microscopes, help with imaging and image processing. We are grateful to Leica Microsystems, Mannheim, Germany for allowing access to the Leica SP8 STED 3X microscope unit. W.W. was a PhD student of the Hartmut Hoffmann-Berling International Graduate School for Molecular and Cellular Biology (HBIGS), University of Heidelberg, Germany.