PLK1 controls centriole distal appendage formation and centrobin removal via independent pathways

Centrosomes are the main microtubule nucleating and organizing centres of metazoan cells. Each centrosome is composed of two barrel-shaped centrioles, composed of nine microtubule triplets surrounded by pericentriolar material (PCM) (Gönczy and Hatzopoulos, 2019; Vásquez-Limeta and Loncarek, 2021). Like DNA, centrosomes replicate in a semi-conservative manner once per cell cycle. In G1, human somatic cells contain two centrioles that have become dis-engaged – one older centriole, called here the ‘grandmother’ centriole, and one younger daughter centriole. As both centrioles give rise orthogonally to one ‘daughter’ centriole during S-phase, the G1 daughter centriole becomes the ‘mother’ centriole. The two new-born daughter centrioles elongate in G2, followed by centrosome separation and maturation at mitotic onset. Therefore, mitotic cells always contain three different centriole generations – one grandmother centriole, one mother centriole and two daughter centrioles.

One key difference between the grandmother centriole and the mother centriole is the presence of distal and subdistal appendages on the grandmother centriole that emerge as spikes from the centriolar barrel (Tischer et al., 2021). The transition from nascent daughter centriole to grandmother centriole via mother centriole encompasses two mitoses, and one and a half cell cycles (Kong et al., 2014). During the second mitosis, the mother centriole gradually recruits the full set of distal and subdistal appendage proteins, driving the formation of both appendages in the next G1 to complete the transformation into a grandmother centriole. Each type of appendage ensures key centrosomal or ciliary functions (Loncarek and Bettencourt-Dias, 2018). Subdistal appendages enable centrioles to anchor and focus microtubules directing intracellular trafficking, cell motility, cell adhesion or cell polarity during interphase (Bornens, 2002; Delgehyr et al., 2005). At its core, sitting on the centriole wall, resides cenexin (also known as ODF2), which is required for the recruitment of all the proteins that are located in the outer parts of subdistal appendages, such as CEP128, centriolin and ninein (Chong et al., 2020; Mazo et al., 2016). Distal appendages are required for the hook-like insertion of the older centriole into the plasma membrane during ciliogenesis (Tanos et al., 2013). The binding of OFD1 is the first building step for the formation of distal appendages (Wang et al., 2018). This protein plays a pleiotropic role regulating distal appendage formation, centriole length and ciliogenesis (Singla et al., 2010). In terms of distal appendage formation, OFD1 binding enables the hierarchical recruitment of all subsequent proteins at mother centrioles, starting with CEP83 and ending with FBF1 and CEP164 (Wang et al., 2018). Super-resolution microscopy further revealed that distal and subdistal appendages proteins mutually influence each other's position relative to the centriole, indicating that they are structurally partially inter-dependent (Chong et al., 2020).

A central player in the current regulatory model of distal appendage formation is centrobin. This protein accumulates in human epithelial cells on daughter centrioles (Zou et al., 2005); it is required for efficient centriole elongation during centrosome duplication and essential for cilia formation (Gudi et al., 2014; Ogungbenro et al., 2018). In the context of distal appendage formation, centrobin has been proposed to act as a placeholder for the recruitment of future distal appendage proteins, and its removal in human epithelial cells under the control of the centriole protein Talpid3 (also known as KIAA0586) is thought to allow the recruitment of OFD1 (Wang et al., 2018). This raises the question as to the exact timing of centrobin removal during the centrosome cycle, and the nature of the regulatory pathways controlling this process. The organism in which centrobin has been best studied is Drosophila melanogaster, where its functions and exact localization can vary from cell type to cell type. Centrobin inhibits basal body function in sensory neurons, but promotes basal body elongation in spermatocytes (Gottardo et al., 2015; Reina et al., 2018); likewise centrobin promotes the recruitment of interphase pericentriolar material in neuroblasts but not in epithelial wing cells (Januschke et al., 2013). The best-studied regulator of centrobin in flies is the polo-like kinase 1 (PLK1). In interphase Drosophila neuroblast, centrobin is exclusively present on the younger centrosome containing the daughter centriole; as this daughter centriole becomes itself a mother centriole, centrobin is removed in a PLK1-dependent manner during mitosis (Gallaud et al., 2020; Januschke et al., 2013, 2011). In human epithelial cells, PLK1 can phosphorylate centrobin in mitosis (Lee et al., 2010). Moreover, PLK1 kinase activity regulates directly or indirectly the levels of several distal appendage components throughout the cell cycle: its activity is required for the reduction of CEP164 on the grandmother centriole as cells reach prophase, but it is also required for the appearance of CEP164 on the mother centriole, which starts as cells progress through mitosis (Kong et al., 2014). Therefore, it is thought that PLK1 promotes the formation of distal appendages including the recruitment of CEP164 via the removal of centrobin from the mother centriole. This hypothesis has, however, never been tested in one coherent model system or cell line, which is important given the cell type-to-cell type variability in centrobin function. Here, we tested the localization pattern of centrobin over the cell cycle, its potential regulation by PLK1, and its role in the recruitment of distal appendage protein recruitment, in non-malignant diploid human retina pigment epithelial RPE1 cells as this cell line has been used to propose the placeholder model (Wang et al., 2018). Furthermore, we tested how subdistal appendage proteins contribute to centrobin dynamics, as the subdistal appendage protein cenexin is a key regulator of PLK1 activity (Colicino et al., 2019). We show that centrobin is removed from the mother centriole as cells reach metaphase, and that this removal depends on the newly emerging daughter centrioles, possibly because of higher affinity binding sites. Centrobin removal and recruitment of distal appendage proteins depend both on PLK1 activity and the presence of cenexin. Surprisingly, these two centrosome cycle steps occur independently of each other. Indeed, in multiple siRNA or CRISPR/Cas9 knockout backgrounds, distal appendage protein recruitment occurs in RPE1 cells without centrobin removal. We propose that in RPE1 cells PLK1 and subdistal appendage proteins regulate the removal of centrobin and the build-up of distal appendages via separate pathways, pointing for a need to revise the regulatory model for centriole maturation.

Centrobin was first described as a daughter centriole-specific protein that localizes to the daughter centriole in G1 before appearing on the two new daughter ‘pro’-centrioles that arise from the grandmother and the new ‘mother’ centriole in S-phase (Zou et al., 2005). How and when centrobin is removed from the mother centriole in human cells is, however, unknown. To address the when, we studied the cell cycle-dependent localization of centrobin in human retina-pigment epithelial cells immortalized with telomerase and expressing the centriole marker eGFP–centrin1 (hTertRPE1-eGFP-centrin1). Cells were fixed for immunofluorescence, stained for centrobin, the S-phase marker PCNA and the DNA marker DAPI, and assigned to: (1) G1, if the cell was PCNA-negative and contained two centrioles; (2) S-phase, if the cell was PCNA-positive (note that only cells with four distinct centrioles were considered); or (3) G2, if the cell was PCNA-negative and contained four centrioles. In addition, an increasing abundance of eGFP–centrin1 on daughter, mother and grandmother centrioles allowed the classification of each centriole in one of those age categories (Tan et al., 2015). Centrobin localization on the different type of centrioles was assessed both in a qualitative manner (present or not present), as well as in a quantitative manner (mean intensity per centriole type) to reveal the full spectrum of localizations. In most G1 cells, centrobin was only visibly present on the daughter centriole (91±4%; mean±s.e.m.), with mean intensity that was at least 3 times higher than on the grandmother centriole (Fig. 1A–C). In S-phase cells that had initiated centriole duplication (with the daughter centriole becoming a mother), and hence contained four eGFP–centrin1 dots, most cells displayed centrobin on both daughter centrioles and the mother centriole (64.5±7%; Fig. 1A,B). The signal on mother centrioles was roughly two times brighter than on the daughter centrioles (Fig. 1C). Finally, in G2 we observed a mix of three populations: 27.4±5.6% cells with centrobin on all four centrioles, 34.3±4.7% of cells with centrobin on the mother and the two daughter centrioles, and 38.2±3.7% of cells with centrobin only on the two daughter centrioles (Fig. 1A,B; the apparent increase in the percentage of centrobin-positive grandmother centrioles was not statistically significant). The signal intensity on daughter and mother centrioles was roughly equivalent. We conclude that centrobin localization is more dynamic than previously estimated, that centrobin is gradually recruited to daughter centrioles and that centrobin can simultaneously be present on daughter and mother centrioles or even the grandmother centriole up to G2 phase.

To determine the exact time-point at which centrobin is removed from mother centrioles, we next analyzed its localization pattern in the different mitotic phases, using DAPI to determine the mitotic stage. In prophase and prometaphase, roughly half the cells still displayed centrobin on the daughter centrioles and the mother centriole, while the other half contained only two centrobin dots on the daughter centrioles; centrobin levels on daughter centrioles were on average twice as high as on mother centrioles (Fig. 1D–F). From metaphase on, 79% of the cells only displayed two centrobin dots on the daughter centrioles, with average levels that were four times higher than mother centrioles; these values did not significantly change once cells reached anaphase and telophase (Fig. 1D–F). Staining with a second, independent, centrobin antibody led to equivalent results (Fig. S1). We conclude that centrobin is gradually removed from the mother centriole to accumulate on daughter centrioles, and that the last step in this process occurs at the prometaphase/metaphase transition.

To better understand the pathways controlling centrobin localization, we next tested whether centrobin removal was linked to the formation of fully elongated daughter centrioles in mitosis. To eliminate daughter centrioles, we inhibited the master regulator of centriole duplication PLK4, with the chemical inhibitor centrinone for 18 h (Wong et al., 2015). This resulted in cells that entered mitosis with only one centriole at each centrosome (called ‘1:1’ cells). When compared to DMSO-treated ‘2:2’ cells or to 2:2 cells that were only treated for 2 h with centrinone, 1:1 cells formed structurally normal spindle poles in mitosis as measured by the recruitment of γ-tubulin and pericentrin, two of the main components of the pericentriolar matrix (Fig. S2A–D). Moreover, these centrosomes were as functional as normal centrosomes with two centrioles; 1:1 cells had the same mitotic timing and the same chromosome segregation error rate when released from a prometaphase arrest induced with the centrosome separation (Eg5; also known as KIF11) inhibitor STLC (DeBonis et al., 2004) as did DMSO-treated or 2:2 cells treated with centrinone for only 2 h (Fig. S2E–H). This contrasted with RPE1 cells lacking centrioles on one (1:0 cells) or both poles (0:0 cells), which we and others have found to spend more time in mitosis and to have an elevated chromosome segregation error rate (Dudka et al., 2019; Wong et al., 2015).

When we analyzed the localization of centrobin in metaphase 1:1 cells, we found that 79.9±6.3% (mean±s.e.m.) contained a single centrobin-positive centriole (Fig. 2A,B). Using antibodies against the subdistal appendage protein cenexin (grandmother centriole marker; Lange and Gull, 1995), we established that this centriole was, as a rule, the mother centriole (Fig. 2A,C). The persistence of centrobin on the mother centriole in metaphase was not due to centrinone per se, since a short 2-h centrinone treatment in control 2:2 cells did not maintain centrobin at the mother centriole (Fig. S2I). Moreover, inhibition of centrosome duplication by depletion of the centriole duplication seed SAS-6 (also known as SASS6) for 24 h (Leidel et al., 2005), gave identical outcome to an 18-h inhibition of PLK4, confirming the specificity of our result (Fig. 2E; Fig. S2J).

The presence of centrobin on the mother centriole in metaphase 1:1 cells could be explained by two hypotheses: either centrobin removal requires the younger centriole to become a ‘mother’ that gave birth to a daughter centriole, or it is removed from mother centrioles due to the presence of a higher affinity site on the fully elongated daughter centrioles. To differentiate between the two possibilities, we looked at the cells depleted for SAS-6 for 24 h, which in our experience can lead to the formation of 2:1 cells, in which only the grandmother centriole gives rise to a daughter pro-centriole (Fig. S2J; Tan et al., 2015). The vast majority of 2:1 cells contained one centrobin-positive daughter centriole associated to the grandmother centriole, and no centrobin on the mother centriole (Fig. 2E). These data indicate that the mother centriole can ‘lose’ centrobin despite not having given rise to a daughter centriole, and are consistent with a model in which centrobin is removed from the mother centriole due to a higher affinity for this protein on daughter centrioles.

To further substantiate this hypothesis, we next compared the centrobin levels on the mother centriole of 1:1 cells, to the centrobin intensities of daughter centrioles in 2:2 cells by quantitative immunofluorescence. Our results indicated that mother centrioles in 1:1 cells contained nearly twice as much centrobin (Fig. 2F), consistent with re-distribution of the mitotic centrobin pool on both daughter centrioles in 2:2 cells. In addition, by partially inhibiting PLK4 with 25 nM centrinone, we also generated cells, in which grandmother and mother centrioles gave rise to two daughter centrioles. This counterintuitive result arises because partial inhibition most likely suffices to block the trans-phosphorylation-dependent PLK4 degradation without blocking its ability to initiate centriole biogenesis, as has been previously seen after the overexpression of catalytically inactive PLK4 (Guderian et al., 2010). Therefore, after 30 h of 25 nM centrinone treatment, we frequently observed spindle poles containing either a mother or a grandmother centriole each surrounded by two daughter centrioles (Fig. 2G). Quantification of centrobin intensities in such 3:3 cells indicated that both daughter centrioles only contained on average ∼60% of the centrobin levels seen on spindle poles with single daughter centrioles in 2:2 cells, suggesting again a re-distribution on all centrioles. Although the intensities showed a strong cell-to-cell variability, the overall pattern was again consistent with an affinity model in which centrobin is redistributed on daughter centrioles during mitosis.

Recent studies postulated that distal appendage formation at mother centrioles requires the step-wise removal of the daughter-specific CEP120, centrobin and Neurl4 in early S-phase to enable the recruitment of OFD1 (Wang et al., 2018). This first step is followed later, at around mitotic onset, with the progressive recruitment of other distal appendage proteins, including CEP164 (Fig. 3A). Thus, in a normal metaphase cell, OFD1 and CEP164 are present on both the grandmother and mother centriole. When looking at 1:1 cells that were co-stained with antibodies against centrobin and OFD1 or CEP164, we found that centrobin-positive mother centrioles were often also OFD1 or CEP164-positive (Figs 2F and 3B). This suggested that the presence of centrobin at the mother centriole in 1:1 cells did not perturb the recruitment of OFD1 or CEP164. Consistent with this, when we stained for OFD1 in a large number of 1:1 cells, we found that a majority of them (75±2.5%; mean±s.e.m.) still displayed OFD1 at both grandmother and mother centrioles, as compared to 91±1.3% in 2:2 cells (Fig. 3C,D). CEP164 localization was even less affected, as it was present at grandmother and mother centrioles to the same extent in 2:2 and 1:1 cells (Fig. 3E,F). Consistent with the idea that the presence of centrobin is not incompatible with the loading of OFD1 and CEP164, we also even found examples of late G2 wild-type (WT) 2:2 cells in which CEP164 and centrobin or OFD1 and centrobin could be found on the same mother centriole (Fig. 3G).

In Drosophila melanogaster neuroblasts, the PLK1 kinase ortholog Polo controls the specific enrichment of centrobin on the younger centrosome (Gallaud et al., 2020; Januschke et al., 2013). We therefore tested whether mitotic phosphorylation also controls the removal of human centrobin in late prometaphase. Specifically, we tested the contribution of PLK1 and Aurora A, another key mitotic kinase controlling the centrosome cycle (Barr and Gergely, 2007). RPE eGFP–centrin1 cells were treated for 2 h with a either mock control (DMSO), 25 or 50 nM of the PLK1 inhibitor BI2536 (Lénárt et al., 2007), or 100 nM of the Aurora A inhibitor Alisertib (Görgün et al., 2010). Most control-treated cells (67±2%) displayed centrobin on the two daughter centrioles, whereas 32±2% (mean±s.e.m.) of the cells displayed three centrobin-positive centrioles (Fig. 4A,B). PLK1 inhibition increased the proportion of cells displaying a third centrobin-positive centriole to, respectively, 52±3% and 55±1%, and resulted in 10% of the cells with four centrobin-positive centrioles (Fig. 4A,B). After Aurora A inhibition, only 29±5% of the cells harbored centrobin at the two daughter centrioles, 36±5% on three centrioles, and 35±3% on all four centrioles (Fig. 4C,D). This indicated that centrobin removal from older centrioles depends both on PLK1 and Aurora A.

Since PLK1 is activated in G2 by Aurora-A to promote mitotic entry and centrosome separation (Seki et al., 2008), we next aimed to inhibit these two kinases only once cells had entered mitosis. We first arrested the cells in mitosis with STLC for 4 h, before inhibiting either PLK1 or Aurora-A for 2 h. STLC treatment and the resulting monopolar spindles did not affect centrobin localization, since we found the same pattern as in untreated cells – most mother centrioles had lost centrobin (Fig. 4E,F). PLK1 inhibition led to the rebinding of centrobin on older centrioles in STLC-treated cells, whereas Aurora A inhibition had no effect (Fig. 4G,H). This suggested: (1) that Aurora-A does not affect centrobin localization once PLK1 is fully active; (2) that PLK1 is the primary kinase controlling centrobin; and (3) that PLK1 inhibition can bring centrobin back to mother centrioles that had already lost it. To confirm the dynamicity of this regulation, we washed out BI2536 in STLC-treated cells after 2 h and replaced it either with DMSO or again BI2536. Washing out PLK1 inhibition restored normal centrobin localization on two centrioles, confirming the dynamic nature of this regulation in mitosis (Fig. 4E,F).

Given that PLK1 activity also enables the recruitment of distal appendage proteins at the mother centriole (Kong et al., 2014), we next investigated the epistasis of this centrosome cycle step in relation to centrobin removal. Normally, CEP164 is present on both the grandmother and mother centriole in >90% of mitotic cells (Fig. 5A,B). Inhibition of PLK1 for 2 h in STLC treated cells did not disrupt CEP164 localization, if anything, it led to higher CEP164 levels at both centrioles (Fig. 5A,B). In contrast, PLK1 inhibition for 24 h resulted in 69±2% (mean±s.e.m.) of cells having CEP164 only at one centriole (Fig. 5C,D), consistent with a role of PLK1 in rendering centrioles compatible with the recruitment of distal appendage proteins (Kong et al., 2014). In the vast majority of cells, the CEP164-positive centriole was the grandmother centriole, identified by the higher intensity of the centrin1 signal (Fig. 5E; Tan et al., 2015). This suggests that PLK1 activity is required for the recruitment of CEP164 on the mother centriole prior to mitosis, but not for its maintenance during mitosis. Immunofluorescence analysis of cells treated with BI2536 for 24 h also indicated that centrobin was more frequently retained on the mother centriole (Fig. 5D,F). Nevertheless, OFD1 was present in 95±1% of cells on both spindle poles (versus 91±1% in control-treated cells; Fig. 5G,H). This confirmed that centrobin removal is not required as a pre-requisite for the loading of OFD1 on mother centriole and suggested that PLK1 controls CEP164 recruitment further downstream in the formation of distal appendages. To directly test whether PLK1 inhibition affects CEP164 recruitment via the presence of centrobin, we combined PLK1 inhibition with the siRNA-mediated depletion of centrobin. Although centrobin was efficiently depleted (Fig. 5I,J), its absence did not rescue the recruitment of CEP164 at the mother centriole; whether centrobin was present or not, PLK1 inhibition led to cells with only one CEP164-positive centriole in 64% of the cases (Fig. 5J,K). We conclude that PLK1 regulates the recruitment of CEP164 and the removal of centrobin via separate pathways.

Finally, we investigated the role of the core subdistal appendage protein cenexin in the regulation of the centrobin and the CEP164 pathways. Cenexin has been reported to modulate PLK1 activity at centrosomes (Colicino et al., 2019); moreover, in mouse embryonic F9 carcinoma cells, cenexin is required for the formation of both distal and subdistal appendages (Ishikawa et al., 2005; Tateishi et al., 2013). In RPE cells, however, it was reported to be only required for the formation of subdistal appendages (Kuhns et al., 2013; Tanos et al., 2013). While using two different siRNAs against cenexin to rule out any off-target effect, we found that efficient cenexin depletion (>80%, see Fig. S3A,B) led to the same configuration as the one seen after PLK1 inhibition in most cells: (1) only the grandmother centriole retained CEP164 (Fig. 6A,B, Fig. S3C; note that grandmother centrioles were identified based on the highest intensity of the eGFP-centrin1 signal); (2) centrobin was present on the mother centriole (Fig. 6C,D, Fig. S3D); and (3) OFD1 recruitment was not affected (Fig. 6E,F). This indicated that, as was the case for PLK1, cenexin is required for the removal of centrobin and the recruitment of CEP164 at the mother centriole, but that it does not affect OFD1 localization. To test for the epistasis of CEP164 recruitment and centrobin removal, we co-depleted both cenexin and centrobin and found that centrobin depletion did not rescue CEP164 binding to the mother centriole in cenexin-depleted cells (Fig. 6G,H; note that the co-depletions were as efficient as the single depletions; Fig. 3E–H). We conclude that cenexin regulates centrobin and CEP164 separately, and in an OFD1-independent manner.

We next extended our dependency analysis to the centrosomal proteins CEP128 and centriolin, which act downstream of cenexin in terms of building up subdistal appendages (Mazo et al., 2016). Using the published hTert-RPE1 cenexin, CEP128 or centriolin CRISPR/Cas9 knockout cell lines, we found that only cenexin deletion partially prevented the recruitment of CEP164, and that no defects could be seen in CEP128 or centriolin knockout cells (Fig. 7A,B; note that the percentage of cells with only one CEP164-positive centriole was lower compared to cenexin-depleted cells; see the Discussion). Our analysis further indicated that centrobin was present on all four centrioles in all three knockout cell lines, but that OFD1 localization was unchanged (Fig. 7C–F). The picture was more differentiated in G1 when only two centrioles are present (G1 cells were obtained by treating for 24 h with the CDK4 and CDK6 inhibitor palbociclib): in cenexin knockout cells, centrobin was present on both centrioles in 82% of the cells versus 18% in WT cells; in contrast in CEP128 and centriolin knock-out cells, centrobin was present on both centrioles in 60–63% of the cells, with a clear enrichment on the daughter centriole (Fig. S4A,B). We conclude that the mitotic removal of centrobin from older centrioles is not a pre-requisite for the recruitment of distal appendage proteins and that the removal of centrobin from maturing centrioles requires subdistal appendage proteins, in particular cenexin. In contrast, distal appendages, appear not to be required, since the depletion of the distal appendage protein CEP164 had no effect of centrobin localization (Fig. S4C–E).

In this present study, we investigated how the localization of the daughter-centriole-specific protein centrobin is controlled over the cell cycle in human epithelial cells, and whether centrobin acts as a placeholder protein for distal appendage proteins on the mother centriole. Our results imply that centrobin is removed from mother centriole in a dynamic manner and transferred on to daughter centrioles once cells reach mitosis, possibly due to a higher affinity. Centrobin removal requires the activity of the mitotic kinase PLK1 and the presence of subdistal appendage proteins on the mature centrioles. Finally, our results imply that centrobin does not act as a placeholder for the distal appendage proteins, but rather that centrobin removal and distal appendage protein recruitment are promoted by PLK1 via separate pathways.

Previous work in Drosophila melanogaster neuroblasts had shown that centrobin is transferred in metaphase in a PLK1-dependent manner from the mother centriole to the daughter centriole, potentially via a direct relocalization mechanism (Gallaud et al., 2020). Here, we show in human epithelial cells that removal from the mother centriole until metaphase is dynamic, as centrobin pools can co-exist on mother and daughter centrioles in G2, prophase and early prometaphase cells; moreover, we demonstrate that the presence of at least one daughter centriole is essential to remove centrobin from the mother centriole. Our quantification of centrobin pools on mother centrioles in 1:1 cells, on normal daughter centrioles in 2:2 cells and in cells with extra daughter centrioles lead us to propose a speculative model in which higher-affinity binding sites on daughter centrioles compete against weaker-affinity binding sites on mother centrioles for a dynamic but limited centrobin pool that is rapidly exchanged via the cytoplasm. In such a model, as centrobin is dynamically exchanged with the cytoplasm, it will gradually leave mother centrioles and accumulate on maturating daughter centrioles; if daughter centrioles are absent, centrobin will remain on mother centrioles. We emphasize, however, that demonstrating such an affinity model will require more direct biochemical evidence with recombinant proteins and purified centrioles.

Although our data indicate that PLK1 inhibition leads to a reversible association of centrobin to the mother centriole, it is unclear at which level this kinase acts. In theory PLK1 activity could decrease the affinity of the centrobin-binding site on the mother centriole, increase it on the daughter centrioles, or switch the binding preference by directly phosphorylating centrobin itself, in line with the fact that centrobin is a PLK1 substrate (Lee et al., 2010). Our present resolution also does not allow us to distinguish whether, in PLK1 inhibited cells, centrobin re-associates to the same distal end of mother centrioles as control S-phase cells. The subdistal appendage proteins cenexin, CEP128 and centriolin, which in metaphase are predominantly present at the grandmother centriole (Colicino et al., 2019; Gasic et al., 2015; Ishikawa et al., 2005; Kong et al., 2014), are also required or contribute to centrobin removal from the mother centriole. Given the vast distances between the two spindle poles at this stage, this control is most likely indirect, although one cannot completely exclude that subdistal appendages on the grandmother centriole may affect the mother centriole in the preceding interphase. One likely pathway is that the subdistal appendages regulate PLK1 activity, consistent with previous studies showing that the presence of cenexin enhances PLK1 activity (Colicino et al., 2019); whether CEP128 and centriolin play a similar role, remains to be investigated.

Finally, our results indicate that centrobin does not act as a placeholder for distal appendage proteins, in contrast to a previously proposed model (Wang et al., 2018). Our data identify multiple experimental conditions in which centrobin presence at mother centriole does not prevent the recruitment of OFD1, which is recruited early, and of CEP164, which is recruited last during distal appendage formation. This co-existence can even be seen in WT G2 cells, with examples of centrobin-positive mother centrioles that have already begun to recruit OFD1 or even CEP164. Our investigation of subdistal appendage proteins further indicates that cenexin is partially required for the presence of CEP164, but not of OFD1, on the mother centriole in metaphase. This role is specific for cenexin, as knockouts of CEP128 and centriolin did not reveal any changes in CEP164 localization, confirming that centrobin removal and CEP164 recruitment are independent. We further note that the acute depletion of cenexin by both tested siRNAs, even when not fully effective, led to a stronger reduction in CEP164 levels on metaphase mother centrioles than in the cenexin knockout cells. This could suggest that knockout cells have adapted to cenexin loss, as has been seen for other deletions (Rossi et al., 2015), allowing them to nevertheless efficiently form distal appendages. Finally, our data confirm that the presence of CEP164 at mother centrioles in metaphase requires PLK1 activity, as seen in other studies (Kong et al., 2014; Tanos et al., 2013). This dependency is, however, only transient, since PLK1 inhibition does not affect the maintenance of distal appendages at later stages. This is consistent with the idea that PLK1 enables distal appendage protein recruitment on the mother centriole by promoting centriole maturation (Kong et al., 2014). It also fits with the idea that distal appendage formation does not solely depend on PLK1 but is under the complex control of several other centrosomal kinases, such as Nek2 and Aurora A (Bowler et al., 2019; Viol et al., 2020). Overall, we postulate that, in human epithelial cells, PLK1 controls centriole maturation via separate pathways, one involving centrobin removal, the second enabling recruitment of distal appendage proteins on the mother centriole. This raises the question of the reasons for centrobin removal since its presence does not impact distal appendage formation. In Drosophila neuroblasts, centrobin localization has to be controlled since it regulates the fate of the young centrosome, enabling it to organize microtubules and to be retained by the stem cell during asymmetric cell division (Januschke et al., 2013). In mammalian cells, however, the reasons for controlling centrobin still remain to be uncovered. Our characterization of 1:1 cells indicate that the presence of centrobin at the mother centriole on one pole neither affects mitotic progression nor change the incidence of chromosome segregation errors, indicating that its removal is not acutely necessary for centrosome or spindle pole function in mitosis. Rather, we speculate that timely removal of centrobin in (pro-)metaphase might be linked to the precision of centriole elongation during the centrosome duplication cycle and to cilia formation, since it has been shown that overexpressing centrobin leads to overelongated centrioles and may affect the axoneme structure during ciliogenesis (Gudi et al., 2015; Ogungbenro et al., 2018).

hTERT-RPE1 cells, hTERT-RPE1 eGFP-centrin1 cells, hTERT-RPE1 eGFP-Centrin1/eGFP-CENP-A cells (both gifts of Alexey Khodjakov, New York State Department of Health, Wadsworth Center, USA; Magidson et al., 2011), hTERT-RPE1 cenexin knockout cells, Cep128 knockout cells, centriolin knockout cells and their parental cell line (gift of Brian Tsou, Memorial Sloan Kettering Cancer Center, Sloan Kettering Institute, USA; Mazo et al., 2016), were all grown in Dulbecco's modified Eagle's medium (DMEM; Thermo Fisher Scientific, Switzerland), supplemented with 10% fetal calf serum (FCS), 100 U/ml penicillin, and 100 mg/ml streptomycin (all Life Technologies, Switzerland) at 37°C and 5% CO₂. All cell lines were regularly tested for contamination. For live-cell imaging, cells were kept in their normal medium if a CO₂ chamber was used, or otherwise cultured in Leibovitz's L-15 medium without Phenol Red (Thermo Fisher Scientific), supplemented with 5% FCS. To inhibit PLK1, 25 nM or 50 nM BI2536 (Axon Lab AG, Switzerland) was used, for either 2 h or 24 h. To inhibit Aurora-A, 100 nM Alisertib (Selleckchem, USA) was used for 2 h. For the PLK1 inhibition washout experiments, cells were treated with 5 µM STLC for 4 h to maintain them in prometaphase, 50 nM BI2536 was added for 2 h before adding back medium containing only STLC for 2 h. To obtain 1:1 cells, cells were treated with 250 nM centrinone (a PLK4 inhibitor; Tocris, UK) for 18 h. To obtain cells with excessive numbers of daughter centrioles, cells were treated with a partial dose (25 nM) of centrinone for 30 h. Alternatively, a short 2 h treatment was added as a negative control. To obtain WT cells, cenexin knockout, CEP128 or centriolin knockout cells in G1 phase, cells were treated for 24 h with the CDK4 and CDK6 inhibitor Palbociclib (Sigma-Aldrich, Switzerland).

In general siRNA treatments were performed for 48 h by applying 40 nmol siRNAs and 1% Lipofectamine RNAimax (Invitrogen) following the manufacturer's protocol. To deplete cenexin, we applied 180 nmol siRNA and 3% Lipofectamine RNAimax (Invitrogen) instead; to prevent centriole duplication and obtain a mix of 2:1 and 1:1 cells, Sas-6 was depleted for only 24 h. The target sequences of the siRNAs used are: siCEP164 (mix of 5′-CAGGTGACATTTACTATTTCA-3′ and 5′-ACCACTGGGAATAGAAGACAA-3′), siCnb (mix of 5′-TGGAAATGGCAGAACGAGA-3′, 5′-GCATGAGGCTGAGCGGACA-3′, 5′-GCCCAAGAATTGAGTCGAA-3′ and 5′-CTCCAAACCTCACGTGATA-3′), siCnx #1 (5′-TGGCTGAGACTGAGCACGA-3′), siCnx #2 (5′-GGCACAACATCGAGCGCAT-3′), siCtrl (5′-AAGGACCTGGAGGTCTGCTGT-3′) and siSas-6 (5′-GCACGTTAATCAGCTACAA-3′).

Cells were grown on acid-etched glass coverslips and either fixed for 6 min at −20°C with ice-cold methanol, or for 15 min with a formaldehyde fixative (0.05 M PIPES, 0.01 M EGTA, 1 mM MgCl₂, 0,2% Triton X-100 and 4% formaldehyde) at room temperature. Coverslips were washed with phosphate-buffered saline solution (PBS) three times before adding a blocking buffer for 1 h at room temperature (7.5% bovine serum albumin, 0.25% sodium azide in PBS). Primary antibodies were diluted in blocking buffer and added for 1 h at room temperature followed by three PBS washes. The following primary antibodies were used: rabbit anti-cenexin (1:1000, Abcam, UK, ab43840), goat anti-cenexin (1:1000, Abcam, ab121023), mouse anti-centrin1 (1:2000, clone 20H5 Merck Millipore, Switzerland), mouse anti-centrobin (1:1000, Abcam ab70448) or, alternatively, mouse-anti-centrobin (1:1000 Sigma-Aldrich, SAB1408254), mouse anti-Sas6 (1:200, Santa Cruz Biotechnology, USA, sc-81431), rabbit anti-CEP164 (1:1000, gift of Erich Nigg, Biozentrum Basel, University of Basel, Switzerland; Graser et al., 2007), rabbit anti-OFD1 (1:1000, Abcam ab222837), mouse anti-pericentrin (1:1000, Abcam ab28144), rabbit anti-γ-tubulin (1:2000; Wilhelm et al., 2019), and rabbit anti-PCNA (1:1000, Abcam ab18197). Cross-absorbed secondary antibodies tagged with Alexa Fluor fluorophores were diluted in blocking buffer (1:400, Invitrogen, Switzerland) and applied for 30 min at room temperature. Vectashield with DAPI (Vector Laboratories, Switzerland) was used to mount the coverslips, and images were taken with a wide-field Olympus Deltavision (GE Healthcare, USA), with a 60×1.4 NA oil objective and a DAPI/FITC/TRITC/Cy5 (Chroma, USA) filter set. The z-stack images were recorded in 0.2 µm steps with a Coolsnap HQ2 CCD camera (Roper Scientific) and the Softworx software (GE Healthcare). Image stacks were deconvolved and protein levels at centrioles were quantified based on maximal intensity projections using Softworx Explorer (GE Healthcare). To quantify pericentriolar protein levels, we applied with Imaris (Bitplane, Switzerland) a 0.5 µm diameter sphere centered on the eGFP–centrin1 signal and quantified the summed intensity within this sphere.

Centrin was also used to distinguish between the old and young centrosomes in experiments where CEP164 localization was studied.

The data is presented as the mean of multiple independent experiments and error bars represent the s.e.m. unless otherwise specified. Significance was determined using Graphpad Prism, and are represented presented in this work according to the following codification: *P<0.05, **P<0.01, ***P<0.001 and ****P<0.0001. An unpaired two-tailed t-test assessed the depletion efficiency for cenexin, centrobin or CEP164. For all the data describing the changes in centrobin and CEP164 localization, a two-way ANOVA with Tukey's multiple comparison test was chosen. For the comparison of the relative centrobin levels on grandmother, mother and daughter centrioles a one-way ANOVA test with Sidak's multiple comparison test was applied. All illustrations and graphs were created using ImageJ Fiji, Graphpad Prism and Adobe Illustrator.

Authors are grateful to A. McAinsh (University of Warwick, UK), A. Khodjakov (NY State University, USA), B. Tsou (Cornell University, NY, USA) for cell lines, E. Nigg (University of Basel, Switzerland) for CEP164 antibodies, Monica Gotta (University of Geneva, Switzerland) and members of the Meraldi laboratory for helpful discussions and support, especially E. Galster for her help in cell lines’ maintenance and A.-M. Olziersky, L. Romanens, D. Dudka, and T. Wilhelm for critical comments on the manuscript.

The peer review history is available online at https://journals.biologists.com/jcs/article-lookup/doi/10.1242/jcs.259120.