The centrosome of somatic cells is the main microtubule organising center [1]. It is required to organise the cytoplasmic microtubule network during interphase and the mitotic spindle during mitosis. This organelle consists of two centrioles embedded in an amorphous pericentriolar material (PCM). In proliferating cells before division, the centrosome needs to be duplicated precisely once so that the mitotic cell harbours two centrosomes, each comprising two centrioles. These two centrosomes will be used to nucleate the microtubules required to assemble the mitotic bipolar spindle during mitosis [2]. The centrosome duplication cycle is divided into several key steps. First, during mitosis, the two paired-centrioles inherited by the daughter cell lose their orthogonal configuration, a process called disengagement. Then, procentrioles (daughter centrioles) nucleate in G1/S phase at an orthogonal angle next to each mother centriole. The two newly synthesised centrioles elongate during the S and G2 phases. In late G2, as cells prepare for mitosis, centrosomes increase in size and recruit additional PCM to enhance their ability to nucleate microtubules. This process is referred to as centrosome maturation. At the end of the G2 phase, the two newly duplicated centrosomes separate to organise a bipolar mitotic spindle, enabling each daughter cell to inherit one centrosome after cell division.

In contrast to normal cells, tumour cells frequently show centrosome number defects that are thought to be the consequence of abnormal regulation of the centriole duplication machinery. Thus, the restriction to a single round of centriole duplication per cell division cycle contributes to the prevention of aberrant centrosome numbers, multipolar spindles and chromosomal instability [3], [4].

Several proteins are critical for the canonical centriole assembly. Based on genome-wide RNAi and genetic screens in C. elegans, a molecular hierarchy for the recruitment of five essential proteins for centriole assembly has been established. SPD-2 is necessary to recruit ZYG-1 at the centriole. Then, both proteins in turn recruit SAS-5 and SAS-6, followed by SAS-4 [5], [6], [7]. Most of these proteins have conserved roles in centriole duplication in other systems. Cep192, the human homolog of C. elegans SPD-2 protein, is a pericentriolar protein required for both centriole duplication and centrosome maturation [8], [9]. Plk4, related to C. elegans ZYG-1, as well as CPAP and HsSAS-6, human homologs of C. elegans SAS-4 and SAS-6, have been identified as key regulators of centriole duplication [8], [10], [11], [12]. For example, overexpression of Plk4 leads to centriole amplification; conversely, mutations or inhibition of the DmSAK/PLK4 gene dramatically impairs centriole duplication [10], [11]. Of these proteins, Plk4 is the only enzyme. It is a short-lived protein kinase that localises at centrioles throughout the cell cycle. Its degradation and activity are tightly regulated during the cell cycle, these two processes being crucial for the centriole biogenesis process [13], [14].

SAS-6, Chk2, CDC25c and Hand1 remain the only identified substrate of Plk4/SAK/ZYG-1 [15], [16], [17], [18], [19], [20]. The physiological relevance of CDC25c and Chk2 phosphorylation events by Plk4 is unknown. SAS-6 phosphorylation by ZYG-1 in C. elegans triggers its targeting to the newborn centriole. The Plk4-mediated phosphorylation of the Hand1 transcription factor inhibits its sequestration in the nucleolus. Hand1 is then released in the nucleus where it activates the transcription of genes required for cell differentiation. Interestingly, a growing list of studies in different models shows that Plk4 centrosomal levels and activity are maximal during mitosis, suggesting the kinase also plays a key role during mitosis, possibly to licence the centriole for subsequent duplication in S phase. However, this possible role remains speculative and difficult to challenge experimentally [14], [21].

Cyclin-dependant kinase 11 (CDK11) mRNA produces a 110 kDa isoform protein expressed throughout the cell cycle and a 58 kDa isoform protein which is specifically translated from an internal ribosomal entry site sequence during G2/M [22]. The small isoform has been shown to be an essential regulator of mitosis. Indeed, CDK11^p58 is required for centrosome maturation, bipolar spindle assembly, maintenance of sister chromatid cohesion and cytokinesis [23], [24], [25].

In this article we describe a new role for CDK11^p58 in the duplication of centrioles. Our data reveal that CDK11 is necessary to ensure proper centriole duplication. Furthermore, we identify Plk4 as a novel partner of CDK11^p58 and show that CDK11^p58 depletion impairs the proper recruitment of Plk4 to mitotic centrosomes.

We have shown previously that CDK11^p58 is required for centrosome maturation (Petretti et al., 2006). A growing list of studies reports that failure to properly orchestrate centrosome maturation can trigger centriole duplication defects. As an example, Cep192 is a major regulator of both centrosome maturation and centriole duplication in C. elegans and in mammalian cells [7], [8], [26]. These data prompted us to investigate whether CDK11^p58 was also required for centriole duplication. To do so, we first looked at the effects of a three-day siRNA-mediated depletion of CDK11 on centriole numbers in asynchronously growing mitotic cells (Figure 1 A). HeLa cells stably expressing GFP-centrin protein [27] were transfected with control or CDK11 specific siRNAs [24] and centrioles were quantified on mitotic figures after fixation of the cells (Figure 1 D). As expected, 90.6% of control metaphase cells showed four centrioles with two centrin dots at each pole of the mitotic spindle (Figure 1 and Table S1). By contrast, after depletion of CDK11, as shown previously, only 37.9% of the spindles were bipolar [24]). Amongst these, only 10.8% of the cells looked normal with two centrin dots at each pole whereas 26.9% of the cells showed spindles with less than two centrioles at one pole. The remaining mitotic cells appeared to form very short or monopolar spindles (54.4%) and all of them showed a reduction in centriole numbers (i.e. less than four centrioles in total). Plk4 RNAi treatment, used as a positive control (Figure 1 C, D and Table S1), yielded highly similar phenotypes with only 14.4% of the cells showing bipolar mitotic spindles with four centrioles, two at each pole. The vast majority of the mitotic cells displayed monopolar and disorganised spindles and showed less than four centrioles. Interestingly, co-depletion of Plk4 and CDK11^p58 proteins resulted in an 8- to 9-fold decrease of the number of cells showing bipolar spindles with four centrioles (10.8% and 14.4% respectively in CDK11- and Plk4-depleted cells vs. 1.5% in co-depleted cells). We observed a dramatic increase in the number of cells with only one centriole when both CDK11 and Plk4 were knocked down (13,9% and 17.8% respectively in CDK11 and Plk4-depleted cells vs. 67.8% in co-depleted cells) (Figure 1 C, D and Table S1).

Most of the proteins required for centrosome duplication localise to this structure. To localise CDK11^p58 protein kinase within the cell, we decided to generate a HeLa “Tet-ON” cell line which allows the temporally controlled expression of Myc-tagged CDK11^p58 (Figure S1). The Myc antibody readily recognised one protein of the expected size in the above cell line induced to express the transgenic protein (data not shown) and also in transiently transfected cells (Figure 2 A) but did not reveal any specific bands in non-induced or control cells validating our Tet-inducible expression system. Immunofluorescence experiments were performed after a very short period of induction (1 to 2 hours), when the protein was expressed at low level. We found that Myc-tagged CDK11^p58 protein kinase was cytoplasmic in interphase (Figure S1 A and B). During early prophase, CDK11^p58 was principally detected around the separating centrosome and on newly nucleated microtubules. In metaphase cells, the protein localised in the vicinity of the centrosome, and decorated spindle microtubules (Figure S1 and C). Following depolymerisation of microtubules by cold treatment, no obvious localisation of CDK11^p58 was observed at the mitotic centrosomes suggesting that the centrosomal localisation is microtubule dependent (data not shown). CDK11^p58 persisted on microtubules emanating from the poles and on the central spindle during telophase and was finally enriched on both sides of the midbody during cytokinesis (Figure S1). Interestingly enough, longer induction to express high levels of Myc-tagged protein increased the cytoplasmic pool resulting in the masking of the signal on mitotic microtubules. This latter result might suggest the existence of a limited number of CDK11^p58-binding sites on the mitotic spindle.

The RNA interference of known regulators of centriole duplication leads to a decrease of the centriole number. By contrast, only a very low number of these molecules, including the Plk4 protein kinase, triggers centriole amplification via their overexpression (reviewed in Bettancourt 2009). To continue the investigation on the function of CDK11^p58 in centriole biogenesis, we decided to check whether the overexpression of a Myc-tagged wild-type CDK11^p58 protein, a kinase-dead variant (CDK11^p58DN) and Plk4 (as a positive control) could modify centriole duplication (see Figure 2 A). CDK11^p58, CDK11^p58DN, Plk4 were transiently expressed in HeLa cells stably expressing the GFP-centrin protein. A day after transfection, aphidicolin was added for 48 hours on these cells to arrest them in S phase (Figure 2 B). Cells were then immuno-stained with Myc antibody and the centrioles of Myc-positive arrested cells were counted (Figure 2 C). As expected, the vast majority (71%) of control cells had duplicated their centrioles during this arrest and thus exhibited four centrioles per cell (two pairs of two closely-linked centrioles). When compared to controls, cells overexpressing CDK11^p58 did not show significant differences in their centriole numbers (∼70%, Figure 2 D). By contrast, the overexpression of the catalytically inactive CDK11^p58DN caused a decrease in the number of cells with four centrioles. Indeed, only 38% of these cells contained four centrioles, while 58% failed to duplicate their centrioles, resulting in the presence of cells with either two or three centrioles (Figure 2 C and D). In a parallel assay, the ability of these cells to amplify their centrioles after Plk4 overexpression was assessed. As published before [11], most of the Myc-Plk4 positive cells amplified their centrioles (Figure 2 C, right panels, and D). Therefore, the dead variant of CDK11^p58 has the ability to inhibit centriole duplication, suggesting it behaves as a dominant negative toward the endogeneous active kinase to prevent centriole duplication. CDK11^p58 kinase activity would thus be required for this process.

In order to further corroborate our observation that CDK11^p58 is required for centriole duplication, we used a centriole overduplication assay in U2OS cells. In this cell line, a prolonged S phase arrest induces multiple rounds of centriole duplication. As a consequence, inhibition of proteins required for centriole duplication leads to a decrease in the number of cells with more than two centrosomes (Habedanck, 2005). U2OS cells were transfected with control, Plk4 (as a positive control) or CDK11 siRNA for 24 hours before being arrested during S phase with aphidicolin for 48 hours (Figure S2 A). Cells were fixed and centrosomes were stained with anti-γ-tubulin antibodies (Figure S2 B). In control S-phase arrested cells, 29.0±1.4% of the cells showed more than two centrosomes (Figure S2 C). A noticeable decrease was observed in Plk4 and CDK11-depleted cells, with a percentage of cells displaying more than two centrosomes of 9.0±2.1% and 11.9±1.7% respectively, indicating a ∼3-fold inhibition of the spontaneous centriole over duplication process observed in U2OS control cells. This latter result further highlights the requirement of CDK11^p58 in the centriole duplication process.

Previous experiments showed that CDK11-depletion diminished γ-tubulin and Plk1 recruitment at the centrosomes during centrosome maturation [24]. Interestingly, Plk1 depletion abolishes Cep192 recruitment at centrosomes [8], [26]. We first decided to analyse the levels of Cep192 protein at mitotic centrosomes (containing 2 centrioles) following control or CDK11 RNAi (Figure S3 A). Quantification of the signal revealed that centrosomal Cep192 was reduced by 3-fold at centrioles in mitotic CDK11-depleted cells (Figure S3 B) but the interphase centrosome level was unchanged (data not shown). SPD-2, the worm counterpart of Cep192 is also required for ZYG-1, the related orthologue of Plk4, to be recruited at centrosomes [5], [7]. Plk4 is a centriole-associated protein kinase. The protein kinase levels in the whole cell and at the centrosome are elevated during mitosis, by contrast to interphase. We next monitored whether CDK11^p58 was also implicated in Plk4 recruitment to the centrosomes in mitosis. Control and CDK11-RNAi treated cells were stained with Plk4, centrosome (Figure 3 A) or centriole probes antibodies (Figure 3 B). Strikingly, CDK11 depletion also impaired Plk4 accumulation at mitotic centrosomes (Figure 3 A–C) whereas staining of Plk4 at centrosomes in interphasic cells was not reduced (Figure 3 D and E). In parallel, we also checked whether the Plk4 protein was stable in CDK11 knock down cells to rule out the possibility that the low amount of centrosomal Plk4 after CDK11 RNAi was caused by an overall Plk4 protein disappearance. As endogenous Plk4 protein cannot be detected by Western blot, we examined the stability of full length Plk4 or a shorter Plk4 fragment also containing its phosphodegron motif [13]. For this purpose, we co-transfected control or CDK11 siRNA together with the Myc-tagged Plk4 constructs and examined by Western blot exogenous Plk4 protein levels. Exogenous Plk4 proteins were very unstable and difficult to detect in the control cells 48 h following RNAi. However, we found that Myc-Plk4 proteins were more stable in CDK11 knock down cells (Figure 3 F), suggesting that more Plk4 protein was present in this case but not targeted to the mitotic centrosomes. Taken together, the above results point out that CDK11^p58 is essential for the correct recruitment of multiple centrosomal proteins including Plk1 and γ-tubulin, but also Cep192 and Plk4.

To date, very few proteins have been found to associate with Plk4. We further tempted to investigate whether CDK11^p58 could physically interact with this master regulator of centriole duplication. As endogenous CDK11^p58 and Plk4 proteins are difficult to detect in cell lysates, we decided to co-express these two proteins with different tags in COS-7 cells and we carried out immunoprecipitation experiments (Figure 4 A, B and C). We were able to detect Myc-Plk4 in GFP-CDK11^p58 pull down (IP GFP, Figure 4 B) and GFP-CDK11^p58 was reciprocally found in Myc-Plk4 pull-down (IP MYC, Figure 4 B). We then performed the same experiment using truncated Myc-Plk4 proteins containing either the N-terminal or the C-terminal end of the protein (Figure 4 C). Interestingly enough, the interaction was only observed between CDK11^p58 and the N-terminal domain of Plk4 (NT), which contains the kinase catalytic domain and the phosphodegron motif required for protein stability [13]. We didn't find any interaction of CDK11^p58 with the Plk4 C-terminal fragment containing the polo box (Figure 4 C). To verify that Plk4 and CDK11^p58 were able to interact directly, we performed an in vitro pull down assay using Maltose-Binding Protein (MBP)-tagged CDK11^p58 immobilised on an amylose column and a recombinant V5-tagged full length Plk4 (See Method section and Figure 4 D). Although the immobilised CDK11^p58 protein kinase was inactive (not able to phosphorylate any of the in vitro CDK kinase substrates like histone H1, MBP or casein), it was able to interact with Plk4 in vitro, revealing a direct interaction between the two proteins (Figure 4 D).

As CDK11^p58 and Plk4 are protein kinases that physically interact, it was tempting to speculate that one could be the substrate of the other. Unfortunately, soluble recombinant CDK11^p58 is inactive so the CDK11^p58 kinase assay on Plk4 was not possible. However, we were able to check that active Plk4 could not phosphorylate a recombinant MPB-tagged CDK11^p58 (Figure S4), since after incubation of these proteins together we couldn't observe any phosphorylated band, whereas Plk4 was able to phosphorylate b casein (positive control). This result suggests that CDK11^p58 is not a substrate of Plk4.

Our study reveals that the mitotic specific CDK11^p58 is able to associate with Plk4 during mitosis and is required to target the Plk4 protein kinase to the centrosome in mitosis. In a last experiment, we decided to check whether mitotic centrioles that had passed mitosis with decreased CDK11 activity were able to respond to high Plk4 activity in the following interphase by amplifying their centrioles. For this purpose, we performed CDK11 RNAi on HeLa cells for two days (to allow CDK11 depletion) and we overexpressed Myc-Plk4 by transient transfection to trigger centriole amplification in the presence of aphidicolin (Figure 5 A). As expected in S-phase arrested control cells, a majority of the cells (75.9±10.4%) displayed 4 centrioles whereas only 52.1±10.0% of the CDK11–depleted cells exhibited this phenotype (Figure 5 C). In this population, we also observed a strong increase of cells with less than 4 centrioles (43.8±15.6%) compared to controls (13.8±4.9%). When we overexpressed Myc-Plk4 in control siRNA treated cells, we observed a strong increase in the number of cells showing a severe centriole amplification (52.1±9.8%) whereas a limited percentage of CDK11 knock down cells overexpressing Plk4 were able to amplify their centrioles (16.7±0.4%). However, in these interphase-arrested cells, a strong recruitment of Plk4 could be observed (Figure 5 B, right panel). Thus, overexpressed Plk4 protein kinase is strongly recruited to interphase centrioles but cannot induce their amplification following CDK11 RNAi.

Our study reveals for the first time a role for CDK11^p58 during the centriole duplication process, and our data suggest that the protein acts in helping the recruitment of master regulators of this process at centrosomes during mitosis. The kinase activity of CDK11^p58 seems necessary for its function in the centriole duplication process, since the overexpression of the kinase-dead variant phenocopies CDK11 RNAi and inhibits centriole duplication.

In agreement with such a mitotic role, the CDK11^p58 isoform is of low abundance and generated through the use of an internal ribosome entry site on the CDK11 mRNA regulated in a mitosis specific manner, leading to this isoform being only present in mitosis [22].

To explain this new function of CDK11^p58, two hypotheses can be made (see model Figure S5). First, we have shown previously that CDK11^p58 is necessary to recruit Plk1 and γ-tubulin during centrosome maturation, which in turn can affect the recruitment of Cep192 and thus possibly Plk4 as well [24]. Indeed, Haren et al. have described that Cep192 recruitment was abolished in Plk1-depleted cells [26]. Moreover, the Cep192 worm counterpart, SPD-2, is required for ZYG-1, the related orthologue of Plk4, to be recruited at centrosomes [5], [7]. Furthermore, Plk1 and separase have been recently involved in the licensing mechanism [28], [29]. The “licensing model” suggests that centrioles (like DNA) are marked during mitosis to allow their duplication in the following interphase [2], [14], [30]. In agreement with this “licensing” hypothesis, electron microscopy studies of amplifying centrioles revealed that only the mother centrioles (generated from previous cell cycle) can be used as a template to generate daughter centrioles [6]. After down regulation of Plk1, centrioles fail to disengage and subsequently centriole duplication is blocked [28], [29]. These last results demonstrate that a complex network of proteins might be necessary to trigger the centrosome biogenesis and that a strong correlation exists between maturation of centrosomes during mitosis and their duplication in interphase (Figure S5).

The second hypothesis would be that CDK11^p58 might also mediate its role via its direct interaction with Plk4. However, CDK11^p58 associates with mitotic microtubules but is not localised at the centrioles itself and was not, so far, characterised as a core component of purified human or Drosophila centrosomes [31], [32]. This might suggest that the interaction with Plk4 occurs mostly in the cytoplasm of the cell or on spindle microtubules. Interestingly, the level of Plk4 protein kinase is tightly regulated throughout the cell cycle by ubiquitin-mediated proteolysis via SCF^Slimb in Drosophila [13], [14] and human cell lines [33]. Plk4 degradation and maintenance of appropriate protein levels in the cell requires phosphorylation of a phosphodegron motif present in the N-terminal fragment of the protein (Figure 4 A). Plk4 itself but possibly other kinases are required for phosphorylation of this important motif [33]. This phosphorylation of the phosphodegron is crucial to allow the binding of SCF^Slimb and in turn for the degradation of the protein because defective Plk4 degradation triggers Plk4 stabilisation and centriole amplification [13], [14], [33], [34]. We first hypothesised that Plk4 was a substrate of CDK11^p58 on this phosphodegron motif. Indeed, we found elevated levels of exogenous tagged-Plk4 proteins following CDK11^p58 depletion reflecting a possible stabilisation of endogenous Plk4 following CDK11 RNAi. However, the global effect following CDK11 RNAi was a centriole duplication defect and not an over duplication as expected in case of Plk4 overexpression. In addition, centrioles from CDK11-depleted cells were not able to over duplicate even if forced Plk4 overexpression was induced. This result suggests that, in CDK11-depleted cells, the centrioles lacked essential mitotic modifications allowing their subsequent duplication in S phase (Figure S5). To conclude, if Plk4 does play a role during mitosis, this possible Plk4 mitotic function cannot be compensated by elevation of Plk4 protein levels during interphase after CDK11 depletion.

To date, CDK11^p58 was suggested to play key roles at various stages of the cell division, during centrosome maturation, chromatid cohesion, and cytokinesis. Our data show for the first time a new role for CDK11^p58 during centriole duplication, probably mediated to some extent by the centrosomal direct or indirect recruitment of Plk1 and Plk4 protein kinases at centrosomes during mitosis. The emerging dual function of the Cep192, Plk1 and CDK11^p58 proteins during centrosome maturation and centriole duplication suggests that these two mechanisms and the machinery involved in these processes are closely connected [8], [26], [28]. Further studies will be required to understand the mechanism of action of CDK11^p58 to potentially validate key centrosomal proteins as substrates of this protein kinase.