Stability of the small γ-tubulin complex requires HCA66, a protein of the centrosome and the nucleolus

The centrosome constitutes a major microtubule-organizing centre in animal cells. Microtubules are nucleated and anchored at the surface of the centrosome, at the pericentriolar material. γ-Tubulin is a major component of the pericentriolar material and supports microtubule nucleation. It is in dynamic exchange with a free cytoplasmic pool (Khodjakov and Rieder, 1999), and is found in two major protein complexes (Oegema et al., 1999): a small `γ-TuSC' (γ-tubulin small complex), containing two molecules of γ-tubulin associated with one molecule each of the γ-tubulin complex proteins GCP2 and GCP3; furthermore, a large `γ-TuRC' (γ-tubulin ring complex), consisting of multiple γ-TuSCs and additional proteins, including GCP4, GCP5 and GCP6 (for a review, see Raynaud-Messina and Merdes, 2007). The amount of γ-tubulin at the centrosome is regulated during the cell cycle and increases sharply at the beginning of mitosis, when spindle formation requires an increase in microtubule nucleation activity (Zheng et al., 1991; Lajoie-Mazenc et al., 1994; Khodjakov and Rieder, 1999). After completion of mitosis, the amounts of centrosome-bound γ-tubulin are reduced to interphase levels. So far, the mechanisms that regulate γ-tubulin-dependent activity are only partly understood. Several kinases have been implicated in regulating the recruitment of γ-tubulin to the centrosome, such as Aurora A and Plk1 (Lane and Nigg, 1996; Hannak et al., 2001; Berdnik and Knoblich, 2002; Terada et al., 2003). Moreover, the cell cycle-dependent recruitment of γ-tubulin complexes depends on proteins of the pericentriolar material such as pericentrin, ninein or ninein-like protein (Takahashi et al., 2002; Casenghi et al., 2003; Chen et al., 2003; Zimmerman et al., 2004; Delgehyr et al., 2005). Whereas pericentrin recruits increased amounts of γ-tubulin complexes to the mitotic centrosome via GCP2 and GCP3, ninein and ninein-like protein anchor γ-tubulin preferably during interphase but are displaced during mitosis in a phosphorylation-dependent manner. Recruitment of γ-tubulin in mammalian cells may be further supported by a protein that associates with the γ-tubulin ring complex and that attaches to the centrosome, termed GCP-WD or NEDD1 (Lüders et al., 2006; Haren et al., 2006).

To identify novel proteins that are responsible for cell cycle-dependent regulation of γ-tubulin, we compared the composition of the pericentriolar material at different phases of the cell cycle. We characterized HCA66 as a protein of the nucleolus that associates with the centrosome specifically from S-phase to mitosis. HCA66 has initially been identified as an autoimmune antigen in hepatocellular carcinomas, and has recently been found to bind to the protein Apaf-1 of the apoptosis pathway (Wang et al., 2002; Piddubnyak et al., 2007). In the present study, we demonstrate that HCA66 is required for the stability of γ-TuSC proteins, and that silencing of HCA66 expression produces defects in centriole duplication and spindle microtubule assembly.

To study changes at the centrosome during the cell cycle, we compared the protein composition of the pericentriolar material from unsynchronized Jurkat cells (66% in G1 phase, 25% in S phase, as verified by flow cytometry), and from Jurkat cells arrested in S phase by a double aphidicolin block (85% in S phase, 10% in G1). Centrosomes were isolated after lysis using a sucrose gradient, and centrosome-containing fractions were pooled and extracted with 1 M potassium iodide to solubilize the pericentriolar material. The soluble pericentriolar material obtained from unsynchronized cells (`async') or from S-phase-arrested cells (`S') was then compared by gel electrophoresis (Fig. 1A). Bands with significantly increased intensity in S phase were investigated by MALDI-tof mass spectrometry. One of these bands (Fig. 1A) was identified as hepatocellular carcinoma-associated antigen 66 (HCA66), a protein of 597 amino acids. Database searches revealed highly homologous sequences from ESTs in mouse, Drosophila and budding yeast. Alignment of HCA66 protein sequences (supplementary material Fig. S1) showed that the N-terminal half of HCA66 (amino acids 1-202) is the most conserved region within the protein (30% identity + 31% conservative exchanges between budding yeast and human), suggesting an important role of this region for the function of the protein. Structure prediction software identified seven HAT repeats in HCA66 (Fig. 1B). These are `half-a-tetratrico-peptide' repeats with structural similarities to TPR and HEAT repeats; each repeat is predicted to form two short amphipathic α-helices connected by a loop (Preker and Keller, 1998). HAT repeats in HCA66 were found between amino acids 87-119, 121-153, 156-188, 304-335, 452-486, 488-520 and 524-557. This type of repeats is thought to be involved in protein-protein interactions.

An antibody raised against a bacterially expressed fragment of HCA66 recognized a single protein band of ∼62 kDa on immunoblots of HeLa cell lysates (Fig. 1C). Equivalent immunoblotting results were obtained in U-2 OS cells (Fig. 1C). Moreover, our antibody recognized a higher molecular weight band in lysates from U-2 OS cells overexpressing GFP:HCA66, corresponding to the GFP-tagged protein (Fig. 1C). Fractionation of cells with salt and detergent revealed that HCA66 is largely insoluble. Most of HCA66 was found in the pellet after extraction and centrifugation (Fig. 1D), and visual inspection revealed that all nuclei accumulated in these fractions. Immunofluorescence experiments with our HCA66 antibody revealed a strong staining of the nucleolus, colocalizing with the marker nucleophosmin (Fig. 1E). Consistently, proteomic analysis identified HCA66 as a nucleolar component (Andersen et al., 2005). Our immunofluorescence data further revealed one or two discrete dots in the cytoplasm that colocalized with the centrosomal marker γ-tubulin (Fig. 1E). Expression of a GFP:HCA66 fusion construct confirmed the dual localization of HCA66 at the nucleolus and at the centrosome (Fig. 1E). We then tested by microscopy whether the association of HCA66 with the centrosome is cell cycle dependent, as indicated by our biochemical data on purified centrosomes (Fig. 1A). We found that U-2 OS cells that were synchronized in S phase and that were pulse labelled with bromo-deoxyuridine displayed HCA66 localization at the centrosome in 91% (±5, n=250) of the cells, whereas in cultures synchronized in G1 phase only 24% (±7, n=250) of the cells showed detectable centrosomal staining (Fig. 2A). This suggests that HCA66 localizes to the centrosome in a cell cycle-dependent manner, and that most of HCA66 at the centrosome is recruited at the G1-S transition. HCA66 remains at the centrosome until metaphase. From anaphase onwards, centrosomal localization of HCA66 is lost, and in telophase HCA66 relocalizes to the nucleoli (Fig. 2B). Experiments using nocodazole indicated that centrosome localization does not depend on polymerized microtubules (Fig. 2C). Moreover, HCA66 does not bind to acentriolar microtubule asters induced by taxol treatment of mitotic cells (Fig. 2C). Taken together, these results suggest that HCA66 is a bona fide centrosomal protein. Deconvolution microscopy at high magnification revealed that centrosomal HCA66 in S phase is mainly concentrated in an area between the diplosomes, but not directly associated with the centrioles (Fig. 2D). Besides centrosomal staining, HCA66 also displays staining of the perichromosomal layer (Fig. 2B), an area where a subset of nucleolar proteins localizes during mitosis (van Hooser et al., 2005). Accordingly, HCA66 was detected in the proteome analysis of the chromosome scaffold (Gassmann et al., 2005).

To map domains of HCA66 that mediate centrosomal and nucleolar targeting, we generated deletion constructs tagged with GFP and examined their distribution in U-2 OS cells. GFP:HCA66_151-597, which lacks the most conserved N-terminal region, displayed diffuse cytoplasmic and nuclear staining (Fig. 3A). To verify whether the N terminus of HCA66 is sufficient for centrosomal and nucleolar targeting, we generated GFP:HCA66_1-149. Consistently, GFP:HCA66_1-149 localized both to the nucleolus and the centrosome, although the centrosome staining appeared to be very weak (Fig. 3A). A smaller fusion protein encoded by GFP:HCA66_1-86 was absent from the nucleus, but colocalized with γ-tubulin at the centrosome (Fig. 3A), indicating that the first 86 amino acids of HCA66 are sufficient to mediate centrosomal targeting. When expressed at low or moderate levels, GFP:HCA66_1-86 localized to the centrosome to a similar degree in G1- and S-phase-arrested cells (in 60% of cells blocked in G1 with 1 mM mimosine for 24 hours, or in 57% of cells released into S-phase from thymidine block, respectively). Moreover, GFP:HCA66_1-86 was found at the spindle poles in mitosis (Fig. 3A). This indicated that the centrosome localization of this HCA66 fragment is not cell cycle dependent. The cell cycle-dependent localization of endogenous, full-length HCA66 is thus probably regulated outside the region of amino acids 1 to 86.

In a following step, we wanted to test whether HCA66 binds to any previously characterized centrosome component. Efforts to identify HCA66 interactors by immunoprecipitation and biochemical methods were fruitless, due to the high insolubility of HCA66 and its deletion mutants throughout the cell cycle. We therefore investigated the effect of overexpression of full-length GFP:HCA66, GFP:HCA66_151-597 and GFP:HCA66_1-86 on the localization of centrosomal proteins. Experiments with GFP, GFP-tagged full-length HCA66, or GFP:HCA66_151-597 had no visible effect on centrin, PCM-1, or γ-tubulin (Fig. 3A,B; supplementary material Fig. S2A). However, high overexpression of GFP:HCA66_1-86 led to reduced centrosomal localization of γ-tubulin in 75% of asynchronous interphase cells, and induced the formation of cytoplasmic aggregates (Fig. 3B; supplementary material Fig. S2A). Closer inspection of cells synchronized either in G1 or in S-phase revealed that the numbers of cells lacking centrosomal γ-tubulin was similar in both cases, 74% of cells in G1 and 68% of cells in S-phase, after high overexpression of GFP:HCA66_1-86. We found that increasing amounts of GFP:HCA66_1-86 at the centrosome lowered the immunofluorescence signal of γ-tubulin at the centrosome below detection level (Fig. 3B, graph, right). The same HCA66 fragment also had a strong effect on the localization of the γ-tubulin complex proteins GCP2 and NEDD1, and weaker effects on PCM-1 and centrin (Fig. 3B; supplementary material Fig. S2A). Consistent with loss of γ-tubulin complex proteins from the centrosome, cells overexpressing GFP:HCA66_1-86 were defective in microtubule re-growth from a centrosomal organizing centre after cold-induced depolymerization (Fig. 3C). Whereas 88% of control cells expressing GFP, or 75% of cells expressing GFP-tagged full-length HCA66 re-grew microtubules within two minutes after recovery, only 35% of the cells expressing GFP:HCA66_1-86 were able to form microtubule asters during this time. Following prolonged incubation to five minutes, centrosomal microtubule asters grew in 46% of GFP:HCA66_1-86-expressing cells (Fig. 3C, graph). As a consequence of expressing high levels of GFP:HCA66_1-86 for two days, the cell cycle was affected, yielding only 7% of diploid cells in G1, 11% in S-phase, 20% in G2, and more than 50% aneuploid cells, as seen by flow cytometry (supplementary material Fig. S2B). Whereas mitotic figures were still visible in 0.5% of the cells after 12 hours, no mitotic cells could be identified any more by immunofluorescence from 24 hours onwards, suggesting that highly overexpressed GFP:HCA66_1-86 arrests the cell cycle. Immunoblot analysis of cells sorted for GFP fluorescence revealed that the amounts of γ-tubulin were unchanged in cells transfected with GFP:HCA_1-86 as compared to controls, indicating that the observed reduction of γ-tubulin at the centrosome was due to displacement of the protein (supplementary material Fig. S2C). GFP:HCA66_1-86 did not displace endogenous HCA66 from the centrosome or from the nucleolus (Fig. 3B).

Because full-length HCA66 localizes to the centrosome in a cell cycle-dependent manner, we wanted to test whether HCA66 is involved in any aspect of centrosome function. For this reason, we performed RNA-silencing experiments using two different oligonucleotides against HCA66 (Fig. 4A). Treatment with the oligonucleotide HCA-4 allowed reproducible depletion of 75% or more of HCA66 protein after 48 hours, as determined by serial dilution and blot densitometry (data not shown). Silencing of HCA66 inhibited the duplication of centrioles, as two or less centrin signals were seen in 50% of the depleted cells during mitosis (n=80) (Fig. 4B, graph), in contrast to controls that showed four centrioles in 70% of all mitotic cells. Consistently, silenced U-2 OS cells that were arrested with hydroxyurea failed to re-duplicate centrioles efficiently (38% of the treated cells), whereas 69% of control cells showed four or more centrioles (Fig. 4C). Overexpression of GFP-tagged full-length HCA66, however, did not alter centriole numbers. Cells lacking HCA66 showed aberrant microtubule organization in mitosis, mostly in the form of monopolar spindles, and failure of chromosome alignment (Fig. 4D). Chromatin condensation and immunolabelling of phosphorylated histone H3 suggested that these cells were in prometaphase (data not shown).

Depletion of HCA66 led to an increase of mitotic figures after 48 hours, from ∼5% in controls to ∼13.4% in depleted cells, suggesting a delay in mitosis (Fig. 4E). Further analysis showed that 80% of the mitotic cells were accumulating in prometaphase, compared with ∼55% of mitotic cells in controls (Fig. 4E), with the vast majority containing monopolar spindles. We also observed an increased number of cells with micronuclei upon HCA66 siRNA treatment (31%, compared with 12% in controls) (Fig. 4F), suggesting that depletion of HCA66 led to chromosome segregation defects. Prolonged siRNA treatment up to 96 hours significantly reduced the number of cells (Fig. 4G) and led to an almost complete disappearance of mitotic figures (<0.7%), indicating that HCA66 is essential for viability.

The defects that we observed upon siRNA treatment against HCA66, e.g. monopolar spindle formation or failure of centriole duplication, were reminiscent of the defects obtained from γ-tubulin depletion (Sunkel et al., 1995; Strome et al., 2001; Hannak et al., 2002; Dammermann et al., 2004; Haren et al., 2006). As overexpression of GFP:HCA66_1-86 affects centrosomal γ-tubulin (Fig. 3B), we decided to investigate the behaviour of proteins of the γ-TuRC in cells treated with HCA66 siRNA. In interphase, γ-tubulin at the centrosome was reduced to 45% of the respective control levels (Fig. 5A,B). In mitotic control cells, γ-tubulin at the centrosome increased fourfold compared with interphase, in good agreement with findings by Khodjakov and Rieder (Khodjakov and Rieder, 1999). However, in HCA66-depleted mitotic cells, centrosome-bound γ-tubulin was drastically reduced to 7% of mitotic control levels (Fig. 5A,B), implying that HCA66 activity might be particularly important during centrosome maturation. Although the formation of centrosomal microtubule asters after depolymerization and regrowth was delayed in HCA66-depleted cells (supplementary material Fig. S3A), photometric analysis of interphase and mitotic cells revealed that the overall amount of microtubule polymer was not significantly affected prior to depolymerization (Fig. 5C). Thus, the reduction of γ-tubulin at the centrosome after HCA66 depletion does not significantly alter steady state levels of microtubule polymer in our cells. This is consistent with Strome et al. (Strome et al., 2001) and Hannak et al. (Hannak et al., 2002), who showed that microtubules can form despite depletion of γ-tubulin, although their nucleation from the centrosome is kinetically disadvantaged. In addition to γ-tubulin, the immunofluorescence signals of GCP2, GCP4 and Nedd1/GCP-WD at the centrosome were reduced after HCA66 siRNA (Fig. 5A,B). Other proteins of the centrosome and of the spindle pole such as pericentrin, centrin, TPX2, Aurora A or Plk1 were not significantly affected (Fig. 5A,B; supplementary material Fig. S3B), suggesting that HCA66 siRNA affects specifically the centrosomal localization of γ-TuRC proteins.

Subsequently, immunoblot analysis was performed to distinguish between problems of centrosomal recruitment of these proteins and reduction in protein amounts. Fig. 6 shows that depletion of HCA66 led to a decrease of the protein levels of γ-tubulin, GCP2 and GCP3. These three proteins are known to form the γ-TuSC. HCA66 and the γ-TuSC protein levels were reduced to a comparable degree after HCA66 siRNA, to around 30 to 40% of control levels (graph, Fig. 6). Moreover, the levels of GCP2 were also reduced following depletion of γ-tubulin, suggesting that the expression of γ-TuSC components is interdependent (Fig. 6). By contrast, silencing of γ-tubulin did not reciprocally affect the levels of HCA66. Experiments involving reverse transcription of RNA, followed by PCR for γ-tubulin, GCP2 and GCP3 showed that transcription of γ-TuSC genes was unaffected by HCA66 silencing, indicating that HCA66 is involved in regulating the protein levels but not the mRNA of γ-TuSC components (supplementary material Fig. S3C). Whereas silencing of HCA66 diminished the amounts of all γ-TuSC components, protein levels of γ-TuRC-specific components such as GCP4 or Nedd1/GCP-WD, or of the pericentriolar protein PCM-1 were not reduced. Moreover, the levels of the kinases Aurora A and Plk1 also remained constant (Fig. 6). Because Aurora A and Plk1 still localize to the centrosome in the absence of HCA66 (Fig. 5; supplementary material Fig. S3B), we conclude that HCA66 is not involved in regulating the expression, the stability or the recruitment to the centrosome of these mitotic kinases.

Because HCA66 showed similarities to the yeast pre-ribosome assembly factor Utp6p, and because depletion of HCA66 reduced the levels of γ-TuSC proteins, we reasoned that translation of their mRNA might have been inhibited. Alternatively, we reasoned that the γ-TuSC proteins might be more susceptible to degradation in the absence of HCA66. To test these ideas, we compared the effects of the translation inhibitor cycloheximide and of the proteasome inhibitor MG132 with the phenotypes obtained after silencing of HCA66. We noticed that the phenotypes produced by these two drugs differed from the effects of HCA66 siRNA: more than 90% of the cells treated with either inhibitor still contained γ-tubulin at the centrosome after 2 days, and we failed to observe any accumulation of monopolar spindles in these treated cells (supplementary material Fig. S3D). We therefore conclude that the phenotype of HCA66 depletion is neither due to a general block of translation, nor due to an unspecific inhibition of proteasome-dependent degradation.

We characterize HCA66 as a novel component of the nucleolus that localizes to the centrosome in a cell cycle-dependent manner. We demonstrate that HCA66 is necessary for centriole duplication and bipolar spindle assembly. Furthermore, we show that HCA66 plays a role in regulating the protein levels of the γ-TuSC components γ-tubulin, GCP2 and GCP3. By contrast, HCA66-depletion does not affect the protein levels of γ-TuRC-specific proteins such as GCP4 or GCP-WD/NEDD1. Nevertheless, these proteins are found less concentrated at centrosomes in mitosis, potentially because their proper recruitment to the pericentriolar material requires fully assembled γ-TuRCs and thus depends on the presence of γ-TuSCs (Raynaud-Messina and Merdes, 2007). Even though HCA66 is necessary for γ-tubulin localization to the centrosome both in interphase and mitosis, depletion of HCA66 affects centrosomal amounts of γ-tubulin most drastically in mitosis. This correlates with the cell cycle-specific localization of HCA66, which binds to the centrosome only between S-phase and metaphase of mitosis, raising the possibility that HCA66 regulates γ-tubulin complex proteins while centrosome bound.

Because failure in centriole duplication has previously been described to result from the depletion of γ-tubulin (Dammermann et al., 2004; Haren et al., 2006), we think that the centriole defects seen after removal of HCA66 are probably due to the loss of γ-tubulin. Likewise, monopolar spindle formation in HCA66-depleted cells can be explained by the loss of functional γ-tubulin or γ-TuSCs, as seen in knock-down experiments or mutants in various species (Sunkel et al., 1995; Barbosa et al., 2000; Strome et al., 2001; Hannak et al., 2002; Barbosa et al., 2003; Raynaud-Messina et al., 2004; Yuba-Kubo et al., 2005; Colombié et al., 2006; Haren et al., 2006). It is likely that the spindle defects in HCA66-depleted cells lead to chromosome segregation defects and micronuclei, explaining the increased cell death observed in HCA66-depleted cultures, but we cannot exclude the possibility that HCA66 plays additional roles in interphase that are important for the survival of the cells.

So far, the significance of the nucleolar localization of HCA66 during interphase remains unclear. Interestingly, several other nucleolar proteins have been described that also fulfil roles at the mitotic spindle or at the centrosome, such as NuSAP and nucleophosmin. NuSAP localizes to the spindle during mitosis and crosslinks microtubules (Raemaekers et al., 2003; Ribbeck et al., 2006). Nucleophosmin binds to the centrosome and is implicated in controlling centriole duplication (Okuda et al., 2000). In addition, nucleophosmin participates in ribosome biogenesis (Frehlick et al., 2007). At the molecular level, nucleophosmin acts as a chaperone, potentially preventing protein aggregation in the nucleolus and assisting nucleo-cytoplasmic shuttling and protein assembly (Szebeni and Olson, 1999; Yu et al., 2006). Likewise, HCA66 may play a dual role at the centrosome and in ribosome assembly, similar to nucleophosmin. This would be consistent with our finding that HCA66 shows homology with the budding yeast protein Utp6p, a protein that is part of the 90S pre-ribosomal particle (Dragon et al., 2002; Dosil and Bustelo, 2004). Depending on the cell cycle, HCA66 may associate with various protein complexes, such as pre-ribosomal particles or γ-TuSCs. Our data indicate that overexpression of the N-terminal region comprising amino acids 1 to 86 of HCA66 displaces γ-tubulin from the centrosome, arguing for an interaction between HCA66 and γ-tubulin. Because overexpression of this N-terminal fragment does not displace endogenous HCA66 and does not alter γ-tubulin protein levels, this fragment probably exerts a dominant effect on centrosome integrity. Morover, this HCA66 fragment binds to the centrosome in a constitutive manner throughout the cell cycle, whereas the full-length, endogenous HCA66 protein only accumulates at the centrosome between S-phase and mitosis. This suggests that the localization of HCA66 to the centrosome is regulated outside its N-terminal region, and that the overexpressed short fragment of amino acids 1 to 86 behaves abnormally. Yeast two-hybrid and biochemical assays failed to reveal binding of HCA66 to γ-TuSC proteins (data not shown), leading us to conclude that HCA66 and γ-tubulin may interact indirectly, or in a regulated, transient manner that is not reproduced in vitro. When trying to identify binding partners of HCA66 in biochemical and immunoprecipitation assays, we noticed that HCA66 was highly insoluble. This complicated further investigation of regulated transient interactions or low-affinity interactions with potential binding partners.

We speculate that a possible mechanism by which the γ-TuSC proteins might be regulated may include a chaperone activity by HCA66, analogous to nucleophosmin function. Because no direct binding of HCA66 to γ-TuSC proteins was seen, such an activity might involve additional partner proteins. HCA66 could either protect the entity of the γ-TuSC, or protect individual γ-TuSC components before assembly. Loss of individual γ-TuSC proteins might affect the translation of their partners via feedback or, alternatively, monomeric γ-complex proteins that fail to assemble into γ-TuSCs might be degraded more rapidly than the assembled ones. These two scenarios would explain the specific loss of all γ-TuSC proteins after silencing of HCA66 expression, and they explain our observation of reduced GCP2 levels in cells in which γ-tubulin was silenced. Other chaperone proteins are known that interact with γ-tubulin and bind to the centrosome, including TCP1, UXT and co-factor D (Melki et al., 1993; Zhao et al., 2005; Cunningham and Kahn, 2008). Interestingly, co-factor D contains a series of HEAT sequence repeats, reminiscent of the HAT repeats found in HCA66, and silencing also results in spindle pole defects (Cunningham and Kahn, 2008). However, silencing of co-factor D does not alter γ-TuSC protein levels.

We propose that HCA66-dependent stability of γ-TuSC proteins represents a novel element of the complex regulatory machinery that controls microtubule nucleation, centrosome duplication and mitotic spindle assembly. In the future, more knowledge on the potential interactors of HCA66 will be needed to understand the exact molecular mechanisms.

HCA66 cDNA was generated by reverse transcription of RNA isolated from Jurkat cells. Full-length clones were prepared by PCR, using KOD DNA polymerase (Novagen) and primers CCGGGGTACCATGGCAGAGATAATTCCAGGA (HCA66fwd) and CGCGGATCCTAAATGGCCAGTCTGATGCA (HCA66rev). pRSET-HCA66_87-366 was generated by cutting full-length HCA66 with EcoRI and HindIII, and cloning it into pRSET-C (Invitrogen). GFP:HCA66 was generated by cloning full-length HCA66 into KpnI and BamHI sites of pEGFP-C1 (Clontech). GFP:HCA66_1-86 and GFP:HCA66_1-149 were generated by PCR using HCA66fwd with AAACTGCAGTCAATTCTCTCAATCTCATCCTTCTT or AAACTGCAGTTGGCTGCCATAATCCACAA, respectively. The PCR products were cloned into KpnI and XhoI sites of pEGFP-C1. GFP:HCA66_151-597 was generated by PCR using primers AGAATTCAATGGAAGATCGATTGTCTTC and HCA66rev. The PCR product was cloned into EcoRI and BamHI sites of pEGFP-C1.

6×His-HCA66_87-366 fusion protein was expressed in E. coli and affinity purified on nickel agarose beads under denaturing conditions. The eluted protein was used for antibody production in rabbits. Other primary antibodies used in this study were: mouse anti-alpha-tubulin (DM1A, Sigma-Aldrich), anti-γ-tubulin (mouse GTU-88 or rabbit AK-15, Sigma-Aldrich), mouse anti-actin MAB1501 (Chemicon), mouse anti-pericentrin, rabbit anti-PCM-1 (Dammermann and Merdes, 2002), mouse anti-centrin 20H5 (gift from Dr J. Salisbury, Mayo Clinic, Rochester, MN), rabbit anti-GCP2 (gift from Dr T. Stearns, Berkley, CA), rabbit anti-GCP4 (Fava et al., 1999), rabbit anti-GCP3 (gift from Dr M. Bornens, Paris, France), rabbit anti-Nedd1 (Haren et al., 2006) and rabbit anti-CPAP (against a bacterially expressed, GST-tagged human CPAP fragment containing amino acids 1 to 295).

U-2 OS cells were cultured in Dulbecco's modified Eagle's medium. Jurkat cells were cultured in RPMI 1640. All media were supplemented with 10% foetal calf serum, 2 mM L-glutamine, 50 IU penicillin and streptomycin. For immunofluorescence experiments, cells were synchronized by mitotic shake-off and replated for further 2 hours to enter G1 phase. Cells in S-phase were synchronized by a double thymidine block and released for 3 hours, before addition of BrdU. Synchronization was verified by immunofluorescence of BrdU (rat anti BrdU, Harlan Scientific), indicating 17.5±2.5% of BrdU-positive cells in G1 populations versus 83.8±5.0% in S-phase populations. For centrosome purification, cells were arrested in S-phase using a double aphidicolin block: 1 μg/ml of aphidicolin was added to the culture medium for 16 hours, washed off and the cells were grown for 9 hours under normal conditions. A second block was then performed for another 16 hours. The percentage of cells in S phase was determined by flow cytometry. For this, cells were washed in cold PBS and fixed in cold 70% ethanol. Cells were then washed and incubated for 30 minutes at 37°C with RNase A (10 μg/ml). Propidium iodide (40 μg/ml) was added to the cells before analysis with a FACSCalibur instrument and CellQuest software (Becton Dickinson). To assay for centriole overduplication, U-2 OS cells were treated first for 12 hours with HCA66 siRNA or control RNA, then 2 mM hydroxyurea was added and incubation continued for additional 40 hours.

GFP constructs were transfected using FuGene-6 (Roche) according to the manufacturer's instructions. Cells were fixed 48 hours after transfection. Double-stranded siRNA oligomers were transfected into U-2 OS cells using a Nucleofector apparatus, program U-24 and nucleofection solution V (Amaxa), according to the manufacturer's instructions. Two siRNAs targeting HCA66 mRNA were used (Xeragon). Results presented here correspond to the targeting of nucleotides 421-432 (HCA-4; CCAGCUUUGUGGAUUAUGGdTdT). Targeting of nucleotides 965-983 (HCA-2; CAGAGGCCAUGUGGAAGUGdTdT) induced similar depletion levels and cellular phenotypes. Control depletion was carried out using oligomers targeting Luciferase. Depletion of γ-tubulin was performed as described (Haren et al., 2006).

Cells grown on glass coverslips were transferred into pre-cooled medium on ice, then into pre-warmed medium at 37°C. Regrowth was stopped by methanol fixation.

Gel electrophoresis and immunobloting were performed using standard protocols. Serum against HCA66 was used at 1:1000 dilution. For immunofluorescence, cells were grown on glass coverslips, fixed with methanol at –20°C, and processed using standard protocols.

The content of microtubule polymer in cells was determined by adapting a protocol of Zhai et al. (Zhai et al., 1996). Briefly, cells were pre-extracted with 0.2% Triton X-100 in PHEM (60 mM PIPES, 25 mM HEPES, 1 mM EGTA, 2 mM MgCl₂) at 37°C. Subsequently, cells were fixed with 4% formaldehyde in the same buffer supplemented with 1 μM taxol, and stained for immunofluorescence of α-tubulin. The amount of microtubule fluorescence was quantified in the whole cell after background subtraction. Images were acquired with an Axiocam camera on an Axioskop 2 microscope (Carl Zeiss) using a 100× NA 1.30 objective, or with a COOLSNAP HQ ICX285 camera (Roper Scientific) on an OLYMPUS IX-70 microscope controlled by DeltaVision Softworx (Applied Precision), using a 100× NA 1.40 objective. After deconvolution, image files were projected using the maximum intensity function. Image processing was carried out using Photoshop (Adobe).

Centrosomes were purified from Jurkat cells as described previously (Bornens et al., 1987). Centrosome fractions were assayed for their ability to stimulate aster formation, by incubation with concentrated mitotic frog egg extract or with pure porcine brain tubulin. In both cases, rhodamine-labelled tubulin was added to visualize microtubules by fluorescence microscopy. Purified centrosomes were incubated in 1 M KI at 4°C in the dark, then centrifuged at 120,000 g for 30 minutes at 4°C. The KI-soluble material was then concentrated and filtered using a Centricon YM-10 (Millipore) device. The retained proteins were recovered, boiled for 5 minutes in protein sample buffer and stored at –80°C until loading onto 7.5% Tris-glycine polyacrylamide gels. Protein bands were visualized by silver staining. Individual bands were cut and digested with Trypsin (Promega) (Shevchenko et al., 1996). The MALDI-tof mass spectra were acquired on a PerSeptive Biosystems Voyager DE STR instrument (Applied Biosystems) and analyzed using MS-Fit tool (http://prospector.ucsf.edu/ucsfhtml4.0/msfit.htm).