Control of Aurora-A stability through interaction with TPX2

The Aurora-A kinase (officially known as Serine/threonine-protein kinase 6) is an important regulator of cell division and acts in several aspects of spindle formation and function (for reviews, see Barr and Gergely, 2007; Vader and Lens, 2008). The levels of Aurora-A are abnormally high in many tumour types, and altered abundance is thought to be of relevance for oncogenic transformation (for reviews, see Gautschi et al., 2008; Vader and Lens, 2008). Indeed, Aurora-A has recently been proposed as a potential target in anti-cancer therapy; inhibitors of its activity have been synthesised, some of which are currently being tested in clinical trials (for reviews, see Gautschi et al., 2008; Pérez de Castro et al., 2008; Karthigeyan et al., 2010).

In recent years, several factors have been shown to interact with Aurora-A and to modulate its activity (for a review, see Carmena et al., 2009), among which the microtubule (MT)-binding protein TPX2 (‘targeting protein for Xklp2’) has a prominent role (Gruss and Vernos, 2004). The mechanism through which TPX2 activates Aurora-A has been extensively characterised: TPX2 binding induces a conformational change in Aurora-A, modifying the position of a key residue (Thr288) in the three-dimensional structure of the kinase and rendering it inaccessible to the PP1 phosphatase. Thus, TPX2 binding stabilises Thr288-phosphorylated Aurora-A, which represents the active form of the kinase (Bayliss et al., 2003; Eyers et al., 2003; Tsai et al., 2003). TPX2 is also required to target Aurora-A to the spindle MTs (Kufer et al., 2002; De Luca et al., 2006); a truncated form of TPX2 lacking the first 33 amino acids and unable to bind to Aurora-A cannot restore Aurora-A localisation to MTs in a TPX2-deficient background (Bird and Hyman, 2008). Despite of the well-documented effects of TPX2 on Aurora-A, functional studies in Xenopus extracts (Brunet et al., 2004; Tsai and Zheng, 2005; Sardon et al., 2008) and human cells (Bird and Hyman, 2008) have not yet provided a complete understanding of the actual role of the complex in spindle assembly and mitotic control.

At the end of mitosis, the abundance of Aurora-A is downregulated through APC/C–Cdh1-dependent proteasome-mediated proteolysis (Taguchi et al., 2002; Lindon and Pines, 2004); this downregulation is important for the organisation of the anaphase spindle (Floyd et al., 2008). The molecular determinants of Aurora-A degradation have been characterised: a canonical destruction box (D-box) in the C-terminal region and a novel motif in the N-terminus (A-box) are required for APC/C–Cdh1-dependent destruction of human Aurora-A. The phosphorylation state of a serine residue (Ser51) in the A-box modulates degradation of Aurora-A as mutants mimicking constitutive phosphorylation of this site cannot be degraded by the APC/C–Cdh1 (Crane et al., 2004). Recent studies suggest that the PP2A phosphatase is responsible for dephosphorylation of Ser51 (Horn et al., 2007). Abnormal phosphorylation of Ser51 has been observed in head and neck cancer, suggesting a link between control of the stability of Aurora-A and tumorigenesis (Kitajima et al., 2007). Interestingly, addition of the Aurora-A-binding region of TPX2 to Xenopus oocyte extracts impairs APC/C–Cdh1-dependent degradation of Aurora-A (Sardon et al., 2008).

Over the past few years, we contributed to the demonstration that both Plk1 (at centrosomes) and TPX2 (at MTs) regulate the localisation of Aurora-A in mitotic cells (De Luca et al., 2006). We now report that TPX2 is also required for regulation of the stability of Aurora-A protein in human cells: we show that Aurora-A protein levels decrease in cells lacking TPX2, in a proteasome- and Cdh1-dependent manner, and that the interaction between Aurora-A and TPX2 is required for protecting Aurora-A from degradation. This novel mechanism of Aurora-A regulation is relevant to the kinetics of accumulation and disappearance of Aurora-A during the cell cycle and hence for the proper execution and exit from mitosis; in addition, it might indicate poorly explored routes to increased kinase abundance in tumours.

In previous experiments, we noticed that the Aurora-A signal decreased in cells silenced for TPX2 by RNA interference (RNAi) (TPX2i) compared with that of controls, possibly hinting at a novel level of control exerted by TPX2 on the abundance of Aurora-A protein. That prompted us to design a series of experiments to substantiate that observation. First, we quantified the Aurora-A-specific immunofluorescence (IF) signal in U2OS single mitotic cells with or without TPX2 after RNAi. Given that TPX2-defective cells reach prometaphase – the stage at which Aurora-A levels are highest – but cannot progress any further (Garret et al., 2002; Gruss et al., 2002; De Luca et al., 2006), we limited our analysis to prometaphase cells from control (GL2i) and TPX2i cultures (examples are shown in supplementary material Fig. S1). Placing selections in different areas of the cells (Fig. 1A), we observed that (i) the intensity of Aurora-A signal at spindle poles was lower in TPX2i compared with control cells, and this was not associated with a decrease in MT density at poles; (ii) there was no concomitant increase in the Aurora-A cytoplasmic fraction; consistent with this, (iii) the total amount of Aurora-A was reduced by ~50% compared with control prometaphases. Comparable results were obtained using independent TPX2-targeting small-interfering RNA (siRNA) oligonucleotides: Fig. 1A shows results obtained with TPX2₁₄₄, used throughout the paper; results obtained with TPX2₁₆₈, previously characterised by Bird and Hyman (Bird and Hyman, 2008), are shown in supplementary material Fig. S2. Thus, inactivation of TPX2, in addition to preventing Aurora-A localisation at MTs, also prevents its accumulation to high levels.

To confirm that conclusion, we measured the overall abundance of Aurora-A protein by western blot (WB) analysis in control and TPX2-silenced U2OS cultures. To obtain comparable cell populations, GL2i cultures were treated with monastrol (MON), which inhibits the kinesin Eg5 and arrests mitotic cells in prometaphase with monopolar spindles but unaltered MT dynamics (supplementary material Fig. S1A), with no effect on Aurora-A levels (supplementary material Fig. S3A); U2OS mitotic cells were collected by shake-off from monastrol-treated and TPX2i cultures, and virtually all were found to be in prometaphase (supplementary material Fig. S3B), with comparable expression of cyclin B1 (Fig. 1B and supplementary material Fig. S1B). The abundance of Aurora-A selectively decreased in the TPX2i samples, and this was specific for Aurora-A, whereas Plk1, which undergoes similar accumulation–degradation kinetics during mitosis (Lindon and Pines, 2004), was unaffected (Fig. 1B). The levels of Aurora-A similarly decreased in non-transformed hTERT-RPE1 epithelial cells silenced for TPX2 (data not shown). To ensure that the decrease in Aurora-A levels did not arise as a consequence of defective MT assembly in TPX2i prometaphase cells (Gruss et al., 2002), we repeated the Aurora-A WB analysis in TPX2i cultures with taxol-stabilised MTs (typical cells are shown in supplementary material Fig. S1A, and the distribution of mitotic stages is shown in supplementary material Fig. S3B): no difference in Aurora-A levels was observed in untreated or taxol-treated TPX2i cultures (Fig. 1B), indicating therefore that the Aurora-A decrease observed in the absence of TPX2 takes place regardless of the amount of stable MTs.

In the reciprocal experiment, we analysed Aurora-A-silenced (Aurora-Ai) mitoses (see mitotic progression in supplementary material Fig. S3B) and found that the levels of TPX2 were unchanged (Fig. 1C): this finding indicates that removing Aurora-A does not influence the abundance of TPX2 in prometaphase cells and that we are not observing a general destabilising effect caused by the disassembly of a multiprotein complex.

We next wanted to rule out the possibility that the decreased abundance of Aurora-A was in fact a side-effect of the abnormally long prometaphase arrest induced by TPX2 RNAi in asynchronous cultures. To test this, we combined the RNAi procedure with a synchronisation protocol (Fig. 1D and supplementary material Fig. S3B) and followed TPX2-silenced cells as they entered mitosis, then collected mitotic cells by shake-off from these and control cultures. The mitotic phase distribution and cyclin B1 levels were both comparable under these conditions (supplementary material Fig. S3B and Fig. 1D, respectively). Aurora-A levels were again downregulated in cultures lacking TPX2 analysed soon after reaching prometaphase arrest (Fig. 1D). Together, these results show that loss of TPX2 induces a decrease in the abundance of Aurora-A in prometaphase. We conclude that the abundance of Aurora-A is normally modulated in early mitotic stages, thus uncovering a novel level of control that requires TPX2.

We wondered whether the decrease in Aurora-A levels in TPX2i prometaphase cells involves proteasome-dependent degradation. To address this, we inhibited the proteasome activity using MG132 (Fig. 2A). Aurora-A levels were comparable in mitoses from thymidine-presynchronised cultures, either untreated or treated with monastrol or MG132 alone (Fig. 2A, left panels), indicating that inhibiting the proteasome in prometaphase per se has no major effect on Aurora-A. Remarkably, however, MG132 counteracted the quantitative decrease of Aurora-A in TPX2i cultures (Fig. 2A, right panels), suggesting that loss of TPX2 function entails proteasome-dependent degradation of Aurora-A in prometaphase.

In a normal cell cycle, Aurora-A is degraded in anaphase in an APC/C–Cdh1-dependent manner. The bulk of Cdh1 is active after the metaphase-to-anaphase transition, but a fraction of Cdh1 is reported to be active in earlier mitotic stages (reviewed in Baker et al., 2007). Indeed, simultaneous RNAi to TPX2 and Cdh1 restored Aurora-A to levels comparable to those seen in control U2OS prometaphase cells (Fig. 2B). No effect was observed when we depleted Cdh1 alone. These experiments indicate that TPX2 protects Aurora-A from Cdh1-mediated proteasome-dependent degradation in early mitosis.

In human cells, the onsets of accumulation of TPX2 and Aurora-A were found to fall within S phase (Gruss et al., 2002) and G2 phase (Bischoff et al., 1998), respectively. These kinetics would be consistent with a requirement for TPX2 function for accumulation of Aurora-A during interphase. To investigate this more directly, we synchronised U2OS cultures by thymidine treatment, then analysed cells after release from synchronisation in parallel WB assays. A weak TPX2 signal was already present at the time of release (G1–S boundary), which increased at 6 hours post-release, when cells were progressing through S phase, whereas Aurora-A was detected 12 hours after release, when most (>60%) cells were in G2 and M phases and cyclin B1 levels were highest (Fig. 3A, left panels). IF analysis (Fig. 3A, right panels) confirmed that TPX2-positive cells are already present at the time of release from thymidine arrest, with an increased signal intensity thereafter. By contrast, Aurora-A was barely detectable at the time of release, with the earliest detectable signals found around centrosomes in isolated cells 6 hours after release (arrowed in Fig. 3A), indicating that the earliest pool of newly synthesised Aurora-A is recruited promptly at centrosomes. Twelve hours after release, Aurora-A signals accumulated both around centrosomes and within nuclei (Fig. 3A); high TPX2 levels were detected in these same cells. We wondered whether TPX2 inactivation would also influence Aurora-A accumulation in interphase cells during G2 (i.e. adherent cells with an intact nuclear envelope examined at 12 hours after thymidine release). Both WB and IF analyses showed that inactivation of TPX2 by RNAi hinders the accumulation of Aurora-A (Fig. 3B), while affecting neither cell cycle progression (60.1±4.1% G2–M cells in GL2i cultures and 65.7±0.7% in TPX2i by FACS analysis) nor the levels of cyclin B1 (Fig. 3B). Thus, TPX2 exerts a protective effect against Aurora-A degradation already in interphase and contributes to its accumulation during G2 phase.

TPX2 binds to and activates Aurora-A through its N-terminal 43 amino acids (Bayliss et al., 2003). Furthermore, amino acids 1–33 of TPX2 are required for localisation of Aurora-A to spindle MTs in human cells (Bird and Hyman, 2008). We wanted to establish whether the formation of the Aurora-A–TPX2 complex is required for control of Aurora-A stability. To address that question, we used constructs encoding fluorescently tagged TPX2-deleted versions (Fig. 4A; for details, see the Materials and Methods section). Full-length TPX2 was encoded by pEGFP-TPX2res, expressing an RNAi-resistant transcript (Fig. 4A). That construct was compared with pEGFP-TPX2res/Δ43, also RNAi resistant, encoding a TPX2 version lacking the first 43 amino acids and thus unable to bind to Aurora-A (Bird and Hyman, 2008), and with pEYFP-TPX2/1–43, coding for the first 43 amino acids of TPX2 only and not including the siRNA target sequence. In a TPX2-silenced background, exogenously expressed TPX2res and TPX2res/Δ43 did both accumulate at spindle poles, as expected, whereas TPX2/1–43, which lacks the MT-binding regions of TPX2, was distributed diffusely in the mitotic cytoplasm (Fig. 4A). We then were in a position to investigate whether these TPX2 mutants would affect the localisation and stability of Aurora-A by quantifying both total and pole-associated Aurora-A IF signals, as done for Fig. 1 (Fig. 4B). Expression of full-length TPX2 (TPX2res) restored normal levels of Aurora-A compared with TPX2-silenced cells expressing empty vector (GFP lanes), both at poles and MTs (POLES/MTs) and in the whole cell. By contrast, TPX2res/Δ43 failed to stabilise either Aurora-A population. Importantly, the N-terminal TPX2 fragment (TPX2/1–43) was sufficient to obtain a recovery of the overall Aurora-A abundance, although it was unable to restore the signal at spindle MTs. These data indicate that the TPX2 region of interaction with Aurora-A is necessary and sufficient to mediate the stabilising effect of TPX2.

The findings that the first 43 amino acids of TPX2 are sufficient to restore normal levels of Aurora-A, but not MT localisation, in a TPX2-deficient background (Fig. 4) suggest that the mechanistic roles of TPX2 in regulating the stabilisation and localisation of Aurora-A are independent. To address this question directly, we took the approach of experimentally altering each one of these effects individually and analysing the other one. First, we stabilised Aurora-A levels in TPX2-silenced prometaphase cells by inhibiting proteasome activity by treatment with MG132 (Fig. 5A): under these conditions, Aurora-A became more abundant in the cytoplasm yet failed to associate with MTs (Aurora-A signals at spindle poles and throughout mitotic cells are quantified in supplementary material Fig. S4). Next, we prevented the localisation of Aurora-A along the spindle by treating cultures with nocodazole, so as to inhibit MT assembly, in a TPX2-proficient background (Fig. 5B). This yielded Aurora-A dispersal in the cytoplasm, yet the overall abundance of Aurora-A remained similar to that seen in prometaphases with properly assembled MTs. These results demonstrate unambiguously that TPX2 regulates the MT localisation and stability of Aurora-A through two independent mechanisms; furthermore, Aurora-A protection from proteasome-dependent degradation is not conferred through MT binding of the Aurora-A–TPX2 complex.

Given that the first 43 amino acids of TPX2 that regulate the overall abundance of Aurora-A (Fig. 4B) also regulate its kinase activity (Bayliss et al., 2003), we asked whether catalytic activation per se confers stability to Aurora-A. To test this, we compared wild-type (wt) and catalytically inactive [K162R, kinase-dead (KD)] (Meraldi and Nigg, 2001; Meraldi et al., 2002) Aurora-A versions in U2OS cells. Both constructs were myc-tagged and showed a comparable transfection efficiency. We found that exogenous wild-type and KD proteins are expressed at comparable levels in both WB (Fig. 5C, myc row) and IF assays (Fig. 5D), both in interphase and mitotic cells. This yielded a comparable overall abundance of the Aurora-A protein, with a similar cell cycle phase distribution, as revealed by cyclin B1 levels (Fig. 5C) in the cell populations transfected with either construct. These assays indicate that the loss of catalytic activity per se does not decrease the stability of Aurora-A, suggesting therefore that the stabilising role of TPX2 is not necessarily dependent on its ability to activate the kinase.

Proteasome- and Cdh1-dependent downregulation of Aurora-A physiologically occurs at exit from mitosis (Taguchi et al., 2002; Lindon and Pines, 2004). Based on the data obtained at this point, we surmised that the protective effect of TPX2 on Aurora-A would need to be lost or attenuated at that time. TPX2 is itself a substrate of degradation at mitotic exit (Stewart and Fang, 2005), and single studies of TPX2 and Aurora-A indicate that both proteins disappear within 2–4 hours of release from nocodazole-dependent prometaphase arrest in HeLa cells – that is, a time corresponding approximately to telophase (Lindon and Pines, 2004; Stewart and Fang, 2005; Floyd et al., 2008). To gain insight into their dependence for degradation, we first characterised their timing of downregulation in U2OS cultures exiting mitosis. U2OS cells were pre-synchronised by thymidine followed by nocodazole treatment during thymidine release; mitoses were collected at round-up by shake-off, replated to terminate mitosis and analysed at intervals of 30 minutes until re-entry in the next G1 phase (for details, see Ciciarello et al., 2010): this showed that the levels of TPX2 indeed begin to decrease before Aurora-A (Fig. 6A). In order to evaluate whether that event is causally required for Aurora-A degradation, we analysed the endogenous levels of Aurora-A in U2OS cells that terminated mitosis with overexpressed TPX2/1–43 (Fig. 6B): that fragment, which contains the region interacting with Aurora-A, as described above, is not ubiquitylated in vitro (Stewart and Fang, 2005) and lacks the KEN-box sequence required for degradation of TPX2 at the end of mitosis. We presynchronised transfected cultures by thymidine, as above, accumulated cells in prometaphase in the presence of monastrol after thymidine release, then released prometaphase-arrested cells in drug-free medium to terminate mitosis and analysed Aurora-A by IF during completion of mitosis. The levels of Aurora-A decreased to an almost undetectable level in telophase cells compared with prometaphase and metaphase cells, when either vector-transfected (Fig. 6B) or non-transfected (data not shown). In sharp contrast with this, Aurora-A signals showed no significant variations in TPX2/1–43-transfected prometaphase, metaphase and telophase cells (Fig. 6B), indicating that forced expression of the Aurora-A-interacting region of TPX2 at mitotic exit prevents degradation of Aurora-A. These results are consistent with the notion that degradation of TPX2 is a prerequisite to convey Aurora-A towards proteolysis when cells complete mitosis.

The levels of Aurora-A are regulated through the cell cycle, with a peak of abundance in early mitotic stages, which reflects cell-cycle-dependent transcriptional activation taking place in G2 phase and proteasome- and APC/C-mediated proteolysis at the end of mitosis (reviewed in Vader and Lens, 2008). In this study, we show that inactivating the Aurora-A regulator TPX2 by RNAi yields a premature decrease in the levels of Aurora-A in prometaphase – that is, when they are normally due to reach highest abundance – and that this decrease is proteasome dependent. These findings suggest that the abundance of Aurora-A does not simply increase steadily during cell cycle progression but that it is actually subjected to a dynamic turnover resulting from the balance between increased synthesis and degradation, which is counteracted in the presence of TPX2. The protective effect of TPX2 on Aurora-A is not reciprocal because TPX2 stability is not altered in the absence of Aurora-A; in line with this observation, accumulation of TPX2 precedes that of Aurora-A during progression from S phase to mitosis in a normal cell cycle.

We have found that the stabilising effect of TPX2 on Aurora-A begins in G2 phase. TPX2 localises to the nucleus during that phase, where a large fraction of Aurora-A also colocalises. Aurora-A fails to accumulate in nuclei of TPX2-silenced cells, revealing that TPX2 is required for the nuclear accumulation of Aurora-A just before mitotic entry. Actually, the earliest Aurora-A signals detected during interphase progression are first seen at centrosomes, which at this point are devoid of TPX2. This suggests that centrosomal Aurora-A (which accounts for 1 in 50 to 1 in 100 of the entire intracellular pool of Aurora-A, based on measurements of IF signals) does not require TPX2-dependent stabilisation. It is possible that one or more centrosomal interactors of Aurora-A (e.g. Ajuba, PAK1 or HEF1) (for a review, see Carmena et al., 2009) exert a local stabilising effect or, non-mutually exclusively, that active stabilisation is more crucial to nuclear than to centrosomal Aurora-A accumulation owing to their significant differential abundance; furthermore, the centrosomally localised proteasome fraction might have limited activity in G2 in order to enable centrosomes to undergo maturation and accumulate large amounts of proteins required for mitotic entry, whereas the nuclear proteasome activity is expected to be high at all cell cycle times to ensure the continuous turnover of transcription, replication, chromatin-modifying and other nuclear factors.

The Cdh1-regulatory subunit of the APC/C ubiquitin ligase directs Aurora-A towards proteasome-dependent degradation at anaphase in a normal cell cycle (Taguchi et al., 2002; Lindon and Pines, 2004). The experiments presented here also implicate Cdh1 in degradation of Aurora-A in prometaphase cells lacking TPX2 activity. These data suggest that Aurora-A is intrinsically prone to proteasome-dependent degradation by what low levels of Cdh1 are present at early mitotic stages (see review by Baker et al., 2007), but binding by TPX2 prevents premature disappearance of Aurora-A until anaphase. Indeed, our time-course analysis of human cells completing mitosis has revealed that downregulation of TPX2 physiologically precedes that of Aurora-A. Previous work with Xenopus extracts showed that adding the 1–39 TPX2 fragment prevented Cdh1-mediated Aurora-A degradation in an in vitro assay (Sardon et al., 2008). Here, we find that forced expression of the TPX2 N-terminal fragment (1–43), containing the Aurora-A interaction domain, prevents the physiological degradation of Aurora-A in human cells exiting mitosis, supporting the idea that downregulation of TPX2 is a prerequisite to render Aurora-A available to the degradation machinery and to ensure timely degradation of both proteins in late mitosis.

The exact molecular mechanism through which TPX2 binding protects Aurora-A from degradation remains to be ascertained. The data at this stage suggest a model whereby binding of the TPX2 N-terminal fragment can modify the conformation or the interactions of Aurora-A, so as to inhibit recognition by the proteolytic machinery or ubiquitylation at crucial sites, the locations of which are currently unknown. Aurora-A harbours a D-box required for recognition by the degradation machinery, which lies just at the end of the catalytic domain. Although no specific change in the conformation of the D-box-containing region has been described upon formation of the Aurora-A–TPX2 complex in crystallisation studies (Bayliss et al., 2003), it is still possible that TPX2 binding to the catalytic domain of Aurora-A interferes with interactions occurring nearby and required for degradation.

In an effort to clarify how TPX2-mediated stabilisation of Aurora-A is regulated, we have sought to analyse its possible links to other established roles of TPX2 on Aurora-A – that is, kinase activation and localisation to the spindle MTs. We find that Aurora-A localisation to MTs is not required for stabilisation by TPX2 as (i) preventing MT formation by nocodazole does not influence the levels of Aurora-A in a TPX2-proficient background, whereas (ii) the TPX2 truncated form (1–43) that binds Aurora-A but not MTs, and therefore fails to localise Aurora-A therein, is capable of protecting Aurora-A from degradation in prometaphase in a TPX2i background. Thus, the interaction between the two proteins, but not the association with MTs, protects Aurora-A from proteolysis. Given that the 1–43 fragment of TPX2 retains the ability to activate Aurora-A (Bayliss et al., 2003), these experiments still leave open the possibility that Aurora-A catalytic activity confers stability to the kinase. To discriminate these functions, we have assessed an Aurora-A KD mutant that has impaired kinase activity (Meraldi et al., 2002) but can still interact with TPX2. We have found that the mutant is not intrinsically any less stable than wild-type Aurora-A, indicating that the stabilising effect of TPX2 is not necessarily exerted through activation of the kinase.

The findings reported here that the stability of Aurora-A is modulated throughout the cell cycle and that TPX2 is required for this control might have profound implications for understanding the origin of Aurora-A deregulation in tumour cells. It has been proposed recently that abnormal protein stabilisation can represent an additional route towards increased Aurora-A activity in cancer cells (Kitajima et al., 2007), in addition to the more extensively explored cases of gene amplification and upregulation of mRNA levels. Understanding the underlying mechanisms regulating the stability of Aurora-A can therefore help to understand fully how control might be lost during tumorigenesis. TPX2 is frequently overexpressed in tumour cells, suggesting that increased or unscheduled levels of TPX2 might represent a novel route to Aurora-A deregulation in tumours (Asteriti et al., 2010). In conclusion, therefore, the present findings that the abundance of Aurora-A protein is regulated by TPX2 highlight a novel level of control of TPX2 on the kinase and underscore the importance of the regulated stability of Aurora-A in mitotic control.

The U2OS cell line was grown at 37°C in a 5% CO₂ atmosphere in DMEM with 10% foetal bovine serum. Nocodazole (0.5 μg/ml), taxol (1 μM), monastrol (100 μM) and MG132 (10 μM) were used for the indicated times. For synchronisation, cells were subjected to a single or double block in 2 mM thymidine (as indicated); cultures were then released from the G1–S arrest by washing away the thymidine and adding fresh medium containing 30 μM deoxycytidine. For the analysis of mitotic exit, nocodazole (0.1 μg/ml) or monastrol (100 μM) were added to the thymidine-released cultures after 5 hours and, (i) for WB analysis, following a further 12 hours, mitotic cells were collected by shaking them off from the dishes, replated in normal medium and harvested after the indicated times; whereas (ii) for IF analysis, following a further 8 hours, monastrol was washed away and normal medium was replaced, then coverslips were fixed after 45, 90 and 120 minutes to follow progression from prometaphase to telophase. For FACS analysis, DNA content was determined by measuring propidium iodide incorporation, using a Coulter Epics XL cytofluorimeter (Beckman Coulter).

The pEGFP-N1 vector was from Invitrogen. pEGFP-hTPX2 (Gruss et al., 2002) (a kind gift of I. Vernos and O. Gruss) was mutagenised (MilleGen) by substituting nucleotides A150, T153 and A156 for T, G and T, respectively (conserved substitutions, pEGFP-TPX2res). pEGFP-TPX2res/Δ43 is a deleted version of pEGFP-TPX2res, lacking the first 43 amino acids (MilleGen). The cDNA region coding for the first 43 amino acids of TPX2 was PCR amplified and subcloned into HindIII–KpnI sites of pEYFP-C1 (pEYFP-TPX2/1–43). Myc-tagged Aurora-A (wild-type and the K162R catalytically inactive mutant, a kind gift of E. A. Nigg) cloned in the pBK-CMV vector have been described previously (Meraldi and Nigg, 2001).

The sequences targeted by siRNA (small interfering RNA) oligonucleotides (QIAGEN or Ambion) were: 144-GAATGGAACTGGAGGGCTT-162 (TPX2₁₄₄), 168-GGGCAAAACTCCTTTGAGA-186 (TPX2₁₆₈), 201-TGAGAAGTCTCCCAGTCAG-219 (Cdh1) and 725-ATGCCCTGTCTTACTGTCA-743 (Aurora-A); the GL2 oligonucleotide, targeting the gene encoding luciferase, was used as a control. Oligofectamine (Invitrogen) was used as transfection reagent following the manufacturer's instructions. siRNAs were used at 80 nM (GL2, TPX2 and Aurora-A) or 100 nM (Cdh1). When TPX2 and Cdh1 siRNAs were used together, the final concentrations were adjusted to 180 nM using GL2 to balance. Cultures were analysed 40 hours after transfection (asynchronous cultures) or at the indicated times after thymidine release (synchronised cultures, with siRNA transfection during thymidine treatment). For co-transfection of siRNA oligonucleotides and plasmids, the following amounts of plasmids (equimolar final concentrations) were used, with 80 nM siRNAs: pEGFP-N1 or pEYFP-TPX2/1–43, 1.3 μg; and pEGFP-hTPX2, pEGFP-TPX2res and pEGFP-TPX2res/Δ43, 2 μg (22.5 cm² dishes). Transfection was performed using Lipofectamine 2000 (Invitrogen; ratio of Lipofectamine:nucleic acid=0.6:1). Cells were harvested 30 hours after transfection and analysed by WB or IF. For IF analysis, in order to enrich the culture in prometaphase figures, monastrol was added for the last 5 hours before harvesting; release into normal medium for 15 minutes was performed before fixation.

Plasmid transfections in dishes (9 cm²) were performed using the following amounts: pEYFP-TPX2/1–43 and pEGFP-N1, 1 μg; myc-tagged Aurora-A constructs (or empty vector in control cultures), 4.5 μg, together with 0.5 μg of GFP vector. Lipofectamine was used at a ratio of 0.6:1 with DNA.

Cells grown on coverslips were fixed in 3.7% PFA, permeabilised in 0.1% Triton-X100–PBS and processed for IF as described previously (De Luca et al., 2008). Primary antibodies are indicated in supplementary material Table S1. Samples were analysed using an Olympus AX70 or Nikon Eclipse 90i microscope equipped with a CCD camera (Diagnostic Instruments and Qimaging, respectively). Colour-encoded images were acquired using IAS 2000 (Deltasistemi) or Nis-Elements AR 3.1 (Nikon) and processed with Adobe Photoshop CS 8.0 and Nis-Elements AR 3.1.

Signals were measured using either Adobe Photoshop CS 8.0 (bmp file format) or Nis Elements AR 3.1 (nd2 file format). Analysis of endogenous Aurora-A in mitotic cells was performed as follows: (i) poles and MTs (POLES/MTs): average pixel intensity, corrected for cytoplasmic background; (ii) cytoplasm: average pixel intensity, corrected for external background; (iii) total: average pixel intensity, corrected for external background and multiplied for the total cell pixels to take into account variations in cell size. Analysis of myc-tagged exogenous Aurora-A and GFP in interphase and mitotic cells was performed as follows: (i) nucleus (interphase) and total (mitoses), average pixel intensity, corrected for external background and multiplied for the total measured area to take into account size variations; (ii) cytoplasm:average pixel intensity, corrected for external background. The ‘myc’ values were corrected for the corresponding GFP values to normalise for the level of transfection of each cell. Images for quantification of mitotic signals were maximum-intensity projections from z-stacks (0.6 μm, 6–8 steps). In all types of analyses, values were statistically analysed using the InStat3 software, using either (i) the unpaired Student's t-test (for Gaussian distributions), applying the Welch correction when required, or (ii) the Mann–Whitney test, when the populations did not follow a Gaussian distribution.

Cells were lysed in culture dishes (total cell extracts) or harvested by shaking-off non-adherent cells (mitotic extracts) and then lysed. For the analysis of G2-enriched cultures, mitotic cells were removed by shake-off before the lysis. Lysis and WB analysis were performed as described previously (De Luca et al., 2008) using the antibodies listed in supplementary material Table S1. Where indicated, a quantitative analysis on scanned films was performed using Adobe Photoshop CS 8.0.

We thank A. Paiardini and D. Bilbao-Cortès for help and advice, and I. Vernos and O. Gruss for reagents. We particularly thank E. A. Nigg for sharing the Aurora-A expression constructs. This work was supported by grants from MIUR–Italian Ministry for University and Research (PRIN 200879X9N9 to P.L.) and AIRC–Italian Association for Cancer Research (to P.L.), Assicurazioni Generali Spa and Fondazione Roma-Terzo Settore (to G.G. and P.L.), Cancer Research UK and an MRC Career Development Award (to C.L.). M.G. is supported by MIUR (Fondo Giovani 2008, strategic area Human Health). Deposited in PMC for release after 6 months.