Regulation of Xenopus Aurora A Activation by TPX2*

The oncogenic protein kinase Aurora A is a critical regulator of meiotic and mitotic cell cycles in eukaryotic cells. Aurora A autoactivation by autophosphorylation is promoted by specific non-catalytic binding proteins. One such protein is TPX2, a required spindle assembly factor in higher eukaryotes whose ability to activate Aurora A by direct binding to the kinase catalytic domain has been established by biochemical and structural analysis. In this report we clarify the autoactivation mechanism of Aurora A by demonstrating that of seven amino acids which become autophosphorylated by Aurora A, only Thr-295 is required for activity. Association of Aurora A with TPX2 leads to activation of the kinase, in parallel with phosphorylation of TPX2. We identify the sites as three Ser residues in the N terminus of TPX2; however, mutation of these residues does not affect Aurora A activation by TPX2. In contrast, the mutation of a putative Aurora A-binding motif in TPX2 abolishes both phosphorylation of TPX2 and activation of Aurora A. We have also investigated the interaction between Xenopus p53 and Xenopus Aurora A. p53 blocks the activity of either full-length Aurora A or the isolated catalytic domain. Interestingly, inhibition is blocked by TPX2, suggesting that the ability of Aurora A to transform cells could be regulated by p53, TPX2, or other binding proteins.

Reversible protein phosphorylation is a regulatory mechanism employed during all eukaryotic cell cycles, and several classes of protein kinase tightly regulate the essential processes of DNA replication, mitosis, and cytokinesis (1). Metazoans contain three classes of Aurora Ser/Thr protein kinase, termed A, B, and C, and each exhibits diverse subcellular localization during the cell cycle, reflecting the discrete role of each kinase during mitosis (2). Aurora was originally described in budding yeast (3) and in the fly (4), where it regulates centrosome separation and is required to generate spindle bipolarity, a precondition for accurate chromosome segregation. More recently, Aurora A homologs in higher eukaryotes have been shown to be critical for multiple steps in cell division, including regulation of the G 2 /M transition, the centrosome cycle, and spindle assembly and stability (5)(6)(7)(8). Significantly, overexpression of either Xenopus or human Aurora A causes cell transformation, and Aurora A is overexpressed in numerous tumors, suggesting that its activity is critical in the etiology of cancer (9 -15). Aurora A activity is regulated by phosphorylation at a conserved Thr residue (Thr-295 in Xenopus, Thr-288 in human) in the activation segment of the catalytic lobe (11, 16 -18). Both Aurora A and Aurora B exist in distinct multiprotein complexes that ensure correct spatial localization and activation (2). Moreover, Aurora B has been shown to regulate its own activator, INCENP, by direct phosphorylation (19 -21). Activation of Aurora A is thought to occur by an autocatalytic phosphorylation event in which partner proteins stimulate autoactivation by inducing precise alignment of the catalytic residues located between the two clefts of the active site (17,22). Recently, two putative Aurora A activators have been described, the LIM domain-containing protein Ajuba (8) and the microtubule-associated protein TPX2 (17,23). Interestingly, because Thr-295 phosphorylation can be rapidly reversed by associated phosphatases, leading to inactivation, TPX2 has also evolved a second regulatory role, the protection of phosphorylated Thr-295 from catalytic attack by protein phosphatases such as PP1 1 (17,23). A recent crystal structure has elucidated the mechanism of Aurora A activation by TPX2 at the molecular level (22). TPX2 binding at two sites on Aurora A locks the catalytic domain of the kinase into an active configuration, generating favorable alignment of catalytic residues and unmasking the substrate-binding site. Moreover, TPX2 binding swings the domain containing phosphorylated Thr-295 into a phosphatase-inaccessible conformation by burying the labile phosphothreonine side chain (22).
Several putative physiological Aurora A substrates and binding proteins have been identified. These include Ajuba, TPX2, centrosomin, D-TACC, and p53 (8, 24 -27). In some cases the binding of Aurora A to these proteins also localizes the kinase to the correct subcellular organelle, where it presumably phosphorylates substrates required for the orderly progression of mitosis. TPX2 is one of the most efficiently phosphorylated Aurora A substrates identified to date and has been shown to be crucial for Aurora A localization to spindle microtubules (24). TPX2 contains multiple consensus phosphorylation sites for Cdc2 and mitogen-activated protein kinase as well as seven putative nuclear localization signals (28). It is required for spindle assembly in Xenopus egg extracts and in HeLa cells and for the generation of stable spindle poles (28 -30). During interphase TPX2 resides in the nucleus due to sequestration into complexes containing the proteins importin ␣ and ␤. However, after nuclear envelope breakdown, the generation of high Ran-GTP levels adjacent to chromosomes stimulates release of TPX2 from the importins and binding to Aurora A (23,31,32). The subsequent molecular events driven by activated Aurora A and phosphorylated TPX2 are poorly understood. TPX2 is phosphorylated upon isolation from mitotic cell extracts (28). However, the specific sites and relative importance of these modi-fications has not been explored. In this paper we investigate the regulation of Aurora A activity and identify several TPX2 phosphorylation sites. We analyze the role of this phosphorylation in TPX2-mediated Aurora A activation and also identify several mutations in the N terminus of TPX2 that uncouple Aurora A activation and TPX2 phosphorylation. The dynamic association of phosphorylated protein complexes containing Aurora A and TPX2 is likely to be important for both Aurora A and TPX2-dependent processes during the cell cycle.

EXPERIMENTAL PROCEDURES
Recombinant Proteins-Hexahistidine-tagged Xenopus laevis Aurora A or TPX2 was expressed in Escherichia coli BL21 (DE3) as described (17). Site-directed mutagenesis was performed using pET30 Aurora A or pET30 TPX2 plasmid DNA as templates using standard mutagenesis procedures with Deep Vent DNA polymerase (NEB) to generate the required mutant cDNA. All cDNAs were fully sequenced to confirm mutations and to verify the absence of secondary point mutations. Aurora A activation by phosphorylation was investigated by mutation of several recently identified sites of autophosphorylation (18). The regulatory T-loop residue (Thr-295, equivalent to Thr-288 in human Aurora A) was mutated to Ala, Glu, or Asp. Xenopus TPX2 was mutated at three putative Aurora A consensus phosphorylation sites in the N terminus (Ser-48, Ser-90, or Ser-94). Mutation to either Ala or Asp was accomplished by changing the cognate triplet codon to GCA or GAC, respectively. TPX2 and Aurora A deletion mutants (including a catalytic fragment of Aurora A (11), encompassing residues 127-407) were generated by PCR and cloned into the vector pET30, which encodes an N-terminal hexahistidine tag to enable rapid purification. Recombinant His-tagged Aurora A and TPX2 mutants were induced for 16 -18 h with 100 M isopropyl-1-thio-␤-D-galactopyranoside and purified from the soluble bacterial fraction on Talon beads (Clontech). After dialysis proteins were separated in to aliquots and stored at Ϫ80°C. GST-p53 (X. laevis) was purified using glutathione-Sepharose 4B beads (Amersham Biosciences) from BL21 (DE3) cells induced for 20 h with 400 M isopropyl-1-thio-␤-D-galactopyranoside. The purified protein was Ͼ80% pure as judged by Coomassie Blue staining and was stored at Ϫ80°C before assay. X. laevis Survivin (20) was cloned from a Stage VI Xenopus cDNA library, and the cDNA was inserted into the vector pET41, which encodes N-terminal glutathione S-transferase and His 6 affinity tags. GST-Survivin was purified from strain BL21 (DE3) using glutathione-Sepharose beads. The antibody that recognizes Aurora A only when phosphorylated at Thr-295 has been described previously as has the relative importance of Anderson-or Laemmli-type gels for analysis of Aurora A or TPX2 after phosphorylation (18).
Assay of Aurora A Activation by TPX2-Inactive (500 ng) or active (200 ng) recombinant Aurora A were preincubated on ice with 2 g of recombinant TPX2-(1-364) or TPX2-(1-364) mutants in kinase buffer (50 mM Tris-HCl, pH 7.4, 0.1 mM EGTA, 0.1% (v/v) ␤-mercaptoethanol, 0.01% Brij 35, 100 nM okadaic acid, 10 mM MgCl 2 ), 0.5 mg/ml histone H3 and then assayed in the presence of 100 M [␥-32 P]ATP for 15 min at 30°C. Aurora A activity was assessed by TPX2 phosphorylation, histone H3 phosphorylation, or by Western blotting with an antibody that recognizes the active Thr-295-phosphorylated Aurora A. To quantify TPX2 and histone H3 phosphorylation, the Coomassie-stained bands were excised from the gel and analyzed by Cerenkov counting in a scintillation counter. Activation of Aurora A by TPX2-(1-364) or TPX2-(1-364) mutants was assessed using an assay that exhibited linear kinetics with respect to time, permitting direct comparison between the wild-type and mutant proteins. To assay Aurora A activation in the presence of TPX2 and/or histone H3 and GST-p53 or GST-Survivin, 5 g of p53, 5 g of Survivin, or 5 g of BSA were preincubated for 10 min with Aurora A and TPX2 or histone H3 before phosphorylation for 15 min at 30°C in the presence of 100 M [␥-32 P]ATP. Phosphorylated TPX2 or histone H3 was visualized by autoradiography.
Protection of Aurora A Dephosphorylation by TPX2-Recombinant, active Aurora A (2 g) was incubated in phosphatase buffer with 500 ng of PP1␥ with or without preincubation with 2 g of TPX2 or 2 g of the TPX2-(1-364)/TPX2-(365-715) fragments. After 30 min the PP1␥ was inactivated with okadaic acid (10 M final concentration), and 200 ng of the Aurora A was assessed for activity using histone H3 (and TPX2) as substrate in the presence of 100 M [␥-32 P]ATP. The reaction was terminated with SDS sample buffer, and the TPX2-mediated protection of Aurora A inactivation by PP1␥ was demonstrated by the simultaneous phosphorylation of histone H3 and TPX2. 32 P-Labeled histone H3 and TPX2 proteins were visualized by autoradiography after SDS-PAGE on an 12% gel.
Determination of Stoichiometry of TPX2 Phosphorylation by Aurora A-To determine the stoichiometry of TPX2 phosphorylation by Aurora A, TPX2 (1-715, full-length), residues 365-715 (C-terminal half) or the N-terminal half of TPX2 containing amino acids 1-364 or mutant TPX2-(1-364) (S48A, S48A/S94A, and S48A/S90A/S94A, 60 pmol each) were incubated with 500 ng of recombinant active Aurora A for the indicated times at 30°C in kinase buffer containing 100 M [␥-32 P]ATP of known specific activity (400 cpm/pmol). At each time point 10% of the reaction was withdrawn, and the reaction was terminated with SDS sample buffer. Covalent incorporation of phosphate into TPX2 was calculated by Cerenkov counting of the 32 P-labeled Coomassie-stained TPX2 gel bands after separation from Aurora A by SDS-PAGE.

Importance of Thr-295 Phosphorylation for Aurora A Activity-Aurora
A is as an essential mitotic phosphoprotein (8), and a significant stoichiometric phosphorylation of Xenopus Aurora A occurs at seven Ser and Thr residues after synthesis and autophosphorylation in bacteria (17,18). The incubation of inactive Aurora A with TPX2, Ajuba, or Xenopus M-phase extracts, which contain a poorly defined array of Aurora A-activating proteins, also promotes the phosphorylation of Aurora A (11,17,23). These findings raise the possibility that phosphorylation of multiple amino acids may be required for the activity of Aurora A toward exogenous substrates. To determine the relative importance of these phosphorylated residues, we individually mutated the seven previously determined Aurora A phosphorylation sites (18) to Ala residues and expressed the mutant proteins in bacteria, where Aurora A stimulates its own autophosphorylation and autoactivation. This activation can be conveniently monitored by changes in the electrophoretic mobility of the kinase (17,18). As a control we expressed kinase dead (D281A) or partially active (K169R) Aurora A mutants (18). As shown in Fig. 1 the only site of phosphorylation that is required for autoactivation is the Tloop residue, Thr-295 (11,16,17), because mutation of any of the other six residues to Ala does not abolish the decreased electrophoretic mobility of the kinase or block the activity of the kinase toward histone H3 or TPX2, two Aurora A substrates. Interestingly, mutation of the other Ser or Thr phosphorylation sites to Ala actually generates a minor enhancement of enzyme activity (e.g. compare T122A with wild type Aurora A), possibly due to reduced competition of autophosphorylation sites with the exogenous substrate histone H3. All kinase assays were performed so that substrate phosphorylation followed linear kinetics, to allow an accurate comparison of the activity of each mutant.
The requirement for Thr-295 phosphorylation for Aurora A activity is well documented (11,17) as is the oncogenic phenotype of cells expressing a T295D mutant of Aurora A, which presumably leads to constitutive (phosphatase-insensitive) activity (Ref. 9, but see Ref. 11). Interestingly T295D or T295E mutants did not exhibit any detectable activity toward a nonspecific substrate, histone H3, as previously described (Fig. 1, middle panel, and Ref. 11), but if the same mutants were preincubated with TPX2 and then assayed for phosphorylation of TPX2 (Fig. 1, bottom panel) or histone H3 (not shown), T295D and T295E exhibited significant phosphotransferase activity. This is consistent with recent crystallographic evidence that provides a molecular explanation for the extensive increase of Aurora A catalytic activity generated by association with TPX2 (22). Intriguingly, the loss of activity generated by mutation of T295 to Ala (or Val, not shown) can be partially rescued by binding of TPX2, but only using TPX2 phosphorylation as a marker of activity (compare Fig. 1, middle and  bottom panels). A more significant rescue occurs with the partially active K169R mutant. However, no activation is evident if the D281A (kinase dead) mutant is assayed in the presence of TPX2, confirming the specificity of the reaction (Fig. 1, bottom  panel). These findings emphasize the significant enhancement of Aurora A activity that occurs upon TPX2 binding and suggest that Aurora A activity should be examined against several unrelated substrates to assess activity in vitro (17,23). Ser-53 and Ser-349 are conserved in human and Xenopus Aurora A, and it is likely that Ser-53 phosphorylation regulates Aurora A degradation by the anaphase promoting complex/cyclosome in both species (33). However, Ser-53 phosphorylation is not required for kinase activity ( Fig. 1 and Ref. 11). The mutation of Ser-12, Ser-53, Thr-103, Ser-116, or Thr-122 to Asp generated enzymes exhibiting wild-type activity after isolation from bacteria (data not shown); however, we confirmed a previous finding that mutation of Ser-349 to Asp markedly reduces activity (11). This seemingly contradictory result can be rationalized by structural data suggesting that this mutation may disrupt a conserved ion pair in the kinase, thus preventing stabilization of the active conformation (11,22). Ser-349 phosphorylation may also be involved in the regulation of PP1 binding (11,22), although this hypothesis requires further evaluation.
The N Terminus of TPX2 Is Sufficient for Aurora A Activation in Vitro-TPX2 stimulates Aurora A activity by both direct (enhancement of catalytic activity) and indirect (prevention of phosphatase-mediated inactivation) mechanisms (17,22,23). As shown in Fig. 2A, the N-terminal half of TPX2 (encompassing residues 1-364, NT) is sufficient to prevent PP1␥-mediated Aurora A inactivation. TPX2 blocks the ability of PP1␥ to remove phosphate from Thr(P)-295 in the activation loop, thus preserving Aurora A activity as assessed by both histone H3 and TPX2 phosphorylation. Interestingly, a C-terminal fragment of TPX2 encompassing residues 365-715 (CT) is unable to prevent PP1␥-mediated inactivation of Aurora A nor is it a substrate for the kinase (Fig. 2A). To investigate these findings further two additional TPX2 deletion constructs were prepared encompassing amino acids 1-115 or 116 -364 of the N-terminal half of the protein. The N-terminal 115 amino acids of TPX2 are the most highly conserved between human and Xenopus and, therefore, might represent the primary site of interaction between TPX2 and Aurora A (22). As shown in Fig. 2B, these amino acids were sufficient to induce activation of a completely inactive Aurora A molecule (top panel), and this 115-amino acid fragment was as good a substrate for the activated Aurora A as the full-length TPX2 (Fig. 2B, bottom panel). However, amino acids 116 -364 were inefficient at activating Aurora A nor was this segment a proficient substrate (Fig. 2B). These data dem-onstrate that the N-terminal half of TPX2 and, more specifically, the N-terminal 115 amino acids both protects Aurora A from inactivation by PP1␥ ( Fig. 2A, Ref. 22, and data not shown) and directly stimulates Aurora A activity (Fig. 2B). These experiments also demonstrate that the Aurora A-mediated sites of phosphorylation on TPX2 are located in the Nterminal 115 amino acids.
Determination of TPX2 Phosphorylation Sites-To analyze the phosphorylation of the N terminus of TPX2 by Aurora A and to elucidate the site(s) of phosphorylation we directly assessed the covalent incorporation of [ 32 P]phosphate into TPX2. As shown in Fig. 3A, phosphorylation of TPX2 by active Aurora A reached a plateau after 120 min, with the incorporation of ϳ1.2 mol of phosphate/mol into either full-length TPX2-(1-715) or TPX2-(1-364). Prolonged incubation did not lead to further phosphate incorporation, suggesting that phosphorylation had reached completion and establishing that fully phosphorylated TPX2 contained Ͼ1 mol of phosphate/mol of protein. Consistent with its inability to bind to Aurora A, the C-terminal half of TPX2 was not phosphorylated (Fig. 3A), as previously described (17). To identify the specific amino acids that were modified by Aurora A, we examined the sequence of Xenopus TPX2 and identified three putative Aurora A consensus motifs (RX(S/T); Ref. 34) at positions 48, 90, and 94. As shown in Fig.  3B, all three of these sites become phosphorylated, because sequential mutation to Ala decreased phosphorylation of the mutant proteins. Ser-48 and Ser-94 were phosphorylated more efficiently than Ser-90, although mutation of all three sites was required to completely prevent phosphorylation of TPX2 by Aurora A (Fig. 3B). These data are consistent with Kufer et al. (24), who observed that phosphorylation of human TPX2 by Aurora A occurred principally on Ser residues.
TPX2 Phosphorylation Is Not Required to Activate Aurora A-The Aurora A-related chromosomal passenger protein kinase, Aurora B, regulates its activator, INCENP, by phosphorylation. INCENP phosphorylation then regulates the activation of Aurora B, generating a positive feedback loop (19,21). These findings suggested that TPX2 phosphorylation might also regulate activation of its cognate kinase, in this case Aurora A.
To examine this possibility mutant TPX2-(1-364) proteins containing either Ala (to prevent Aurora A phosphorylation) or Asp (to introduce negative charge and mimic phosphorylation) at Ser-48, -90, or -94 or combinations thereof were assessed for their ability to reactivate Aurora A, which had been inactivated by PP2A. When inactive Aurora A was incubated with TPX2-(1-364) or any of the phosphorylation site mutants, its ability FIG. 1. Analysis of Aurora A autophosphorylation sites. Recombinant His-tagged Aurora A or the indicated point mutants (2 g) were denatured in SDS sample buffer, separated on a 12% Anderson gel, and visualized with Coomassie Brilliant Blue to assess purity and electrophoretic mobility, which is dependent upon phosphorylation (Ref. 18, top panel). The activity of each Aurora A mutant was assessed using two well established substrates, histone H3 (0.5 mg/ml, middle panel) or the N terminus of TPX2 (amino acids 1-364, 0.1 mg/ml, bottom panel). 200 ng of the indicated Aurora A mutant was assayed in kinase buffer as previously described (17), and histone H3 or TPX2 phosphorylation was assessed by autoradiography after SDS-PAGE. Autoradiograms of the 32 P-labeled histone H3 or TPX2 substrates are shown. WT, wild type.
to catalyze Aurora A autophosphorylation at Thr-295 was not diminished regardless of the mutation (Fig. 4A). Fig. 4B demonstrates that equal amounts of recombinant mutant were used in each experiment and illustrates how mutation of different residues affects TPX2 migration after SDS-PAGE and Coomassie staining of an Anderson gel (top panel). Non-phosphorylated recombinant TPX2 (1-364, predicted molecular mass, ϳ45 kDa) migrates highly anomalously (ϳ60 kDa), and Aurora A phosphorylation induces a further retardation of the protein. This shift is dependent on phosphorylation at Ser-48, because a S48A mutant fails to shift in the gel upon phosphorylation, and TPX2 mutants containing the S48D mutation are constitutively retarded in the gel, even in the absence of phosphorylation (Fig. 4B, middle panel). These findings suggest that Asp mimics Ser-48 phosphorylation. Further evidence that the phosphorylation of TPX2 does not influence its activating ability is shown in Fig. 4B. The mutation of all three phosphoacceptor sites (Ser-48, Ser-90, and Ser-94) to either Ala or Asp does not prevent TPX2-dependent hyper-activation of recombinant Aurora A, as assessed by histone H3 phosphorylation. A robust Aurora A activation was also observed if 4-fold less of each TPX2 mutant was analyzed using either PP2Ainactivated or active Aurora A (data not shown), indicating that phosphorylation does not affect the dose dependence of TPX2 stimulation of Aurora A activity. Interestingly, the same phosphorylation site mutants of TPX2 were also equally effective in preventing PP1-mediated Aurora A inactivation (data not shown), consistent with the regulatory effect of TPX2 on Thr-295 accessibility (22).
To further investigate the binding of Aurora A to the N FIG. 2. The N terminus of TPX2 regulates Aurora A activation. A, protection of bacterial Aurora A from dephosphorylation by the N terminus of TPX2. Recombinant, active Aurora A (2 g) was incubated in phosphatase buffer with 500 ng of PP1␥ with or without preincubation with 24 pmol of TPX2 or 24 pmol of the indicated TPX2 mutant (FL, full-length; CT, C-terminal residues 365-715; NT, N-terminal residues 1-364). After 30 min, the PP1␥ was inactivated with okadaic acid, and 500 ng of Aurora A was assessed for activity using histone H3 (and TPX2) as substrate in the presence of 100 M [␥-32 P]ATP. The reaction was terminated with SDS-sample buffer, and 32 P-Labeled histone H3 and TPX2 proteins were visualized by autoradiography after SDS-PAGE on an 8% gel. B, fine-structure mapping of the activating segment of TPX2. PP2A-treated (inactivated) Aurora A was incubated with equal molar concentrations of recombinant His-tagged TPX2, either full-length (1-715, 2 g) or mutant (1-364, 365-715, or 116 -364, 1 g) or (1-115, 300 ng) or BSA (2 g, extreme left lane) and assayed in the presence of 100 M [␥-32 P]ATP. The activation of Aurora A was determined by SDS-PAGE and Western blotting with a phosphospecific Aurora A antibody that recognizes only the active, Thr-295-phosphorylated (pThr 295) enzyme (top panel) or a pan-Aurora A antibody that recognizes both active and inactive Aurora A equally well (middle panel). The phosphorylation of TPX2 or TPX2 fragments by activated Aurora A was assessed by SDS-PAGE on a 12% Anderson gel followed by autoradiography of the 32 P-labeled TPX2 proteins (lower panel). Similar results were obtained in several independent experiments. terminus of TPX2, we exploited a recent crystallographic analysis of TPX2 and its interaction with the catalytic domain of Aurora A. Three conserved amino acids that closely contact Aurora A are Tyr-8, Tyr-10, and Asp-11, and these residues are conserved between human and Xenopus TPX2 (22). We reasoned that mutation of these residues to a small amino acid might prevent TPX2 binding and, therefore, block activation of Aurora A. As shown in Fig. 5, mutation of all three conserved residues to Ala eliminated activation of Aurora A by TPX2, as evidenced by a lack of both increased histone H3 kinase activity and phosphorylation of TPX2 (lanes 13 and 14). The mutation of Tyr-8 or Tyr-10 alone to Ala was sufficient to prevent Aurora A activation, as assessed by histone H3 phosphorylation, although TPX2 still became phosphorylated at a reduced level (lanes 5-8). However, combined mutation of both Tyr-8 and Tyr-10 to Ala severely blocked phosphorylation of TPX2 and completely abolished Aurora A activation, indicating that interaction with the kinase had been prevented (lanes 9 and 10). Mutation of Asp-11 to Ala alone had much less effect on the ability of TPX2 to activate Aurora A (lanes 11 and 12), demonstrating the critical importance of Tyr-8 and Tyr-10 in mediating binding and activation of Aurora A by TPX2.
TPX2 Prevents Aurora A Inhibition by p53-We are interested in investigating the regulation of Aurora A by distinct binding partners; one such Aurora A-interacting protein is the tumor suppressor p53 (27). The inhibition of Aurora A activity by p53 may be important for tumorigenesis, as demonstrated by p53-dependent inhibition of aneuploidy and transformation by Aurora A (27,35). Human p53 has been reported to inhibit the activity of Aurora A through a transactivation-independent binding interaction with the non-catalytic region of the kinase (27). We examined the effect of p53 on Aurora A using purified Xenopus proteins. As shown in Fig. 6, Xenopus Aurora A activity was inhibited by wild-type Xenopus p53; however, inhibition was not dependent on binding of p53 to the N terminus of Aurora A (27), because p53 was still able to inhibit the isolated catalytic domain of Xenopus Aurora A. Inhibition was dose-dependent, and we observed maximal inhibition with 5 g of GST-p53 but no inhibition with 5 g of either GST-Survivin or BSA as controls (Fig. 6A). Strikingly, the ability of p53 to block either full-length or catalytic domain alone-mediated phosphorylation was blocked by TPX2-(1-364) (Fig. 6B), as assessed by the ability of Aurora A to phosphorylate histone H3 (or TPX2, not shown). These data establish that Xenopus Aurora A catalytic domain is inhibited by Xenopus p53 and support a model in which TPX2 and p53 compete for similar binding sites within the Aurora A catalytic domain.

DISCUSSION
Xenopus Aurora A is a functional homolog of human Aurora A as assessed by its ability to generate tumors in nude mice, effects on the centrosome cycle, spindle assembly, and the conservation of several regulatory phosphorylation sites (6,7,11). We have shown that sites of autophosphorylation besides Thr-295 are not required for generation of an active kinase. As demonstrated previously (11, 16 -18), Thr-295 phosphorylation (Thr-288 in human Aurora A) is required for activation, but we now show that the mutation of Thr-295 to a negatively charged amino acid does indeed mimic activation, albeit only when the kinase is assayed in the presence of TPX2. In cells, Aurora A is likely to be complexed to PP1, Ajuba, and/or TPX2, depending upon the status of the cell cycle. It is, therefore, expected that in the presence of an activating protein, Aurora A mutants in which the T-loop Thr is changed to Asp have sufficient activity to generate cellular transformation (9) even though the isolated kinase appears inactive when assayed with a nonspecific substrate in vitro (11,35). In a similar vein, although K169R Aurora A does not cause tumors in nude mice (11), the same mutation does appear to generate aneuploidy in mammalian cells (35), whereas mutation of lysine to methionine blocks the ability of human Aurora A to override the mitotic spindle assembly checkpoint in cells exposed to taxol (15). It has been suggested that tetraploidization and centrosome amplification generated by Aurora A K162R (equivalent to K169R in Xenopus) is due to a kinase-independent function of Aurora A in human cells (35). However, we find that K169R Aurora A is partially active either alone or in the presence of TPX2 (Ref. 18 and Fig. 1), making it unlikely that known functions of Aurora A are kinase-independent. These findings could be reconciled by our finding that K169R Aurora A has at least 25% wild-type activity in the presence of TPX2, a physiological kinase substrate (Fig. 1), and this partial activity may be highly enriched at specific intracellular sites during mitosis, causing apparent defects in cytokinesis after overexpression (35). In this study we have further investigated the regulation of Aurora A by TPX2 (17,22,23). We show that the N-terminal 115 amino acids of Xenopus TPX2 are sufficient for Aurora A activation (Fig. 2). These findings are in agreement with those of Bayliss et al. (22), who have identified amino acids 1-43 of human TPX2 as functional for both kinase activation and protection from dephosphorylation. These 43 amino acids, which include Tyr-8, Tyr-10, and Asp-11, are the most highly conserved between several metazoan species (22). A nonphosphorylatable mutant of TPX2-(1-364) containing either Ala or Asp was still able to activate Aurora A in vitro (Fig. 4). These mutated residues lie outside the minimal activation motif reported by Bayliss et al. (22). It is, therefore, consistent that phosphorylation of TPX2 at Ser-48, Ser-90, or Ser-94 does not regulate the binding or ability of TPX2 to activate Aurora A (Fig. 4). A similar crystallographic analysis of Aurora B and its activators, Survivin and INCENP, will be required to explain the recent finding that a nonphosphorylatable INCENP Cterminal mutant is a poor activator of Aurora B (19,21). These studies emphasize fundamental differences between the feedback regulation of Aurora A and B by their substrates, in keeping with other differences between Aurora paralogs.
Our data strongly suggest that phosphorylation of Xenopus TPX2 at Ser-48 causes the marked band shift identified in previous experiments assessing endogenous TPX2 in mitotic extracts (28), since mutation of Ser-48 to Ala abolishes this shift and mutation to Asp constitutively mimics it (Fig. 4). Ser-48 phosphorylation could, therefore, likely cause a major regulatory conformational change in TPX2 during mitosis, perhaps bringing about changes in the conserved NLS or coiled coil domains that may affect binding to the importins or other regulatory proteins (28). Multiple Aurora A phosphorylation sites are also present in human TPX2, and human TPX2 is an efficient Aurora A substrate (17,24); it is, therefore, likely that FIG. 4. Aurora A-mediated TPX2 phosphorylation is not required for Aurora A activation in vitro. A, mutant TPX2 proteins retain the ability to stimulate autoactivation of inactive Aurora A. PP2A-treated inactivated Aurora A was incubated with 1 g of the indicated TPX2 protein in the presence of 100 M ATP, and the incorporation of phosphate at Thr-295 (pThr 295) was determined by SDS-PAGE on a 12% Anderson gel followed by Western blotting using the phosphospecific Aurora A antibody (top panel). Aurora A activation was also established by the decreased electrophoretic mobility of Aurora A visualized with a pan-specific Aurora A antibody (bottom panel). B, TPX2 proteins that cannot be phosphorylated by TPX2 are still able to stimulate Aurora A activity. phosphorylation fulfills similar unidentified functions as with Xenopus TPX2. The importance of TPX2 in the cell cycle has been highlighted by several recent studies in which TPX2 levels were knocked down by degradation of TPX2 mRNA using specific small interfering RNA. In HeLa cells TPX2 is required for spindle formation, because it promotes microtubule assembly around chromatin, presumably by activating Aurora A (30,31). TPX2 is also required for spindle pole integrity during mitosis (29), although the extent of "cross-talk" between TPX2 and Ajuba, a physiological centrosomal Aurora A activator (8), is currently unknown. TPX2 is an essential target of Ran-GTP for microtubule assembly in Xenopus extracts, raising the possibility that TPX2 phosphorylation mediated by Aurora A may be important for this process. TPX2 phosphorylation might also regulate reversible binding to inhibitory importin molecules, binding to microtubules or to motors that bundle microtubules (32,36,37). Interestingly, TPX2 was originally identified as a targeting protein for Xenopus kinesin-like protein 2 (38); therefore, a critical function for TPX2 phosphorylation may be the modulation of motor protein localization or activation. It will, therefore, be interesting to assess the effects of our phosphorylation site mutants on spindle formation and stability in Xenopus egg extracts.
Our finding that TPX2 regulates Aurora A by a phosphorylation-independent mechanism (Fig. 4) led us to investigate the effects of TPX2 on Aurora A in the presence of the regulatory molecule p53. In this study we have demonstrated that p53 inhibits the catalytic activity of Aurora A in vitro, although by a different mechanism than previously described (27). It will be interesting to determine the effect of p53 on Aurora A after immunoprecipitation from cellular extracts, which contain multiple Aurora A-binding proteins such as PP1, TPX2, and Ajuba. TPX2 binds to the catalytic domain of Aurora A, but Ajuba interacts directly with the N-terminal non-catalytic domain (8,24). The association of p53 with Aurora A in the presence of these molecules could, therefore, change during mitosis. Consequently it will be important to assess Aurora A activity during the cell cycle in cells exposed to a variety of cellular stresses, which increase p53 protein levels, or in cells that are null for the p53 gene. Fibroblasts lacking p53 are more sensitive to transformation by Aurora A than wild-type cells, and the ability of Aurora A to cause tumors is opposed by p53, indicative of a direct functional interaction between the two molecules (27,35). Given the transforming ability of Aurora A and its consistent overexpression in many aggressive human tumors, chemically designed small molecules that regulate Aurora A activity either by blocking TPX2 or Ajuba binding or mimicking p53 function may be useful anti-proliferative drugs.