The centrosomal protein Lats2 is a phosphorylation target of Aurora-A kinase

Authors


  • The first two authors contributed equally to this work.

    Communicated by: Eisuke Nishida

* Correspondence: E-mail: hnjm2606@biken.osaka-u.ac.jp

Abstract

Human Lats2, a novel serine/threonine kinase, is a member of the Lats kinase family that includes the Drosophila tumour suppressor lats/warts. Lats1, a counterpart of Lats2, is phosphorylated in mitosis and localized to the mitotic apparatus. However, the regulation, function and intracellular distribution of Lats2 remain unclear. Here, we show that Lats2 is a novel phosphorylation target of Aurora-A kinase. We first showed that the phosphorylated residue of Lats2 is S83 in vitro. Antibody that recognizes this phosphorylated S83 indicated that the phosphorylation also occurs in vivo. We found that Lats2 transiently interacts with Aurora-A, and that Lats2 and Aurora-A co-localize at the centrosomes during the cell cycle. Furthermore, we showed that the inhibition of Aurora-A-induced phosphorylation of S83 on Lats2 partially perturbed its centrosomal localization. On the basis of these observations, we conclude that S83 of Lats2 is a phosphorylation target of Aurora-A and this phosphorylation plays a role of the centrosomal localization of Lats2.

Introduction

The onset and exit of mitosis are tightly regulated by the phosphorylation and dephosphorylation of numerous proteins. These mitotic phosphorylations are carried out by various mitotic serine/threonine kinases such as Cdk, Polo, NIMA and Aurora (reviewed in Nigg 2001). Members of these kinase families are highly conserved from yeast to human and participate in centrosome maturation and separation, spindle assembly, nuclear envelope breakdown, chromosome condensation and cytokinesis. According to recent studies in yeast and Drosophila melanogaster, mutations in these kinases can lead to phenotypes that are characteristic of defects in mitosis, including monopolar spindles, unequally separated bipolar spindles and failure in cytokinesis. This suggests that inaccurate controls of mitotic processes can lead to aneuploidy or genetic instability, finally causing tumour formation or apoptosis (Nigg 2001). Therefore, to understand oncogenesis and tumour progression, it is crucial to define the signalling pathways of these mitotic kinases.

The Drosophila Aurora gene, which is highly homologous to Saccharomyces cerevisiae IPL1, was identified in a search for the gene that regulates the formation of a functional centrosome and mitotic spindle (reviewed in Bischoff & Plowman 1999; Giet & Prigent 1999). Amorphic alleles of aurora result in pupal lethality and in mitotic arrest in which the condensed chromosomes are arranged on circular monopolar spindles. The loss of function of the Aurora kinase causes failures in centrosome separation and bipolar spindle formation (Glover et al. 1995). In mammals, the Aurora kinase family consists of at least three members, including Aurora-A, Aurora-B and Aurora-C (Adams et al. 2001; Nigg 2001). Aurora-A is prominent both at the centrosomes and in the nucleus of G2 phase cells and at the half-spindle (a zone between the kinetochore and the spindle pole) in metaphase and telophase cells (Crosio et al. 2002; Hirota et al. 2003). Therefore, Aurora-A is considered to be involved in centrosome maturation and mitotic spindle assembly (Dutertre et al. 2002; Blagden & Glover 2003). Aurora-B is prominent in the nucleus during interphase, at the midzone during anaphase and in post-mitotic bridges during telophase, and is considered to be involved in chromosomal events and cytokinesis (Adams et al. 2001). Aurora-C is localized to the centrosome during the later stages of mitosis but its function is not yet clear (Kimura et al. 1999). Therefore, these Aurora kinase homologues are implicated in mitotic regulation including centrosome duplication, centrosome maturation, chromosome segregation and cytokinesis (reviewed in Dutertre et al. 2002; Blagden & Glover 2003).

Human Aurora-A is reported to be located on chromosome 20q13, a region that is frequently amplified in breast cancers and in diverse cancer cell lines. Over-expression of Aurora-A leads to abnomal centrosome amplification, chromosomal instability and transformation of NIH3T3 cells (Bischoff et al. 1998; Zhou et al. 1998). In p53−/– cells, over-expression of Aurora-A also causes extra centrosomes through defects in cytokinesis and consequent tetraploidization (Meraldi et al. 2002). This effect is blocked by p53, which directly binds and inhibits Aurora-A (Chen et al. 2002). Moreover, in primary cultures of mouse embryo fibroblasts, over-expression of Aurora-A inaccurately enters anaphase despite defective spindle formation, indicating that the elevated Aurora-A expression overrides the mitotic spindle assembly checkpoint (Anand et al. 2003). Recently, some attractive studies on the novel phosphorylation targets of human Aurora-A have been reported. One of the reports showed that human Aurora-A can phosphorylate PP1, a protein phosphatase type 1 and phosphorylations of PP1 by Aurora-A inhibit its phosphatase activity in HeLa cells (Katayama et al. 2001). Moreover, TPX2, a prominent component of the spindle apparatus, has been shown to be required for recruiting Aurora-A to spindle microtubeles in HeLa cells and to be the likely regulator of Aurora-A activity at mitotic spindle in Xenopus eggs, albeit the biological function of phosphorylation of TPX2 by Aurora-A remains unclear (Kufer et al. 2002; Eyers et al. 2003; Tsai et al. 2003). TACC3, a human homologue of the centrosomally associated protein D-TACC, has also been shown to be phosphorylated by Aurora-A (Giet et al. 2002). Phosphorylation of D-TACC by Aurora-A is probable because the recruitment of D-TACC to the centrosome requires the phosphorylation of additional centrosomal substrates (Blagden & Glover 2003). These observations suggest that Aurora-A protein is crucial for genomic stability and the maintenance of the cell cycle progression in mammalian cells.

Mammalian Lats2, a novel serine/threonine kinase, is a member of Lats kinase family to which the Drosophila tumour suppressor lats/warts also belongs (Yabuta et al. 2000). Mutation of the lats gene leads to dramatic overproliferation of phenotypes and diverse developmental defects in Drosophila mosaic animals and homozygous mutants (Justice et al. 1995; Xu et al. 1995). Similar to the Drosophila mutants, mice deficient in LATS1 gene, a counterpart of LATS2, have been shown to develop soft-tissue sarcomas and ovarian stromal cell tumours (St. John et al. 1999). However, no reports on mice deficient in the LATS2 gene have been shown.

Structural comparisons between mammalian Lats1 and Lats2 have revealed that the overall sequence similarity in the N-terminus between these proteins is much lower than in the kinase domain, except for two stretches of highly conserved sequence (Hori et al. 2000; Yabuta et al. 2000; Li et al. 2003). Therefore, the structural diversity or different modifications of the N-terminus between Lats1 and Lats2 are important to execute their independent functions during cell cycle. Enhanced expression of LATS1 in human tumour cell lines caused cell cycle arrest at G2/M through inhibition of Cdc2 kinase activity or induced apoptosis by up-regulating the level of pro-apoptotic proteins such as Bax or Caspase-3 (Yang et al. 2001; Xia et al. 2002). Moreover, Lats1 is phosphorylated in early prophase to function as a negative regulator of Cdc2 kinase by interacting with Cdc2 during mitosis (Tao et al. 1999) and is localized to the mitotic apparatus (Nishiyama et al. 1999; Hirota et al. 2000). Human LATS2 is located at chromosome 13q11-12, in which a loss of heterozygosity has been frequently observed in many primary cancers (Yabuta et al. 2000). Over-expression of human Lats2/Kpm in HeLa cells has also shown to cause G2/M arrest through inhibition of Cdc2 kinase activity and to induce apoptosis (Hori et al. 2000; Kamikubo et al. 2003), whereas over-expression of mouse Lats2 in v-ras-transformed NIH3T3 cells has shown to inhibit G1/S transition through down-regulation of Cyclin E/Cdk2 kinase activity and to suppress tumorigenicity of NIH3T3/v-ras cells (Li et al. 2003). These observations suggest that Lats kinases act as tumour suppressors by inhibition of cell cycle progression or apoptosis. Although Lats1 was recently found to be phosphorylated by Cdc2 (Morisaki et al. 2002), the kinase that phosphorylates Lats2 has not yet been identified.

In this study, we show that Lats2 is phosphorylated by human Aurora-A kinase and that Lats2 and Aurora-A co-localize to the centrosome during the cell cycle. We also show that the Aurora-A-induced phosphorylation on Lats2 plays a role in its centrosomal localization.

Results

Lats2 is phosphorylated during the cell cycle

Lats2 has been reported to be phosphorylated in the M phase (Hori et al. 2000). We therefore examined, here, the phosphorylation state of Lats2 during the cell cycle by arresting HeLa cells at the G1/S phase by using the thymidine-aphidicoline double block method (Fig. 1A). Cells were harvested at the indicated times after the blocks had been removed and analysed by Western blotting with an anti-human Lats2 monoclonal antibody (3D10) (Yabuta et al. 2000). In synchronized cells, we observed that Lats2 migrated as two bands, one that migrated slowly and another that migrated more quickly. The slowly migrating band was particularly prominent in lysates of cells that had progressed to the M phase (8 h after release from the G1/S phase). We also observed this band shift in the lysates of mitotic HeLa cells that had been treated by the microtubule-depolymerizing agent nocodazole (Fig. 1B, lanes 7 and 8). To examine whether this band shift is due to phosphorylation of Lats2, nocozazole-treated HeLa cell lysates were incubated with various concentrations of lambda-protein phosphatase (Fig. 1B, lanes 9–11). The slowly migrating Lats2 band was converted into the fast-migrating band after the phosphatase treatment in a dose-dependent manner. This conversion was completely blocked by the addition of phosphatase inhibitors (Fig. 1B, lane 12), indicating that Lats2 is indeed phosphorylated at the M phase, in which nocodazole-treated cells are arrested by the spindle assembly checkpoint. Interestingly, we observed that the band corresponding to the Lats2 protein in the G1/S-arrested cell lysates (the same sample as cell lysate of 0 h in Fig. 1A) was also converted by the phosphatase treatment into the fast-migrating band, which showed a similar mobility as the band in lane 11 of Fig. 1(B) (Fig. 1B, lanes 1–5). This conversion was also completely blocked by the addition of phosphatase inhibitors. These results indicate that Lats2 is phosphorylated not only at the M phase but also at the G1/S phase. This G1/S-phosphorylated band was also observed during interphase (Fig. 1A; 0, 1.5, 3, 10 and 12 h), indicating that Lats2 is phosphorylated during interphase. Therefore, these findings suggest that Lats2 is regulated by at least two distinct phosphorylation events during the cell cycle.

Figure 1.

The phosphorylation of Lats2 is dependent on the cell cycle phase. (A) HeLa cells were arrested in the G1/S phase by the thymidine-aphidicoline double block method. Cells were released from the block by replacing the medium with fresh medium without drugs. Thirty micrograms of each extract prepared at the indicated time (h) after the release were analysed by Western blotting for Lats2 or Cyclin B. (B) HeLa cells were arrested in the G1/S phase as time = 0 in A (left side) or the M phase by nocodazole-treatment (right side). The nuclear extract (Yabuta et al., 2000) from each cell was incubated at 4 °C or 30 °C with the indicated concentrations (units) of lambda protein phosphatase (New England Biolabs) in the absence (–) or presence (+) of phosphatase inhibitors. These extracts were analysed by Western blotting with the 3D10 antibody. (C) Cyclin D1/Cdk4, Cyclin B/Cdc2, Cyclin E/Cdk2 and Cyclin A/Cdk2 kinase complexes purified from Sf-9 insect cells were used in in vitro kinase reactions. The substrates used were Rb C-terminus (for only Cyclin D1/Cdk4), Histone H1, GST-Lats2N (79–621), GST-Lats2Ckd (622-1088, kinase dead) or GST-Nop10 (as a negative control). These proteins were resolved by SDS–PAGE and 32P-labelled proteins were visualized by autoradiography. The arrowheads and arrows indicate the locations of the GST-Lats2N (c. 96 kDa) and GST-Lats2C kDa (c. 86 kDa) proteins, respectively. Coomassie Brilliant Blue (CBB) staining shows a loading control (bottom panels). (D) An in vitro kinase reaction was performed with three different kinases and two Lats2 substrates. The kinases were GST-DNA-PK, GST-Nek2 and GST-Aurora-A. The substrates were GST-Lats2N (79–621) and GST-Lats2Ckd (622–1088, kinase dead). As a negative control substrate, GST-Nop10 was used. The proteins were resolved by SDS–PAGE and 32P-labelled proteins were visualized by autoradiography. CBB staining shows a loading control (bottom panels).

Lats2 is phosphorylated by Aurora-A in vitro

This cell cycle-dependent phosphorylation of Lats2 suggests that the regulation of Lats2 may be directly involved in cell cycle regulation. We wondered, therefore, whether the cell cycle-dependent phosphorylation of Lats2 could be mediated by the Cyclin/Cdk kinases. To test this notion, we performed in vitro kinase assays with four classes of Cyclin/Cdk kinases, namely, Cyclin D1/Cdk4, Cyclin B/Cdc2, Cyclin E/Cdk2 and Cyclin A/Cdk2, which act in G1, G2/M, G1/S and S, respectively (Kitagawa et al. 1996). The substrates used were two truncated forms of glutathione-S-transferase (GST)-fused Lats2, namely, Lats2N (amino acids 79–621) and Lats2Ckd (amino acids 622–1088) (Fig. 2A) because full-length Lats2 was too unstable for recovery. We did not prepare the 1–78 amino acids region of Lats2 for a substrate of Cdk kinases because there is no consensus sequence of the Cdk phosphorylation site (S/T-P-X-R/K) (Kitagawa et al. 1996) in this region. GST-Lats2Ckd, which is defective in kinase activity due to the substitution of the catalytically essential lysine 687 residue with methionine (data not shown), was used rather than the intact Lats2C protein to prevent the autophosphorylation of Lats2C. As shown in Fig. 1(C), Cyclin D1/Cdk4 and Cyclin B/Cdc2 kinases could not phosphorylate either of the GST-Lats2 constructs, although these Cyclin/Cdk kinases could phosphorylate the substrates used as positive controls (Rb C-terminus or Histone H1). Although Cyclin E/Cdk2 and Cyclin A/Cdk2 appeared to slightly phosphorylate the degraded products of Lats2N, these may not be the major phosphorylation events to which Lats2 is subjected during the cell cycle because they are of similar intensity as the phosphorylation of the negative control substrate Nop10, which is a protein involved in the pseudouridylation of pre-rRNAs (Henras et al. 1998). These results suggest that the cell cycle-dependent phosphorylation of Lats2 is due to kinase(s) other than the Cyclin/Cdk kinases. In particular, it appears that the mitosis-dependent phosphorylation of Lats2 is not due to Cyclin B/Cdc2 kinase, which functions during mitosis.

Figure 2.

Recombinant Aurora-A kinase phosphorylates the S83 residue of Lats2 in vitro. (A) Schematic diagram of Lats2 that shows the location of the kinase domain. Beneath are the GST-fusion proteins used in the kinase reactions, namely, those containing Lats2 amino acids 1–79, 1–118, 79–118, 79–151, 113–151, 79–257, 79–621 (Lats2N) and 622–1088 (Lats2Ckd, kinase dead). (B) In vitro kinase reactions were performed with GST-Aurora-A and truncated GST-Lats2 proteins. The truncated GST-fusion proteins used were those containing amino acids 1–78, 1–118, 79–118, 79–151, 113–151, 79–257 and 622–1088 (Lats2Ckd) of Lats2. CBB staining shows a loading control (bottom panel). (C) An in vitro kinase reaction was performed with GST-Aurora-A and GST-Lats2-79-118 proteins that harbour the indicated substitutions (left panel). CBB staining shows that equal amounts of proteins were present (right panel). (D) In vitro kinase reactions were performed using GST-Lats2-79-257 as a substrate with the kinases GST-Aurora-A, GST-Aurora-B and GST-Aurora-C. CBB staining shows a loading control (right panel).

As the kinase that phosphorylates Lats2 is not known, we searched for candidates with the in vitro kinase assay. When we tested DNA-PK (Kim et al. 1999), Nek2 (Fry et al. 1995) and Aurora-A, we found that only Aurora-A could phosphorylate Lats2N efficiently in vitro (Fig. 1D). Aurora A could not, however, phosphorylate GST-Lats2Ckd. Aurora-A is reported to be a centrosomal protein kinase whose kinase activities are regulated in a cell cycle-dependent manner (Nigg 2001) and peak at the G2/M phase (Bischoff et al. 1998; Farruggio et al. 1999; Hirota et al. 2003).

Aurora-A kinase phosphorylates the serine 83 of Lats2 in vitro

To determine more specifically the region of Lats2 that is phosphorylated by Aurora-A, we produced several deletion mutants of Lats2 as GST-fusion proteins (Fig. 2A) and used them as substrates in the in vitro kinase assay. As shown in Fig. 2(B), GST-Lats2 proteins containing amino acids 1–118, 79–118, 79–151, 79–257 and 79–621 were readily phosphorylated by Aurora-A, whereas the Lats2 proteins consisting of amino acids 1–78 and 113–151 were very poorly phosphorylated. This suggests that Aurora-A directly phosphorylates the amino acid 79–118 region of Lats2. To further map the Aurora-A phosphorylation site(s) on Lats2, we mutated serine 83 (S83), S91, threonine 93 (T93) or S94 in GST-79-118 to cysteine (S83C), cysteine (S91C), aspartic acid (T93D) or alanine (S94A), respectively. In vitro kinase assays with Aurora-A were performed with these mutated sequences. A loading control experiment with Coomassie Blue staining showed that an equal amount of each protein had been applied (Fig. 2C, right panel). Whereas phosphorylation of the S91C, T93D or S94A mutants was not significantly altered, the phosphorylation of the S83C mutant was remarkably diminished (Fig. 2C, left panel). In addition, we found that the S83 residue is conserved not only in human Lats2 but also in mouse Lats2, mouse Lats1 and human Lats1 (data not shown). We also asked whether Aurora-B and -C can phosphorylate Lats2 as well as Aurora-A. An in vitro kinase assay was performed with these purified Aurora kinases using purified Lats2-79-257 as a substrate. Lats2 was phosphorylated by Aurora-B and -C but much more inefficiently than Aurora-A (Fig. 2D). These results indicate that Lats2, at least the 79–257 region, is phosphorylated predominantly by Aurora-A and weakly by Aurora-B and Aurora-C in vitro.

Aurora-A kinase phosphorylates the serine 83 of Lats2 in vivo

To confirm that S83 of Lats2 is phosphorylated by Aurora-A, we raised a specific monoclonal mouse anti-phospho-S83 antibody (3B11) by immunizing mice with a Lats2 peptide whose S83 residue is phosphorylated. The in vitro kinase assay was performed with wild-type Aurora-A or the kinase-dead form of Aurora-A with GST-Lats2-79-118 (WT) or GST-Lats2-79-118 (S83C) in the absence of [γ-32P]ATP and analysed by Western blotting with the 3B11 antibody (Fig. 3A). When wild-type Aurora-A was used, the 3B11 antibody specifically detected the phosphorylated form of WT Lats2-79-118 (lane 2). However, no phosphorylation of the S83 site was detected when kinase-dead Aurora-A was used (lane 3). Therefore, the 3B11 antibody is specific for the phosphorylated form of S83.

Figure 3.

Lats2 is phosphorylated on S83 and interacts with Aurora-A in vivo. (A) Specificity of the anti-phospho-S83 monoclonal antibody (3B11). In vitro kinase reactions were performed with GST-Lats2-79-118-WT (lanes 1–3) or -S83C (lanes 4 and 5) using GST-Aurora-A (WT; lanes 2 and 4) or GST-Aurora-A-kinase-dead (KD; lanes 3 and 5) in the absence of [γ-32P]ATP. The assays were analysed by Western blotting with the 3B11 antibody (top panel). That equal amounts of GST-Lats2-79-118 or GST-Aurora-A proteins were present was confirmed by Western blotting with anti-GST antibody (middle and bottom panels). (B) Phosphorylation state of endogenous Lats2 protein at the M phase. Endogenous Lats2 was immunoprecipitated from nocodazole-arrested M phase (Noc.) or asynchronous (AS) cell extracts with the anti-Lats2 monoclonal antibody (3D10; lanes 1 and 3) or the anti-phospho-Ser83 monoclonal antibody (3B11; lanes 2 and 4), followed by Western blotting with the 3D10 antibody. (C) Specificity of the anti-Aurora-A polyclonal antibody. 293T cells were transfected with HA-tagged Aurora-A, Aurora-B, Aurora-C or vector alone. The whole cell lysates were resolved by SDS–PAGE, followed by Western blotting with anti-HA (left panel) or anti-Aurora-A antibodies (right panel). (D) In vivo interaction of Lats2-1-393 with Aurora-A. 293T cells were co-transfected with 6Myc-Lats2-1-393 (lanes 2, 3, 5, 6, 8 and 9) and/or GFP-Aurora-A (lanes 1, 3, 4, 6, 7 and 9). As controls, the lysate of 293T cells transfected with empty vectors were used (V; lanes 1, 2, 4, 5, 7 and 8). GFP-Aurora-A or 6Myc-Lats2-1-393 was immunoprecipitated with the anti-GFP antibody (lanes 4–6) or anti-Myc antibody (lanes 7–9), respectively. Whole cell lysates (lanes 1–3, about 7 µg each) and immunoprecitates were resolved by SDS–PAGE, followed by Western blotting with the anti-Myc (top and 2nd panels from top) or anti-Aurora-A (3rd panel from top) or anti-GFP antibodies (bottom panel). An asterisk indicates IgG light chain.

The 3B11 antibody was used to show that the S83 residue in endogenous Lats2 is phosphorylated in vivo in a cell cycle-dependent manner. The 3D10 antibody that recognizes Lats2 was also used. Therefore, the Lats2 proteins in lysates from nocodazole-treated or asynchronous HeLa cells were immunoprecipitated with either the 3D10 or the 3B11 antibody, followed by Western blotting using the 3D10 antibody (Fig. 3B). The 3B11 antibody immunoprecipitated the phosphorylated-S83-Lats2 protein from the nocodazole-treated cell lysate (lane 2) but not from the asynchronous cell lysate (lane 4). In contrast, the 3D10 antibody brought down equivalent amounts of Lats2 protein from each cell lysate (lanes 1 and 3), although the 3D10 antibody could not recognize the M phase-specific phosphorylated form of Lats2 in the immunoprecipitate of the 3D10 antibody (lane 1). The 3D10 antibody may be able to recognize the M phase-specific phosphorylated form of Lats2 in Western blotting but not to immunoprecipitate it under this experimental condition. Therefore, the S83 residue in endogenous Lats2 is the primary phosphorylation site in vivo. Taken together, we conclude that Aurora-A is the phosphorylating kinase of Lats2 in the M phase because Aurora-A predominately phosphorylates S83 on Lats2 in vitro and this site is also phosphorylated in vivo.

Lats2 interacts with Aurora-A

Next, we raised a polyclonal rabbit antibody against GST-Aurora-A protein. 293T cells expressing HA-tagged Aurora-A, -B, -C or vector alone were lysed and analysed by Western blotting with either anti-Aurora-A (Fig. 3C, left panel) or anti-HA (right panel) polyclonal antibody. The anti-Aurora-A antibody specifically recognized only HA-Aurora-A but not HA-Aurora-B or HA-Aurora-C. In addition, the anti-Aurora-A antibody also recognized a band corresponding to the endogenous Aurora-A protein in 293T cells and untransfected HeLa S3 cells. Therefore, it appears that the anti-Aurora-A antibody specifically recognizes the Aurora-A protein and does not crossreact with other proteins. Using this antibody, we performed co-immunoprecipitation experiments to know whether Aurora-A and Lats2 interact in vivo. As we were unable to confirm the interaction of endogenous Aurora-A with Lats2 by immunoprecipitation assays using anti-Aurora-A or 3D10 antibodies (data not shown), we co-transfected 293T cells with GFP (green fluorescent protein)-Aurora-A and/or 6Myc-Lats2-1-393 (Fig. 3D, lanes 1-3). When we performed immunoprecipitation experiments with each transfected cell extract by using the anti-GFP antibody, we detected 6Myc-Lats2-1-393 in the GFP-Aurora-A immunoprecipitate (Fig. 3D, top panel, lane 6). In reciprocal immunoprecipitation experiments, GFP-Aurora-A was detected in the 6Myc-Lats2-1-393 immunoprecipitate (Fig. 3D, third panel from top, lane 9). This weak interaction between Lats2 and Aurora-A is reminiscent of an unstable complex that is commonly observed between an enzyme and a substrate. These results indicate that Aurora-A interacts with the N-terminus (1–393 amino acids) of Lats2 in vivo, which supports the notion that Lats2 is a phosphorylation target of Aurora-A in vivo.

Lats2 co-localizes with Aurora-A at the centrosome

If Lats2 is a phosphorylation target of Aurora-A in vivo, it is likely that their subcellular distributions are similar during the cell cycle. To test this possibility, we examined whether Lats2 co-localizes with Aurora-A during various cell cycle stages. As the 3D10 antibody, which was raised against the N-terminal portion (amino acids 78–256) of Lats2, could not detect any endogenous Lats2-specific signals, HeLa S3 cells were transiently transfected with GFP-fused full-length human Lats2 or the GFP-vector alone, followed by immunofluorescence staining with anti-Aurora-A antibody (Fig. 4A). GFP-Lats2 was observed as one or two bright spots beside the nucleus in interphase (left panels, i and ii; green). In prophase, these bright spots translocated toward the opposite poles of the cell (left panel, iii). A similar pattern was also observed for Aurora-A, except that Aurora-A was located within the nucleus (middle panels; red). The nuclear localization of Aurora-A was detected with both of the anti-Aurora-A antibodies and we observed no signal with the secondary antibody alone in our experiments (data not shown). Therefore, it is unlikely that the localization of Aurora-A in the nucleus as determined by these two antibodies is as a result of cross-reaction of anti-Aurora-A antibodies. The yellow spots in the merged images indicate that GFP-Lats2 co-localizes with Aurora-A during interphase, prophase and telophase (right panels). However, when the cells enter metaphase, the GFP-Lats2-specific signal was diffusely distributed throughout the cell (left panel, iv), whereas Aurora-A was localized to two predominant bright spots that are reminiscent of the spindle microtubules and the spindle poles. A similar change in subcellular distribution was also observed in the cells expressing the 6Myc-tagged Lats2 construct (Fig. 4B,i and C,i and ii), and the diffuse distribution of Lats2 was also observed during anaphase (Fig. 4B,ii). Interestingly, when cells enter cytokinesis, GFP-Lats2 was observed again as bright spots and the spots were distributed to daughter cells and localized at each pole (Fig. 4A,v), which is suggestive of re-localization to the centrosomes of GFP-Lats2 when the cells enter cytokinesis. Moreover, GFP-Lats2 was also found at the midbody during cytokinesis (Fig. 4A,v). Together with the previous reports that Aurora-A is localized to the interphase and mitotic centrosomes as well as to the spindle poles (Bischoff et al. 1998; Zhou et al. 1998), these results indicate that Lats2 co-localizes with Aurora-A at the centrosomes during interphase, early prophase and cytokinesis. That Lats2 localizes to centrosomes during interphase was also confirmed by its co-localization with γ-tubulin (Fig. 4C).

Figure 4.

Lats2 co-localizes with Aurora-A at the centrosome. HeLa S3 cells were transiently transfected with GFP-fused full-length Lats2 (A) or 6Myc-tagged full length Lats2 (B and C). The transfected cells were synchronized at the S phase by thymidine-single block, released from the block and then fixed with formaldehyde at various cell cycle stages: interphase (A-i, -ii, B-i and C-i), prophase (A-iii), metaphase (A-iv and C-ii), anaphase (B-ii) and telophase (A-v). 6Myc-Lats2 was visualized by immunofluorescence staining of the fixed cells with anti-Myc antibody followed by incubation with Alexa-Fluor 488-conjugated anti-mouse immunoglobulin G (B and C, green). Aurora-A or centrosomes were visualized by immunofluorescence staining with anti-Aurora-A (A and B, red) or anti-γ tubulin (C, red) antibody, respectively, followed by Texas Red or Alexa-Fluor 594-conjugated anti-rabbit immunoglobulin G. DNA was visualized by staining with Hoechst 33258 (B and C, blue). DNA and merged images are shown in the right panels. The yellow signals reflect the co-localization of GFP- or 6Myc-tagged Lats2 and the indicated proteins (Aurora-A or γ-tubulin). Scale bar, 10 µm.

The S83 of Lats2 is phosphorylated at the centrosome, the mitotic spindle pole and the midbody

To confirm that S83 on Lats2 is phosphorylated at the centrosomes in vivo, the spatial and temporal distributions of the S83-phosphorylated (pS83) Lats2 in HeLa S3 cells were analysed with the 3B11 antibody. As shown in Fig. 5(A), the phosphorylation signals of S83 by the 3B11 antibody were observed at the centrosomes of interphase and prometaphase cells (Fig. 5A,i and ii, left panels) and at the spindle poles of metaphase, anaphase and telophase cells (Fig. 5A,iii,iv and v, left panels). The centrosomal localization of pS83 was also confirmed by its co-localization with γ-tubulin (Fig. 5A,i–vi, right panels; yellow). The S83-phosphorylations of endogenous Lats2 and the subcellular localizations of GFP- and 6Myc-tagged Lats2 were observed at both a single centrosome and duplicated centrosomes during interphase (Figs 4 and 5, and data not shown). Centrosomes are duplicated during early S phase and mature during late S phase. Centrosomes then separate during early mitosis (reviewed in Doxsey 2001; Nigg 2002). Therefore, these results suggest that both the centrosomal localization and the initial S83-phosphorylations of Lats2 occur before S phase. The S83-phosphorylations at the centrosomes or the spindle poles were more prominent at prometaphase and metaphase (Fig. 5 and B,ii and iii, green spots) than at the other stages. These observations are similar to the subcellular localization of Aurora-A during the cell cycle (Fig. 5B). However, Aurora-A localizes to both the spindle poles and half-spindles, while the S83-phosphorylated Lats2 localizes only to the spindle poles during mitosis. Interestingly, during cytokinesis, while the S83-phosphorylation disappeared from the centrosomes or the spindle poles, it was detected at the midbody (Fig. 5A,vi and 5B-vi, left panels). It is noteworthy that this localization of pS83 during cytokinesis is more similar to that of Aurora-B than Aurora-A (Crosio et al. 2002). These results suggest that the endogenous Lats2 is localized to the centrosomes/the spindle poles, where it is phosphorylated on S83 by Aurora-A during the cell cycle.

Figure 5.

S83-phosphorylated Lats2 localizes to the centrosomes/spindle poles during the cell cycle besides cytokinesis. (A and B) HeLa S3 cells were synchronized at the S phase by thymidine-single block, released from the block and then fixed with formaldehyde at interphase (i), prometaphase (ii), metaphase (iii), anaphase (iv), telophase (v) and cytokinesis (vi). S83-phosphorylated Lats2 was visualized by immunofluorescence staining with the 3B11 antibody followed by incubation with Alexa-Fluor 488-conjugated anti-mouse immunoglobulin G (left panels, green). Centrosomes/spindle poles or Aurora-A were respective visualized with anti-γ-tubulin (A) or anti-Aurora-A (B) antibodies (2nd panels from left, red). DNA was visualized by staining with Hoechst 33258 (blue). DNA and merged images are shown in the right panels. The yellow signals indicate the co-localization of S83-phosphorylated Lats2 with the centrosomes/spindle poles. Scale bar, 10 µm.

Phosphorylation of S83 plays a role of the centrosomal localization of Lats2

To explore the significance of S83 phosphorylation in Lats2 localization, HeLa S3 cells were transfected with 6Myc-tagged full-length human Lats2 containing S83C or S83E (S83 mutated to glutamate), and the centrosomes were detected by staining with γ-tubulin (Fig. 6A and B, 2nd panels from left, red). The subcellular localization of both the S83C and S83E mutants during the cell cycle frequently showed mislocalization of Lats2 at the centrosome(s) of interphase HeLa S3 cells (Fig. 6A and B, top-left panel, arrows) in comparison with the 6Myc-Lats2 wild-type (Figs 4C and 6A and B; these experiments were performed in equal conditions). In mitotic cells, there is no detectable staining of the S83 mutant proteins (S83C, S83E) (Figs 4A and B, bottom-left panels), as well as wild type (Fig. 4C). Therefore, we could not obtain any data on the centrosomal localization by comparison between each S83 mutant and wild type during all phases of mitosis. Next, among the cells expressing each 6Myc-Lats2 -WT, -S83C or -S83E, the numbers of cells in which the 6Myc-Lats2 obviously localized to the centrosome(s) during interphase were counted to assess the percentages of the centrosomal localization of Lats2 WT and two S83 mutants. As shown in Fig. 6(C), in 55% of cells expressing Lats2 WT, the 6Myc-Lats2 localized to the centrosome while, in the case of the S83C and S83E mutants, the percentage of cells harbouring the centrosomal Lats2 were reduced by less than 30 and 43%, respectively. These results indicate that the non-phosphorylation of the Ser 83 residue disturbs the centrosomal localization of Lats2.

Figure 6.

The phosphorylation of S83on Lats2 plays a role of its centrosomal localization. HeLa S3 cells were transiently transfected with 6Myc-tagged full-length Lats2-S83C n(B). The transfected cells were fixed with formaldehyde at interphase (A-top and B-top panels), prometaphase (lower in B-bottom panel), metaphase (lower in A-bottom panel), anaphase (upper left in A-bottom panel) and telophase (upper right in A-bottom panel). 6Myc-Lats2 was visualized by immunofluorescence staining of the fixed cells with anti-Myc antibody followed by Alexa-Fluor 488-conjugated anti-mouse immunoglobulin G (A and B, green). Centrosomes were visualized by immunofluorescence staining with anti-γ-tubulin antibody (A and B, red), followed by Alexa Fluor 594-conjugated anti-rabbit immunoglobulin G. The positions of centrosomes are indicated by white arrows. DNA was visualized by staining with Hoechst 33258 (A and B, blue). Merged images are shown in the right panels. The yellow signals reflect the co-localization of the 6Myc-Lats2 proteins and γ-tubulin. (C) Histogram shows the percentages of cells in which the indicated 6Myc-tagged proteins localize to the interphase centrosomes in HeLa S3 cells. These results were obtained from four independent experiments (more than 50 positive cells in 500 cells each) and bars indicate standard deviations. All data were statistically analysed based on the Student's t-test. *P = 0.001.

Discussion

In this study, we have shown that Lats2 is phosphorylated in at least two distinct stages of the cell cycle, G1/S phase and M phase containing nocodazole arrest (Fig. 1A and B), suggesting that Lats2 is regulated by multiple phosphorylations throughout the cell cycle. Cyclin E/Cdk2 and Cyclin A/Cdk2 kinases alone could produce very weak phosphorylation signals on the degradation products of Lats2N (Fig. 1C). This observation might suggest that one of the phosphorylated forms of Lats2 during interphase may be due to Cyclin E/Cdk2 and Cyclin A/Cdk2 kinases. Because it is probable that mouse Lats2 also regulates the G1/S transition through down-regulation of Cyclin E/Cdk2 kinase activity in NIH3T3 cells (Li et al. 2003), the G1/S-dependent phosphorylation(s) of Lats2 that we have shown in Fig. 1(B) may be implicated in some functions of Lats2 on G1/S transition. We identified a centrosomal kinase, Aurora-A, as one of the candidate kinases for phosphorylation of S83 residue on Lats2. Immunostaining data with the anti-phosphorylated S83 antibody reveals that S83 on Lats2 is phosphorylated in vivo during the cell cycle and it is more prominent during prophase and metaphase than interphase, anapahse and telophase, which is similar to the expression pattern of Aurora-A during the cell cycle (Fig. 4A and Bischoff et al. 1998). These results suggest that S83 on Lats2 is a phosphorylation target of Aurora-A in vivo. As Aurora-B and -C also phosphorylate Lats2, although very weakly (Fig. 2D), it will be important to examine in future whether Aurora-B or Aurora-C can more efficiently phosphorylate S83 or other sites of Lats2 by using the full-length protein of Lats2. In fact, subcellular localization of phosphorylated S83 is observed at the midbody of cytokinesis cells, which is similar to that of Aurora-B (Fig. 5). Therefore, the phosphorylation of S83 on Lats2 may be regulated by not only Aurora-A but also Aurora-B.

Immunofluorescence data of exogenous GFP or 6Myc-tagged Lats2 protein indicate that Lats2 and Aurora-A co-localize at the centrosome during the cell cycle, except for metaphase and anaphase. We previously showed that Lats2 exists in the ‘nuclear fraction’ that was prepared from cell lysates by Western blot analysis (Yabuta et al. 2000). Probably, the centrosome was distributed in this fraction during our subfractionation procedure. Although a recent report of Li et al. has shown that ectopically expressed mouse Lats2 localized in the cytoplasm of NIH3T3 cells and that the majority of endogenous Lats2 protein is located in the cytoplasm in their fractionation experiments using lung cancer cells (Li et al. 2003). Moreover, the ectopic over-expression of Lats2 in cells tends to localize diffusely to the cytoplasm (Figs 4 and 6), which may be due to the degradation of exogenous Lats2 protein. However, our immunostaining data using the 3B11 antibody showed that endogenous Lats2 protein is located not only at centrosomes but also in the nucleus during interphase (Fig. 5A,i and B,i). These observations are consistent with the previous reports that Aurora-A localizes to both the centrosome and the nucleus during the G2 phase (Crosio et al. 2002; Hirota et al. 2003). Recently, Aurora-A was reported to be required for mitotic entry of human cells in concert with its interacting activator Ajuba, a LIM protein. It is likely that Aurora-A is initially activated at the G2 phase and its activity is required for the recruitment of the Cyclin B1/Cdc2 complex to centrosomes (Hirota et al. 2003). Although Lats2 localized to the centrosome before its duplication in early S phase, it is not possible that Lats2 is involved in the regulation of centrosomal duplication, because we could also observe two close spots of γ-tubulin, as well as typical and normal duplications of the centrosome, in HeLa cells in which the wild-type or the S83 mutants of Lats2 was over-expressed ectopically. Moreover, in accordance with a previous report on a role of Drosophila Aurora-A in centrosome maturation, Aurora-A promotes the recruitment of D-TACC to centrosomes and phosphorylates it (Giet et al. 2002). Because the phosphorylation of S83 on Lats2 is one of requirements for centrosomal localization of Lats2 in this issue, Aurora-A may also promote the recruitment of Lats2 to centrosomes as well as Cyclin B1/Cdc2 complex and D-TACC for the centrosome maturation. The accumulated Lats2 kinase at the centrosome may rapidly phosphorylate other centrosomal components, together with some centrosomal kinases including Cdc2 and Plk1, polo-like kinase, in order to progress the centrosome maturation efficiently. γ-Tubulin is one of the components that are recruited to the MTOC (microtubule-organizing centre) during the centrosome maturation. When the wild-type or the S83 mutants of Lats2 were over-expressed in HeLa cells, we could observe one or two spots of γ-tubulin in these cells (Figs 4 and 6), which is suggestive of the recruitment of γ-tubulin to MTOC. Therefore, it is unlikely that Lats2 is involved, at least in the recruitment of γ-tubulin. Moreover, the over-expression of the wild-type or the S83 mutants of Lats2 did not cause abnormal chromosome alignment and aberrant mitotic spindle formation (Figs 4 and 6, and data not shown). To date, several structurally different protein kinases, including Aurora kinases and Cdks, have been shown to localize at the centrosome regulating the centrosomal function during the cell cycle (Mayor et al. 1999). Among these proteins, Nek2, as a NIMA-like kinase and Plk1 are representative centrosomal kinases, although Nek2 could not phosphorylate Lats2 in vitro. Although as yet untested, it is intriguing whether Plk1 also phosphorylates Lats2 because Plk1 is involved in not only controlling centrosomal functions but also DNA damage checkpoint (Smits et al. 2000). However, the GFP-Lats2 or 6Myc-Lats2-specific signal was diffusely distributed throughout the cell but not at the centrosome during mitosis (Fig. 4). By using the 3B11 antibody, we could observe that endogenous Lats2 localizes at the centrosome or the spindle pole during mitosis. Therefore, the diffusely distribution of GFP- and 6Myc-Lats2 may also be due to the effects of over-expression of these proteins in a mitotic cell, including the protein degradation.

When the kinase activities of these Lats2 mutants were assessed by examining the auto-phosphorylation of immunoprecipitates generated by the anti-GFP antibody from extracts expressing GFP-Lats2 wild-type or S83C mutant, we were unable to observe remarkable differences in their phosphorylation levels (data not shown). A previous report has shown that over-expression of either wild-type or kinase inactive form of Aurora-A in HeLa cells triggered mitotic defects including aberrant cytokinesis and the formation of tetraploid cells, but not centrosome amplification in S phase (Meraldi et al. 2002). Therefore, we suppose that the Aurora-A-dependent phosphorylation of S83 on Lats2 is not implicated in the aberrant cytokinesis and the formation of tetraploid cells caused by over-expression of Aurora-A in HeLa cells. On this issue, we showed that S83 phosphorylation plays a role of the centrosomal localization of Lats2 (Fig. 6). It is noteworthy that there is no detectable staining of the S83 mutant proteins in telophase cells, although this is seen in cells expressing the wild-type protein (Figs 6A and B). Therefore, it can be speculated that the phosphorylation of S83 may be involved in not only its centrosomal localization but also its midbody localization during cytokinesis. Moreover, we showed that all of the cells expressing S83C mutant are not localized to the centrosome (Fig. 6C). The result suggests that the phosphorylation of S83 is not the only cause for centrosomal localization of Lats2 and therefore further studies are required. Recently, a report has shown that Lats2 kinase activity and two LATS conserved domains (LCD1 and LCD2), two stretches of highly conserved sequence in amino-terminus between Lats1 and Lats2, are required for Lats2 to suppress tumorigenicity and to inhibit cell proliferation (Li et al. 2003). It is notable that S83 locates in LCD1. The phosphorylation of S83 by Aurora-A may be important for Lats2 to suppress tumorigenicity and to inhibit cell proliferation via centrosomal regulations.

Taken together, our data suggest that Lats2 is a novel centrosome-associated kinase that may be involved in regulating the centrosome and/or mitotic spindle downstream of Aurora-A.

Experimental procedures

Cell culture, cell cycle synchronization and transfections

All cell lines were maintained in Dulbecco's modified Eagle's medium (DMEM) with 10% foetal calf serum (FCS, HyClone, Logan, UT, USA), 100 U/mL penicillin and 100 µg/mL streptomycin. HeLa cells were synchronized to enter the G1/S phase by the thymidine-aphidicoline double block and release protocol (Tsuruga et al. 1997). Cells at the G2/M phase were collected 9 h after release. Mitotic cells were only obtained by shaking-off after incubation for 18 h in medium containing 80 ng/mL nocodazole. Cell synchrony was monitored by Western blotting with anti-Cyclin B antibody or by FACS analysis (Becton-Dickinson, Franklin Lakes, NJ, USA). Transient transfection of HeLa S3 and 293T cells were carried out using LipofectAMINE or PLUS reagents according to the manufacturer's instructions (Invitrogen, Carlsbad, CA, USA).

Plasmids and site-directed mutagenesis

To isolate the complete human LATS2 cDNA, we screened a human placenta cDNA library (Clontech), using as a probe HindIII-PstI fragments from the pAP3neo-HsLATS2 plasmid that contains partial Lats2 cDNA (Yabuta et al. 2000). The nucleotide sequences of both strands of the isolated clone were determined by the dideoxy chain termination method. To construct the pCMVmyc-Lats2 full plasmid, we prepared BamHI-XcmI fragments by polymerase chain reaction (PCR) using the isolated clone as a template. These fragments were inserted into the BamHI and XcmI sites of the pCMV-HsLATS2 plasmid (Yabuta et al. 2000). For expression in bacteria, full Lats2 cDNA was released from the pCMVmyc-Lats2 full plasmid by BamHI and XhoI cleavage and recloned into the pGEX4T vector to produce pGEX4T-Lats2 (Amersham Pharmacia Biotech, Piscataway, NJ, USA). pGEX-Lats2N was constructed by ligating the EcoRI-HpaI fragment from pAP3neo-HsLATS2 into EcoRI and SmaI sites of pGEX4T2. pGEX-Lats2C was constructed by ligating the BamHI-XhoI fragment from another plasmid, pCMVmyc-Lats2C, into the BamHI and XhoI sites of pGEX4T. The pCMVmyc-Lats2C plasmid had been constructed by ligating the HincII-NotI fragment from pAP3neo-HsLATS2 into the EcoRV and NotI sites of pCMVmyc. pGEX-Lats2-79-257 was constructed by digesting pGEX-Lats2N with NotI followed by self-ligation. pCMV6myc-Lats2 full was constructed by ligating the BamHI-XhoI fragment from pCMVmyc-Lats2 full into the BamHI and XhoI sites of pCMV6myc. pCMV6myc-Lats2-1-393 was constructed by ligating the BamHI-PmlI fragment from pGEX-Lats2 into the BamHI and EcoRV sites of pCMV6myc. The other truncated Lats2 mutants were constructed in pGEX4T by PCR with the following primers which contain either an EcoRI site or a XhoI site: Lats2-1-78, F11 (5′-CCGGAATTCATGAGGCCAAAGAGTTTTCCT-3′) and R9 (5′-ATACTCGAGCCTCAAGGCTTTCTGATAAGG-3′); 1-118, F11 and R3 (5′-ATACTCGAGGCCAGCCATCTCCTGGTC-3′); 79-118, F2 (5′-GCGGAATTCGAAATCAGATAT TCCTTGTTG-3′) and R3; 79-151, F2 and R2 (5′-ATACTCGAGCGCACAATCTGCTCATTCC-3′); 113-151, F3 (5′-ATAGAATTCGACCAGGAGATGGCTGGC-3′) and R2. All point mutants of Lats2 and Aurora-A KD (kinase dead: D273E) (Shindo et al. 1998) were generated by site-directed mutagenesis using the QuickChange Site-Directed Mutagenesis Kit (Stratagene, La Jolla, CA, USA) according to the manufacturer's instructions. Aurora-A (Aik2) (Zhou et al. 1998), Aurora-B (AIM1) (Tatsuka et al. 1998) and Aurora-C (AIE2) (Tseng et al. 1998) cDNA were generated by PCR from the human placenta cDNA library. All amplified sequences were confirmed by DNA sequencing.

Expression and purification of recombinant proteins

For the expression of GST-fused Lats2 mutants, Aurora-A, Aurora-B, Aurora-C and Aurora-A KD, pGEX plasmids with the appropriate cDNAs were introduced into bacteria. The cultures were induced with 0.5 mm isopropyl β-d-thiogalactopyranoside (IPTG) and incubated at 37 °C for 6 h. Cells were collected and lysed in PBS containing 1% Triton X-100, 2 µg/mL leupeptin, 10 µg/mL aprotinin, 1 mm PMSF, 1 mm benzamidine, 1 mm NaF and 1 mm Na3VO4 by brief sonication. After centrifugation, the clear lysate was adsorbed to Glutathione Sepharose 4B (Amersham Pharmacia Biotech) and eluted with 10 mm reduced glutathione. GST-DNA-PK, GST-Nek2, GST-Nop10 and active Cyclin-Cdk kinase complexes (Cyclin D1/Cdk4, Cyclin B/Cdc2, Cyclin E/Cdk2 and Cyclin A/Cdk2) were obtained from MBL Co. Ltd (Japan). Rb C-terminus (amino acids 701-928) was purchased from New England Biolabs (Beverly, MA, USA).

Antibodies

The generation and specificity of the 3D10 anti-human Lats2 monoclonal antibody has been previously described (Yabuta et al. 2000). To generate an anti-Aurora-A polyclonal antibody, rabbits were injected with a recombinant GST-fused full-length Aurora-A protein. The antisera were then affinity-purified against the protein. Anti-γ-tubulin polyclonal antibody (Sigma, St Louis, MO, USA) was purchased. Anti-HA polyclonal, anti-Cyclin B, anti-Myc and anti-GST monoclonal antibodies were obtained from MBL Co. Ltd.

Generation of anti-phospho-Ser83 monoclonal antibody and Western blotting

To establish a mouse hybridoma that produces anti-phospho-S83-Lats2 antibodies, mice were immunized subcutaneously with the KLH-conjugated phosphopeptide CREIRYS(PO3H2)LLPF (amino acids 78–87) emulsified in Freund's complete adjuvant. Thereafter, the mice were boosted four times at biweekly intervals with the KLH conjugate in Freund's incomplete adjuvant. B-cell hybridomas were generated from the spleen cells of these mice and the 3B11 anti-phospho-S83 antibody was affinity-purified by the phospho-antigen-peptide column. To eliminate non-specific antibodies reacting with the unphosphorylated antigen peptide, the antibody preparation was passed through a non-phospho-Lats2-peptide (CREIRYSLLPF) column. The specificity of the 3B11 antibody was confirmed by both Western blotting (shown in Fig. 3C) and ELISA (data not shown). To detect the in vivo phosphorylation of S83, immunoprecipitations were performed with 5 µg of either the 3D10 or 3B11 antibody, after which the immunoprecipitates were resolved by SDS–PAGE and transferred to polyvinylidene difluoride (PVDF) membranes (Millipore Corporation, Bedford, MA, USA). Western blotting was performed with 5 µg of 3D10 antibody in TBST (100 mm Tris-Cl, pH 7.5, 150 mm NaCl, 0.05% Tween-20) containing 1% BSA. Western blotting using antibodies other than 3D10 and 3B11 was performed in TBST containing 5% non-fat milk.

In vitro kinase assays and immunoprecipitations

In vitro kinase assays were performed with 1 µg of GST-purified kinases and 2 µg of GST-purified substrates for 30 min at 30 °C in kinase buffer (20 mm Tris-HCl, pH 7.5, 10 mm MgCl2, 5 mm MnCl2, 1 mm DTT, 1 mm NaF, 0.1 mm Na3VO4, 10 mmβ-glycerophosphate) containing 20 µm ATP and 10 µCi [γ-32P]ATP. To examine the interaction between 6Myc-Lats2 and 1-393 and GFP-Aurora-A, the transfected 293T cells were lysed in lysis buffer A (50 mm Tris-HCl, pH 7.5, 250 mm NaCl, 1 mm EDTA, 0.2% NP-40, 1 mm PMSF, 1 µg/mL aprotinin, 2 µg/mL leupeptin, 1 µg/mL pepstatine A, 1 mm NaF, 1 mm Na3VO4). After centrifugation, the clear lysates were immunoprecipitated with 2 µg of anti-Myc or anti-GFP antibodies for 3 h at 4 °C. The immune complexes were collected by adding 30 µL of 50% protein G sepharose (Amersham Pharmacia Biotech) slurry. The complexes were washed five times with lysis buffer B (50 mm Tris-HCl, pH 7.5, 50 mm NaCl, 1 mm EDTA, 0.1% NP-40).

Immunofluorescence staining

HeLa S3 cells that transiently expressed GFP-full length Lats2 or GFP alone were fixed by sequential incubations with 4% formaldehyde in PBS, 0.1% Triton X-100 in PBS and then 0.05% Tween-20 in PBS, each for 10 min at room temperature. To prepare mitotic cells, the transfected cells were blocked at the S phase by the addition of 2.5 mm thymidine for 22 h. They were then released from the block by replacing the medium with fresh medium without drugs. They were fixed 12 h later. After being washed, cells were incubated with anti-Aurora-A antibody or anti-γ-tubulin antibody, followed by incubation with Texas Red (Amersham Pharmacia Biotech) or AlexaFluor 594 (Molecular Probes, Eugene, OR, USA)-conjugated anti-rabbit/mouse immunoglobulin G as previously described (Tsuruga et al. 1997). To visualize 6Myc-Lats2 or its derivatives, cells expressing each 6Myc-Lats2 derivative were fixed and stained with anti-Myc antibody followed by Alexa-Fluor 488-conjugated anti-mouse immunoglobulin G. DNA was stained by Hoechst 33258 (Sigma). Immunofluorescence staining with the 3B11 antibody was performed as similar way. The stained cells were observed with an Axiophot microscope (Zeiss) or BX51 microscope (Olympus).

Acknowledgements

We thank Dr P. Hughes for critically reading the manuscript. This work was supported by a Grant-in-aid for Scientific Research on Priority Areas from the Ministry of Education, Culture, Sports, Science and Technology of Japan, and grants from the Osaka Cancer Society, the Yasuda Medical Research Foundation, the Welfide Medical Research Foundation, the Japanese Foundation for Multidisciplinary Treatment of Cancer and the Osaka Cancer Research Foundation.

Ancillary