Roles of Aurora Kinases in Mitosis and Tumorigenesis

Aurora kinases represent a novel family of serine/threonine kinases crucial for cell cycle control. The first Aurora kinase was discovered in Drosophila (1). Because its mutations resulted in a failure of centrosome separation, leading to the formation of monopolar spindles, it was given the name “Aurora,” reminiscent of the North Pole. Since then, homologues of Aurora have been identified in different species. Saccharomyces cerevisiae has a single Aurora kinase: Ipl1 (increases in ploidy 1; ref. 2). In Caenorhabditis elegans (3, 4), Drosophila (1, 5), and Xenopus (6, 7), there are two types of Aurora kinases: Aurora-A and Aurora-B. Mammals own at least three Aurora kinases: Aurora-A, Aurora-B, and Aurora-C (for nomenclature, see ref. 8).

Special interest in Aurora kinases has arisen since people discovered that defects in these kinases lead to severe mitotic abnormality. Silencing of Aurora-A leads to abnormal spindle morphology in C. elegans, Drosophila, and human cells (1, 9, 10), whereas disruption of Aurora-B causes chromosome misalignment and cytokinesis failure in both Drosophila and HeLa cells (11, 12), indicating their multiple roles in centrosome function, chromatid separation, and cytokinesis. Little is known about Aurora-C until recent reports show that Aurora-C is also a chromosomal passenger protein, and that it binds directly to INCENP and survivin in vitro (13-15). Given the facts that Aurora-C-INCENP can phosphorylate endogenous histone H3 in mammalian cells, and that Aurora-C could rescue the Aurora-B–silenced multinucleation phenotype in human cells, Aurora-B and Aurora-C may be functionally overlapping (13-15). Here, we will refer primarily to Aurora-A and Aurora-B because the expression of Aurora-C is largely limited to the testes, and because there is a paucity of Aurora-C–related data.

Aurora kinases are overexpressed in a variety of tumor cell lines (16-18), suggesting that these kinases might play a role in tumorigenesis. In particular, Aurora-A can transform certain cell lines when overexpressed (17, 19, 20). Inhibitors of Aurora kinases are used in cultured cells and xenograft models to treat cancer, and an encouraging effect has been shown (21, 22).

In this article, we mainly review the roles of Aurora kinases in the regulation of mitotic progression and tumorigenesis. We will also discuss these kinases as potential targets for cancer diagnosis and therapy.

Aurora kinases comprise mainly two domains: a regulatory domain in the NH₂ terminus and a catalytic domain in the COOH terminus. The regulatory domain is diverse largely, whereas the catalytic domain with a short segment of diverse COOH terminus shares >70% homology among Aurora-A, Aurora-B, and Aurora-C (Fig. 1A). There is a D-Box in the COOH terminus and an A-Box in the NH₂ terminus of Aurora kinases, which are responsible for degradation (23-26). Sequence analysis and point mutations revealed that phosphorylation at the threonine site within the catalytic domain is required for the kinase activity (27, 28).

Crystal structures of the Aurora-A and Aurora-B catalytic domain have been resolved recently (28-30). Like other serine/threonine protein kinases, Aurora kinases contain two lobes in their catalytic domain joined up by a hinge. The N lobe is responsible for positioning the ATP phosphate group through an αC helix, whereas the activation loop within the C lobe is able to harbor substrates (28, 31). Only when the two lobes arrive at a certain conformation can the Aurora kinases fulfill their kinase functions. Thus far, the NH₂-terminal domain has not yet been explored in the tertiary structure level.

Despite great similarities in sequences and structures, Aurora kinases are completely different in their subcellular distribution (Fig. 1B). Aurora-A localizes to the pericentriolar material from the end of S phase to the beginning of the next G₁ and spreads to the pole proximal ends of spindle microtubules during mitosis. In contrast, Aurora-B remains in the nucleus and moves to centromeres from prometaphase to metaphase. After anaphase begins, Aurora-B relocates gradually to the midzone and persists at the midbody until cytokinesis is completed. Moreover, Aurora-B undergoes different localizations along with at least three other partners: INCENP, survivin, and Borealin. They form a tight complex within the cell during mitosis and are named “chromosomal passengers” as they move precisely from site to site at specific times.

Aurora-A and Aurora-B exhibit divergent functions in mitotic control, corresponding to different subcellular distribution of the two kinases. Aurora-A is mainly involved in centrosome function, mitotic entry, and spindle assembly, whereas Aurora-B participates in chromatin modification, microtubule-kinetochore attachment, spindle checkpoint, and cytokinesis. Along with these two kinases, different partners and substrates take part in these processes.

The enrichment of Aurora-A either to the centrosome or microtubule compartment suggests that Aurora-A may play dual roles in mitotic spindle assembly (Fig. 2).

In G₂ phase, Aurora-A first localizes to pericentriolar material and is activated by the LIM protein Ajuba (32), which is required for centrosome maturation. Centrosome maturation takes place every cell cycle, referring to the recruitment of several pericentriolar material proteins that enhance microtubule nucleation ability in the late G₂ phase, such as γ-tubulin and some coiled-coil proteins (33). Depletion of Aurora-A decreases centrosomal γ-tubulin intensity as well as two other pericentriolar material components in C. elegans embryos (9). It has been corroborated in Drosophila and mammalian cells that Aurora-A is required to localize centrosomal protein centrosomin, which could bind γ-tubulin and other centrosomal components (34, 35). Moreover, Aurora-A is necessary to recruit D-TACC/maskin to the centrosome and phosphorylates D-TACC/maskin at the centrosome in Drosophila and Xenopus (36-38). D-TACC/maskin forms a complex with Msps/XMAP215 and stabilizes microtubules around the centrosome by antagonizing the depolymerization activity of MCAK (37, 39). Therefore, it is convincing that Aurora-A facilitates centrosome maturation in microtubule nucleation ability by recruiting and/or phosphorylating centrosomal components, sequentially contributing to mitotic spindle assembly.

The second role of Aurora-A in spindle assembly is exerted when it localizes to the pole proximal ends of microtubules after nuclear envelope breakdown, prevailingly considered as a part of the Ran-spindle assembly pathway. Ran is a small GTPase and has been implicated in multiple mitotic events. During spindle assembly, high concentration of RCC1 near the chromosomal region transforms Ran-GDP into a GTP-bound state, which later releases the spindle assembly factors, such as TPX2 and NuMA, by inhibiting importin-α/β (40). The Ran-spindle assembly pathway was explored more extensively after the relation between TPX2 and Aurora-A was uncovered. The NH₂ terminus of TPX2 directly interacts with the COOH terminus of Aurora-A (41), triggering conformational changes in the activation segment of the kinase, thus not only stimulating the autophosphorylation but also burying the phosphothreonine site from protein phosphatase 1 dephosphorylation (28). TPX2 is also required for targeting Aurora-A to the spindle (41, 42). The interaction between Aurora-A and TPX2 undergoes the regulation of Ran-GTP gradient (42). Recently, profound evidence was shown that Aurora-A–coated magnetic beads could function as MTOCs in the presence of Ran-GTP in Xenopus egg extracts and induce bipolar spindle formation without centrosomes and chromatin, confirming the significance of Aurora-A in centrosome function and spindle assembly (43). It might be a hint that Ran directs spindle assembly by transforming its gradient of active form to the activity of protein kinases, such as Aurora-A, hence passing the commands on to the yet unknown downstream substrates. Therefore, identifying the substrates of Aurora-A will be an exciting task in investigating the mechanism of spindle assembly.

Interestingly, three newly published articles identified HURP as a Ran-imporin-β–regulated protein involved in spindle assembly (44-46). HURP mainly concentrates at the chromosome proximal ends of mitotic spindle (45, 46), where it stabilizes kinetochore fibres and promises a normal interkinetochore tension and efficient kinetochore capture (45, 46). Such HURP distribution exhibits a Ran-GTP dose-dependent characteristic and is damaged by excessive importin-β or Ran-GDP (44, 45). It has been reported that Aurora-A phosphorylates and stabilizes HURP in vitro and in vivo during mitosis (47). Given the reverse gradient of their distribution on microtubules, it would be really important to investigate how Aurora-A and HURP correspond to the signal of the Ran pathway.

Duplicated centriole pairs separate at the G₂-M transition, marching to the opposite poles of the cell for proper establishment of bipolar spindle. There is evidence that centrosome separation requires functional Aurora-A. RNA interference or microinjection of Aurora-A antibody in HeLa cells could prevent centrosome separation (10). In C. elegans embryos, Aurora-A knockdown by RNA interference could interrupt centrosome separation and spindle assembly (9). In Xenopus, a dominant-negative mutation of Aurora-A results in monopolar spindle (6, 48), a phenomenon similar to the inhibition of a kinesin-related protein XlEg5 (48). Because XlEg5 is phosphorylated by Xenopus Aurora-A both in vivo and in vitro (48), it might be possible that Aurora-A regulates centrosome separation by phosphorylating the kinesin-related proteins.

Apart from the spindle assembly and centrosome separation, Aurora-A contributes to transition from G₂ to M phase. Suppression of Aurora-A by RNA interference results in G₂-M arrest of HeLa cells and promotes apoptosis (49). Phosphorylation of CDC25B at the centrosome by Aurora-A was discovered in HeLa and U2OS cell lines, and this phosphorylation is an important step to relocalize cyclin B1 to the nucleus and activate cyclin-dependent kinase 1 for the cell to enter mitosis (50). It was also found that ectopic expression of Aurora-A could bypass the G₂-M checkpoint activated by DNA damage (50).

It has been found that Aurora-B phosphorylates histone H3 at Ser¹⁰ and Ser²⁸ (51, 52) as well as its variant CENP-A at Ser⁷ (53, 54) in mitosis. H3-S10 phosphorylation is now widely used in Aurora-B kinase assay; yet, the function remains largely unknown. In Tetrahymena, H3-S10 phosphorylation is required for chromosome condensation (55, 56); however, the connection between these two events is still elusive in mammalian cells. Moreover, a recent experiment showed that H3-S10 phosphorylation by Aurora-B allows the routine disassociation of HP1 from heterochromatin (57, 58), but the biological effect of HP1 release in mitosis is still unclear. Interestingly, Aurora-A also phosphorylates CENP-A (54). Phosphorylation of CENP-A by Aurora-A at Ser⁷ in prophase is required for Aurora-B concentration to inner centromeres in prometaphase. Maintenance of later CENP-A phosphorylation state depends on Aurora-B activity and is required for proper microtubule-kinetochore attachment (54). It would be meaningful to explore how Aurora-A and Aurora-B cooperate in understanding the subtle regulation of mitosis.

Accurate separation of chromatids requires attachment of each kinetochore to microtubules from the opposite poles. Two mechanisms in cells are raised to regulate this process: one is the correction of error attachment, and the other is the spindle checkpoint. Because Aurora-B inhibition causes a higher frequency of syntelic and/or merotelic attachments in cells during prometaphase and drives these cells to enter anaphase with no arrest and misaligned chromosomes (12, 22), it has long been proposed that Aurora-B plays a crucial role in both pathways.

Recently, two laboratories have discovered that Aurora-B phosphorylates centromeric MCAK to inhibit its microtubule depolymerization activity during the mitotic spindle assembly (59, 60). They found that, in mitosis, if microtubule-kinetochore attachment is in a bipolar fashion, Aurora-B may phosphorylate MCAK to keep the kinetochore microtubules in a more stable form. Otherwise, the phosphorylation of MCAK by Aurora-B is abolished, and the kinetochore microtubules are depolymerized to allow other microtubules to catch the unattached kinetochores (Fig. 3A). Phosphorylated and unphosphorylated MCAK proteins exhibit slight difference in localization, with the former closer to the inner centromere where Aurora-B is enriched and the latter closer to inner kinetochore bound of protein phosphatase 1 (59, 61, 62). It is conceivable that the dynamics of phosphorylation/dephosphorylation modification of MCAK protein might be the way that Aurora-B and protein phosphatase regulate the kinetochore-microtubule connection.

The other surveillance mechanism is the spindle checkpoint, which checks whether unattached kinetochores exist. The checkpoint proteins are located on unattached kinetochores and inhibit the APC/C activity to prevent cells from entering anaphase. Aurora-B is involved in the recruitment of checkpoint proteins to kinetochores (Fig. 3B). Cytostatic factor extracts depleted of Aurora-B show diminishment of Mps1, Bub1, CENP-E, Bub3, Mad1, and Mad2 to kinetochores (63). When cells treated with ZM447439, an inhibitor of Aurora-B, or depleted of Bub1 were done with checkpoint inducement, Bub1 and Aurora-B function in two different signaling pathways and converge on the mitotic checkpoint complex, which includes BubR1, Bub3, Mad2, and Cdc20 (64). However, whether Aurora-B is directly involved in checkpoint activation, or whether it just guarantees recruitment of checkpoint proteins, remains a mystery. Experiments using temperature-sensitive kinetochore mutants in budding yeast revealed that when kinetochores undergo tension defects, Aurora-B creates unattached kinetochores to activate the spindle checkpoint (65), but in higher eukaryotes, this has not yet been proven.

Aurora-B plays an essential role in cytokinesis. Depletion of Aurora-B produces polyploidy due to a failure in cytokinesis in C. elegans embryos and Drosophila S2 cells (4, 66). Although the mechanism has not yet been elucidated, several substrates of Aurora-B were identified, including the type III intermediate filament desmin and vimentin (67, 68), midbody component ZEN-4/MKLP1 (69, 70), and other proteins like TACC1 (71) and MgcRacGAP (72). Aurora-B and Rho, a prominent regulator of cytokinesis, share certain substrates but exert phosphorylation independently, suggesting that Aurora-B and Rho might coordinate in cytokinesis control (73). Moreover, Aurora-B activates Rho by converting MgcRacGAP to a RhoGAP during cytokinesis (72), indicating that Aurora-B is multifunctional in cytokinesis regulation.

The activity of Aurora kinases undergoes strict regulation during mitosis. Aurora-A reaches the maximum activity at the G₂-M transition, whereas the activity of Aurora-B persists from metaphase to the end of mitosis (74). This mechanism of regulation turns out to be a complicated process of phosphorylation and degradation.

Different from most kinases within the cell, Aurora kinases are activated through autophosphorylation at the activation loop (29).

Aurora-A is activated mainly by Ajuba (32) and TPX2 (42) as well as protein phosphatase inhibitor-2 (75), the focal scaffolding protein HEF1 (76), and a newly identified Bora protein (77). Phosphorylated Aurora-A is dephosphorylated by protein phosphatase 1 (78). Up to now, TPX2 obtains the best understanding as an Aurora-A activator. In the absence of TPX2, the Aurora-A activation site is exposed to protein phosphatase 1 and is dephosphorylated in a crucial threonine; after binding to TPX2, the activation segment of Aurora-A rotates so that the phosphothreonine is buried inside and protected from protein phosphatase 1 (28).

A recent report raised a new question in screening for TPX2 homologue in C. elegans. The report discovered that a protein named TPXL-1 has homology with the NH₂-terminal domain of TPX2 (79). Surprisingly, although TPXL-1 could bind to and activate Aurora-A, it has neither a NLS signal nor the ability of chromatin-dependent microtubule nucleation (79), which have been shown to be the basic characteristic of TPX2. Thus, it has been thought that TPXL-1 is not a true functional orthologue of TPX2 in C. elegans (80), and who will be the real activator of Aurora-A?

Less is investigated of Aurora-B regulators. Prevailing view believes that Aurora-B is regulated by other members of the passenger protein complex, especially activated by INCENP (81, 82). Crystal structure analysis showed that INCENP forms a crown around the N lobe of Aurora-B (30). Different from Aurora-A/TPX2 complex, INCENP induces the active conformation of the T loop allosterically. INCENP activates Aurora-B in a two-step process: it partially activates the kinase at the first step, with the N lobe and C lobe of Aurora-B in an open position; then, full activation of the kinase carries on after INCENP phosphorylation in a conserved TSS motif near the COOH terminus, and the two lobes of Aurora-B close (30). Up to the present, whether survivin bound to Aurora-B stimulates Aurora-B kinase activity remains contradictory (81, 82). Recombinant survivin can enhance an Aurora-B IP kinase assay in Xenopus mitotic extracts for >10-fold, and overexpression of survivin in HEK293 cells also elevates the Aurora-B kinase activity (81, 83). Nevertheless, another in vitro kinase assay using recombinant proteins showed that survivin does not detectably influence the activity of Aurora-B kinase (82).

Aurora kinases are barely detected after cells exit mitosis and enter G₁ phase. Destruction of both Aurora-A and Aurora-B is mediated by APC/C^Cdh1 (23, 25, 84). Several domains within their sequences have been identified for degradation, including KEN-box, D-box, and A-Box (23, 24, 26). Nevertheless, they are not all necessary for the degradation of both Aurora-A and Aurora-B. Aurora-A destruction requires D-box and A-box but not KEN-box (26), whereas Aurora-B degradation needs intact KEN-box and A-box (23). This difference might partially explain the distinct temporal regulation of their kinase activity.

It is worth pointing out that Aurora-A and Aurora-B share little regulators or substrates despite their high similarity in sequence. How are they discriminated, for example, by TPX2? Using structure-based mutation, Byliss and Eyer found that a single amino acid mutation in Aurora-A to the homologue site of Aurora-B abolishes TPX2 recognition and vice versa (31, 85). It provides us a view that Aurora kinases experience delicate regulation in regulating mitosis events.

Aurora kinases are overexpressed in a wide range of human cancers. Human Aurora-A was first isolated as the product of gene BTAK (breast tumor amplified kinase, also named STK15) on chromosome 20q13, a region that is commonly amplified in primary breast tumors, colorectal cancers, and other cancer cell lines, including breast, ovarian, colon, prostate, neuroblastoma, and cervical cell lines (17, 19, 86). More strikingly, overexpression of Aurora-A transforms NIH 3T3 cells and Rat1 fibroblasts in vitro; when these cells were injected into nude mice, they grew into tumors (17, 19, 20). However, whether Aurora-A is a potential oncogene remains controversial because no more cell lines were found to be transformed by overexpression of Aurora-A, and overexpression of Aurora-A in mouse model did not result in malignant tumor after a long latency (87). These data suggest that Aurora-A might be a promoting factor for tumorigenesis, and other factors are also needed in cell transformation process. For Aurora-B, the case is even more complicated. Aurora-B is also expressed at high level in primary human colorectal cancers and other tumor cell lines (18). Overexpression of Aurora-B results in multinucleation and polyploidy in human cells (16, 88). However, down-regulation or overexpression of an inactive form of Aurora-B yields the same polyploidization phenomenon. How Aurora kinases are involved in tumorigenesis would be discussed further below (Fig. 4).

Overexpression of Aurora-A causes centrosome amplification both in cell cultures and in rat mammary model (89-91). This kind of centrosome amplification seems not due to centrosome duplication but as a result of cytokinesis failure and consequent multinucleation (89). Up until now, no evidence has shown that Aurora-A functions in cytokinesis directly, and it would be more likely that Aurora-A causes cytokinesis failure in an indirect way, possibly by affecting Aurora-B or other proteins that directly regulate cytokinesis. Aurora-A–induced polyploid cells would be arrested by a “G₁ phase post-mitotic checkpoint” dependent on the p53-Rb pathway, and most of the cells would die through apoptosis. However, when these cells unfortunately lose p53 functions as in p53^−/− cells, they could not detect the hyperploidy and go on DNA replication and cell division, which might cause aneuploidy and abnormal centrosome number (89, 92).

Overexpression of Aurora-B also leads to centrosome amplification and increased ploidy, and the absence of p53 exacerbates this phenotype (88, 89). Nevertheless, Aurora-B–induced centrosome amplification is possibly due to other pathways. It has been reported that Aurora-B overexpression induces chromosomes lagging in metaphase, chromosome segregation error, and errors in cytokinesis, and thus may play a role in carcinogenesis (93).

Cells overexpressing Aurora-A override the mitotic spindle checkpoint and enter anaphase despite abnormal spindle formation (91). Interestingly, the checkpoint protein Mad2 remains at the kinetochores from prometaphase to anaphase, suggesting a persistently activated spindle checkpoint (91). Therefore, it might be possible that Aurora-A disrupts the interaction between Mad2 and cdc20 or prematurely activates APC/C proteolytic functions. It has been found that, when Aurora-A overexpression overrides the spindle checkpoint triggered by nocodazole, a microtubule destabilizer, BubR1 is significantly reduced, and a subunit of APC/C cdc27 could not be phosphorylated (94).

Depletion of Aurora-B or overexpression of an inactive form in cells would also compromise the spindle checkpoint because the activity of Aurora-B is required for checkpoint protein recruitment. Because cytokinesis would also fail without Aurora-B activity, it makes a good explanation why down-regulation of Aurora-B generates increased ploidy.

Aurora-A interacts with several tumor suppressor gene products, including p53 (92, 95, 96), BRCA1 (97), NM23-H1 (98), and Chfr (99), and it might be a powerful way to promote cell viability and tumorigenesis. Thus far, the interaction between Aurora-A and p53 has been well investigated. On one hand, p53 suppresses Aurora-A–induced centrosome amplification and cellular transformation (100), whereas on the other hand, Aurora-A phosphorylates p53 and thus inhibits p53 activity and promotes degradation of this protein. Two Aurora-A phosphorylation sites have been identified in p53. By phosphorylating Ser²¹⁵, Aurora-A abrogates the DNA-binding and transactivation activity of p53 and subsequently inhibits downstream target genes of p53, such as p21 and PTEN (96). Phosphorylation of Ser³¹⁵ will enhance the affinity of p53 with Mdm2, which promotes ubiquitination and degradation of the p53 protein (95). Besides, Aurora-A directly binds to BRCA1 and phosphorylates it at Ser³⁰⁸, and abortion of this phosphorylation process can decrease the number of cells in M phase in mouse embryo fibroblasts (97). In addition, Aurora-A is regulated by Chfr, a mitotic checkpoint protein and tumor suppressor. Chfr interacts with and ubiquitinates Aurora-A both in vitro and in vivo (99). Taken together, Aurora-A and tumor suppressor gene products are antagonistic in cell viability control. The balance of quantity and activity between Aurora-A and tumor suppressor gene products might be important to maintain the normal growth rate of cells.

Recent studies found that Aurora-A is a downstream target of mitogen-activated protein kinase 1/ER2 in pancreatic cancer (101). In human carcinogenesis, the Ras/Raf/MEK/ERK/MAP kinase pathway would be enhanced, and the cell response to growth signals is known to increase (102). Therefore, Aurora-A overexpression might drive cells into a faster growing state.

In addition, Aurora-A induces telomerase activity and human telomerase reverse transcriptase by up-regulation of c-Myc (103). Because increase of telomerase activity is frequently seen in many cancer cells and would elongate cell survival time, Aurora-A might promise cell a longer life.

Furthermore, both Aurora-A and Aurora-B promote inappropriate cellular mobility (23, 104), which underlies invasion and metastasis of tumor cells (105). Aurora-A enhances collagen I–induced cell migration and anchorage-independent growth in Aurora-A–stable cell line by activating RalA (104). In the case of Aurora-B, overexpression or nondegradation mutants could promote anchorage-independent growth in soft agar (23). However, the molecular mechanism has not yet been elucidated.

Cancers vary greatly in their cause, symptoms, response to treatment, and possibility of cure, which cause difficulty in their diagnosis and therapy. Considering that Aurora kinases play a critical role in tumorigenesis, it might be of great importance to evaluate Aurora kinases in cancer detection and therapy.

Elevated expression of Aurora-A has been detected in >50% of colorectal, ovarian, and gastric tumors and in 94% of invasive duct adenocarcinomas of the breast (21). Could Aurora-A serve as a clinical tumor prognosis marker? Microarray-based screening for molecular markers in medulloblastome revealed positive staining for Aurora-A and identified it as an independent prognostic factor for overall survival (106). However, experiments comparing the level of Aurora-A expression in human breast tumor specimens with clinical variables revealed no association between Aurora-A staining and tumor size, lymph node status, or hormone receptor status (107).

Despite the above debate, studies of the polymorphism of Aurora-A have proven that Aurora-A is a candidate low-penetrance tumor-susceptibility gene in mouse and human (108), raising the possibility that it might be useful as a caution of predisposition to cancer. Two functional coding single nucleotide polymorphisms as 91T>A and 169G>A are associated with occurrence of esophageal squamous cell carcinoma and increase the risk of esophageal cancer and gastric cancer (109-111). More clinical evidence is needed to answer if the genotype of Aurora-A would be involved in determining “cancer susceptibility.”

Cancers are characterized by the unrestrained growth of cells. Therefore, modulation of abnormal cell cycle regulation would be a valuable therapeutic strategy to many tumor types. RNA interference targeting Aurora-A suppresses tumor growth and enhances the sensitivity to chemotherapy and UV irradiation–induced apoptosis in human cells (112, 113), showing its potential as a target for cancer therapy.

Thus far, several inhibitors of Aurora kinases have been reported (for review, see refs. 114, 115). One of them, VX-680, was designed against the ATP-binding site of Aurora kinases and was proven a high potent, selective, and reversible inhibitor. It greatly inhibits tumor growth in three xenograft models in vivo, including leukemia and colon and pancreatic tumors (21). This molecule ensures the possibility to use Aurora kinases as a target for cancer therapy. Other inhibitors of Aurora kinases have been analyzed, such as ZM447439 (22) and Hesperadin (12). Although it is not clear whether Hesperadin would cause tumor cell death, ZM447439 has shown its ability. In colony formation assay, ZM447439 is more toxic to proliferating cells than nondividing cells (22), indicating that it might also be a selective drug to kill tumor cells. These inhibitors provide us a powerful tool to explore the relationship among molecular mechanisms, clinical symptom, and disease therapy.

The Aurora kinase family provides a new concept to understand the mitosis process, tumorigenesis, and the possible relationship between them. It is known that cyclin-dependent kinases provide a regulatory system of the cell cycle. They require cyclins as binding subunits to achieve their kinase activity and affect downstream substrates. Likewise, Aurora kinases also bind to their partners, such as TPX2 and INCENP, to maintain kinase activity. Interestingly, Aurora-A and Aurora-B show much difference in subcellular distribution and carry out divergent functions that are tightly related to their distribution, even if they share >70% homology in their catalytic domain. It might be intriguing to establish and compare the regulation networks of different Aurora kinases. At the mean time, it might be meaningful to investigate the relationship between the cyclin-dependent kinase regulatory system and the Aurora-like regulatory systems.

It is striking to note that deregulation of a single mitotic regulator might lead to huge risk of tumorigenesis. Of particular interest, Aurora kinases are shown to be candidate targets to inhibit tumor cell growth. We believe that uncovering the molecular function of Aurora kinases will bring great rewards in understanding cell cycle control and provide new strategies for drug design in cancer therapy.

We thank all the other members in our laboratory for valuable comments and discussion.