Targeting Aurora-2 Kinase in Cancer¹

Introduction

Advances in cancer research over the past 30 years have greatly increased our knowledge and understanding of the genetics of cancer, including the multistep basis of tumorigenesis (1). Simply stated, the multistep model suggests that a series of genetic alterations is necessary for the transformation of a normal cell into a cell with a malignant tumorigenic phenotype. Even though great progress has been made in identifying the genetic alterations responsible for the overall process, the underlying factors driving these genetic changes are not well understood. As we learn more about the abnormal physiology of cancer cells and about the underlying causes of these abnormalities, we should be able to find new targets for therapies that specifically affect cancer cells.

Two distinguishing characteristics of most cancer cells are (a) a change in the amount or organization of DNA compared with normal cells and (b) the loss of control of key regulatory cell cycle checkpoints (2, 3). It is arguable whether these two characteristics are acquired during the early events of tumorigenesis and are largely responsible for late-stage cancer phenotypes. It is likely that cancer is a disease of the cell cycle, but there are additional characteristics that contribute to the overall cancer phenotype.

Changes in DNA amount are most commonly due to gains or losses of whole chromosomes (tumor cell aneuploidy). The causes of aneuploidy are not fully understood, but the improper segregation of chromosomes during mitosis may be an important factor (4). Precise segregation of chromosomes is an important event in the progression of the cell cycle and is a highly complex process. There are many potential causes of improper segregation during mitosis. Recently, much attention has been focused on the centrosome, which in normal cells appears to orchestrate the changes required for proper chromosome segregation (5). This orchestration includes: (a) serving as an anchor for microtubule nucleation; (b) serving as an organizing center for transformation of the microtubules into a mitotic spindle apparatus; and (c) serving as a platform for recruitment of structural, motor, and catalytic proteins that constitute the centrosome complex (6). One set of centrosome-associated proteins is the family of kinases called AIRKs,³ which are thought to be key regulators of centrosome duplication, chromosome segregation, and cytokinesis (7). The nomenclature of AIRK family members in various organisms is complicated and sometimes confusing. Table 1 shows a summary of AIRK family members and the nomenclature found throughout the scientific literature.

An additional function of AIRKs, which is just beginning to be understood, is their role as regulators of entry into and exit from mitosis. This putative function is based primarily on pioneering work in Xenopus, where it has been shown that Eg2, a Xenopus AIRK, is responsible for signaling entry into the G₂-M phase of meiosis II in developing oocytes, through control of the expression of cell cycle-regulatory proteins in a transcription-independent manner (8, 9). Considering this finding, it is tempting to speculate a role for AIRKs in regulating entry into mitosis in mammalian cells as well, thereby suggesting an additional role for AIRKs in tumorigenesis. Combined with the role that AIRKs play in controlling the integrity of chromosome separation, this implicates AIRK family members as possible components that provide the underlying driving force in the progression of tumorigenesis. The remainder of this review will focus on describing the oncogenic role that AIRKs (specifically aurora-2) play in tumorigenesis as regulators of both centrosome function and direct mitotic control.

AIRKs

AIRKs are serine/threonine kinases that show conservation of both structure and function throughout eukaryotic organisms. AIRKs have a COOH-terminal catalytic domain that is highly conserved within the family and an NH₂-terminal domain that is variable among organisms (10). The most conserved motif is the putative activation loop, where many protein kinases are activated by phosphorylation. The activation loop contains a highly conserved threonine residue that is likely phosphorylated by regulatory kinases. Mutating this threonine (Thr²⁸⁸ in aurora-2) to an acidic residue results in an enzyme with hyperactivity (11), suggesting that the threonine residue is a target for kinase activation.

The AIRKs are associated with centrosome and microtubule dynamics. They localize to the centrosome around the pericentriolar material, microtubules, and the mitotic spindle, in keeping with their role in chromosome segregation and cell division (12). A temperature-sensitive mutant of the Saccharomyces cerevisiae yeast AIRK, Ipl1, demonstrates abnormal chromosome number, indicating that Ipl1 kinase regulates chromosome segregation events (10).

Three AIRK genes have been identified in Homo sapiens. The first human AIRK family member is aurora-2, also known as aurora A, aik1, btak, stk15, and ark1 (Refs. 11–13; Table 1). Aurora-2 maps to chromosome 20q13.2-q13.3, a region frequently amplified in human malignant tumors (14), and its protein product localizes to the centrosome and spindle poles (12). Specific inhibition of aurora-2 kinase by mutation leads to formation of a monopolar mitotic spindle, indicating its requirement for centrosome distribution and formation of the bipolar spindle (Refs. 15 and 16; Fig. 1). A second human AIRK gene is aurora-1, also called aurora B, aik2, ark2, and stk12 (11, 17), which maps to chromosome 17p13 and encodes a protein that localizes to the midbody of the microtubules. The inhibition of aurora-1 kinase results in the formation of multinucleated cells, suggesting a role in cytokinesis (18). The third human gene is aurora-3, also referred to as aurora C, aie2, and aik3 (19, 20), which maps to chromosome 19q13.3. Its gene product localizes to the centrosome only during anaphase (10). The expression of aurora-3 kinase has been shown to be testis specific (21).

Aurora-2 Kinase in Tumorigenesis

AIRKs are key regulators of centrosome maturation, chromosome segregation, and cytokinesis. Of the three human AIRKs, aurora-2 has been most strongly implicated in carcinogenesis. Aurora-2 is overexpressed and/or amplified in various malignancies such as primary colorectal, breast, and gastric tumors as well as breast, ovarian, colon, prostate, neuroblastoma, cervical, and gastric cancer cell lines (Table 2). In addition to being overexpressed, the gene encoding aurora-2 kinase is often amplified in many tumor types, suggesting one possible mechanism for its overexpression (11, 13). Using microarray studies, we recently showed that aurora-2 is overexpressed and its gene is amplified in human pancreatic tumors and pancreatic cancer cell lines (Refs. 22 and 23; Fig. 2); this has recently been confirmed by others (24). Aurora-2 kinase is of considerable interest as a potential drug target because it shows both gene amplification and up-regulation in cancer tissue. These findings are quite reminiscent of the situation in the early exploration of HER-2/neu as a target in breast cancer in the mid-1980s (25).

Aurora-2 kinase has been shown to play an oncogenic role in cell culture studies (11). Overexpression of aurora-2 transforms rat fibroblasts, and ectopic expression of aurora-2 in mouse NIH 3T3 cells leads to transformation in vitro and to the appearance of abnormal centrosome numbers (26). Overexpression of aurora-2 kinase in near-diploid human breast epithelial cells results in centrosome abnormalities with induction of aneuploidy (26). Overexpression of aurora-2 and centrosome amplification have been shown to be early events in tumorigenesis in a rat mammary carcinogenesis model (27). These findings lend support to the hypothesis that aurora-2 amplification and overexpression help drive the multistep acquisition of genetic alterations required for tumorigenesis. In this review, we propose that the oncogenicity of aurora-2 results from two related yet distinct functions of the protein, namely, (a) chromosome segregation and control of genomic stability and (b) regulation of entrance into mitosis. The remainder of the review will discuss the involvement of aurora-2 in these processes and how changes in aurora-2 expression might lead to tumorigenesis.

Aurora-2 in Genomic Instability

Genomic instability, manifested by changes in the amount and organization of DNA compared with normal cells (2), has proven to be a hallmark of most cancer cells. Genomic instability is grouped into four categories of DNA alterations: (a) subtle DNA sequence changes; (b) whole gains and losses of chromosomes; (c) chromosome translocations; and (d) gene amplifications (28). The molecular mechanisms responsible for mediating each type of genomic instability are now beginning to be understood. The most common manifestation of genomic instability in cancer cells involves the gain or loss of whole chromosomes, resulting in aneuploidy. In colorectal cancers, for example, aneuploidy occurs 10–100 times more frequently than any other type of genomic instability (29). The causes and outcomes of aneuploidy are still areas of active investigation and disagreement. As shown in many tumor types, problems in chromosome segregation can be caused by improper mitotic checkpoint function (4), allowing a cell to proceed with mitosis before formation of the appropriate mitotic machinery. A further cause of aneuploidy involves the loss of regulation of centrosome duplication and function, resulting in cells with abnormal centrosome numbers (4).

A key aspect of chromosome segregation during mitosis is the duplication of the centrosome and the organization of the mitotic spindle poles. A centrosome is typically composed of two centrioles encompassed by a relatively undefined electron-dense pericentriolar material. A centriole is composed of a nine-triplet array of microtubules made up of α-, β-, δ-, and ε-tubulin together with centrin proteins (30). The pericentriolar material is a network of filamentous polymers speckled with small ring-shaped structures of γ-tubulin. γ-Tubulin is believed to serve as the nucleation site for microtubule elongation (31).

There are numerous stages during centrosome duplication and function that can be regulated. During the G₁ phase of the cell cycle, the two centrioles of a single centrosome remain together to function as a single unit, and then they separate into two components at the onset of S phase. The individual centrioles give rise to new centriole partners, resulting in two distinct centrosomes capable of microtubule assembly and mitotic spindle organization. Centrosomes organize microtubule polymerization during interphase; however, during mitosis, there is a dramatic increase in microtubule nucleation potential. The increase in microtubule nucleation is believed to assure the random contact between centrosome microtubules and chromosome kinetochores (32). Upon establishment of the mitotic spindle, the spindle microtubules shorten, and sister chromatids segregate and move toward opposite spindle poles. Motor proteins such as dynein (33) and kinesins are involved in the dynamic behavior of the mitotic spindle (34, 35). Furthermore, phosphorylation or dephosphorylation of key centrosomal proteins has been shown to regulate the rate of microtubule assembly and the organization of the mitotic spindle (36). The precise mechanism by which the centrosomal cycle or microtubule assembly and disassembly is regulated has not been fully elucidated, but members of the aurora family appear to play central roles in the entire process.

Whereas the precise role that aurora-2 plays in centrosome duplication has yet to be determined, some interesting observations have recently been made. Overexpression of aurora-2 in cell culture has been shown to cause centrosome amplification (Fig. 3), although it appears to do so in a kinase-independent manner, as evidenced by the fact that overexpression of a kinase-dead mutant of aurora-2 in cell culture still results in an amplified centrosome phenotype (37). However, aurora-2 kinase activity is necessary to obtain a transformed phenotype by overexpression of aurora-2. Therefore, in addition to its role in centrosome duplication, genetic and biochemical studies have shown that aurora-2 kinase activity is crucial for proper centrosome separation (15). Again, the precise mechanism is not understood, but aurora-2 has been shown to phosphorylate substrates thought to be involved in centrosome separation, such as the kinesin-related protein Eg5 in Xenopus (38). Other specific signaling pathways that involve tightly regulated phosphorylation events involving aurora-2 remain unknown, and determining protein-protein interactions with aurora-2 is currently an active area of investigation.

Aurora-2 in Mitotic Control

As shown in Xenopus laevis oocytes, aurora family members play an important role in regulating the entry into mitosis (8). Oocyte development is halted at the G₂-M transition of meiosis II, where it rests until maturation is triggered by the steroid hormone progesterone. Progesterone binds to an unknown surface-associated receptor and signals the resumption of the meiotic cell cycle without the expression of newly transcribed genes. This “reawakening” is achieved by the translation of mRNAs that have been stored in the oocytes since the meiotic arrest. The stored mRNAs have short poly(A) tails and are not translated until modified to have an extended poly(A) tail. The signaling pathway proposed for this process begins with the activation of the Xenopus aurora kinase, Eg2, after progesterone binds its receptor. The activated Eg2 phosphorylates the CPEB, which in turn promotes the addition of a long poly(A) tail onto specific strands of mRNA, resulting in their translation (9).

This signaling scheme has also been shown in mouse oocytes (39). After the induction of oocyte maturation, ARK1, the murine equivalent of aurora-2, becomes active and phosphorylates CPEB on Ser¹⁷⁴. Either blocking ARK1 activity or truncating CPEB such that it cannot be modified by ARK1 inhibits cytoplasmic polyadenylation and leads to the inhibition of meiotic progression. Interestingly, cyclin B1 mRNA is a target for cytoplasmic polyadenylation during mouse oocyte maturation (39). The expression and formation of an active cdc2/cyclin B1 is known to be required for progression into mitosis. Thus, in summary, members of the aurora family appear to play essential roles in regulating meiosis and possibly meiotic progression in vertebrates.

Aurora-2 amplification and overexpression have been linked to cancer development and early progression for several years (11). However, the fact that aurora-2 continues to be overexpressed in late-stage tumors has been somewhat enigmatic. It is clear that up-regulation of aurora-2 results in abnormal mitotic spindles and gives rise to genomic instability and aneuploidy. This increases the rate of acquisition of the necessary genetic mutations that give rise to a cancer phenotype. However, after the genetic mutations are present, how do cancer cells cope with continued overexpression, and why does there appear to be a strong selective pressure to maintain aurora-2 overexpression in many tumor types?

We speculate that besides inducing genomic instability, aurora-2 and possibly other aurora family members play important roles in regulating the commencement of mitosis in somatic cells, in addition to their roles in developing oocytes. This could be accomplished by regulating the cytoplasmic polyadenylation of transcripts that code for key regulatory proteins such as cyclin B1 (Fig. 4). Consistent with this view, recent results indicate that cyclin B1 mRNA is polyadenylated in somatic cells before entry into mitosis (40) Additional mechanisms could be through the association of aurora-2 kinase with cell cycle-regulatory proteins such as the cdc2/cyclin B1 complex (41), p53 (42), the GTPase-activating protein (RasGAP; Ref. 43), or other protein interactions that have yet to be discovered. These functions would require aurora-2 kinase to be continuously expressed throughout tumorigenesis.

Aurora-2 as a Molecular Therapeutic Target

Taken together, the above findings make aurora-2 a potential therapeutic target of great interest. Significantly, aurora-2 appears to be largely and specifically up-regulated and amplified in cancer cells compared with normal cells. Northern blotting studies have shown the expression of some of the aurora kinase family members in normal tissues. More specifically, aurora-2 kinase has been described as high in thymus, with low levels in small bowel, testis, colon, spleen, and brain (44). Another study showed AIRKs to be elevated only in the testes (12). This information suggests that an inhibitor targeting aurora-2 kinase would have some selectivity for tumor cells versus normal cells.

A homology model and the crystal structure of aurora-2 have recently been published (45). The homology model was constructed based on sequence identity and homology to other serine/threonine kinases. In particular, aurora-2 showed good sequence alignment with cAMP-dependent kinase. The homology model was similar to the crystal structure obtained by X-ray crystallography using a truncated version of aurora-2 (residues 107–403; Ref. 46). The three-dimensional structure of aurora-2 reveals an ATP binding site that is relatively well conserved among other kinases. Despite the overall conservation, there are several unique features of the active site that can be exploited for drug design. Unique to the aurora family of kinases is the presence of two lysine residues in the glycine-rich nucleotide binding motif (45). Such differences may prove useful in designing inhibitors with aurora kinase selectivity. The solved structure confirms the unique conformation of the activation loop that appears to be conserved in AIRKs. Sequence and structural similarities among the three human AIRKs suggest it will be a challenge to design small molecules with absolute specificity for a single AIRK family member.

Conclusions

The discovery that genomic instability plays a critical role in the onset and progression of human tumors has led to a search for factors controlling genomic stability. Of the four main categories of genomic instability, aneuploidy is most frequent and is a defining characteristic of most human cancers. One of the major mechanisms governing aneuploidy is the proper segregation of chromosomes during cell division. A family of serine threonine kinases that regulates the segregation of chromosomes has emerged as potential mediators of aneuploidy and therefore potential novel targets for antitumor drug development. The finding that aurora-2 kinase plays a role in oncogenic transformation further demonstrates that these kinases may be attractive potential drug targets.

The putative role that AIRKs play in regulating entry into and exit from mitosis suggests an additional mechanism behind the oncogenicity of aurora-2. It also provides a model to explain continued overexpression of aurora-2 after the development of a primary tumor. Cancer is partly defined by a cell’s ability to continuously cycle through the mitotic process, therefore requiring the disruption of key regulatory proteins that monitor and signal the transitions required throughout the cell cycle. Based on known interactions of aurora kinase family members with cell cycle-regulatory proteins, such as cdc2/cyclin B1, it is probable that aurora-2 plays a central role in regulating the mitotic process.

Currently, several drug development efforts are under way to further validate aurora kinases as drug targets and to identify inhibitors for cancer therapy. It is too soon to know whether inhibitors of the aurora kinases will make good anticancer agents, but it is certain that these proteins represent new entry points for therapies targeting abnormal centrosome function in cancer.