The evolutionary conserved chromosomal passenger complex (CPC) is essential for faithful transmission of the genome during cell division. Perturbation of this complex in cultured cells gives rise to chromosome segregation errors and cytokinesis failure and as a consequence the ploidy status of the next generation of cells is changed. Aneuploidy and chromosomal instability (CIN) is observed in many human cancers, but whether this may be caused by deregulation of the CPC is unknown. In the present review, we discuss if and how a dysfunctional CPC could contribute to CIN in cancer.
CIN, aneuploidy and cancer
Aneuploidy, a state in which the genomic content of a cell deviates from a multiple of the haploid set of chromosomes, occurs in around 70% of all tumours . Aneuploid cells arise as a consequence of mis-segregation of whole chromosomes during cell division. In many cancer cells, the rate of chromosome mis-segregation is increased, resulting in a high frequency of chromosome gain and loss, a condition known as chromosomal instability (CIN). It has long been debated whether CIN itself can drive tumorigenesis or whether it arises as an indirect consequence of cell transformation. Increasing pieces of evidence, based on the examination of mouse models in which CIN is induced, suggest that chromosome mis-segregations could promote cancer development . Human aneuploid tumours can be categorized into two types: those that exhibit clonal aneuploidy and those that have heterogeneous karyotypes . The former situation implies that cancer cells have gone through a period of CIN or a rare event of chromosome mis-segregation, whereas the latter case is indicative of ongoing CIN . This poses the question whether persistent chromosome mis-segregation provides an advantage to the cancer cell or if it is the aneuploid genome composition that is beneficial. Loss or gain of particular chromosomes could indeed be beneficial for the cancer cell, as it can shift the balance between genes with oncogenic properties and tumour suppressor genes, thereby conferring benefits with respect to proliferation and survival [1,4,5]. Additionally, CIN by itself may be profitable. It will give rise to a cell population with a variety of karyotypes allowing fast adaptation to a changing environment, which could help to develop resistance to anti-cancer drug treatment [3,6,7]. The level of CIN is probably a critical determinant of cellular fitness because elevation of the segregation error frequency results in reduced viability of cancer cells . Moreover, mis-segregation rates may not only affect tumour cell fitness, but also appear to determine whether CIN can promote or inhibit the development of cancer .
Origins of CIN
Faithful chromosome segregation during mitosis requires that kinetochores (multiprotein structures that assemble at centromeres and form microtubule attachment sites) of the duplicated chromosomes bind to microtubules emanating from opposite spindle poles (chromosome biorientation) . Since the interactions between kinetochores and spindle microtubules are stochastic in early mitosis, many sister chromatids initially acquire attachments that are not bipolar. Both kinetochores of the sister chromatids may bind microtubules emanating from one spindle pole (syntelic attachments) or a single kinetochore may attach to microtubules derived from both spindle poles, whereas the other kinetochore is attached to microtubules from just one spindle pole (merotelic attachments). In addition, one kinetochore can become attached to microtubules from one pole, whereas its sister remains unattached (monotelic attachment) . When these kinetochore–microtubule (k–MT) attachment errors persist until anaphase, they can result in whole chromosome mis-segregation and in aneuploidy of the two daughter cells. In addition, lagging chromosomes that are a frequent consequence of merotelic attachments can also cause structural chromosomal aberrations [12–14]. However, the combined activities of the error-correction machinery and the mitotic checkpoint (MC; a surveillance pathway that prevents chromosome segregation until all sister chromatids are attached to the mitotic spindle ) ensure that k–MT attachment errors are resolved before anaphase onset to preserve genome integrity.
A number of defects are thought to cause CIN. First, multipolar mitotic spindles, arising as a consequence of centrosome over-duplication, increase the likelihood of acquiring merotelic k–MT attachments in pro-metaphase. Due to centrosome clustering, cells with multiple centrosomes frequently undergo a pseudo-bipolar anaphase with the merotellicaly attached sister chromatids appearing as lagging chromosomes [16,17]. Secondly, impaired sister chromatid cohesion affects chromosome biorientation and increases the frequency of mis-segregating chromosomes [18–21]. Third, a weakened MC fails to delay anaphase onset sufficiently to allow all chromosomes to achieve proper biorientation before anaphase onset [7,15]. And fourth, excessively stable k–MT attachments lead to persistent incorrect k–MT attachments and lagging chromosomes in anaphase [22,23]. Although centrosome amplification, cohesin defects and hyperstabilized k–MT interactions have been described in cancer or cancer cell lines [16,17,19,22,23], evidence for a weaker function of the MC as a widespread feature in chromosomally unstable cancer cells is controversial [24–28].
The CPC ensures faithful chromosome segregation
An evolutionarily conserved component of the error correction machinery is the mitotic kinase Aurora B . Aurora B functions in complex with three other non-enzymatic proteins; INCENP (inner centromere protein), Borealin and Survivin, together known as the chromosomal passenger complex (CPC). At the onset of mitosis Aurora B is recruited and activated at the inner centromere, where it promotes the destabilization of k–MT attachments through the phosphorylation of (among others) the KMN network, the core microtubule-binding protein complex at the kinetochore [30–32]. Destabilization of incorrect k–MT attachments allows for a new opportunity to establish bipolar attachments [30,33]. As sister-kinetochores bind microtubules from opposite poles, the chromosome orients each kinetochore toward the attached pole greatly favouring appropriate bioriented attachments [34,35]. Also, centromeres of bioriented chromosomes experience increased tension that spatially separates Aurora B from its outer-kinetochore substrates resulting in changes of the phosphorylation status of these substrates [30,36–38]. In addition, Aurora B-counteracting phosphatases, such as protein phosphatase 2A (PP2A) and PP1, are essential for removal of phosphate groups from Aurora B target sites. The combined action of Aurora B kinase and the PP2A and PP1 phosphatases are thought to create a highly dynamic kinetochore microtubule interface that promote the correction of mis-attached k–MTs, and the stabilization of correctly attached microtubules [39–43]. Besides its role in error correction, Aurora B also directly feeds into the MC by controlling the kinetochore recruitment of the MC kinase Mps1 [44–46] and is crucial for cytokinesis [47–53]. This latter function is probably associated with its localization on the central spindle midzone and midbody during anaphase and telophase respectively. Through its multiple activities during cell division the CPC ensures chromosomal stability. The questions if and how the function of the CPC is disturbed in chromosomally unstable cancer cells are discussed in the present review.
Consequences of impaired Aurora B function
A functional CPC requires Aurora B kinase activity and INCENP, Survivin and Borealin for proper activation and localization of the kinase. The level of localized Aurora B activity needs to be tightly regulated to prevent chromosome mis-segregation and to support cytokinesis (Figure 1). In line with this, partial inhibition of Aurora B using low concentrations of a small molecule inhibitor was shown to hamper the resolution of merotelic attachments during (pro-)metaphase and increases the number of lagging chromosomes in anaphase . Higher doses of Aurora B inhibitor or RNAi-mediated kinase depletion perturb chromosome alignment in metaphase, indicative of mal-attachments other than merotelics [52–54]. This is followed by massive chromosome mis-segregations in anaphase and cytokinesis failure [52–54].
Possible fates of mitotic cells with increased or decreased levels of local Aurora B kinase activity
Instead of impairing general kinase activity, local kinase activity can also be affected by specifically disturbing Aurora B localization to the inner centromere or central spindle via expression of truncation and/or point mutants of Borealin, Survivin or INCENP [55–57]. Local perturbation of Aurora B activity can give rise to CIN by at least two different mechanisms: impaired correction of mal-attachments or by centrosome amplification due to failed cytokinesis (Figure 1). Moreover, in contrast with reduced Aurora B activity, hyper-activation or overexpression of Aurora B may also cause problems through, for example, failure of stabilizing bipolar attachments. While under normal circumstances unattached kinetochores, created by ongoing k–MT destabilization, would keep the MC active, a failure to stabilize bipolar attachments could give rise to aneuploidy when the MC is simultaneously weakened (Figure 1). A recent study in budding yeast in which the Aurora B and INCENP homologues Ipl1 and Sli15 were both overexpressed suggests that hyperactivity of Aurora B can indeed prevent stabilization of bipolar attachments in this organism . However, whether bioriented attachments fail to be stabilized in mammalian cells that overexpress Aurora B is not known. Mere overexpression of Aurora B in Chinese hamster embryo cells enhanced phosphorylation of the Aurora B substrate histone H3, indicating that elevated levels of the kinase in mammalian cells can indeed result in increased activity . This was associated with an increased frequency of lagging chromosomes and chromosome bridges in anaphase and the formation of multinucleated cells . Increased multinucleation was also seen upon Aurora B overexpression in human fibroblast and epithelial cell lines . Moreover, in a small fraction of murine mammary gland epithelial cells, multiplication of the genomic content, as judged by FACS analysis and chromosome count, was observed upon overexpression of Aurora B . The observed multinucleation and polyploidization suggests that Aurora B overexpression can lead to cytokinesis failure or mitotic slippage [59–61]. In addition to these phenotypes, Aurora B overexpression has also been reported to induce premature sister chromatid separation in both diploid and tetraploid mouse epithelial cells, although it remains unclear how this cohesion defect is mediated . Moreover, whether cohesion loss could explain the earlier described chromosome segregation defects is unknown . Although these collective findings suggest that overexpression of Aurora B can indeed disturb chromosome segregation and cytokinesis, the mechanism underlying these cell division defects remains poorly understood. Furthermore, the consequence of overexpression of the other CPC subunits for chromosome segregation or cytokinesis has not been extensively studied.
Aurora B deregulation in tumours
Since Aurora B ensures error-free chromosome segregation, a process that is disturbed in many tumour types, the question arises how well Aurora B is functioning in chromosomally unstable cancer cells. Below, we describe several mechanisms that could lead to deregulation of Aurora B and discuss these aberrations in the context of cancer.
Deregulation of Aurora B gene expression
Overexpression of Aurora B has been reported in a variety of tumour types . Moreover, Aurora B is one of the genes in the CIN70 signature, a list of genes whose overexpression is associated with aneuploidy in cancer . However, a more recent paper questioned the validity of the CIN70 profile and posed the possibility that expression of these genes may be related to proliferation rather than aneuploidy . Since Aurora B is mainly expressed during G2 and mitosis [65,66], it is more abundant in cycling cells, raising the question whether the up-regulation is merely a consequence of the proliferative state of cancer cells. Aberrant overexpression of a gene may be caused by gene amplification or by (epi)genetic changes in the gene promoter, neither of which have been reported as a cause for increased expression of Aurora B in tumours. Moreover, a recent study indicated that the increased mRNA levels of many kinetochore genes, including CPC genes, in tumours was strongly correlated with the activation of a general cell division programme . Thus, although overexpression of Aurora B in cultured cells can result in segregation defects, it is questionable whether the high Aurora B levels observed in cancer will be a causal factor for CIN.
On the other hand, reduced levels of Aurora B clearly give rise to segregation errors. Mouse embryonic fibroblasts (MEFs), in which both alleles for Aurora B have been knocked out, exhibit a range of aberrations associated with impaired mitosis or cytokinesis, including multinucleation, formation of micronuclei and multipolar spindles . Homozygous Aurora B knockout mice do not survive beyond embryonic stage E9.5; however, heterozygous Aurora B knockout mice develop normally and have an increased tumour incidence . Although an increase in DNA content was reported for a small number of MEFs derived from these mice, the ploidy status of either healthy or tumour cells from adult animals was not determined . It can, therefore, not be excluded that subtle segregation defects could eventually have driven cancer development . In cultured cells, INCENP, Borealin and Survivin are required for activation and/or localization of Aurora B  and similar to inactivation of Aurora B, loss of these subunits impairs error correction and cytokinesis [56,69–71]. INCENP-, Survivin- or Borealin-null mice also die during embryonic development and MEFs derived from early embryos show signs of defective mitosis [72–74]. There is presently no indication that mitotic errors frequently occur in MEFs derived from mice heterozygous for these CPC members. However, since the phenotype of the heterozygous adult animals was not thoroughly investigated, a low frequency of segregation errors may have been missed. Moreover, it is also not known whether these animals have an increased cancer incidence later in life [72–74].
In cancer cells, reduced levels of Aurora B might result from genetic alteration of one of the coding alleles. The human AURKB (Aurora kinase B) gene is located at the 17p13.1 locus, in close proximity to the gene encoding the tumour suppressor p53, a region that is often deleted in tumours [4,75–79]. Evaluation of a panel of breast carcinomas showed that in 12% of tumours the Aurora B and p53 genes were co-deleted . Interestingly, when this is accompanied with loss or mutation of the second p53 allele, it could result in cells with reduced Aurora B expression and loss of functional p53. Non-transformed cells frequently arrest in G1, in a p53-dependent manner, when they have mis-segregated chromosomes or failed cytokinesis [81,82]. Cells with co-deleted Aurora B and p53 genes may progress through the cell cycle with aberrant chromosome and even centrosome numbers, leading to further chromosome mis-segregation in the next mitosis or to secondary structural genomic alterations [13,14]. It would be interesting to determine in non-transformed human cells to what extend heterozygous loss of the Aurora B gene affects Aurora B protein levels and whether this is sufficient to elevate the rates of chromosome mis-segregation. By doing so, in a p53 proficient and deficient background, the cellular response to potential chromosome mis-segregations as a result of loss of one Aurora B allele can be compared.
Single nucleotide substitutions can give rise to a protein with altered functionality. The numbers of mutations that are present in a tumour vary per tumour type and can range from fewer than ten to over a thousand . A major challenge in interpreting all these genetic alterations is to distinguish mutations that confer a selective advantage to the cell, the so-called driver mutations, from passenger mutations that do not contribute to cellular fitness. Driver mutations exist in two flavours: (1) gain-of-function mutations that impose oncogenic properties on a protein; (2) mutations that abrogate protein function, which in cancer usually occur in tumour suppressor genes . The problem of pin-pointing driver mutations in the cancer genome is often-times dealt with by comparing mutational profiles across tumour samples [85,86]. If a mutation at a specific site is present in a large proportion of tumours, but not in healthy tissue, it probably entails an oncogenic mutation . Loss-of-function mutations that provide a selective advantage to a cell are not necessarily confined to a single nucleotide, but occur with elevated frequency along the length of the gene . These protein-inactivating mutations can be mis-sense mutations that lead to amino acid substitutions, but also frame-shift mutations, alteration of splice sites or mutations that introduce a premature stop-codon . Many major driver mutations have already been identified based on their overall occurrence in cancer. Since methods to analyse high-throughput sequencing data are getting more refined, it also becomes feasible to address more complex issues regarding the contribution of mutations to cancer establishment or maintenance. For example, the classification of genes as cancer-associated genes or neutral genes might not be as black-and-white as previously stated, as there appear to be gradations in the oncogenic- or tumour suppressor capacity that genes harbour . It could also be possible that combinations of mutations in certain genes give a fitness advantage, rather than each genetic alteration having an effect in an isolated manner. The large amounts of tumour sequencing data that are being generated worldwide make it possible to obtain mutation data about any gene of interest as well as the genetic context of the identified mutations. Moreover, as the output of next-generation sequencing data is often made available via different online platforms, a wide audience can obtain access to and make use of this information .
Point mutations in any of the genes that encode members of the CPC could prevent the formation of a functional complex and result in chromosome mis-segregations or cytokinesis failure. As such, mutations in CPC genes might promote tumour development through induction of CIN. These mutations are not classical cancer drivers, as they do not directly provide a benefit with regard to tumour cell survival or proliferation, but could enable the creation of a population in which a certain subset of cells may have acquired a karyotype that supports transformation. Currently there are no indications that the genes encoding proteins of the CPC are mutated with elevated frequency in tumours. However, since mutation of any of the four proteins might result in Aurora B dysfunction, the cumulative mutation rate of the CPC across tumours might encompass all cases with mutations in Aurora B, INCENP, Borealin or Survivin. Characterization of how mutations in the CPC genes that occur in cancers affect chromosome segregation could reveal whether this may indeed be a mechanism by which CIN in tumours can arise.
The Aurora B regulatory network
Aurora B function relies on many different factors. Numerous proteins are involved in the localization and activation of the kinase. In addition, Aurora B antagonizing phosphatases influence the phosphorylation status of its substrates. The list of proteins that are phosphorylated by Aurora B is extensive [32,88–90] and together they determine the different functions of Aurora B with respect to error correction, MC activation and cytokinesis. The intricate network of Aurora B regulators and effectors denotes that perturbation of any of these proteins might, in fact, result in aberrant Aurora B function (Figure 2). Below we have listed a number of examples in which CIN and tumorigenesis are in some way linked to a change in Aurora B functionality.
Schematic overview of the Aurora B regulatory network
Overexpression of the MC protein Bub1 in mice caused Aurora B hyperactivation by a thus far unknown mechanism and resulted in aneuploidy and increased tumour formation . The observed CIN in MEFs derived from these mice could be reversed by Aurora B inhibition, indicating that hyperactivation of Aurora B caused the CIN upon Bub1 overexpression . However, whether Aurora B hyperactivation was also responsible for the increase in tumour incidence remains to be demonstrated.
Mice heterozygous for an acetylation-defective mutant of BubR1, another component of the MC are also chromosomally unstable and prone to tumour development . Experiments on MEFs from these mice indicated that these cells not only had difficulties maintaining the MC, but also exhibited reduced binding of BubR1 to the Aurora B counteracting PP2A–B56 phosphatase and had increased levels of phosphorylated Ndc80 . As such, the BubR1 acetylation-defective mutants may affect chromosome segregation via dual effects on MC and k–MT interactions. The effect of BubR1 on phosphorylation of Aurora B substrates via regulation of PP2A–B56 is also observed in human cells [42,43,93]. Depletion of BubR1 prevents the formation of stable k–MT interactions and this phenotype can be rescued by Aurora B inhibition [93,94]. Interestingly, genetic mutations in BubR1 are associated with mosaic variegated aneuploidy (MVA), a disease characterized by growth retardation, microcephaly, mosaic aneuploidies and pre-disposition to cancer . Bi-allelic BubR1 mutations identified in MVA patients caused defects in checkpoint functioning and/or chromosome alignment via specific effects of the mutations on certain protein interaction domains or as a result of reduced BubR1 protein stability . In some MVA patients, the chromosome segregation defects may thus be explained by a disturbed balance between PP2A–B56 and Aurora B as a consequence of BubR1 dysfunction .
Overexpression of Mad2 promotes aneuploidy and tumorigenesis in mice . Interestingly, Mad2 overexpression in human cells reduces Aurora B levels at centromeres and leads to hyper-stabilization of k–MT interactions and lagging chromosomes in anaphase . This indicates that the aneuploidy induced by Mad2 overexpression might be mediated via delocalization of Aurora B. How Mad2 affects the signalling pathways involved in CPC centromere recruitment remains to be elucidated. In addition, experiments in human cell lines showed that depletion of the PP2A regulatory subunit Aα (PPP2R1a) resulted in increased Mad2 phosphorylation and reduced growth of Mad2 overexpressing cells . Knockdown of Aurora B inhibited Mad2 phosphorylation and abolished the PPP2R1a depletion-induced growth arrest, providing a different regulatory link in which Aurora B acts upstream of Mad2 .
The final example of how Aurora B activity could be changed as a result of altered functionality of regulatory proteins comes from the recently discovered link between the ARF (alternative reading frame) tumour suppressor and Aurora B . The CDKN2a (cyclin-dependent kinase inhibitor 2A) locus, which encodes both INK4a and ARF in different reading frames, is frequently mutated in tumours . Germline mutations of this gene cause a pre-disposition to development of melanomas . It is well established that ARF stabilizes p53 through its interaction with MDM2 (murine double minute 2 homologue) . Loss of ARF resulted in aneuploidy in MEFs and splenocytes from adult mice, however, the segregation defects and weakened MC that probably give rise to the aneuploid state were independent of p53 . Instead, Aurora B protein levels were increased in ARF knockout cells and the phenotypes could be rescued by depletion of Aurora B to near wild-type levels, indicating that Aurora B overexpression is the likely cause for the mitotic defects in ARF knockout cells .
Currently, it is unclear whether perturbation of Aurora B or of the other members of the CPC by aberrant expression or by mutations could mediate CIN in certain tumours. Although present in low frequency in cancer, analysis of the effects of potential tumour-associated mutations of Aurora B, INCENP, Borealin or Survivin could provide further insight into the functionality of the CPC in chromosomally unstable tumours. Moreover, the proper level of activity of Aurora B, at either the centromere in (pro-)metaphase or the central spindle in anaphase is not only dictated by the other CPC members, but is also controlled by a complex regulatory network involving many other proteins  (Figure 2). Mutations or defects in either of these regulators could impair Aurora B function and cause CIN. Taking this into account, the number of tumours in which Aurora B function could be disturbed may be much higher than predicted by mutational analysis of CPC subunits. Yet, it will be a challenge to identify all the potential genetic alterations that may indirectly affect Aurora B function and to determine whether they thereby contribute to CIN and cancer.
We thank Dr N. Jelluma for a critical reading of the paper.
This work is supported by the Dutch Cancer Society [grant number KWF UU2011–5134]; and the Netherlands Organisation for Scientific Research [grant number NWO Vici 918.12.610].
The Dynamic Cell: held at Robinson College, Cambridge University, Cambridge, U.K., 4–7 September 2014.