Aurora B hyperactivation by Bub1 overexpression promotes chromosome missegregation

Abstract

High expression of the mitotic kinase Bub1 is associated with a variety of human cancers and correlates with poor clinical prognosis, but whether Bub1 alone can drive tumorigenesis was unknown. We provided conclusive evidence that Bub1 has oncogenic properties by generating transgenic mice that overexpress Bub1 in a wide variety of tissues, resulting in aneuploidization. Consistently, Bub1 transgenic mice developed various kinds of spontaneous tumors as well as accelerated Mycinduced lymphomagenesis. While the mitotic checkpoint was robust in Bub1 overexpressing cells, misaligned and lagging chromosomes were observed. These defects originated from increased Aurora B activity and could be suppressed by inhibition of Aurora B. Taken together, this indicates that Bub1 has oncogenic properties and imply that aneuploidization and tumorigenesis result from Aurora B-dependent missegregation. Here, we focus on the complex relationship between Bub1 and Aurora B and discuss the broader implications of Bub1-dependent Aurora B activation in mediating error correction.

Introduction

Dynamic mechanisms in the eukaryotic cell exist to ensure high fidelity segregation of genetic material during cell division. Chromosome segregation is delayed by cell cycle checkpoints until DNA replication is complete and sister chromatids are fully aligned at the metaphase plate. Mitotic checkpoint failure, in the absence of cell death, results in aneuploid progeny when chromosome segregation is imbalanced. Aneuploidy refers to an abnormal karyotype that is not a multiple of the haploid complement. Although aneuploidy is a common feature of tumors and correlates with poor clinical prognosis,^1–4 how aneuploidy contributes to the oncogenic process is currently a matter of debate.

Inaccurate chromosome segregation during mitosis is one potential source of aneuploidy in tumors.⁵ The rate at which whole chromosomes are gained or lost is referred to as whole chromosomal instability. The mitotic checkpoint delays anaphase onset until all kinetochores are properly attached to the spindle, thereby protecting against chromosome missegregation.⁶ Recent mouse models have shown that aberrant chromosome segregation not only can correspond to tumorigenesis in certain contexts, but may in fact act as an initiator.^7–13 Moreover, kinetochores that are aberrantly attached to microtubules are detected and resolved by the error correction machinery, of which the principal component is the mitotic kinase Aurora B.^14–18 While reduced Aurora B activity is known to cause both misaligned and lagging chromosomes during mitosis,^19–25 less is known about the impact of increased Aurora B activity on segregation fidelity. More importantly, how these alterations in Aurora B activity impact tumorigenesis are unclear.

Despite the strong correlation between aneuploidy and tumorigenesis in mice, mutations in mitotic checkpoint genes, including Bub1, are rare in human tumors.^26–30 Instead, mitotic regulators may influence chromosome segregation faithfulness by alterations in gene expression. In the case of Bub1, several studies have found increased Bub1 expression in subsets of lymphomas, breast, gastric and prostate cancers.^31–38 Furthermore, multiple independent studies including diverse tumor types have identified Bub1 overexpression as an event that correlates with poor clinical prognosis.^37,39 It is worthwhile to note that like many mitotic regulators, Bub1 expression is cell cycle regulated and consequently, increases with proliferation.⁴⁰ Therefore, increased Bub1 levels in human cancer tissues may simply represent the mitotic index of the tumor compared with normal quiescent tissues. Thus, although expression profiling suggested a strong correlation between increased expression and certain cancers, whether elevated Bub1 alone could drive tumorigenesis required in vivo verification. To thoroughly address the question of whether Bub1 overexpression is causal for tumorigenesis, we generated transgenic mouse strains that ubiquitously overexpress Bub1.⁴¹

Bub1 Overexpression Promotes Chromosome Missegregation through Aurora B Hyperactivation

We sought to generate transgenic mice that overexpressed Bub1 in a wide variety of tissues since Bub1 overexpression correlates with a broad range of human cancers. For this, we drove Bub1 expression from a vector containing the chicken β-actin promoter and the CMV enhancer, otherwise described as the CAGGS promoter. We obtained two independent lines of HA-Bub1 transgenic mice, Bub1^T85 and Bub1^T264 with both moderate and high levels of Bub1 overexpression, respectively.⁴¹ Bub1 transgenic primary MEFs, splenocytes and hepatic lymphocytes were highly aneuploid, with gains and losses of one to three whole chromosomes.⁴¹ The most common mitotic defect for both mild and high Bub1 overexpression was lagging chromosomes,⁴¹ a defect that is a frequent event in human tumor cells.^42,43 Lagging chromosomes originate when kinetochores are attached to microtubules emanating from both spindle poles, a condition referred to as merotely.⁴⁴ Cells with higher levels of Bub1 overexpression also demonstrated an increase in unaligned chromosomes in metaphase.⁴¹ Misaligned chromosomes can occur from monotelic attachment, where one kinetochore lacks microtubule attachment, syntelic attachment, where both sister kinetochores are attached to microtubules emanating from the same pole, or an unbalanced merotelic attachment, where one kinetochore is attached to microtubule bundles from both poles but one side has more microtubules than the other.⁴⁵ Whereas monotelic attachment unambiguously signals to the mitotic checkpoint, whether syntelic attachments and their lack of tension are able to signal the checkpoint is an issue of contention.^45–48 However, Bub1 overexpression has no impact on the strength of mitotic checkpoint signaling.⁴¹ We deduced this from the normal recruitment of mitotic regulators such as Mad2, Cenp-E, Cdc20 and BubR1 to kinetochores in Bub1 transgenic MEFs, by the same duration of the mitotic arrest in the presence of spindle poisons nocodazole or taxol as in wild type, by the absence of prematurely separated sister chromatids, and by the normal mitotic timing of Bub1 transgenic MEFs.⁴¹ Therefore, unlike Mad2 overexpression, which extends mitotic timing and promotes cyclin B1 stabilization,¹² Bub1 overexpression induces aneuploidy and chromosome segregation defects independent of mitotic checkpoint signaling.

Why might it be surprisingly that Bub1 overexpression does not impact mitotic checkpoint signaling? First, Bub1 is a mitotic checkpoint factor that assembles early on kinetochores and is required for the kinetochore recruitment of numerous checkpoint regulators, including those that inhibit APC/C in the event of checkpoint activation.^9,49–51 Cells overexpressing Bub1 do harbor more Bub1 at kinetochores and this, potentially, could have influenced checkpoint signaling by altering recruitment of Bub1-dependent mitotic regulators to kinetochores. Second, Bub1 is a mitotic kinase and one proposed substrate is Cdc20, the APC/C cofactor.^52,53 Then, one might predict that increasing Bub1 activity via overexpression would prolong mitosis, similar to Mad2 and Hec1/Ndc80 overexpression.^11,12 It is important to note that overexpression in this case does equate to increased catalytic activity as Bub1 transgenic MEFs demonstrate considerably elevated phosphorylation of the substrate histone H2A.⁵⁴ One possible explanation is that Bub1 phosphorylation of Cdc20 does not make a major contribution to mitotic checkpoint signaling in mice. Alternatively, perhaps either Cdc20 is limiting relative to Bub1, so increasing Bub1 activity and levels does not alter APC/C inhibition from Cdc20 phosphorylation or the phosphatase that relinquishes the Cdc20-mediated inhibition overwhelms Bub1. It should be pointed out that whether Bub1 kinase activity contributes to the mitotic checkpoint is controversial. For example, there are conflicting reports for the role of Bub1 kinase in checkpoint signaling in budding and fission yeasts.^55–57 In human cells, it has been shown that Bub1 kinase activity is necessary for precise chromosome alignment and a fully optimal checkpoint using an isogenic siRNA complementation system.⁵⁸ On the other hand, in Xenopus, Bub1 kinase-inactive protein was competent for checkpoint signaling⁵⁹ and murine cells expressing a Bub1 kinaseinactive mutant were restored for checkpoint activity.^60,61 It is difficult to resolve whether these differences on the role of Bub1 kinase truly reflect variation in the signaling mechanism of the mitotic checkpoint in numerous organisms or simply reflect the limitations of each experimental system. For example, mutation of the ATP-binding motif in Bub1 has been reported to destabilize protein levels.⁶² In mammalian systems, expressing wild-type or Bub1 mutants to levels comparable to endogenous has been problematic^53,58 and are further limited by residual endogenous protein from partial siRNA knockdown or inadequate Cre recombinase-mediated deletion of the endogenous allele. At this point, full clarification of the role of Bub1 kinase in checkpoint signaling is needed.

Since mitotic checkpoint signaling is intact with Bub1 overexpression, we were intrigued as to why these cells harbored near-diploid aneuploidies and defective chromosome segregation. Given that the segregation defects were unresolved, we then examined whether Bub1 overexpression influenced the error correction pathway. Importantly, the protein stability and localization of Aurora B was unaffected by Bub1 overexpression.⁴¹ On the other hand, we found increased phosphorylation of Aurora B substrates, such as Knl1,⁶³ and an increase in the amount of Bub1 in a complex with Aurora B in Bub1 transgenic cells compared with wild type.⁴¹ More exciting was the finding that Bub1-induced chromosome segregation defects could be suppressed by partial inhibition of Aurora B.⁴¹ Additionally, co-overexpression of the Aurora B in vivo inhibitor, BubR1,⁶⁴ also suppressed Bub1- induced chromosome missegregation.⁴¹ In both instances, aneuploidy was also restored to near wild-type levels with Aurora B inhibition. We conclude that Bub1 overexpression drives Aurora B hyperactivation.

What is the Relationship between Bub1 and Aurora B?

Contradictory reports in the literature have resulted in some confusion over what the molecular relationship is between Aurora B and Bub1. Namely, some reports suggest that Bub1 is upstream of Aurora B, while others have concluded that Bub1 is either downstream or in a parallel pathway as Aurora B. Evidence suggesting that Bub1 is upstream of Aurora B is based on the observation that Bub1 depletion inhibits Aurora B inner centromeric localization and results in the re-distribution of Aurora B to chromosome arms.^65,66 In addition, using fission yeast and human cells, it has been observed that phosphorylation of histone H₂A at threonine 120 by Bub1 kinase contributes to the recruitment of Aurora B to centromeres in a shugoshin- and Cdk1-dependent manner.^54,67 These reports indicate that Bub1 is involved in Aurora B localization and thus, its error correction activity. In direct contrast, others have found that siRNA-mediated depletion of Bub1 does not delocalize centromeric Aurora B,⁵⁰ that Aurora B catalysis is actually required for accumulation of Bub1 on tension-less kinetochores,⁶⁸ or that Bub1 and Aurora B exist in parallel pathways for checkpoint signaling.⁶⁹ One possible explanation for these divergent results is the difficulty in obtaining efficient and complete depletion of Bub1 by siRNA. While it is clear that Bub1 is required for mitotic progression and a robust checkpoint response,^9,40,59 several experiments with Bub1 siRNA have had difficulty replicating this result.^49,69 Finally, a direct link between Aurora B function and Bub1 has been suggested as it was demonstrated that Xenopus Bub1 can phosphorylate INCENP, an Aurora B modulating protein, in vitro,⁷⁰ although the residue(s) involved are not specified. Phosphorylation of INCENP, particularly at the TSS residues, has been implicated in Aurora B activation,^71–73 whereas phosphorylation of INCENP at Thr59 and Thr388 by Cdk1 is necessary for the metaphase to anaphase transition and the recruitment of Plk1 to kinetochores.⁷⁴ Interestingly, Bub1 depletion has been reported to reduce INCENP centromere/kinetochore targeting.⁷⁵

Potential Mechanisms for Bub1-Dependent Aurora B Activation

The observation of increased Aurora B activity in Bub1 overexpressing cells begs the question of how Bub1 influences Aurora B activity. It is thought that Aurora B contributes to the error correction pathway through its localization at inner centromeres and activity against substrates that influence spindle microtubule-kinetochore affinity.^14,17,63,76 Aurora B is the enzymatic component of the chromosome passenger complex (CPC), which includes INCENP, Borealin and Survivin.⁷⁶ These three proteins form a helix bundle that associates with Aurora B through the IN-box on INCENP.^71,77–79 Studies on how Aurora B affects error correction have proven to illustrate the incredible complexity that seems to underlie Aurora B in vivo activation.

One model that has been suggested for Aurora B activity is the spatial separation model.^17,80 Here it has been proposed that tension pulls kinetochores away from a constant Aurora B kinase zone at the inner centromere and into a phosphatase rich zone that de-phosphorylates Aurora B substrates, resulting in stable attachment.⁸⁰ Although chromosomes most likely bi-orient due to geometric constraints,^81,82 stochastic attachment is unlikely to initially result in a perfect geometry on all chromosomes. Whereas correct attachments are stable, incorrect attachments are unstable and promptly destabilized. Trial-and-error should result in the eventual correct attachment as destabilization would provide a new opportunity for bi-orientation.¹⁷ Evidence to support this model comes from data that Aurora B clustering at inner centromeres contributes to activation.^65,83 In this model, one might imagine that for Bub1 overexpression to alter Aurora B activity, Bub1 would need to impact the breadth of the Aurora B zone, such that the equilibrium between Aurora B and phosphatases now favors Aurora B (Fig. 1A). Then hyperactivation of Aurora B via Bub1 would destabilize correct attachments resulting in the perpetual on-off of microtubules from kinetochores, promoting errors like misalignments and lagging chromosomes.

Figure 1.

Two models for how Aurora B and Bub1 overexpression may contribute to microtubule attachment. (A) Chromosomes are pink, microtubules are blue, kinetochores are light blue, phosphatase zone is peach and Aurora B zone is yellow. Dephosphorylated Aurora B substrates are green boxes and phosphorylated Aurora B substrates are red boxes. Left, normal; Right, Bub1 overexpression expands the Aurora B zone to overwhelm the phosphatase rich zone. (B) Chromosomes are pink, microtubules are blue, kinetochores are light blue, and phosphatase(s) are peach diamonds. Aurora B containing complexes are yellow circles. The red or blue outline color represents the various Aurora B subcomplexes. Yellow circles without an outline are unactivated Aurora B complexes. Dephosphorylated Aurora B substrates are green boxes and phosphorylated Aurora B substrates are red boxes. Left, normal; Right, Bub1 overexpression increases the level of a specific Aurora B subcomplex.

The exact molecular details of how Bub1 overexpression directly influences Aurora B activation are unclear. One possibility is that Bub1 phosphorylates the Aurora B modulator INCENP, as Xenopus Bub1 can phosphorylate both histone H2, a bona fide Bub1 substrate, and INCENP in vitro.⁷⁰ C-terminal INCENP phosphorylation has been linked to Aurora B activation,^71–73 while INCENP phosphorylation at other residues has been suggested to contribute to Plk1 kinetochore recruitment.⁷⁴ This is particularly intriguing given that there is biochemical evidence to suggest varying subcomplexes of Aurora B exist in the cell.^84,85 For example, subcomplexes with Aurora B and INCENP have been isolated that are independent of Borealin and Survivin.⁸⁵ It is tempting to speculate that both the composition of Aurora B modulators within the complex and their post-translation status may regulate Aurora B catalysis to determine substrate specificities in vivo. Varying the substrate specificity of Aurora B subcomplexes may be important for mediating resolution to a range of microtubule kinetochore attachment defects. This could also explain why certain mutations that alter Aurora B activity result in syntelic attachments, whereas others promote merotelic attachments. For example, hyperactivation of Aurora B via Mst1 knockdown increased Aurora B auto-phosphorylation and promoted mis-alignments.⁸⁶ This contrasts with the elevated frequency of lagging chromosomes and merotelic attachments observed with Bub1 overexpression.⁴¹ In vivo evidence implies the existence of distinct Aurora B niches. First, deregulating Aurora B activation through haspin kinase depletion preferentially affects certain substrates.⁸⁷ Second, dephosphorylated Hec1, an Aurora B substrate, localizes with Aurora B, implying an Aurora B containing complex with limited activity against its substrate.⁸⁸ Third, low levels of Aurora B kinase activity have been observed in cells depleted for INCENP.⁸⁹ Similarly, Aurora B targets were phosphorylated in borealin-depleted cells in which Aurora B was artificially targeted to inner centromeres.⁶⁵ Together, these observations promote the notion that there are activated Aurora B pools that are INCENP- and borealin-independent. Alternatively, one could interpret that INCENP and borealin contribute to Aurora B activation purely for its localization, which seems unlikely given the evidence that CPC subunits elevate Aurora B catalysis in vitro.^71–73,90 In this model, one might imagine that for Bub1 overexpression to increase Aurora B activity, Bub1 could specifically impact certain Aurora B subcomplexes (Fig. 1B). An interesting aspect is that these subcomplexes could locally change in the presence of improperly attached kinetochores to promote bi-orientation.

Bub1 Overexpression Drives Tumorigenesis

Whether aneuploidy causes cancer has been an issue hotly debated, mainly because cancer is a potential, although not obligatory outcome of aneuploidization.⁹¹ For aneuploidy to be deemed an initiating feature for cancer, it should occur in primary cells and produce aneuploid progeny that propagate and are prone to malignant transformation.⁴⁴ In primary cells, Bub1 overexpression promotes chromosome missegregation that is dependent on Aurora B hyperactivation.⁴¹ One interesting observation was that suppressing Aurora B with moderate amounts of ZM447439 dramatically reduced mis-alignments and lagging chromosomes in Bub1 transgenic MEFS to levels below wild-type MEFs,⁴¹ raising the interesting possibility that once Aurora B hyperactivation is quenched, Bub1 overexpression is actually beneficial for chromosome segregation. Consistent with Bub1 overexpression having oncogenic properties, Bub1 transgenic mice had increased frequency of spontaneous tumors and also demonstrate accelerated Eµ-Myc lymphomagenesis.⁴¹ Taken together, it is reasonable to implicate Bub1-mediated hyperactivation of Aurora B in cellular transformation, although it remains formally possible that Bub1 upregulation may have other additional oncogenic roles.

While a majority of cancer cells demonstrate a functional mitotic checkpoint,⁹² many still undergo chromosomal instability at high frequency. One potential mechanism for the acquisition of an abnormal karyotype in the presence of a robust checkpoint is defective error correction. A defect in the error correction pathway can generate chromosome missegregation such as lagging chromosomes in anaphase, a defect frequently observed in human tumor cells.^42,43 It is tempting to speculate that both hypo- and hyperactivation of Aurora B may be a more widespread mechanism to generate aneuploidies. Many potential alterations could deregulate this pathway. For example, post-translational modification of CPC subunits is known to alter Aurora B in vitro activity.^{71–73,90,93} Moreover, alteration of gene expression of the CPC subunit Survivin has been identified in human tumors.^94,95 Understanding the intricacies in Aurora B function will highlight more about how cells maintain chromosome segregation fidelity and potentially protect from aneuploidization.

It has been known for several years now that Bub1 insufficiency promotes tumorigenesis.^9,13,96,97 Given that Bub1 overexpression has oncogenic properties,⁴¹ we can now appreciate that optimal Bub1 expression is crucial for preventing aneuploidization and tumorigenesis. Unlike Bub1 insufficiency,^9,13 Bub1 overexpression does not deregulate mitotic checkpoint signaling. Indeed, among mouse models with defects in chromosome segregation, Bub1 overexpression activates an oncogenic mechanism that is unique,^7,8,10–13 as the checkpoint is intact.⁴¹ This underscores why it is important to determine the consequences of overexpression of various checkpoint genes, particularly those that are biomarkers for chromosomal instability.⁴ In summary, the finding that Bub1 has oncogenic properties provides in vivo verification for why increased Bub1 expression correlates with a variety of human cancer types and is a biomarker for poor prognosis.^37,39,98