The fidelity of chromosome segregation during mitosis is intimately linked to the function of kinetochores, which are large protein complexes assembled at sites of centromeric heterochromatin on mitotic chromosomes. These key “orchestrators” of mitosis physically connect chromosomes to spindle microtubules and transduce forces through these connections to congress chromosomes and silence the spindle assembly checkpoint. Kinetochore-microtubule attachments are highly regulated to ensure that incorrect attachments are not prematurely stabilized, but instead released and corrected. The kinase activity of the centromeric protein Aurora B is required for kinetochore-microtubule destabilization during mitosis, but how the kinase acts on outer kinetochore substrates to selectively destabilize immature and erroneous attachments remains debated. Here, we review recent literature that sheds light on how Aurora B kinase is recruited to both centromeres and kinetochores and discuss possible mechanisms for how kinase interactions with substrates at distinct regions of mitotic chromosomes are regulated.
Aurora B kinase regulates kinetochore-microtubule attachment stability in mitosis
Equal division of genetic material during mitosis requires that each sister chromatid of a mitotic chromosome stably attach to spindle microtubules emanating from each of the two opposite spindle poles. A complex network of proteins built on the centromere region of mitotic chromosomes, collectively called the kinetochore, mediates these attachments. Successful chromosome segregation also requires the precise regulation of kinetochore-microtubule attachment stability. In early mitosis, the mitotic spindle begins to form and establish its proper geometry at the same time that microtubules begin to dynamically probe for chromosomes, and as a result, erroneous kinetochore-microtubule attachments are likely to form (Figure 1) [1–6]. In order to prevent premature stabilization of kinetochore-microtubule attachments and to limit the accumulation of erroneous attachments, microtubule turnover at kinetochores is high during early mitosis; conversely, as mitosis progresses and chromosomes begin to bi-orient, kinetochore-microtubule turnover decreases and stably bound microtubules accumulate at kinetochores [2,6–10]. These stable attachments allow kinetochores to harness the forces generated from depolymerizing microtubule plus ends to power chromosome movements and to silence the spindle assembly checkpoint, which delays anaphase until all kinetochores are properly connected to microtubules [2,6,8,11–13].
Kinetochore-microtubule attachments in mitosis
The Chromosomal Passenger Complex (CPC) comprises INCENP, Borealin, Survivin and Aurora B kinase, the enzymatic component of the complex that phosphorylates multiple substrates on mitotic chromosomes to ensure proper chromosome segregation [14–17]. One of the numerous functions of Aurora B kinase is to regulate kinetochore-microtubule attachment stability [14,18–21]. For this purpose, Aurora B phosphorylates outer kinetochore-associated substrates, including Hec1 of the NDC80 complex, which directly links kinetochores to microtubules [8,22,23]. Phosphorylation of Hec1 decreases the affinity of NDC80 complexes for microtubules, and as a result, reduces kinetochore-microtubule attachment stability [24–27]. As mitosis progresses, Aurora B kinase-mediated phosphorylation of Hec1 and other outer kinetochore substrates decreases, resulting in increased stabilization of kinetochore-microtubule attachments, which in turn promotes chromosome congression and silencing of the spindle assembly checkpoint [12,13,24–27].
Current models for kinetochore-microtubule attachment regulation by Aurora B kinase
Although proper regulation of kinetochore-microtubule attachments requires that Aurora B phosphorylate substrates at outer kinetochores, the kinase itself, along with the rest of the CPC, resides prominently at inner centromeres during mitosis [28–32]. To explain how the centromere-localized kinase is capable of regulating the function of outer kinetochore proteins, researchers have proposed the “spatial positioning” model (Figure 2), which posits that activated Aurora B kinase emanates from the inner centromere as a diffusible gradient to phosphorylate its substrates [19,33–38]. In this model, Aurora B is recruited to and activated at inner centromeres in early mitosis, prior to formation of kinetochore-microtubule attachments [31,34–50]. Kinetochores in this case lack pulling forces from attached microtubules and are physically close to the inner centromere. As such, kinetochore substrates are situated within the reach of the kinase and are highly phosphorylated. As kinetochore-microtubule attachments are generated, microtubule-based pulling forces stretch kinetochores away from centromeres and outside the boundaries of the Aurora B gradient, which results in decreased phosphorylation of kinetochore substrates and subsequent stabilization of kinetochore-microtubule attachments [19,33–38,51]. In support of this model, Liu et al.  demonstrated that an ectopically targeted FRET sensor capable of detecting Aurora B kinase activity was phosphorylated when positioned at centromeres, but not at kinetochores, when kinetochores were properly bi-oriented. Additionally, ectopically targeting Aurora B to kinetochores using a Mis12-INCENP fusion protein destabilized kinetochore-microtubule attachments and delayed spindle assembly checkpoint silencing. This led the authors to conclude that stabilization of attachments in metaphase results from the spatial separation of outer kinetochore substrates from centromere-localized Aurora B kinase . It is important to point out, however, that other models describing Aurora B regulation of attachment stability (discussed below) predict that irreversibly targeting Aurora B to kinetochores in metaphase, when kinase activity at this region is known to be low, would lead to destabilization of attachments. Thus, the data presented by Liu et al.  do not necessarily rule out other mechanistic models for Aurora B regulation of kinetochore-microtubule attachment stability .
Spatial positioning model for Aurora B kinase-mediated regulation of kinetochore-microtubule attachment stability
The spatial positioning model described above is rooted in the idea that Aurora B kinase emanates from the inner centromere as a steep, diffusible gradient capable of differentially phosphorylating substrates within a short distance (∼50–100 nm) [26,36,37,51–53]. The presence of such a fine-tuned gradient is debated, and the mechanisms for how the proposed gradient is established and maintained remain unknown [21,38,51,52,54,55]. Similar to the spatial positioning model, the “dog leash” model accounts for differential activity of Aurora B kinase towards its substrates based on their distance from the centromere but does not rely on a diffusible gradient of the kinase. In this proposed mechanism, Aurora B's “zone” of activity is restricted by its interaction with the CPC component INCENP, which contains a long single α-helix (SAH) domain that may be capable of extending up to 80 nm to reach kinetochore substrates in early, but not late mitosis [21,45,47]. Consistent with this model, expression of a mutant version of chicken INCENP containing a shortened SAH domain in human cells resulted in decreased phosphorylation of outer kinetochore-, but not centromere-localized Aurora B substrates .
A third model to describe Aurora B regulation during mitosis suggests that spindle microtubules promote the activation of Aurora B kinase and facilitate phosphorylation of kinetochore substrates [31,40–42,44–50]. While the available data suggest a role for microtubules in activating Aurora B kinase, it seems unlikely that this is the major mechanism for regulating kinase activity at outer kinetochores in response to kinetochore-microtubule attachment, since the Hec1 tail domain and other kinetochore-associated Aurora B substrates remain highly phosphorylated when cells enter mitosis in the presence of the microtubule depolymerizing drug nocodazole [24,37,56].
A growing number of studies in both budding yeast and human cells have demonstrated that Aurora B kinase localizes not only to centromeres, but also to kinetochores, suggesting an alternative regulatory mechanism for controlling kinetochore-microtubule attachment stability, in which Aurora B kinase is recruited directly to kinetochores in early mitosis to phosphorylate its substrates and in turn, is evicted from kinetochores as stable attachments form (Figure 3) [21,24,43,56–62]. In the sections below we discuss recent studies that shed light on the localization and functional properties of the CPC at both centromeres and kinetochores and how these new findings may lead to refinement of the current models for Aurora B kinase regulation of kinetochore-microtubule attachments during mitosis.
Direct recruitment model for Aurora B kinase-mediated regulation of kinetochore-microtubule attachment stability
Aurora B localization and activity at centromeres
Aurora B and the CPC are recruited to centromeres via phosphorylated histones
Aurora B kinase and its CPC cofactors are recruited to the centromere region of mitotic chromosomes just prior to nuclear envelope breakdown [14,40,63], and this recruitment depends on phosphorylation of histones H3 and H2A (Figure 4) . A significant body of work has demonstrated that Haspin kinase phosphorylates histone H3 at Thr3 (pH3-T3), which recruits the CPC component Survivin [64–67]. The BIR domain of Survivin directly interacts with pH3-T3 [28,56,66,68,69], while a separate helical domain of Survivin forms a three-helix bundle with Borealin and INCENP, which is connected to Aurora B through the C-terminal IN-Box of INCENP [32,69]. It has also been demonstrated that Bub1 phosphorylates histone H2A at Thr120 which recruits the Shugoshin proteins Sgo1 and Sgo2 to centromeres, which in turn recruit the CPC [64,70–73]. In metazoans, a number of studies have suggested that this linkage is mediated through Borealin [64,70–73]. Antibodies to both pH3-T3 and pH2A-T120 localize to centromeres, and loss of either phospho-mark reduces centromeric CPC localization, which has led to a model in which Aurora B kinase and the CPC are recruited to regions of centromeric chromatin where the two marks overlap [64,68,72,74]. This concentrated pool of centromere-localized Aurora B kinase is proposed to phosphorylate both centromere and kinetochore substrates to ensure proper chromosome congression and segregation [34,36,37,64,66,71,75,76].
Recruitment pathways for centromere-localized Aurora B kinase
pH2A-T120 and pH3-T3 recruit the CPC to distinct locations within the centromere
In a series of recently published studies, three groups carried out experiments to directly test the above model for CPC localization; that is, to determine if CPC recruitment to inner centromeres requires the overlap of pH3-T3 and pH2A-T120 in human cells [56,61,62]. Using antibodies to both phospho-marks, the authors of all three studies reported that pH2A-T120 and pH3-T3 did not show significant overlap in cells; instead, immunostaining revealed that pH3-T3 localized distinctly as a single focus at the inner centromere, while pH2A-T120 localized as a pair of dots flanking the inner centromere [56,61,62]. This is consistent with previous data suggesting distinct localization patterns of the two histone marks [21,58,64,66,71,77]. Line-scans and two-color localization experiments further revealed that pH2A is localized inside of the inner kinetochore protein CENP-C, on the order of ∼100 nm, in both prometaphase and metaphase cells, which places this mark within the “kinetochore-proximal outer centromere” region, distinct from the pH3-T3-marked chromatin at the inner centromere [56,61,62]. Importantly, all three studies reported that each histone mark individually was sufficient to recruit Aurora B kinase and the CPC. Each group analyzed U2OS cells containing an ectopic Lac operator (LacO) array stably integrated in the short arm of chromosome one that were expressing fusions of either LacI-Haspin or LacI-Bub1 . In cells expressing LacI-Bub1, the chromatin surrounding the LacO array was positive for pH2A-T120 but not pH3-T3, and the single phosphorylation mark (pH2A-T120) was sufficient for recruitment of Aurora B kinase in a manner dependent on Sgo1 [56,61,62]. Similarly, when Haspin was directed to the ectopic locus through expression of a LacI fusion protein, the local chromatin was positive for pH3-T3 but not pH2A-T120, and this single modification was also sufficient to recruit Aurora B kinase and its CPC partners [56,61,62]. Moreover, each histone mark was sufficient to recruit a population of the CPC to spatially distinct regions within the centromere region of mitotic chromosomes in human cells (Figure 4). While depletion of Haspin or inhibition of its kinase activity resulted in loss of both the pH3-T3 mark and accumulation of Aurora B at inner centromeres, a population of Aurora B remained localized to the kinetochore-proximal outer centromere, coincident with pH2A-T120 [56,58,61,62]. Conversely, inhibition of Bub1 kinase activity resulted in loss of the pH2A-T120 mark, but Aurora B kinase and components of the CPC remained localized at the inner centromere coincident with pH3-T3 [56,61,62,74]. Inhibition of both Bub1 and Haspin kinase activities; however, resulted in no detectable Aurora B and CPC components at centromeres [49,56,61,62,74].
In the studies described above, Aurora B kinase localized prominently to inner centromeres as a single focus, but was not clearly discernable at pH2A “marked” kinetochore-proximal outer centromere regions in control cells [56,61,62]. This population of the CPC became readily detectable however, when phosphorylation of H3-T3 was prevented through Haspin knockout or Haspin inhibition [56,58,61,62], suggesting cross-talk between the two centromere-localized populations of the CPC. These results point to the possibility of a multifaceted loading process whereby a pool of Aurora B kinase is loaded directly to the inner centromere binding sites provided by pH3, and a second population of the complex is initially recruited to the kinetochore-proximal outer centromere by pH2A/Sgo1 and subsequently relocated to the inner centromere region. This mechanism is similar to what has been suggested for Sgo1, which first loads to kinetochores in early mitosis, which is required for its subsequent relocalization to the inner centromere, where it functions to protect cohesion and prevent premature sister chromatin separation [38,73,77,79]. Furthermore, authors from the recent Liang et al.  study suggest that relocalization of Aurora B kinase from the pH2A-T120 binding sites to the inner centromere pH3-T3 binding sites in metaphase may be required to silence the spindle assembly checkpoint in response to kinetochore-microtubule attachment. The authors reported that experimentally induced retention of Aurora B and the CPC at the kinetochore-proximal outer centromere in metaphase cells resulted in a small increase (by ∼20%) in Aurora B kinase-mediated phosphorylation of the kinetochore scaffolding protein KNL1 that led to sustained checkpoint signaling . In sum, three recent studies report the identification of discrete populations of Aurora B kinase within the centromere region that are recruited by distinct histone modifications. These studies suggest that the different populations of the CPC functionally interact and cooperatively contribute to the robust accumulation of Aurora B kinase at the inner centromere of mitotic chromosomes.
A role for centromere-localized Aurora B kinase in chromosome segregation
In light of the finding that each of the two histone marks recruits a distinct population of Aurora B and the CPC to centromeres, the authors of the studies described above [56,61,62] tested if either population is required for Aurora B kinase activity at kinetochores or for proper chromosome segregation. The three studies reported that inhibition of either pathway alone did not result in chromosome segregation errors or reduced phosphorylation of kinetochore Aurora B kinase substrates [56,61,62]. However, in cells inhibited for both Bub1 and Haspin kinase activities, chromosome segregation was compromised, although it was noted that the defects were less severe than those observed in cells inhibited for Aurora B kinase itself [56,62]. Thus, the inner centromere and kinetochore-proximal outer centromere populations of Aurora B likely have redundant roles in ensuring proper chromosome segregation. Strikingly, in cells inhibited for both Bub1 and Haspin kinase activities, which resulted in a complete loss of centromere-localized CPC, Aurora B kinase localization remained high at kinetochores, and kinetochore substrate phosphorylation by Aurora B kinase was not largely reduced when compared to control cells (Figure 4) [56,61,62]. As such, the chromosome segregation defects resulting from loss of centromere-localized Aurora B could not be attributed to large-scale loss of phosphorylation of kinetochore substrates [56,61,62]. Importantly, these results provide evidence that centromere accumulation of the Aurora B kinase is not strictly coupled to Aurora B activity at kinetochores. This idea is consistent with a number of previous studies, the first of which demonstrated that centromere-localized Aurora B is not required for the regulation of kinetochore-microtubule attachments in chicken DT40 cells . Here, cells depleted of endogenous Survivin and expressing a mutant version of Survivin defective for centromere localization completed mitosis normally with no detectable chromosome segregation defects . In addition, kinetochore-microtubule attachment defects observed in HeLa cells depleted of the large kinetochore-associated scaffolding protein KNL1, which resulted in loss of kinetochore-associated Aurora B kinase activity, could not be rescued by ectopic targeting of the CPC to centromeres . Finally, several studies have suggested that centromere accumulation of the CPC is uncoupled from kinetochore-associated Aurora kinase activity in budding yeast. Campbell and Desai  reported that budding yeast cells expressing INCENP/Sli15 mutants that fail to localize to centromeres exhibited normal chromosome bi-orientation and Aurora/Ipl1 kinase-mediated error correction. Recent studies have further demonstrated that the CPC is recruited directly to kinetochores in budding yeast, and this population is sufficient for Aurora kinase/Ipl1 activity at kinetochores and error-free chromosome segregation in the absence of centromere-localized CPC [59,60].
An obvious question that emerges from the recent studies in human cells [56,62] is what causes the observed chromosome segregation errors in the absence of centromere-localized Aurora B if the kinase is still able to phosphorylate kinetochore-associated substrates? A plausible explanation is that centromere-localized Aurora B phosphorylates centromere-localized substrates to promote proper chromosome segregation. Previous studies have demonstrated that Aurora B kinase regulates the activity and localization of MCAK, a centromere-localized kinesin-13 motor that promotes microtubule depolymerization, an activity that is implicated in correcting erroneous kinetochore-microtubule attachments [3,9,75,76,81–83]. Interestingly, the centromere localization of MCAK is perturbed in cells that are depleted of Haspin or inhibited for Bub1 kinase activity [62,66,67,84,85], consistent with the notion that alterations in MCAK activity and/or localization might contribute to the chromosome segregation errors observed in cells lacking centromere-localized Aurora B kinase. Further investigation into the role of MCAK and other centromere-localized Aurora B kinase substrates is needed to resolve this issue. It is interesting to note that while antibodies to CPC components localize prominently to the inner centromere in early mitosis, antibodies to phosphorylated, active Aurora B (pT232) and phosphorylated, active INCENP (pS893/pS894) show minimal inner centromere localization in early mitosis, but levels significantly increase as mitosis progresses [24,61]. Consistent with this idea, Sgo1/2-PP2A antagonizes Aurora B activity at the centromere in early mitosis, which may explain this change in activity as mitosis progresses . These results suggest that centromere substrates of Aurora B kinase that contribute to proper chromosome segregation may be phosphorylated and perhaps activated in late mitosis rather than in early mitosis.
Finally, it is important to point out that recent studies demonstrating that centromere-localized CPC is not explicitly required for phosphorylation of kinetochore substrates do not rule out the possibility that the centromere- and kinetochore-localized pools of the CPC may exhibit cross-talk and impact each others’ localization or activity. In fact, studies have demonstrated that while delocalization of the CPC from centromeres did not result in decreased activity of Aurora B at kinetochores in human cells and in M-phase Xenopus egg extracts, the regulation of kinase activity in response to kinetochore-microtubule attachment was compromised [27,65,87]. Why this is the case is not clear, but in the future it will be important to resolve how the centromere pool of the CPC contributes to proper regulation of Aurora B kinase substrate phosphorylation at kinetochores in response to microtubule attachment.
Aurora B localization and activity at kinetochores
Aurora B kinase localizes to kinetochores
As previously mentioned, in addition to its centromere localization, Aurora B has also been detected at the kinetochore. Antibodies directed to active, phosphorylated Thr232, which resides in the T-loop of the kinase domain of Aurora B and is required for full activity of the kinase, recognize kinetochores in early mitosis, centromeres in late prometaphase and metaphase, and to the spindle midzone in anaphase [24,54,88–90]. A similar localization pattern was observed for phosphorylated Aurora B Ser331, a site whose phosphorylation is required for optimal Aurora B activity . Furthermore, antibodies to phosphorylated, active INCENP (pS893/pS894) recognize both kinetochores and centromeres in early mitosis, and this localization shifts primarily to centromeres in late prometaphase and metaphase and the spindle midzone in anaphase . Together, these studies raise the possibility that a population of Aurora B kinase, and likely the entire CPC, is recruited directly to kinetochores in early mitosis where it phosphorylates its kinetochore substrates to promote kinetochore-microtubule turnover. In such a model, as mitosis progresses and kinetochores accumulate bound microtubules, Aurora B is evicted from kinetochores, resulting in decreased kinetochore substrate phosphorylation, stabilization of attachments, and silencing of the spindle assembly checkpoint (Figure 3) [12,13,35,61,92–94].
How is Aurora B kinase recruited to the kinetochore to phosphorylate kinetochore substrates?
A major task that remains to be tackled is identifying the kinetochore binding sites for the CPC. In budding yeast, two research groups have made considerable headway on this front by demonstrating that inner kinetochore COMA (Ctf19/Okp1/Mcm21/Ame1) complex recruits the CPC through a direct interaction between INCENP/Sli15 and Ctf19 [59,60]. Importantly, this kinetochore-associated population is sufficient to support Ipl1/Aurora kinase activity and normal chromosome segregation in the absence of centromere-localized CPC [59,60]. In metazoan cells, however, the kinetochore binding sites for Aurora B and the CPC remain unknown. In a recent study discussed above , authors approximated the location of Aurora B kinase in early mitotic cells to ∼22 nm outside of the inner kinetochore protein CENP-C, which places it ∼8 nm inside of the N-terminus of the outer kinetochore protein Hec1. Many kinetochore proteins localize in this region of the kinetochore, making it difficult to predict specific binding sites. A previous study reported that Aurora B kinase localization to kinetochores is dependent on the kinetochore protein KNL1 , and Broad et al.  demonstrated that Aurora B localization is at least partially dependent on Bub1. Moreover, a recent study found that KNL1 undergoes significant conformational changes upon kinetochore-microtubule attachment . Together, these findings make it tempting to speculate that KNL1 may directly or indirectly provide binding sites for CPC components in early mitosis. As mitosis progresses and kinetochore-microtubule attachments accumulate, KNL1 may undergo conformational changes that occlude these sites, leading to eviction of Aurora B kinase and its CPC cofactors (Figure 3). This speculative model remains to be tested.
Importantly, the available data suggest that Aurora B kinase may be recruited to multiple locations within the kinetochore to facilitate different functions. Based on the recent studies described above, we predict that Aurora B is recruited directly to outer kinetochores to phosphorylate outer kinetochore substrates involved in kinetochore-microtubule attachment regulation [21,24,57,61,89]. However, as mentioned above, in budding yeast, the CPC is recruited to the inner kinetochore COMA complex, whose homolog in humans is the CENP-O/P/Q/U complex, a component of the Constitutive Centromere Associated Network (CCAN) [59,60]. Indeed, Aurora B kinase has substrates at the inner kinetochore that are important for mitotic progression. For example, the Mis12 complex component Dsn1 is phosphorylated by Aurora B in early mitosis to promote kinetochore assembly by facilitating an interaction between the inner-kinetochore protein CENP-C and the Mis12 complex [95–104]. A recent study from Bonner et al.  demonstrated that delocalization of the CPC from centromeres in M-phase Xenopus egg extracts resulted in decreased Dsn1 phosphorylation and consequently, failure to assemble the outer kinetochore. The authors found that the SAH domain of INCENP was required, in a microtubule-independent manner, for Aurora B kinase-mediated phosphorylation of Dsn1 and kinetochore assembly. Ectopic targeting of INCENP lacking its central SAH domain to the inner kinetochore protein Nsl1 and subsequent recruitment of the CPC to this region rescued both Dsn1 phosphorylation and kinetochore assembly . The authors concluded that the INCENP SAH domain is critical for localizing the CPC to the inner kinetochore, in close proximity to the Mis12 complex so that Aurora B is able to efficiently phosphorylate Dsn1 to promote outer kinetochore assembly . These results bring to light the idea that multiple kinetochore functions rely on kinetochore-associated Aurora B kinase activity, and the CPC may be recruited to different locations within the kinetochore to support these functions.
A growing number of studies have demonstrated that centromere-localized Aurora B kinase is not explicitly required for Aurora B kinase activity at kinetochores. Based on these studies and data from numerous model systems, the current model for Aurora B kinase-mediated regulation of kinetochore-microtubule attachment stability (i.e. the “spatial positioning model”) should be re-evaluated. Numerous studies have mapped Aurora B kinase to kinetochores, and ectopically targeting the CPC to kinetochores in multiple cell types rescues loss of the centromere-localized population. In budding yeast, at least one kinetochore binding site for the CPC has been identified, which suggests that Aurora/Ipl1 kinase localizes to kinetochores to specifically phosphorylate kinetochore substrates. This straightforward mechanism for CPC function at budding yeast kinetochores, in which the kinase is recruited to the region of mitotic chromosomes where its substrates are located, is likely utilized in metazoan cells as well. The next major challenge is to identify the binding site, or more likely, binding sites, for the CPC at metazoan kinetochores.
Aurora B kinase, a component of the centromere-localized Chromosomal Passenger Complex, is known as the “master regulator” of kinetochore-microtubule attachment stability.
Individual histone phospho-modifications recruit distinct populations of Aurora B kinase to discrete regions of the centromere.
The centromere-localized population of Aurora B is not required for activity or localization of the kinase at outer kinetochores.
Aurora B kinase has been localized to multiple regions within centromeres and kinetochores of mitotic chromosomes, which calls for a re-evaluation of current models describing the spatial regulation of kinetochore-microtubule attachment stability.
The authors declare that there are no competing interests associated with the manuscript.
Work in the DeLuca lab related to the topic of the article is funded by the National Institutes of Health (NIH) [grant number R35GM130365].
The authors thank members of the DeLuca lab for helpful discussions.