Aurora-PLK1 cascades as key signaling modules in the regulation of mitosis

See allHide authors and affiliations

Science Signaling  14 Aug 2018:
Vol. 11, Issue 543, eaar4195
DOI: 10.1126/scisignal.aar4195


Ever since its first observation under the microscope over a century ago, mitosis—the process wherein a cell divides into two genetically identical daughter cells—has captured the attention of biologists and physicians alike owing to its beauty, sophistication, and importance for the life cycle of eukaryotic organisms. The key factors that control mitosis primarily do so through protein phosphorylation and regulated degradation. In this review, which contains six figures, one table, and 432 references, we discuss the mechanisms by which only a handful of mitotic protein kinases control various mitotic events in distinct spatiotemporal contexts. We highlight the central role in this control of the two-tiered Aurora-PLK1 cascades that are organized by several multifunctional scaffold proteins and integrated within larger mitotic regulatory networks.


Mitosis is controlled by reversible protein phosphorylation involving specific kinases and phosphatases. A handful of major mitotic protein kinases, such as the cyclin B–CDK1 complex, the Aurora kinases, and Polo-like kinase 1 (PLK1), cooperatively regulate distinct mitotic processes. Research has identified proteins and mechanisms that integrate these kinases into signaling cascades that guide essential mitotic events. These findings have important implications for our understanding of the mechanisms of mitotic regulation and may advance the development of novel antimitotic drugs. We review collected evidence that in vertebrates, the Aurora kinases serve as catalytic subunits of distinct complexes formed with the four scaffold proteins Bora, CEP192, INCENP, and TPX2, which we deem “core” Aurora cofactors. These complexes and the Aurora-PLK1 cascades organized by Bora, CEP192, and INCENP control crucial aspects of mitosis and all pathways of spindle assembly. We compare the mechanisms of Aurora activation in relation to the different spindle assembly pathways and draw a functional analogy between the CEP192 complex and the chromosomal passenger complex that may reflect the coevolution of centrosomes, kinetochores, and the actomyosin cleavage apparatus. We also analyze the roles and mechanisms of Aurora-PLK1 signaling in the cell and centrosome cycles and in the DNA damage response.

Introduction to mitotic kinases

Mitosis is a highly complex process whereby a parent cell produces two genetically identical daughter cells. Mitosis comprises a host of coordinated events that are temporally divided into four phases— prophase, metaphase, anaphase, and telophase—that are immediately followed by cytokinesis (1). A key event in mitosis is the equal segregation of the duplicated chromosomes by a bipolar microtubule (MT)–based molecular machine termed the mitotic spindle. Spindle MTs are generated through different pathways at each of four distinct structures that are all defined as MT-organizing centers (MTOCs): centrosomes, chromosome arms, kinetochores, and preexisting MTs (25). Mitosis is precisely orchestrated by reversible protein phosphorylation resulting from the concerted action of certain protein kinases and phosphatases. The complex formed by cyclin B and cyclin-dependent kinase 1 (CDK1) is the master regulator of mitosis, around which additional mechanisms have evolved to ensure robustness, precision, and specificity for distinct mitotic events (6, 7). The serine-threonine kinases of the Aurora family and the founding member of the Polo-like kinase (PLK) family, PLK1, are found in all eukaryotic lineages and cooperate with CDK1 in the control of mitosis and cytokinesis, being essential for proper execution of these processes (6, 813). These mitotic kinases are frequently overexpressed in cancers and are considered as attractive anticancer drug targets (14, 15). Together with the Aurora kinases and PLK1, some members of the never in mitosis gene A (NIMA)–related kinase (NEK) family of protein kinases also participate in the regulation of mitosis downstream of CDK1 (1618).

Two highly similar kinases of the Aurora family, Aurora A (AurA) and Aurora B (AurB), perform different functions in mitotic cells. AurA forms distinct complexes with several cofactors that guide its subcellular localization and activity (12, 19). It localizes to spindle MTs and to centrosomes, which are non–membrane-bound organelles consisting of one or two (depending on the cell cycle phase) centrioles surrounded by pericentriolar material (PCM) that serve as major MTOCs (12, 20). AurA plays an essential role in centrosome maturation and separation—two events that occur concomitantly at the G2-M phase transition and represent the earliest stages of mitotic spindle formation (2, 6, 12, 2123). Centrosome maturation involves a substantial increase in size and MT-nucleating capacity due to the recruitment of additional PCM components, including the γ–tubulin ring complex (γ-TuRC), which serves as a template for MT nucleation at MTOCs in animal cells (20, 24). AurA is also required for timely mitotic entry and assembly of the bipolar spindle (7, 12). AurB (in mitotic cells) and its close relative AurC (in meiotic cells) interact directly with the scaffold protein inner centromere protein (INCENP), through which each serves exclusively as a catalytic subunit of the heterotetrameric chromosomal passenger complex (CPC), which also includes the regulatory and targeting subunits Survivin and Borealin (13, 25, 26). All CPC subunits depend on one another for their localization and stability. As the name implies, the CPC changes its mitotic localization from chromatin (in prophase) to centromeres and kinetochores (in prophase and metaphase) to central spindle and midbody (in anaphase and telophase), and it is involved in chromosome dynamics and cohesion, kinetochore-MT attachments, the spindle assembly checkpoint, and cytokinesis (25, 27).

The PLK family comprises five members. PLK1 is a multifunctional kinase implicated in various aspects of mitosis and cytokinesis (11, 28, 29). By contrast, PLK4 functions in S phase as the key centriole assembly–promoting factor, and it is only found in species that have centriolar centrosomes (3033). PLK2, PLK3, and PLK5 are only present in vertebrates and have diverse functions in nonproliferative tissues (11, 29). A characteristic feature of PLKs is the presence of the C-terminal polo boxes, which are organized into polo box domains (PBDs) and are involved in the control of kinase activity and target specificity (34, 35). Through its PBD, which recognizes specific phosphorylated motifs, PLK1 docks onto a myriad of proteins to perform critical mitotic functions (36, 37). The PBD docking causes conformational changes in PLK1 that partially activate the kinase. Full activation of PLK1, however, requires its phosphorylation in a region known as the activation loop or T-loop (34, 35, 38). The functions of PLK1 in mitosis and cytokinesis overlap with those of both AurA and AurB (6, 11, 14). This observation, along with the colocalization of PLK1 with AurA at centrosomes and with AurB at centromeres, kinetochores, the central spindle, and the midbody (Fig. 1), has long suggested a functional link between the Aurora kinases and PLK1 (6).

Fig. 1 Regulation of mitosis and cytokinesis by the four complexes organized by the core Aurora cofactors.

The complexes formed by each of the four core Aurora cofactors with Aurora A (AurA) or Aurora B (AurB) are shown in the middle of the figure along with text indicating the subcellular localization of each complex. All core Aurora cofactors, except TPX2, upon their priming phosphorylation, bind PLK1 and organize the respective Aurora kinase and PLK1 into a two-tiered kinase cascade that controls distinct events during mitotic progression, as illustrated by the diagrams of cells. Filled ovals next to each cell indicate which of the cofactors form functional complexes with Aurora at each stage of the cell cycle, and the subcellular localization of AurA (red), AurB (yellow), and PLK1 (blue) during mitotic progression is shown. Inactive AurA is shown in gray; active forms of the kinases are shown in color. Phosphate groups are depicted as solid black dots. Kinase-activating phosphate groups on the T-loop of AurA, AurB, and PLK1 are depicted as black dots with a yellow center.


How the same protein kinases (the cyclin B–CDK1 complex, the Aurora kinases, and PLK1) work together in different spatiotemporal contexts to guide diverse mitotic events is central to our understanding of the regulation of mitosis. Research in the last decade has shed light on this matter by identifying proteins and mechanisms that control different pools of the Aurora kinases and PLK1 in specific mitotic compartments and by demonstrating that AurA and AurB serve as physiological activating kinases for PLK1 (3948). Here, we review these findings and present evidence that in vertebrates, the Aurora kinases serve as catalytic subunits of the distinct complexes they form with the scaffold proteins Bora, centrosomal protein of 192 kDa (CEP192), targeting protein for Xklp2 (TPX2), and INCENP, which can be defined as “core” Aurora cofactors (Fig. 1). These complexes control crucial aspects of mitosis and all four pathways of spindle assembly. Moreover, three of the abovementioned scaffold proteins (Bora, CEP192, and INCENP) organize an Aurora kinase and PLK1 into two-tiered signaling cascades underlying key mitotic events. These cascades may also involve certain NEK family kinases, which are known to cooperate with the Aurora kinases and PLK1 in promoting centrosome separation, spindle assembly, and cytokinesis (16, 17). The proteins of the NEK family are quite enigmatic in their origin and function because the number of NEK paralogs varies substantially and appears to correlate with the role of cilia or flagella—MT-based organelles formed by specialized centrioles termed basal bodies—in the life cycle of eukaryotes. Whereas the genomes of nonciliated yeasts and molds encode a single NEK gene, humans have 11 NEK family members. By contrast, the genome of Giardia lamblia, a unicellular parasite that has four pairs of flagella, encodes nearly 200 NEK paralogs (~70% of the kinome). It was, therefore, proposed that the ancestral function of the NEK proteins is related to the evolution of the basal bodies (17, 49, 50). The functions of the NEK kinases will not be discussed in this review. For information on this subject, the reader is referred to published reviews (1618, 49).

TPX2 and CEP192: Two major AurA cofactors and mediators of MT nucleation

Among the mitotic protein kinases, AurA has attracted special interest because of its overexpression and amplification in various cancers and its proposed role as an oncoprotein (12, 51). Several proteins have been implicated in the regulation of AurA activity [for details, see (12, 19, 52)]. The first identified and most studied of the AurA cofactors is TPX2, a key player in the chromatin-driven spindle assembly pathway (3, 53, 54). This pathway relies on the production of RanGTP—the guanosine 5′-triphosphate (GTP)–bound form of the small guanosine triphosphatase (GTPase) Ran—by the chromatin-associated Ran guanine nucleotide exchange factor (GEF), regulator of chromosome condensation 1 (RCC1).

In the mitotic cytoplasm, TPX2 and numerous other spindle assembly factors are inhibited through the binding of their nuclear localization signals to the nuclear transport receptors importin β and the importin α/β heterodimer. RanGTP releases this inhibition through binding to importin β, which results in the activation of the spindle assembly factors with consequent nucleation, stabilization, and bundling of MTs near chromosomes (53, 55, 56). RanGTP also promotes the binding of TPX2 to AurA, which enables AurA activation and targeting to spindle MTs (Fig. 2) (5760). The TPX2-AurA complex promotes MT nucleation through a mechanism that has been partly elucidated. Specifically, in the presence of RanGTP, TPX2-AurA oligomerizes and binds to another complex comprising the TPX2 partner protein receptor for hyaluronan-mediated motility (RHAMM) and γ-TuRC (61, 62). AurA then phosphorylates neural precursor cell expressed, developmentally down-regulated protein 1 (NEDD1), the adaptor subunit of γ-TuRC, at a conserved serine residue, thus rendering the TPX2–AurA–RHAMM–γ-TuRC complex competent for MT nucleation (Fig. 2) (62, 63). TPX2 and RHAMM, in addition to their role in MT nucleation, also promote MT focusing at spindle poles by an incompletely understood mechanism that depends on a heterodimeric E3 ubiquitin ligase (and breast and ovarian tumor suppressor) composed of breast cancer susceptibility gene 1 (BRCA1) and BRCA1-associated RING domain protein 1 (BARD1), and, therefore, likely involves protein ubiquitylation (54, 61, 64).

Fig. 2 Mechanism of action of the core AurA cofactors.

In G2 phase, the pericentrin (PCNT)–mediated recruitment of CEP192 complexes initiates the AurA-PLK1 kinase cascade that culminates in the PLK1 phosphorylation-dependent docking of γ-TuRC and CKAP5 onto CEP192 and drives centrosome maturation and separation. In parallel with this cascade, another AurA-PLK1 cascade organized by Bora begins to operate in G2 phase and activates the bulk of PLK1 in the cytoplasm (middle). Before mitotic entry, active PLK1 promotes SCFβTrCP-mediated proteasomal degradation of Bora, thus making itself and AurA available to other mitotic partner proteins and substrates. Upon mitotic entry and nuclear envelope breakdown (M phase), the RanGTP that is produced by the chromatin-associated GEF RCC1 releases TPX2 from its inhibitory interaction with importin α/β (Imp α/β). TPX2 then binds to and activates AurA in a phosphorylation-independent manner. The TPX2-AurA complex oligomerizes (not shown) and interacts with the RHAMM–γ-TuRC complex to form an MT-nucleating TPX2–AurA–RHAMM–γ-TuRC complex, which, after its phosphorylation by AurA, becomes competent for MT nucleation. These processes underlie the chromatin-driven spindle assembly pathway (1). In addition, TPX2 works together with the Augmin complex and γ-TuRC to enable MT-driven spindle assembly (2). CKAP5 cooperates with γ-TuRC in promoting MT nucleation. Inactive forms of AurA and PLK1 are shown in gray; active forms are shown in color. Phosphate groups are depicted as solid black dots. Kinase-activating phosphate groups on the T-loop of AurA and PLK1 are depicted as black dots with a yellow center. The associated adaptor subunit of the γ-TuRC, NEDD1, is not shown for simplicity.

TPX2 has also emerged as a key partner of the eight-subunit Augmin complex and of γ-TuRC in a distinct spindle assembly pathway that enables MT nucleation and branching from preexisting MTs, leading to MT amplification within the spindle (Fig. 2) (3, 4, 65, 66). It was reported that Augmin depletion completely abolishes TPX2-dependent MT nucleation in cell-free Xenopus laevis egg extracts, suggesting that TPX2 promotes MT nucleation exclusively from preexisting MTs and not de novo (67). This finding is important but needs validation because critical TPX2 partners in MT nucleation could be codepleted along with the Augmin complex.

That TPX2 depletion abrogates AurA localization to spindle MTs but not to centrosomes (58, 68) suggests that different factors target the kinase to these two mitotic structures. Although numerous proteins, including Ajuba, p21 (Rac1)–activated kinase 1 (PAK1), actin-related protein 2/3 complex subunit 1B (ARPC1B), nucleophosmin, and Bora, have been implicated in the stimulation of AurA activity at centrosomes, how the kinase is recruited to and activated at these organelles and how it promotes centrosome maturation has remained unclear (41, 6974). A study using Xenopus egg extracts helped to answer these questions by identifying CEP192 [known as spindle-defective protein 2 (Spd-2) in invertebrates] as the major targeting and activating cofactor of AurA at centrosomes (39). This finding was key to deciphering the role of AurA at centrosomes because CEP192 is a master regulator of biogenesis and function of these organelles, and, in vertebrates, it is the most critical of the resident centrosomal proteins for mitotic PCM formation and for centrosome-driven MT nucleation in mitotic and interphase cells (7582).

Subsequent work in Xenopus egg extracts and human cells revealed that the CEP192-mediated AurA recruitment to and activation at centrosomes initiates a multistep signaling cascade underlying centrosome maturation and separation and MT nucleation (47, 48). Morphologically, centrosome maturation presents as an expansion of the outer PCM layer of the interphase centrosomes into a proteinaceous PCM matrix that nucleates and anchors mitotic centrosomal MTs (20, 24, 8386). In vertebrates, centrosome maturation is initiated by the recruitment of CEP192 to centrosomes in G2 phase. This recruitment depends on the large scaffold protein pericentrin, which organizes the PCM in interphase, and likely involves phosphorylation of pericentrin by PLK1 (47, 87, 88). Both PLK1 and CEP192 are present in the inner PCM layer throughout the cell cycle (83), but how PLK1 is initially activated in G2 phase is unknown. In cells, CEP192 is constitutively bound to a fraction of AurA, and it recruits AurA to centrosomes both in G2 and in mitosis (Fig. 2) (39, 47). CEP192 also binds to PLK1 and recruits it to centrosomes (47, 48). Experimental evidence indicates that the accumulation of CEP192 complexes in the PCM promotes their oligomerization (possibly with the involvement of additional factors), which triggers AurA autophosphorylation in the T-loop at Thr288 in the human protein or Thr295 in the Xenopus protein (Thr288/295), resulting in kinase activation (39, 47, 89, 90). AurA then phosphorylates PLK1 at the conserved Thr210 (human) or Thr201 (Xenopus) in its T-loop, thus activating PLK1 and facilitating its docking onto CEP192 at Thr44 (human) or Thr46 (Xenopus). This docking may require a priming phosphorylation of CEP192 Thr44/46 by an as-yet-unidentified kinase (47). The CEP192-bound PLK1 then phosphorylates CEP192 at several serine residues to generate the attachment sites for γ-TuRC in complex with its adaptor protein NEDD1. The phosphorylation of CEP192 by PLK1 also generates bindings site(s) for the cytoskeleton-associated protein 5 (CKAP5) [also known as colonic and hepatic tumor overexpressed protein (chTOG) and Xenopus MT-associated protein of 215 kDa (XMAP215)] and possibly for other proteins (47). CKAP5 functions as both an MT polymerase and an essential cofactor of γ-TuRC in MT nucleation (9196). The Cep192-organized kinase cascade (Fig. 2) culminates in MT nucleation and anchoring to the PCM and has been recapitulated using beads coated with a recombinant N-terminal 1000–amino acid fragment of CEP192 in Xenopus egg extracts (47).

Thus, in vertebrates, AurA serves as a critical catalytic subunit of distinct protein complexes formed by CEP192 and TPX2 that drive MT nucleation in at least two different pathways of spindle assembly mediated by centrosomes and chromatin, respectively. AurA, through its binding to TPX2, is also predicted to associate with the putative TPX2–Augmin–γ-TuRC MT-nucleating complex, but the role of AurA in MT-driven spindle assembly requires further investigation (65, 67). In this regard, AurA has been proposed to promote the affinity of the Augmin complex for MTs by phosphorylating the MT-binding domain of the HAUS augmin-like complex subunit 8 (HAUS8) (97).

Distinct mechanisms of AurA activation at centrosomes and spindle MTs

Given the key role of TPX2 and CEP192 as AurA cofactors in two separate spindle assembly pathways (chromatin-mediated and centrosome-mediated, respectively), it seems important to know whether these proteins use different mechanisms to activate AurA and, if so, whether this difference relates to the distinct modes of MT nucleation by chromatin and by centrosomes. The activation of protein kinases mainly occurs through one of two mechanisms (or both simultaneously): (i) the phosphorylation at a specific residue in the T-loop by the kinase itself (autophosphorylation) or by another kinase, or (ii) the direct interaction of the catalytic domain with another domain of the kinase or with a binding partner protein (98100). The distinction between these two mechanisms in relation to AurA is important because the abundance of AurA phosphorylated at Thr288/295 in the T-loop is often used as a readout of the activity of the kinase in cells (52). Early in vitro studies with recombinant proteins concluded that TPX2 activates AurA both by direct binding and by protecting the T-loop from dephosphorylation by protein phosphatases (57, 59). Notably, the bacterially produced recombinant AurA used in those studies was autophosphorylated in the T-loop, presumably as a result of spontaneous AurA oligomerization or aggregation, and required dephosphorylation by protein phosphatases before being used in in vitro assays (57, 59, 101).

By contrast, multiple lines of evidence indicate that AurA T-loop autophosphorylation is inhibited in the cytoplasm and at spindle MTs and occurs only at spindle poles as a consequence of CEP192-mediated kinase activation, suggesting that AurA is activated by different mechanisms at centrosomes and spindle MTs. First, AurA T-loop phosphorylation depends on the presence of centrosomes, and ablation of the endogenous CEP192-AurA interaction abrogates both the centrosomal localization and the T-loop phosphorylation of AurA in Xenopus egg extracts and in mammalian cells (39, 47). Second, the fraction of AurA that localizes to centrosomes and spindle pole MTs is phosphorylated on the T-loop, whereas the MT-bound pool of the kinase, which is activated by TPX2, is not phosphorylated in mammalian cells (39, 102104). Accordingly, TPX2 does not promote substantial T-loop phosphorylation of endogenous AurA in extracts from metaphase-arrested Xenopus eggs (39). Third, structural studies have provided evidence for the existence of two distinct modes of AurA activation: by dimerization-mediated T-loop autophosphorylation and by TPX2 binding-mediated allostery. Although both modes activate AurA to approximately the same extent (~100-fold), the underlying mechanism appears to be quite different in terms of dynamics of the conformational changes within the T-loop (43, 105, 106). In addition, although the centrosomal AurA cofactor in invertebrates is still unknown, a study in Caenorhabditis elegans showed that T-loop–phosphorylated AurA resides only at centrosomes and is required for centrosome-driven MT nucleation, whereas the MT-associated, Tpx2-bound AurA is nonphosphorylated and essential for chromatin-mediated MT assembly (107).

Thus, CEP192 and TPX2 target AurA to distinct mitotic structures and use fundamentally different mechanisms to activate the kinase: CEP192 enables oligomerization-dependent AurA T-loop autophosphorylation, whereas TPX2 activates AurA through direct binding (Table 1). Each mechanism seems to have adapted to the corresponding spindle assembly pathway. Because the CEP192-mediated mechanism depends on the local concentration of CEP192-AurA complexes but does not depend on RanGTP, it begins to operate in G2 phase, before nuclear envelope breakdown, and couples CEP192 centrosomal recruitment to AurA activation, which triggers the AurA-PLK1 cascade, culminating in MT nucleation and anchoring and resulting in the formation of radial centrosomal MT arrays (39, 47, 48). In contrast, the formation of the MT-nucleating TPX2–AurA–RHAMM–γ-TuRC complex depends on both RanGTP and nuclear envelope breakdown, because TPX2 is a nuclear protein and AurA and RHAMM are predominantly cytoplasmic (12, 61, 62, 108). Thus, TPX2-mediated MT nucleation occurs around the chromatin, in a diffuse manner, and only after nuclear envelope breakdown (Fig. 2).

Table 1 Properties of the major Aurora complexes.
View this table:

Cooperativity between CEP192 and TPX2 in stimulating AurA-PLK1 signaling and bipolar spindle assembly

The mechanisms by which CEP192 and TPX2 activate AurA in cells need further investigation. In particular, it is not clear why the T-loop phosphorylation of endogenous AurA requires its binding to CEP192 and does not occur in complexes with TPX2, although TPX2-bound AurA forms dimers or higher-order oligomers on MTs (39). In addition, it is puzzling why a fraction of the MT-bound AurA in the vicinity of spindle poles, where it is targeted through the dynein-dependent poleward transport of TPX2, is phosphorylated on the T-loop (39, 54, 102104, 109, 110). The answers to these questions may, in part, come from studies on the serine/threonine-specific protein phosphatase 6 (PP6), which is the main T-loop phosphatase of AurA in cells (109). During mitosis, PP6 localizes to the cytoplasm without a substantial association with spindle structures, specifically dephosphorylates AurA in endogenous TPX2 complexes, and is essential for proper spindle assembly and chromosome segregation (109, 111). A recent study revealed that PP6 is inhibited during mitosis by PLK1, which docks onto PP6 and phosphorylates its regulatory subunit, PP6R2, at multiple sites (112). PLK1 docking onto PP6R2 depends on a priming phosphorylation of PP6R2 by CDK1, but it is not essential for PLK1-mediated PP6 inhibition (112). In light of these findings and of the fact that active PLK1 accumulates at centrosomes in CEP192 complexes, it can be inferred that PP6 is less active when localized in the vicinity of centrosomes than around spindle MTs. This local decrease of PP6 activity may enable T-loop autophosphorylation [and, hence, enhanced activation (106, 113)] of AurA in TPX2 complexes targeted to spindle poles, thereby explaining the presence of AurA phosphorylated at Thr288/295 on spindle pole MTs (Fig. 3A).

Fig. 3 Regulation of the Aurora-PLK1 cascades at spindle poles and kinetochores.

(A) AurA-PLK1 signaling is facilitated at spindle poles through multiple PLK1-mediated positive feedback loops. The active PLK1 in CEP192 complexes may (i) phosphorylate PCNT, thereby promoting CEP192 recruitment (double-headed arrow) (87); (ii) phosphorylate CEP215, thereby facilitating expansion of the PCM (as inferred from studies in Drosophila) and the interdependent recruitment of CEP215 and PCNT (double-headed arrow) (88, 151, 152); (iii) abrogate the CEP192-PP1 interaction by phosphorylating the PP1-docking motif of CEP192 (115); and (iv) cooperate with cytoplasmic PLK1 in phosphorylating and inactivating PP6 (112), which may enable activation loop autophosphorylation of the MT-bound AurA in TPX2 complexes (109). (B) The CPC, comprising AurB, INCENP, Borealin, and Survivin, organizes a network of proteins that drives the formation and function of the outer kinetochore. Only a portion of the functional protein interactions occurring at the centromere and kinetochore are shown, without depicting the distinction between these two structures or the kinetochore-MT attachment status. The recruitment of CPCs to centromeres and kinetochores promotes CPC oligomerization and trans-autophosphorylation of AurB, which then phosphorylates multiple substrates, including PLK1. It was proposed that, after the initial CDK1-dependent CPC recruitment and AurB activation at the inner kinetochore, AurB promotes the recruitment of the kinase BUB1, which phosphorylates H2A histone at Thr120 to generate a binding site for SGO1 or SGO2 (SGO1/2) [(158); reviewed in (141)]. SGO1/2 recruits the CPC in a manner that depends on Borealin dimerization and phosphorylation by CDK1 (427429). In a positive feedback loop, active AurB, PLK1, and CDK1 phosphorylate and activate Haspin, thereby triggering the main mechanism of CPC recruitment to centromeres (154158, 430432). This mechanism is counteracted by the PP1 regulatory subunit Repo-Man, which is recruited to chromatin and dephosphorylates histone H3 at Thr3 (256258). See text for further details. Red arrows, phosphorylation events; black arrows, activating effects; black lines ending with a T-bar, inhibitory effects; dashed lines, protein-protein interactions. Phosphate groups are depicted as solid black dots. Kinase-activating phosphate groups on the T-loop of AurA, AurB, and PLK1 are depicted as black dots with a yellow center.

CEP192 complexes have also been implicated in the regulation of PP1, which, along with PP2A, plays a major role in the regulation of mitosis and mitotic exit (114, 115). PP1 dephosphorylates a large spectrum of proteins, and its substrate specificity is determined by docking onto the conserved canonical PP1-binding motif RVxF in target proteins that act as regulatory subunits of the phosphatase (116, 117). It was shown that CEP192 is one of such proteins and that PLK1-mediated phosphorylation of Thr951 in the PP1-docking motif of CEP192 abolishes PP1 binding (Fig. 3A) (115). Because PP1 has wide substrate specificity and antagonizes both the Aurora kinases and PLK1 (57, 114, 118123), one could speculate that PP1 may counteract any step of the CEP192-organized kinase cascade. Thus, CEP192 and TPX2 may cooperate with one another and with protein phosphatases to spatially restrict AurA and PLK1 activity. This would ensure maximal activation of both kinases at spindle poles and moderate activation of AurA (because of the PP6-mediated restraint of its T-loop autophosphorylation) on spindle MTs, which is essential for proper execution of mitosis (39, 47, 48, 109).

Furthermore, CEP192 and TPX2 may work together in promoting spindle pole separation—a process that is integral to bipolar spindle formation. This notion is based on the observation that beads coated with an antibody that recognizes AurA, which recruit both CEP192 and TPX2 through AurA, organize bipolar spindles, whereas CEP192-coated beads, which do not recruit TPX2, promote only the assembly of monopolar MT asters in Xenopus egg extracts (47, 124). Spindle pole separation is driven by the plus end–directed motor proteins kinesin family member 11 (KIF11) (also known as kinesin-5 and Eg5) and KIF15, which act synergistically and redundantly to generate sliding forces on antiparallel MTs between the two poles (125128). TPX2 stimulates the MT binding and activity of KIF11 and KIF15, in part, by promoting the formation of the MT bundles to which KIF15 binds (54, 110, 128131). Conceivably, these TPX2 functions are essential for organizing the two CEP192-generated centrosomal MT asters into a bipolar spindle.

AurA-PLK1 cascades

Both AurA and PLK1, in addition to their roles in centrosome maturation and spindle assembly, are also required for timely mitotic entry (7, 11, 12). Studies of this particular process led to the first demonstration of a direct functional link between AurA and PLK1. Specifically, it was shown that timely mitotic entry relies on PLK1 activation by AurA, which is mediated by the cytoplasmic scaffold protein and AurA cofactor Bora. Bora binds AurA and PLK1 and promotes activating phosphorylation of PLK1 at Thr210 by AurA in G2 phase (Fig. 2) (41, 42, 71). PLK1 then phosphorylates several substrates of the mitotic entry network, leading to CDK1 activation and driving cells into mitosis (7). In a positive feedback loop, CDK1 phosphorylates Bora at the PBD-docking motif and at other sites, thereby facilitating PLK1-Bora binding and PLK1 activation by AurA (132135). PLK1 also phosphorylates Bora to generate a phosphodegron recognized by the S phase kinase-associated protein 1 (SKP1)–cullin-1–F-box β-transducin repeat containing protein (βTrCP) (SCFβTrCP) E3 ubiquitin ligase complex, which targets most of Bora for proteasome-mediated degradation before the onset of mitosis (132, 136). Thus, Bora organizes AurA and PLK1 into a two-tiered kinase cascade that culminates in the activation of PLK1 and CDK1 concomitant with the degradation of Bora (Fig. 2).

Following mitotic entry, cytoplasmic PLK1 remains phosphorylated on the T-loop throughout mitosis (41, 42), and most of it is refractory to dephosphorylation and inactivation, even after AurA depletion (47, 137). It was proposed that the small pool of Bora that escapes degradation continues to promote PLK1 activation during mitosis (137). The observations in Xenopus egg extracts and mammalian cells suggest that additional factors (such as the CPC) may also contribute to PLK1 activation in the mitotic cytoplasm independently of AurA and Bora (47, 138). Although it was initially suggested that Bora promotes PLK1 activation both in the cytoplasm and at centrosomes (41), accumulating evidence indicates that two distinct AurA-PLK1 cascades operate simultaneously, beginning in G2 phase: One is organized by CEP192 at centrosomes and enables centrosome maturation (47, 48), and the other is organized by Bora in the cytoplasm and ensures activation of cytoplasmic PLK1 (Fig. 2 and Table 1) (41, 42, 137). Unlike CEP192 and TPX2, Bora localizes exclusively to the cytoplasm, without association with any cellular structure(s) (42). In addition, Bora does not seem to activate AurA per se; instead, it “unlocks” the T-loop of PLK1, thus rendering it a better substrate for AurA (41, 42).

In addition to the kinase cascades organized by Bora and CEP192, other AurA-PLK1 cascades may exist to allow fine-tuning of the spatiotemporal activity or more specialized functions of both kinases. In this regard, two proteins, Furry and Gravin [also called A-kinase anchoring protein 12 (AKAP12)], also promote PLK1 activation by AurA (139, 140). Furry localizes to centrosomes and spindle poles and is required for spindle pole integrity but not for centrosome maturation and separation (139). Gravin has been proposed to facilitate asymmetric recruitment of AurA and PLK1 to spindle poles, which enables proper spindle orientation in germ-line stem cells (140).

Aurora-PLK1 modules at distinct mitotic signaling platforms

Analogy between the CEP192 complex and the CPC

As noted above, the subcellular localization and roles of PLK1 overlap with those of both AurA and AurB (6, 11, 14). Studies in Drosophila melanogaster and human cells have provided a compelling mechanistic explanation for this overlap by demonstrating that AurB, like AurA, serves as a PLK1-activating kinase (4446). AurB, through binding to INCENP, plays an essential role in kinetochore formation and function and in cytokinesis as a part of the CPC (25, 27, 141143). AurB-mediated phosphorylation of the PLK1 T-loop occurs at centromeres, kinetochores, the central spindle, and the midbody and is required for the establishment of correct kinetochore-MT attachments and for proper cytokinesis (4446). Moreover, INCENP, after undergoing priming phosphorylation at Thr388 by CDK1, binds to PLK1 and, at least in Drosophila, serves as a platform for AurB-PLK1 signaling (Fig. 1) (25, 44, 144, 145).

These observations support the hypothesis that the CEP192 complex and the CPC are functionally similar, although they contain different Aurora paralogs (146). Similar to the interactions of CEP192 with AurA and PLK1 (39, 47, 48), INCENP binds to AurB constitutively and to PLK1 in a manner that depends on phosphorylation of INCENP and appears to promote AurB-mediated activation of PLK1 (25, 44, 144, 145). Notably, AurB may phosphorylate PLK1 docked onto INCENP or other neighboring proteins (4446, 144, 147, 148). In addition, the recruitment of CEP192 complexes to centrosomes (39, 47) and of CPC to centromeres and kinetochores (40, 44, 149) is inherently coupled to AurA or AurB activation, respectively, by an analogous, oligomerization-dependent mechanism that results in kinase T-loop autophosphorylation and initiates Aurora-PLK1 signaling. These observations are in line with the fact that the highly conserved critical threonine residue in the T-loop that is phosphorylated in AurA, AurB, and PLK1 each lies within the consensus phosphorylation motif for the Aurora kinases (150). The recruitment of CEP192 complexes and CPCs and the subsequent Aurora kinase activation are facilitated by the downstream kinase itself (PLK1) through several positive feedback loops. Both the recruitment of all components of the CEP192 complex and centrosome maturation require the PLK1-mediated phosphorylation of pericentrin (47, 87, 88). In addition, PLK1 may facilitate AurA activation at spindle poles by promoting the recruitment of pericentrin, in a complex with CEP215, to centrosomes and by inhibiting PP1 and PP6 (Fig. 3A) (88, 112, 151, 152). Likewise, PLK1 facilitates AurB activation at centromeres or kinetochores through phosphorylation of the CPC subunit Survivin (153) and of Haspin (154, 155), a serine-threonine kinase that phosphorylates histone H3 to generate a binding site for the CPC on centromeric chromatin (Fig. 3B) (156158). Furthermore, the T-loop phosphorylation of both AurA and AurB is normally suppressed by protein phosphatases and molecular chaperones, and this suppression is released upon oligomerization of CEP192 complexes or CPCs, respectively (39, 40, 109, 112, 159). Finally, the active Aurora kinase in CEP192 complexes and CPCs promotes MT assembly, albeit through distinct mechanisms specific for the centrosome-driven and kinetochore-driven pathways, respectively. Whereas the CEP192 complex directly promotes MT nucleation and anchoring (47), CPC stabilizes MTs in the vicinity of kinetochores through AurB-mediated phosphorylation and inhibition of MT-destabilizing factors, such as mitotic centromere-associated kinesin (MCAK) and stathmin 1 [STMN1; also known as oncoprotein 18 (OP18)] (3, 40, 160). Both the CEP192-mediated (39, 47) and the CPC-mediated (40, 149, 161163) mechanisms of MT assembly have been recapitulated by forced oligomerization of endogenous CEP192 complexes and CPCs, respectively, on artificial templates in Xenopus egg extracts.

Thus, the CEP192 complex and CPC appear to operate in a similar fashion, which likely reflects their analogous role in driving the Aurora- and PLK1-dependent recruitment and supramolecular assembly of proteins into the outer PCM layer and the outer kinetochore, respectively (Fig. 4A) (47, 48, 141, 163, 164). Both these mitotic structures serve as MTOCs and signaling platforms and assemble over a constitutive inner network of proteins that are present at centrosomes, centromeres, and inner kinetochores throughout the cell cycle (20, 25, 8386, 141). AurB, and, by inference, the CPC, is the most upstream regulator of the recruitment of the outer kinetochore proteins, including those involved in the spindle assembly checkpoint (141, 163, 164). Likewise, all three core components of the CEP192 complex (CEP192, AurA, and PLK1) (47) are essential for centrosome maturation, with CEP192 lying at the hierarchical top of this process (11, 12, 75, 76, 82).

Fig. 4 The Aurora-PLK1 cascades organized by CEP192 and INCENP operate at the opposite ends of spindle MTs.

(A) The CEP192 complex at centrosomes and the chromosomal passenger complex (CPC) at centromeres and kinetochores promote Aurora and PLK1 phosphorylation-dependent formation of the mitotic outer PCM layer and the outer kinetochore, respectively. (B) By serving as a γ-TuRC anchor, CEP192 localizes exclusively to the minus ends of spindle MTs. INCENP, as a scaffold protein for the CPC, functions in prometaphase and metaphase at the interface between chromatin and spindle MTs. In anaphase, the CPC relocates from chromosomes to MTs of the central spindle and to the actomyosin ring, where it promotes the formation of the central spindle and cleavage furrow and the ingression of the cleavage furrow during cytokinesis. Only a few nonbranched MTs of the anaphase asters are shown for simplicity. Note that, in both metaphase and anaphase, CEP192 complexes and CPCs localize to the spindle poles and the cell equator, respectively.

The analogy between the CEP192 complex and the CPC is consistent with the evidence of the similarity between and interchangeability of AurA and AurB. A single amino acid substitution (G198N) is sufficient to convert AurA into an equatorial (localized to the equatorial plane of the mitotic spindle) kinase with AurB-like properties (165169). Moreover, AurA promotes the kinetochore-MT attachment error correction by phosphorylating kinetochore substrates of AurB, such as the kinetochore complex component NDC80 [also called highly expressed in cancer 1 (HEC1)], which is a key constituent of the NDC80 complex and the core MT-binding component of the kinetochore (170173). Furthermore, AurA appears to be capable of binding to INCENP and, hence, of substituting for AurB in the CPC (172). Consistent with these observations, certain organisms, such as the social amoeba Dictyostelium discoideum and the starfish, have only one Aurora paralog that functions as both a polar and an equatorial Aurora kinase (174, 175). Phylogenetic analyses suggest that Aurora kinases, although present in all eukaryotes, have undergone lineage-specific expansions through gene duplication events to give rise to polar and equatorial kinase(s) (176, 177). Because the duplications occurred independently in plants, invertebrates, and vertebrates, it was inferred that the polar and equatorial Aurora kinases in these lineages are paralogs, rather than true orthologs, although they have similar localization and functions (176). Thus, conceivably, a protein other than the CEP192 ortholog Spd-2 may serve as the centrosomal cofactor of AurA in invertebrates, and this would explain why an interaction between AurA and Spd-2 has not been reported in Drosophila or C. elegans. In summary, the close evolutionary and structural relationships between polar and equatorial Aurora kinases and the functional analogy between the CEP192 complex and the CPC are consistent with the hypothesis that centrosomes and kinetochores have evolved from a common pre-eukaryotic precursor MTOC that was associated with both the cell wall and the chromatin and had a dual function as both centrosome and kinetochore (178).

The CEP192 complex and the CPC as signaling hubs at opposite ends of spindle MTs

Owing to their distinct scaffolding properties, CEP192 and INCENP organize the Aurora-PLK1 cascades at the opposite (minus and plus, respectively) ends of spindle MTs (Fig. 4B). Whereas CEP192, by docking γ-TuRC, serves as a centrosomal anchor for the minus ends of growing MTs (47), INCENP functions during mitosis at the interface between chromatin and spindle MTs by virtue of being targeted to these structures through its centromere-targeting domain that binds Survivin and Borealin and two MT-binding domains, respectively (141, 161, 179181). The dual recognition of chromatin and MTs is essential for AurB activation, spindle assembly, and robust spindle assembly checkpoint arrest (161, 181, 182). The association of the CPC with the MT plus ends is ensured by the CPC-mediated recruitment of the kinetochore null protein 1 (KNL1) complex, MIS12 complex, NDC80 complex (KMN) network—a conserved complex of proteins of the outer kinetochore that serves as a platform for end-on kinetochore-MT attachments—and of the plus-end tracking (+TIP) proteins, such as Duo1 and multipolar spindle protein 1 (Mps1)–interacting protein 1 (Dam1) and spindle and kinetochore-associated (SKA) complexes, cytoplasmic linker protein (CLIP)–associating proteins (CLASPs), CLIP-170, CKAP5, MCAK, and MT end-binding protein 2 (EB2). Through selective binding to plus ends of growing MTs, these proteins control MT dynamics and strengthen the affinity of the KMN network for MT plus ends (25, 173, 183186). Because of their distinct mitotic localization, the CEP192 complex and the CPC engage in phosphorylation of different substrates and, hence, exert specific effects on MT assembly and dynamics.

The CEP192 complex and the CPC in the coevolution of centrosomes, kinetochores, and the actomyosin cleavage apparatus

Centrosomes play a key role in cytokinesis by determining the plane of cell division. In anaphase, when chromosomes move poleward, MTs of the two centrosome-organized asters interact with their plus ends at the spindle center to promote the formation of the central spindle (Fig. 4B). The anaphase MT asters comprise centrosomal MTs interconnected by an MT meshwork generated through branching MT nucleation (187189). The antiparallel plus ends of the MTs of these asters and of the MTs nucleated by chromatin are stabilized, amplified through the Augmin pathway, and bundled by specific proteins to form the central spindle (189191).

Both the astral MTs and the central spindle emit signals to the cell cortex that regulate the activity of RhoA, a small GTPase that promotes contractile ring assembly and cleavage furrow ingression (187, 192, 193). A signal generated by the shorter astral MTs inhibits RhoA at cortical regions adjacent to spindle poles (187). In C. elegans, this RhoA-inhibitory signal is mediated by the active AurA in Tpx2 complexes bound to astral MTs (194). Another signal, which is generated by the central spindle and by a population of stabilized, long astral MTs at the cell equator, activates RhoA at the equatorial cortex, thereby promoting actomyosin ring assembly and contraction followed by the ingression of a cleavage furrow that partitions the cytoplasm of the dividing cell (187, 192, 193).

The CPC localizes to both the central spindle and the actomyosin contractile ring (Fig. 4B) and plays a key role in the formation of these structures and in the coordination of cytokinesis with chromosome segregation (25). As a result of a decrease in CDK1 activity at the onset of anaphase, the CPC dissociates from chromosomes and binds to an MT plus end–directed motor enzyme, the mitotic kinesin-like protein 2 (MKLP2) [also known as kinesin family member 20A (KIF20A)], which transports the CPC to the central spindle (25, 195, 196). The localization of PLK1 to the central spindle also requires MKLP2, suggesting that PLK1 may associate with the CPC at this structure, like it does at centromeres (44, 144, 196). In addition, the CPC is targeted to the equatorial cortex through direct INCENP binding to actin (197199). Notably, the depletion of CEP192 in human cells inhibits Rac1, a member of the Rho GTPase family that promotes actin polymerization and has been implicated in tumor invasiveness and metastasis (200, 201). This finding and the observation that Rac1 localizes to centrosomes (202) raise the intriguing possibilities that CEP192 is involved in the actin-organizing function of the centrosome (203) and that the coupling of the actin cytoskeleton to the tubulin cytoskeleton is yet another common feature of the CEP192 complex and the CPC.

The CPC and AurB- and PLK1-mediated phosphorylation of proteins of the central spindle and of the contractile ring have been implicated in multiple aspects of cytokinesis, such as the bundling and stability of the MTs of the central spindle, the length of the central spindle, and furrow ingression and abscission (25, 28, 29, 187, 193, 199). In Drosophila cells, phosphorylation of the Plk1 T-loop by AurB releases Plk1 from its inhibitor, Map205, which is required both for Plk1 recruitment to the midbody and for successful cytokinesis (45). It is yet unclear whether this mechanism operates in vertebrate cells. Key steps of cytokinesis, including central spindle formation and the MKLP2-dependent recruitment of the CPC and other components of the central spindle and cleavage furrow, have been recapitulated in Xenopus egg extracts using beads coated with an antibody that recognizes AurA, which act as artificial centrosomes (124, 204). On the basis of experiments using this system, Ishihara and colleagues proposed that centrosomes not only nucleate radial arrays of MTs but also may initiate chemical waves that travel across the cytoplasm and induce secondary MT nucleation events underlying the formation of the astral MT meshwork (205, 206). Because the centrosome-like properties of the AurA-binding beads in Xenopus egg extracts are driven by the CEP192-organized AurA-PLK1 cascade (39, 47), it can be inferred that this cascade underlies the formation of the MT asters that guide cytokinesis, at least in this experimental system. It is tempting to speculate that the CEP192-organized kinase cascade also generates a signaling wavefront that guides aster growth and positioning and that the CEP192 complex and the CPC functionally interact with one another by relaying signals through MTs.

Among centrosomal proteins, the presence of CEP192 or its ortholog Spd-2 is the most strongly correlated with the presence of centrosomes in the organism, suggesting that this protein has played a key role in the evolution of these organelles (30, 77, 207, 208). There is no gene encoding an ortholog of CEP192 in the genome of the planarian flatworm Schmidtea mediterranea, which lacks centrosomes but forms centrioles de novo in ciliated cell types (77). This finding is surprising because centrosomes are essential for early embryo development, although they may be dispensable at later developmental stages (209213). What makes S. mediterranea exceptional is the fact that its embryonic development does not depend on stereotyped cleavage patterns generated by oriented cell divisions and precise cleavage plane geometry (77, 214, 215). It was therefore proposed that the main selective force for the evolution of the animal centrosome was the preservation of cell polarity and cell individuation necessitated by the transition to multicellularity (77, 216). Because the cleavage plane geometry in animal cells is determined by coordinated signaling from the centrosomal MT asters and the central spindle to the cell cortex, it was further postulated that the centrosome and the cortical actomyosin cleavage apparatus have coevolved (216, 217). On the other hand, as noted above, centrosomes and kinetochores are thought to have evolved from a common pre-eukaryotic ancestral MTOC (178). These three hypotheses point to an evolutionary link between centrosomes, kinetochores, and the cleavage furrow–based cytokinesis apparatus. It is remarkable that this link is supported by experimental evidence suggesting that the analogous, functionally overlapping kinase cascades organized by CEP192 and INCENP underlie the formation of mitotic centrosomes, kinetochores, and the cleavage furrow.

Aurora and PLK1 regulation at mitotic signaling platforms

It can be assumed from the above analysis that the elucidation of the signaling cascades organized by CEP192 and INCENP may reveal common regulatory mechanisms rooted in the evolutionary relatedness of centrosomes and kinetochores. At present, the side-by-side comparison of the two signaling cascades may not give a complete picture of their similarities because of our limited knowledge of the regulation of the CEP192 complex (as opposed to that of the CPC, which has long been intensely studied). Here, we highlight only the aspects of the CPC regulation that, based on what is known about the CEP192 complex, might also be pertinent to CEP192.

It seems contradictory that PLK1, despite its role as a downstream kinase in the Aurora-PLK1 cascades, facilitates AurA activation at centrosomes and of AurB activation at centromeres and kinetochores (42, 153155, 218). However, this contradiction can be explained by the existence of multiple positive feedback loops in the kinase cascades that converge on Aurora activation. These positive feedback loops are formed because (i) the recruitment of CEP192 complexes and CPC to the sites of their action is inherently linked to oligomerization-driven Aurora activation, and (ii) the circuits involved in the recruitment of CEP192 complexes are themselves activated by PLK1, whereas those involved in the CPC recruitment are activated by AurB and PLK1 (Fig. 3). Because these two regulatory principles also favor interdependence of the positive feedback loops, the enrichment and activation of AurB at centromeres and kinetochores may occur in an autocatalytic, stepwise, and exponential manner. Specifically, a small pool of the kinase that was proposed to be initially activated through an unknown mechanism at inner kinetochores may initiate the CPC recruitment pathway that is mediated by budding uninhibited by benzimidazole protein 1 (BUB1) and the centromeric cohesion protector proteins shugoshin 1 or shugoshin 2 (SGO1/2) and limits AurB localization and activation to kinetochores, which may, in turn, trigger the main, Haspin-mediated pathway that drives the recruitment and activation of most of the centromere-associated AurB (141, 219). A positive feedback loop of AurB activation also facilitates kinetochore-MT attachment error correction and coordinates it with checkpoint signaling. This feedback loop is mediated by the monopolar spindle protein 1 (MPS1), which is activated by AurB (27, 141, 220, 221). MPS1 phosphorylates the key scaffold protein of the KMN network, KNL1, at multiple sites, generating phospho-marks specifically recognized by BUB3, which enables the recruitment of the BUB1-BUB3, Bub1-related kinase (BUBR1)–BUB3, and mitotic arrest deficient-like 1 (MAD1)–MAD2 checkpoint complexes to unattached kinetochores (184, 219). MPS1 also facilitates SGO1/2-mediated recruitment of CPC and AurB activation, in part, by phosphorylating Borealin (Fig. 3B) (222, 223). The KMN network integrates the kinetochore-MT attachment and error correction with spindle assembly checkpoint (173, 183, 184). AurB and PLK1, together with CDK1, play a central role in this integration, as well as in the assembly of the KMN network (141, 184, 224226). The robust, multilayered CPC recruitment process may be essential for the rapid formation of the outer kinetochores at mitotic entry. It would, however, result in persistent, uncontrolled AurB and PLK1 activation and phosphorylation of downstream substrates if it were not counteracted by protein phosphatases.

One of the main roles of Aurora in CEP192 complexes and, possibly also in the CPC, is to activate PLK1 through T-loop phosphorylation (4448). Notably, the cytoplasmic pool of PLK1 is T-loop–phosphorylated at mitosis onset and is maintained in this state throughout mitosis by the Bora-AurA complex and, possibly, by additional factors (41, 42, 137). Thus, it seems counterintuitive that the Aurora-mediated PLK1 activation in CEP192 complexes and CPCs would be essential during mitosis because the cytoplasmic PLK1, which docks to its target proteins, is already active. Hence, the primary selective force for the emergence of the AurA-PLK1 cascades organized by CEP192 and INCENP might have been the involvement of these scaffold proteins in events occurring outside mitosis, when PLK1 is not fully active: in G2 phase (centrosome maturation and outer kinetochore formation) and during and after mitotic exit (cytokinesis). Nevertheless, Aurora activation in both complexes may also play an important role during mitosis, because the Aurora-PLK1 cascades receive other regulatory inputs that may allow rapid switching and fine-tuning of the activity of both kinases.

The regulated phosphorylation of proteins at the kinetochore-MT interface by AurB and PLK1 is essential for the main kinetochore functions: kinetochore-MT attachment and error correction and the spindle assembly checkpoint (25, 138, 141). AurB and PLK1 appear to act antagonistically in controlling these processes: Whereas the AurB-mediated phosphorylation of the KMN network proteins reduces the MT-binding affinity of the kinetochore, which promotes error correction, PLK1 stabilizes kinetochore-MT attachments, in part, by creating a binding site for the AurB-inactivating protein phosphatase PP2A on the kinetochore-associated checkpoint protein BUBR1 (227229). The functional antagonism between AurB and PLK1 raises the question as to how a dynamic switch between the activities of two kinases of the same cascade is achieved (230). PLK1 docks to multiple centromere and kinetochore proteins, and this docking may amplify PLK1 activation by AurB (Fig. 3B) (44, 119, 144, 148, 155, 225, 231242). On the other hand, because the activation of the Aurora kinases, PLK1, and some of their downstream kinases is mediated by T-loop phosphorylation, the activities of any of these kinases could be selectively quenched by protein phosphatases that specifically target them (114, 243, 244).

Several centromere and kinetochore proteins have been shown to recruit serine-threonine protein phosphatases—most notably those of the PP1 and PP2A families. Such proteins, therefore, promote their own dephosphorylation or dephosphorylation of neighboring proteins (Fig. 3B) (114). Among the most studied examples is the recruitment of the PP2A phosphatase, in a complex with its regulatory subunit B56, by SGO1/2 and BUBR1 to centromeres and kinetochores, respectively. The SGO1/2-PP2A-B56 complex counteracts the AurB- and PLK1-mediated removal of centromeric cohesin from chromosomes, which is essential for the enrichment of cohesin and CPC at centromeres [reviewed in (219, 245)]. In addition, SGO1/2-PP2A-B56, together with MCAK (which also binds to SGO1/2), contributes to the kinetochore-MT attachment error correction [(227, 246248); reviewed in (141)]. BUBR1-PP2A-B56 cooperates with a complex formed by PP1 and KNL1 in a regulatory circuit to counteract the activity of AurB at kinetochores, thereby stabilizing the kinetochore-MT attachments (Fig. 3B) (121, 122, 227229, 249). In addition, BUBR1-PP2A-B56 and PP1-KNL1 remove the MPS1-mediated KNL1 phosphorylations necessary for the recruitment of the BUB checkpoint proteins (Fig. 3B). This generates a negative feedback loop that enables spindle assembly checkpoint silencing when proper kinetochore-MT attachments are established (250253).

PP1 also forms a complex with myosin phosphatase-targeting subunit 1 (MYPT1), a regulatory subunit that localizes to both centrosomes and kinetochores. PP1-MYPT1 inhibits PLK1 by dephosphorylating it at Thr210, which destabilizes kinetochore-MT attachments (Fig. 3B) (118, 119, 254). MYPT1, upon its phosphorylation by CDK1, binds to PLK1, and the resulting PP1-MYPT1-PLK1 kinase-phosphatase complex was proposed to regulate the balance between the pools of PLK1 docked to kinetochore substrates that are primed by CDK1 and those docked to kinetochore substrates that are primed by PLK1 itself, such as the centromere protein U (CENP-U) (Fig. 3B) (118, 119, 237, 239, 255). PP1-MYPT1 may also counteract the activity of PLK1 at centrosomes because MYPT1 depletion rescues the centrosomal localization of γ-tubulin in PLK1-deficient cells (118).

Notably, the docking of PP1 to KNL1 and to another PP1 regulatory subunit, Repo-Man, which counteracts Haspin by mediating the dephosphorylation of histone H3 at Thr3, is abrogated by the AurB-mediated phosphorylation of these proteins at a Ser or Thr residue within the consensus PP1-binding motif, in a manner analogous to the PLK1-mediated inhibition of the PP1-CEP192 interaction (Fig. 3) (115, 121, 123, 249, 256260). Moreover, proteins that contain the canonical PP1-binding motif RV[S/T]F are phosphorylated at this motif specifically during mitosis, usually by AurB, and this modification abrogates PP1 docking (260). These findings suggest that PP1 holoenzyme complexes may play an especially important role in regulating the CPC-centered signaling networks. The data also reinforce the view that CDK1, PLK1, and AurB not only target substrates directly but also interfere with the localization or activity of the counteracting phosphatases, or both, at mitotic signaling platforms (112, 121, 260262).

Thus, it has become apparent that the CPC works together with PLK1-docking proteins and phosphatase complexes to organize sophisticated regulatory circuits underlying kinetochore assembly and functions (Fig. 3B). It can be inferred that AurB may phosphorylate PLK1 from a distance either through its release and diffusion from CPCs or through stretching of the single α-helix (SAH) domain of INCENP, which allows AurB to reach more distant substrates by acting as a flexible tether (Fig. 5A) (141, 179). The resulting mixed field of the Aurora and PLK1 activities could be shifted to favor the activity of either kinase by specific protein phosphatases and other regulators (Fig. 5, B and C). For example, a shift toward predominant AurB activity favoring error correction could be established through PLK1 inactivation or removal by PP1 or by an E3 ubiquitin ligase, such as CUL3-KLHL22, which promotes degradation-independent PLK1 dissociation from kinetochores (Fig. 5B) (118, 263). Conversely, a transition toward predominant PLK1 activity favoring the kinetochore-MT attachment could result from AurB dephosphorylation by PP1 or PP2A-B56 phosphatase complexes. An analogous outcome could also be achieved by INCENP degradation or stripping from centromeres through various mechanisms (Fig. 5C) (25, 141).

Fig. 5 Examples of the AurB-PLK1 activity switches at kinetochores.

(A) In the CPC at kinetochores, AurB may perform the activating T-loop phosphorylation of PLK1 that is docked onto the phosphorylated PBD-binding motif of either INCENP or a neighboring protein (X), leading to signal amplification and to the formation of a mixed AurB and PLK1 activity field (green). (B) The balance of the local kinase activities can be shifted toward AurB (blue) through the inactivation of PLK1 or its removal from INCENP or X by either a protein phosphatase (PPase) that dephosphorylates PLK1 and its docking proteins or by CUL3, which promotes degradation-independent PLK1 dissociation from kinetochores. (C) The balance of the local kinase activities can be shifted toward PLK1 (yellow) through PPase-mediated inactivation of AurB (1), stripping of the CPC from kinetochores (2), or degradation of INCENP (3). Red arrows, protein phosphorylation; black arrows, protein dephosphorylation, inactivation, or stripping; gray arrows, protein dislocation. Inactive forms of AurB are shown in gray; active forms of AurB and PLK1 are shown in color. For simplicity, the CPC components Borealin and Survivin are omitted, and the CPC is not shown in its oligomerized state. Note that in the mitotic cytoplasm, PLK1 is active because it is phosphorylated at Thr210 in the T-loop by the Bora-AurA complex. Phosphate groups are depicted as solid black dots. Kinase-activating phosphate groups on the T-loop of AurB and PLK1 are depicted as black dots with a yellow center.

Toward comprehensive understanding of Aurora-PLK signaling

The core Aurora cofactors

We have established that the Aurora kinases serve as upstream activators of PLK1 and appear to mediate most of their mitotic functions as catalytic subunits of four major complexes organized by the scaffold proteins Bora, CEP192, TPX2, and INCENP (Fig. 1 and Table 1). These four proteins control the localization and activity of Aurora in the cytoplasm and at all key mitotic structures that also serve as signaling platforms (2, 187, 206, 264266)—centrosomes, spindle MTs, centromeres, kinetochores, the central spindle, and the midbody—and are integral to all four pathways of spindle assembly. We, therefore, propose to call these proteins core Aurora cofactors. All core Aurora cofactors, except TPX2, have been shown to bind PLK1 in a phosphorylation-dependent manner and to assemble two-tiered Aurora-PLK1 cascades, leading to PLK1 activation and culminating in specific downstream effects at the corresponding mitotic compartments (Fig. 1 and Table 1).

Whereas INCENP serves as a CPC platform and as the only cofactor of the equatorial kinases AurB and AurC (25, 145), the three core AurA cofactors Bora, CEP192, and TPX2 appear to bind the kinase directly and in a mutually exclusive manner and to compete with one another for binding to the kinase (39, 57, 132). On the basis of these observations and on the structural studies of the TPX2-AurA and INCENP-AurB complexes, it can be inferred that the core AurA cofactors bind to a common region in the catalytic domain of AurA corresponding to the region of AurB that mediates binding to INCENP (57, 145). In contrast, several other reported AurA regulators, such as the transforming acidic coiled-coil–containing protein 3 (TACC3), Ajuba, and calmodulin, bind to the AurA N-terminal region outside of the kinase domain (69, 267, 268). The competition between Bora, CEP192, and TPX2 for binding to AurA may have important implications because it allows for control of mitotic processes through reciprocal coordination of the abundance of the core AurA cofactors, which form functionally distinct complexes with the kinase. Experimental interference with the abundance of any core AurA cofactor appears to reciprocally affect the function of the other two, presumably by altering the availability of AurA. For example, expression of recombinant Bora or of its AurA-binding fragment in mammalian cells results in centrosome and spindle defects (132) similar to those caused by the ablation of CEP192 and TPX2 (58, 75, 76), whereas Bora depletion leads to an increase in the density of spindle MTs (132), consistent with enhanced function of the TPX2-AurA complex (269). Likewise, the recombinant AurA-binding fragments of CEP192 or TPX2 displaced endogenous AurA from centrosomes and spindle MTs and inhibited centrosomal MT nucleation in Xenopus egg extracts (39, 269). The fundamental importance of the core AurA cofactors is further underscored by the observations that the abundance of Spd-2 in the cytoplasm and the Plk1 docking to Spd-2 determine centrosome size and spindle length in C. elegans embryos (270, 271) and that the TPX2-AurA interaction is essential for proper length of the mitotic spindle in human cells (272). Thus, the core AurA cofactors, in addition to controlling specific AurA pools, may have a more general mitotic regulatory role, which may explain why either depletion of CEP192 or overexpression of Bora in mammalian cells leads to more severe mitotic defects than those seen after centrosome ablation or PLK1 depletion or inhibition, respectively (75, 76, 132).

Key role of AurA-PLK signaling in the centrosome cycle

Each of the core Aurora cofactors is involved in at least one of the four spindle assembly pathways, although the involvement of Bora appears to be indirect (Table 1). Consistent with this and with the role of PLK1 as a downstream effector of AurA and AurB, it was proposed that PLK1 has evolved also to control spindle MT–related processes, specifically MTOC formation and cytokinesis. This assumption is based on the fact that PLKs, unlike the Aurora kinases, are conspicuously absent in land plants, which lack centrosomes and other defined MTOCs and use a cytokinesis mechanism that differs substantially from that of animals in that it does not involve cleavage furrow formation (11, 273). The major role of PLK1 in centrosome and spindle formation and in cytokinesis is also supported by studies of the PLK1 interactome and phosphoproteome in human cells (112, 147, 218, 274, 275).

Like PLK4 and CEP192, PLK1 is essential for centrosome biogenesis and function. In proliferating cells, centrosomes duplicate once per cell cycle. During this process, a new centriole, termed a procentriole, is assembled orthogonally within the PCM of each of the two parental centrosomes at its proximal end (276281). The canonical centrosome duplication cycle is synchronized with and in many ways analogous to the DNA replication cycle, with both cycles being driven by oscillations of cyclin and CDK activity (20). Centrosome duplication in vertebrates is initiated in G1–early S phase through recruitment of the key regulator of centriole assembly, PLK4, to each of the two parental centrosomes. PLK4 recruitment requires CEP192 and two other centrosomal proteins, CEP63 and CEP152 (282289). The local PLK4 concentration at parental centrosomes promotes PLK4 activation through trans-autophosphorylation in a manner analogous to AurA and AurB activation in CEP192 complexes and CPCs, respectively (39, 40, 290293). Moreover, akin to an active Aurora kinase that initiates protein recruitment to centrosomes or kinetochores, active PLK4 triggers a cascade of sequential recruitment of the core centriolar proteins. This cascade is evolutionary conserved and has been elucidated and covered in many reviews (20, 294297).

Phylogenetic evidence suggests that both CEP192 and PLK4 have evolved with the emergence of the PCM and the “canonical”—that is, containing centrioles and PCM—centrosome (30, 77, 207). Whereas basal bodies are widespread across eukaryotes, orthologs of CEP192 and PLK4 are found only in the Amorphea lineage, almost exclusively in opisthokonts, which have canonical centrosomes (30, 31). The presence of CEP192 appears to correlate with the presence of the MT-nucleating PCM (30, 77, 207, 208). Consistent with this observation, CEP192, PLK4, and certain other PCM proteins colocalize to acentriolar MTOCs (aMTOCs) that are formed—presumably by the PCM—during spindle assembly in mouse oocytes and early embryos (298300). PLK4 likely evolved by duplication of the PLK1 gene specifically to render the PCM capable of centriole assembly during interphase, in synchrony with DNA replication (30, 281). During evolution, PLK4 has acquired the unique abilities to oligomerize through its diverged, cryptic PBD and to self-activate through T-loop trans-autophosphorylation (290293, 301, 302). In contrast, PLK1 has become dependent on Aurora kinases for activation, which fostered the evolution of the Aurora-PLK1 cascades and scaffolding proteins. CEP192 represents the epitome of this evolution, because it contains a multitasking N-terminal domain that organizes the AurA-PLK1 cascade, serves as a platform for the PLK1-mediated docking of γ-TuRC and CKAP5, and is also involved in the recruitment of PLK4 (47, 48, 284, 286, 287).

Following procentriole assembly and elongation, which occur in S and G2 phases, the procentrioles convert into centrosomes through acquisition of the PCM. This process, termed centriole-to-centrosome conversion, begins in G2-M phase and concludes in the following G1 phase (280, 303, 304). By the end of mitosis, each newly made centriole also loses the cartwheel—a structure that forms at the base of the nascent procentriole and determines its ninefold symmetry—and disengages from the parental centrosome while still retaining a connection to the parental centrosome through a flexible tether (294, 297). All three events—PCM acquisition, cartwheel loss, and centriole disengagement—require PLK1 activity and are essential for procentrioles to become centrosomes capable of nucleating MTs and duplicating (279, 280, 303, 305309). Because CEP192 serves as a master regulator of PCM protein recruitment through AurA and PLK1 activation in G2-M and contributes to PLK4 recruitment in G1-S (47, 48, 284, 286), it controls both stages of centrosome duplication: centriole assembly and the centriole-to-centrosome conversion.

Accumulating evidence suggests that the CEP192 complex may also function in interphase. CEP192 is required for MT nucleation by interphase centrosomes (76, 310), and AurA and PLK1 are involved in the suppression of ciliogenesis (268, 311315). In this regard, the recently identified functional partner of CEP192, cancerous inhibitor of PP2A (CIP2A), promotes primary cilia disassembly through AurA activation (316, 317). Moreover, PCM maintenance and centriole stability in interphase cells have been suggested to require PLK1 activity (318). Given these observations and the finding that both CEP192 and PLK1 localize to the inner PCM layer (83), it is tempting to speculate that the CEP192 complex is integral to centrosome maintenance throughout the cell cycle. All these data are consistent with the key role of CEP192 in centrosome evolution and biogenesis.

Aurora-PLK1 signaling in noncanonical pathways of spindle assembly

Evidence suggests that, in addition to the four main spindle assembly pathways discussed above, other spindle assembly pathways may exist, at least in certain cell types. One such pathway, wherein spindle MTs are generated and organized by acentriolar PCM foci, operates in mouse oocytes and early embryos (298, 319, 320). An analogous acentriolar pathway may also contribute to spindle formation in other systems, as evidenced by centrosome ablation experiments in Drosophila, chicken DT40 cells, and monkey kidney cells (321324).

In most animals, centrioles are provided to the zygote by the sperm, as basal bodies of the flagellum. To avoid the formation of a multipolar spindle in the zygote, centrosomes are usually degraded or inactivated in oocytes (325). Centriole inactivation during oogenesis in Drosophila is driven by a reduction in Plk1 expression and the consequent loss of PCM from centrosomes (318). In rodents, unlike in most animals, sperm basal bodies degrade during spermatogenesis or fertilization and, therefore, early embryos lack centrosomes until the blastocyst stage, when centrioles are generated de novo (325, 326). Despite the absence of centrosomes, key PCM proteins form aMTOCs that drive spindle assembly in mouse oocytes and early embryos. These aMTOCs undergo PLK1-dependent remodeling and fragmentation in G2-M phase followed by refocusing into the two spindle poles (298, 319, 320, 327). This fragmentation was proposed to occur through the same PLK1-dependent mechanism that drives centrosome separation (298, 307, 328).

CEP192 is a bona fide marker of the aMTOCs that form in mouse oocytes and early embryos (298). Moreover, CEP192 colocalizes with AurA and PLK1 at aMTOCs and is required for the formation of these structures (298, 300, 317). Furthermore, aMTOCs also contain several CEP192 partner proteins: pericentrin, CEP152, AurA, PLK1, PLK4, and γ-tubulin (298300, 317, 319). Proper aMTOC formation and localization of CEP192 and its partner proteins require CIP2A. CIP2A binds to CEP192, and both proteins cooperatively promote the recruitment of pericentrin, AurA, and PLK1 to aMTOCs, as well as local AurA T-loop phosphorylation. These CIP2A functions depend on its phosphorylation by PLK1, but not on PP2A activity (317). CIP2A also localizes to centrosomes and promotes centrosome separation in a PP2A- and PP1-independent manner, although the mechanism is unclear (329, 330). These findings strongly suggest that the CEP192-organized kinase cascade drives the PCM protein recruitment underlying the formation of both mitotic centrosomal MTOCs and aMTOCs in vertebrates. Consistent with this notion and with the preferential reliance of PCM assembly in flies on the CEP152 ortholog Asterless (Asl) and the CEP215 ortholog Centrosomin (Cnn), it was proposed that the formation of both mitotic centrosomes and aMTOCs in Drosophila is driven by an analogous mechanism involving these two proteins and, to a lesser extent, the CEP192 ortholog Spd-2 (321).

Another putative, noncanonical spindle assembly pathway is related to the spindle matrix, the long-elusive amorphous material that retains some integrity upon MT disassembly and is thought to facilitate spindle formation by concentrating α/β-tubulin heterodimers and MT-organizing factors (331337). It has been proposed that the formation of the spindle matrix and of non–membrane-bound organelles, including centrosomes, involves a liquid-liquid phase separation (coacervation) that occurs as a result of molecular crowding of specific intrinsically disordered proteins or RNAs, or both (335, 338, 339). A study in Xenopus egg extracts identified the low-complexity protein, Bub3-interacting and GLE-2–binding sequence containing ZNF207 (BuGZ), as a putative key spindle matrix component that undergoes phase transitions and promotes spindle assembly (335). Moreover, AurA was shown to bind BuGZ and to become activated upon incorporation into BuGZ coacervates in vitro. On the basis of this observation and on the reduction of AurA T-loop phosphorylation on spindle MTs in BuGZ-depleted cells, it was proposed that BuGZ coacervation promotes AurA activation in mitosis (340). The notion that BuGZ is an AurA cofactor in what might be an additional spindle assembly pathway is intriguing and in line with the concepts outlined above. If BuGZ coacervation and its involvement in AurA activation are confirmed in cells, it will be important to investigate how these processes relate to other spindle assembly pathways and mitotic regulatory networks, whether and how the coacervation contributes to the formation of the mitotic PCM and outer kinetochores, and, if so, whether it is regulated by the Aurora-PLK1 cascades.

Aurora-PLK1 signaling in the control of mitotic entry

Experimental and phylogenetic evidence suggests that the increasing complexity of the mitotic MTOCs and of the chromosome segregation mechanisms—driven, in part, by the transition to multicellularity—has been the primary selective force for the evolution of the Aurora-PLK signaling. However, the Aurora kinases and PLKs also have functions that do not seem to be directly related to the MT cytoskeleton, such as the regulation of the cell cycle checkpoints, DNA replication, the DNA damage response (DDR), and DNA repair (29, 341343). At least some of these functions could be traced to the ancestral role of these kinases in mitotic spindle assembly and chromosome segregation.

One example supporting this notion is the regulation of mitotic entry by the Bora module. This module is present only in animals, because the Bora gene is not found outside the animal lineage (71, 132134). This fact and the observation that land plants, whose cell cycle machinery is otherwise very similar to that of yeast and animals, lack PLKs (273, 344) suggest that the Bora-AurA-PLK1 mitotic entry regulatory module is not a core component of the cell cycle control system and might have evolved secondarily to the mitotic function of PLK1. The phosphorylation of PLK1 at Thr210 in the T-loop is essential for proper execution of mitosis and cell viability in unperturbed cells (11, 29, 345). It is therefore conceivable that to guard the cell against deleterious mitotic errors that may arise due to insufficient PLK1 activity, the activation of the mitotic entry network became dependent on PLK1-mediated phosphorylation events. This could result in a sensor mechanism that monitors the concentration of active PLK1 in the cytoplasm and delays the initiation of mitosis until this concentration reaches a certain threshold that is sufficient to support mitosis. Consistent with this hypothesis, PLK1 activation in the late G2 phase is important for commitment to mitosis during normal cell cycles (346). Furthermore, the putative G2-M PLK1 activity sensor (Fig. 6) might have been co-opted by the G2-M DNA damage checkpoint, which prevents cells with genomic damage from entering mitosis, thereby giving cells a chance to repair the damaged DNA. This checkpoint is initiated by activation of the DNA damage-sensing serine-threonine kinases ataxia-telangiectasia mutated (ATM) and ATM- and Rad3-related kinase (ATR). ATM and ATR, with the assistance of the adaptor protein Claspin, activate their downstream serine-threonine kinases, checkpoint kinase 2 (CHK2) and CHK1, respectively, which, through suppression of PLK1 activity, relay the inhibitory signal to the mitotic entry network (Fig. 6) (7, 35, 347). The activity of ATM or ATR and CHK2 or CHK1 promotes PLK1 inactivation through several mechanisms, including the degradation of PLK1 mediated by APC-CCdh1, the anaphase-promoting complex (also known as the cyclosome or APC/C) in complex with CDC20 and CDC20 homolog 1 (CDH1), the dephosphorylation of PLK1 at pThr210, the SCFβTrCP-mediated degradation of Bora, and the inhibition of AurA activity (347355). The co-option of the PLK1 activity sensor by the DNA damage checkpoint is consistent with the fact that, like PLK1 and Bora, orthologs of several components of the PLK1 G2-M checkpoint module, such as CHK1, CHK2, and CDC25, a protein phosphatase that activates cyclin B–CDK1 by removing inhibitory phosphate groups (Fig. 6), are not encoded in the genomes of land plants. Therefore, ATM and ATR use downstream targets other than CHK2 and CHK1 in the G2-M DNA damage checkpoint in these organisms (344, 356, 357). Thus, the increasing role of PLK1 in mitosis and its involvement in the G2-M DNA damage checkpoint may have been the primary selective forces for the evolution of the Bora module, with its main role being to activate the bulk of PLK1 in the cytoplasm and to make it available to the target proteins before mitosis. In light of these considerations, the decision to enter mitosis may depend on the balance between the AurA- and Bora-mediated phosphorylation of PLK1 at Thr210 and the opposing effects of the DNA damage and replication stress responses. When a threshold concentration of active PLK1 is reached, PLK1 phosphorylates the proteins of the mitotic entry network, thus overriding the G2-M checkpoint and enabling mitotic entry. Specifically, PLK1 activity promotes SCFβTrCP-mediated degradation of Bora, Claspin, and the CDK1 inhibitory kinase WEE1; inhibits MYT1 (another inhibitory kinase of CDK1) and CHK2; and increases the activity of CDC25 (Fig. 6) [(132, 136, 358360); reviewed in (7, 29, 347)]. The antagonism between the Bora signaling module and the DNA damage and replication stress response in activating cytoplasmic PLK1 may explain the indispensability of the Bora module for mitotic entry in the presence of DNA damage or after the repair of damaged DNA (checkpoint adaptation or recovery, respectively) (11, 41, 361).

Fig. 6 The PLK1 module of the G2-M checkpoint.

The PLK1 activity sensor (blue shading) of the mitotic entry network may have been co-opted by the G2-M DNA damage checkpoint. Red arrows, protein phosphorylation events; black arrows, activating effects; black lines ending with a T-bar, inhibitory effects. Phosphate groups are depicted as solid black dots. Activating phosphorylation of PLK1 at Thr210 in the T-loop is shown as a black dot with a yellow center. In positive feedback loops, the cyclin B (Cyc B)–CDK1 complex cooperates with PLK1 to inactivate WEE1 and MYT1 and activate CDC25C (not shown).

Aurora-PLK1 signaling in the control of the DDR and DNA repair

Given the fundamental role of the Aurora kinases and PLK1 in the evolution of MTOCs, the involvement of these proteins in the DDR and DNA repair appears to be consistent with the emerging link between MT cytoskeleton and DNA repair. This link is supported by several lines of evidence, albeit the underlying mechanisms remain poorly understood. DNA double-strand breaks (DSBs) are deleterious lesions that are lethal if left unrepaired. Therefore, in response to DSBs, cells induce an evolutionarily conserved, elaborate DDR program that enables cell cycle checkpoint activation and DNA repair. This program involves sequential recruitment of proteins to DSBs and reversible posttranslational modification of histones and of the recruited proteins by phosphorylation, ubiquitylation, sumoylation, acetylation, methylation, and poly(adenosine 5′-diphosphate)-ribosylation (35, 362364). These events are underpinned by a signaling cascade initiated by the ATM-mediated phosphorylation of the histone 2A variant H2AX at Ser139. The H2AX phosphorylated at Ser139, which is called γ-H2AX, is recognized by the mediator of DNA damage checkpoint protein 1 (MDC1), which, following its phosphorylation by ATM, binds the forkhead-associated (FHA) domain of the E3 ubiquitin ligase RING finger protein 8 (RNF8) [reviewed in (35, 362)]. RNF8, in conjunction with the ubiquitin-activating enzyme (E1) and the heterodimeric ubiquitin-conjugating enzyme (E2) Ubc13, synthesizes Lys63 (K63)–linked ubiquitin chains on linker histone H1 (365). These chains are recognized by another E3 ubiquitin ligase, RNF168, which, in conjunction with E1 and Ubc13, ubiquitylates the core histone H2A and possibly other proteins. RNF8 and RNF168 lie at the apex of the ubiquitin-dependent signaling network involved in most aspects of the DDR, including the recruitment of multiple DSB repair and signaling factors, such as the p53-binding protein 1 (53BP1), Rad18, BRCA1, Rap80, and RIF1, to DNA lesions (362, 363, 365, 366).

Notably, a mammalian paralog of RNF8, checkpoint protein with forkhead and RING finger domains (CHFR), which is inactivated in many tumors, has been implicated in a checkpoint that delays mitotic entry in response to MT damage (367369). This function of CHFR depends on its autoubiquitylation and ubiquitylation of AurA and PLK1 (369372). Moreover, a homolog of RNF8 and CHFR in fission yeast, Dma1, is involved in a checkpoint that delays cytokinesis when chromosomes are not attached to the mitotic spindle. Dma1 prevents the initiation of cytokinesis through ubiquitylation of the scaffold protein Sid4 and the consequent inhibition of the recruitment of the yeast PLK1 homolog Plo1 to spindle pole bodies (373, 374). RNF8 and its yeast homologs Dma1 and Dma2 also inhibit mitotic exit in response to MT damage (375377). Furthermore, experiments in budding yeast revealed that Dma1 and Dma2 are targeted through their FHA domains to the site of cell division, where they promote cytokinesis through ubiquitylation of septins in an evolutionarily conserved manner (375, 378). Septins are GTP-binding proteins that are conserved from yeast to humans (but are notably absent in land plants) and involved in cytokinesis and other cellular processes through polymerization into hetero-oligomeric complexes and filaments (379, 380). On the basis of these observations, Chahwan and co-workers proposed that RNF8 originally evolved to control mitosis and cytokinesis and was only later co-opted to function in the DDR. In both processes, RNF8 seems to target higher-order oligomeric (more precisely, octameric) structures, such as septin filaments and nucleosomes. The authors go even further to suggest that the histone code might have evolved to mimic posttranslational modifications in the cytoskeleton rather than the other way around (378, 381).

The fundamental purpose of the DDR is the repair of DSBs by either of two pathways: The error-prone nonhomologous end-joining (NHEJ) pathway, which ligates breaks irrespective of sequence homology, and the error-free homologous recombination (HR) pathway, which requires a homologous DNA template (382). The choice between the NHEJ- and HR-mediated pathways for DSB repair and replication fork restart depends on the reciprocal balance between the key DNA repair factors 53BP1 and BRCA1. It is thought that in G1 phase, 53BP1 promotes NHEJ while inhibiting BRCA1-mediated HR, whereas in S and G2 phases, BRCA1 promotes HR while suppressing 53BP1-mediated NHEJ (366, 383387). PLK1 appears to play a role in both DSB repair pathways.

Because the preferred substrate for HR is the sister chromatid, HR-mediated error-free DSB repair can occur only in the late S or G2 phases of the cell cycle, the time during which PLK1 is activated in the Bora complex and the CEP192 complex. Consistent with this, PLK1 has been implicated in HR-mediated DSB repair by facilitating the recruitment of BRCA1 and the recombinase Rad51, as essential HR factor, to DSB sites and by enforcing the BRCA1-mediated replication fork restart mechanism (387389).

As cells enter mitosis, DNA repair must be inactivated to prevent potentially deleterious telomere fusions that can lead to aneuploidy (390). 53BP1 plays a major role in this inactivation, in the maintenance of chromosome stability, and in the integration of several aspects of mitosis with the cell cycle and DNA repair machineries. At least some of the 53BP1 mitotic functions are regulated by PLK1 and the Aurora kinases. Specifically, the phosphorylation of 53BP1 by PLK1 and CDK1 inhibits its recruitment to chromatin flanking the DSB, and this inhibition is essential for inactivating DNA repair during mitosis. Loss of this inactivation leads to aberrant DSB repair, causing AurB-dependent sister telomere fusions and aneuploidy (390). Moreover, the phosphorylation of 53BP1 by AurB is critical for optimal association of 53BP1 with kinetochores and efficient MCAK-dependent resolution of erroneous merotelic kinetochore-MT attachments that cause aneuploidy (391, 392). Furthermore, 53BP1 and the ubiquitin carboxyl-terminal hydrolase 28 (USP28) are the key components of the mitotic surveillance checkpoint mechanism that triggers p53-dependent cell cycle arrest or cell death in response to centrosome loss or prolonged mitosis (393395). It is yet unknown whether this mechanism requires the activities of the Aurora kinases or PLK1. It has also been shown that 53BP1 binds AurA and PLK1 in mitosis and that PLK1-53BP1 docking promotes DNA damage checkpoint inactivation (358, 396).

There is growing evidence that MTs may be involved in DNA repair by influencing chromatin dynamics through the linker of nucleoskeleton and cytoskeleton (LINC) complex composed of the Sad1 and UNC-84 (SUN) domain proteins and Klarsicht, ANC-1, and Syne homology (KASH) domain proteins located at the inner and outer nuclear membranes, respectively. The LINC complexes connect the nuclear lamina to the cytoplasmic MTs and also contribute to centrosome attachment to the outer nuclear membrane (397). In mammalian cells, LINC complexes and the intact MT cytoskeleton are required for the 53BP1-dependent increase in chromatin mobility upon induction of DSBs and for efficient NHEJ-mediated repair of damaged chromatin (398). In C. elegans, LINC complexes and the MT cytoskeleton are essential for efficient DNA damage repair and meiotic recombination, and they function with the Fanconi anemia (FA) pathway to inhibit NHEJ in favor of HR. On the basis of these data, it was proposed that LINC complexes have a conserved role in suppressing NHEJ and in promoting HR at DSBs (399). Consistent with this notion, LINC complexes have been shown to cooperate with the DNA repair MRN complex, which consists of meiotic recombination protein 11 (MRE11), the DNA repair protein RAD50, and Nijmegen breakage syndrome protein 1 (NBS1), and with the nuclear actin organizer Formin 2 to promote clustering of the DSB-damaged transcriptionally active genes in G1 phase. This clustering suppresses NHEJ while permitting HR-mediated DNA repair in S-G2 phases (400). Although the role of Aur-PLK1 signaling in the regulation of LINC complexes during DNA repair is yet unknown, the phosphorylation of a component of the LINC complex, SUN1, by PLK1 and CDK1 inhibits SUN1 interaction with the nuclear lamina, which is critical for nuclear envelope breakdown at mitotic entry (401).

A functional link between the MT cytoskeleton and DNA repair is also supported by the findings that several proteins, including BRCA1, the MRN-CtBP–interacting protein (CtIP) complex, ATR, CHK1, CHK2, TPX2, and the NEK kinases, have been implicated both in DDR or DNA repair and in spindle assembly and chromosome segregation and that numerous DNA repair proteins localize to centrosomes (17, 64, 402408). In addition, mutations in several centrosome and spindle pole proteins, including those involved in the DDR, and in ATR cause clinically related autosomal recessive microcephaly syndromes (409). Furthermore, ATR localizes to centromeres in a manner that depends on AurA activity and on the presence of R-loops—structures composed of DNA-RNA hybrids and displaced single-stranded DNA—and promotes AurB activation (presumably through CHK1), thereby ensuring faithful chromosome segregation (408, 410). Thus, there is growing evidence that the Aurora-PLK1 signaling modules and the key DDR proteins may work together in common mitotic regulatory networks.

Conclusions and outlook

In over two decades since their discovery (810), the Aurora family kinases and PLK1 have risen to the forefront of mitosis research and have emerged, along with CDK1, as master regulators of cell division. Research has identified key Aurora complexes and Aurora-PLK1 cascades that operate in conjunction with cyclin B–CDK1 and appear to represent the basic signaling modules in the regulation of mitosis. The studies have revealed a surprising ubiquity and versatility of the Aurora-PLK1 signaling owing to the unique subcellular targeting and scaffolding properties of the four core Aurora cofactors Bora, CEP192, TPX2, and INCENP. One of the important advantages that scaffold proteins confer to regulatory networks is the possibility of achieving new responses from preexisting signaling components (411). Therefore, the identification of several Aurora-PLK1 scaffolding proteins sheds light on the conundrum of how the same protein kinases can cooperatively control diverse cellular events to enable the complexity of mitosis and cytokinesis.

The application of specific inhibitors of the Aurora kinases and PLKs has revolutionized mitosis research (14, 28). The most commonly used inhibitors of the Aurora proteins and PLK1 block the catalytic activity by interacting with the adenosine 5′-triphosphate (ATP)–binding pockets, thus inactivating all pools of these kinases in cells (14). Because such inhibitors may interfere with all mitotic phases, as well as with mitotic entry and exit (Fig. 1), the mechanisms by which they suppress proliferation or induce the death of different cell populations may vary and may be difficult to predict. Although kinase inhibition is a powerful approach, the discovery of distinct Aurora and PLK1 signaling modules highlights the limitations of inhibiting all Aurora or PLK1 activity. Structure-guided design of pool-specific Aurora and PLK1 inhibitors, such as those targeting the kinase-scaffold protein interface, may yield powerful tools for probing mechanisms of mitosis and cytokinesis and may lead to the development of novel anti-mitotic drugs (412417).

Further progress in understanding the full complexity of Aurora and PLK1 signaling will require integration of biochemical, proteomic, single-molecule, imaging, and structural methodologies to elucidate the molecular function of each core Aurora cofactor and identify all components of Aurora-PLK1 cascades. Whereas the CPC and TPX2 have attracted much attention over the past two decades, studies focusing on Bora and CEP192 have been lagging, and our knowledge about how these proteins function at the molecular level remains quite rudimentary. The CEP192-organized signaling cascade appears to involve multiple steps, to require additional factors, and to drive the recruitment of numerous proteins (47). Given the central role of CEP192 in centrosome evolution and biogenesis, a comprehensive functional analysis of this protein will allow us to decipher how the centrosome nucleates MTs and to gain invaluable insights into other functions of this still mysterious organelle. It seems particularly important to decipher the link between centrosomes and cell cycle regulation, DNA repair, proteostasis, and innate immune responses (402, 409, 418421).

The functional analogy between the CEP192 complex and the CPC can be exploited to design similar experimental strategies to study the mechanisms by which these complexes promote protein recruitment and the assembly of mitotic centrosomes and kinetochores, respectively. Such strategies will further our understanding of the functional and evolutionary link between centrosomes, kinetochores, and the actomyosin cleavage apparatus. Full advantage should be taken of the already established techniques of reconstitution of the mitotic centrosome and kinetochore assembly on artificial templates in Xenopus egg extracts (40, 47, 149, 162, 163, 422, 423). The application of this experimental approach will make it possible to decipher the signaling cascades organized by CEP192 and INCENP and lay the groundwork for reconstitution of the mitotic centrosome and kinetochore assembly in vitro from purified components.

Because the core AurA cofactors Bora, CEP192, and TPX2 appear to compete with one another for binding to the kinase, they may operate interdependently in the same regulatory network, which may contribute to the previously described reciprocal relationship between the spindle assembly pathways (19, 132, 322, 424, 425). Hence, it is critical to delineate the mechanisms that control the synthesis and stability of these core AurA cofactors and their cross-talk with other regulators, such as CDK1, the NEK kinases, protein phosphatases, and E3 ubiquitin ligases.

No matter how much we have learned about mitosis, the sheer complexity of this process and its diversity across eukaryotes suggest that we have barely scratched the surface in our understanding of the underlying mechanisms. A myriad of questions regarding the Aurora kinases, PLKs, and their regulators remain unanswered, indicating that we are only at the beginning of a challenging research journey and that many exciting discoveries lie ahead.