Introduction

Faithful chromosome segregation is essential for self-replication in living organisms. Failure of these processes causes chromosomal instability, resulting in aneuploidy and subsequent genetic disorders. Accurate chromosome segregation in eukaryotic cells is achieved by precise bipolar attachment of the spindle microtubules to each sister chromatid. The microtubules-sister chromatids linkage is enabled by the kinetochore, a large protein complex assembled on the specific genomic region in chromosomes named centromere [1,2,3]. The kinetochore also provides a binding platform for spindle assembly checkpoint (SAC) proteins to monitor and correct erroneous microtubule attachment to the kinetochore [1, 4,5,6]. Given that the kinetochore plays these crucial roles, deciphering the mechanism underlying assembly of the complex is significant for understanding the process behind precise chromosome segregation.

The kinetochore consists of more than a hundred proteins in vertebrates [1, 5, 7, 8]. Studies using electron microscopy (EM) or fluorescence microscopy have indicated that morphology of the outer kinetochore structure changes dynamically during mitotic progression [9,10,11,12,13]. The changes occur upon microtubules attachment to the kinetochore, which alters localization of multiple kinetochore proteins, including SAC proteins. The pulling force produced from microtubules to the kinetochore distorts the shapes of centromeric chromatin and inner kinetochore. It is suggested that the shape changes are important for the stable bipolar microtubule attachment and for SAC signal silencing [4, 14,15,16].

The microtubules and centromeric chromatin are connected by two major protein interaction networks in the kinetochore: constitutive centromere-associated network (CCAN) and the Knl1 complex–Mis12 complex–Ndc80 complex network (KMN) [2, 8, 17, 18]. CCAN is a 16-subunit complex in vertebrates and is associated with centromeric chromatin throughout the cell cycle. KMN is recruited onto CCAN in M-phase. As the Ndc80 complex (Ndc80C) of KMN directly binds to the microtubules together with other microtubule-binding proteins [19], the CCAN–KMN interaction forms a key step in building the microtubule–kinetochore interaction surface. We recently observed dynamic interaction changes between centromere, CCAN and KMN during cell cycle progression [20, 21]. In this review, we discuss the changes in the kinetochore morphology and the dynamics of the protein interaction network that provides plasticity to the kinetochore during mitotic progression.

The centromere and kinetochore

The centromere is a specific genomic region in the chromosome recognizable by a primary constriction [3]. In most species, except for in budding yeast, the centromere position is specified by sequence-independent epigenetic mechanisms [3, 22]. A key epigenetic marker for centromere specification is the centromere protein (CENP)-A, a centromeric histone H3 variant. CENP-A assembles a centromere-specific nucleosome with histone H4 and H2A/B [3, 22], forming the centromeric chromatin with interspersed H3-containing nucleosomes (Fig. 1).

Fig. 1
figure 1

The kinetochore in M-phase. a Mitotic chromosome segregation and a trilaminar structural model for the kinetochore. The kinetochore is a large protein complex built on the centromere of each sister chromatid (chromosome). The kinetochores attach to microtubules emanating from each opposing spindle pole, segregating chromatids. Electron microscopy studies revealed that the kinetochore has three layers in its structure: electron-dense inner and outer plates, and an electron-translucent middle layer. b A model of basic kinetochore structure. The main structure of the kinetochore is formed of constitutive centromere-associated network (CCAN) composed of 16 protein subunits and the KMN (Knl1, Mis12, and Ndc80 complexes: Knl1C, Mis12C, Ndc80C) network (KMN). CCAN interacts with the centromeric chromatin epigenetically marked with the CENP-A nucleosome during the entire cell cycle. CCAN recruits KMN, which directly binds to microtubules, in M-phase, forming a linkage between the centromere and microtubules

Full size image

A large number of proteins are assembled onto the centromere, building the kinetochore that provides a bridge between centromeric chromatin and microtubules during mitosis. EM studies have revealed that the vertebrate kinetochore forms a trilaminar structure on the centromere, with electron-dense inner and outer plates, and an electron-translucent middle layer (Fig. 1a) [11, 23,24,25,26,27].

Although the composition of molecular substances in each layer needs to be investigated further, CENP-A and CCAN are known to be components of the inner plate [28]. CCAN binds to the CENP-A-containing centromeric chromatin throughout the cell cycle [8, 17, 22]. In the M-phase, CCAN recruits KMN [18, 29,30,31]. An EM study demonstrated that Ndc80C of KMN is a component of the outer plate [32]. Owing to its microtubule-binding ability, Ndc80C builds a microtubule–kinetochore interaction surface for a stable end-on attachment together with other microtubule-binding proteins (Fig. 1B) [29, 33,34,35].

Dynamic morphological changes in the kinetochore during M-phase

Kinetochore expansion

In addition to the trilaminar structure, EM studies have also depicted that an expanded fibrous structure—known as the fibrous corona—radiates from the distal surface of the outer plate in the kinetochore without attached microtubules (Fig. 2a) [10, 24, 26, 36,37,38,39,40]. This prominent fibrous structure extending over 100 nm to the cytoplasm is diminished in the kinetochores that attach to microtubules, suggesting that the kinetochore morphology alters dynamically upon kinetochore–microtubule attachment during M-phase [10, 41]. The corona includes proteins such as motor proteins (CENP-E, dynein), Rod–Zw10–Zwilch complex (RZZ), SAC proteins, and microtubule-associating proteins, and these components demonstrate equivalent dynamic morphological changes in light microscopic studies [12, 39, 40, 42,43,44,45,46,47]. In cells treated with drugs inhibiting microtubules polymerization, the corona forms a large crescent-like structure on the centromere or expands further, resulting in a ring surrounding the sister centromeres [12, 43, 45,46,47].

Fig. 2
figure 2

Kinetochore expansion and compaction during M-phase. a Fibrous corona. An expanded structure, called “fibrous corona” is formed on the outer kinetochore at unattached kinetochores. Once microtubules attach to the kinetochore, most of the components of the expansion are removed in a dynein-dependent manner. b Rod–Zw10–Zwilch complex (RZZ). RZZ forms a dimer with the antiparallel arrangement of Rod. Spindly and Mps1 kinase activity promote oligomerization of RZZ. Farnesylation of Spindly C terminus is crucial for the RZZ oligomerization by suppression of the inhibitory effect of Spindly N terminus

Full size image

The expanded crescent is also observed during early prophase in unperturbed cells and then becomes compact after microtubule end-on attachment by removal of the corona proteins dependent on dynein (Fig. 2a) [12, 46, 48,49,50,51]. These observations indicate that the outer kinetochore morphology alters dynamically during mitotic progression. The expansion and compaction of the outer kinetochore play important roles in robust spindle assembly [12, 13, 45]. The expanded shape in unattached kinetochores facilitates lateral microtubule attachment that later promotes efficient end-on attachment. Further, the microtubule attachment induces compaction of the outer kinetochore; therefore, the structural change might reduce the risk of binding to microtubules from the opposite pole. Disruption of compaction leads to erroneous microtubule attachment, such as merotelic attachment, as well as chromosome segregation error [47].

RZZ and Spindly oligomerize to form expanded kinetochore

What makes the expanded fibrous structures had been a long-standing question. Recent studies have demonstrated that oligomerization of the RZZ complex and Spindly protein contributes to the formation of the meshwork for expanded outer kinetochore [43, 44, 47].

The RZZ complex consists of three subunits that were originally identified in Drosophila: Rod (Rough Deal), Zw10 (Zest-White 10), and Zwilch, and is conserved in metazoans (Fig. 2b) [52,53,54,55,56,57,58,59]. Based on size exclusion column chromatography analyses, it was suggested that the RZZ complex contains two copies of each subunits [52, 59]. Indeed, structural study on human RZZ demonstrated the 2:2:2 stoichiometric RZZ complex [60]. In the complex, two molecules of Rod, the largest subunit, interact in an antiparallel configuration with beta-propellers in its N terminus at opposite ends (Fig. 2b) [60]. Spindly, which was originally discovered in Drosophila as a dynein adaptor detected on the kinetochore in mitosis, has functions in chromosome alignment and SAC signaling [61,62,63,64]. Human Spindly is farnesylated at its C terminus that is essential for its kinetochore interaction and binding with RZZ via the beta-propellers of Rod [60, 65, 66].

Both Spindly and RZZ are proteins that reside on the expanded crescents [43, 44, 47, 56, 64]. RZZ localization to the kinetochore depends on Knl1 and Bub1, a KMN subunit and SAC protein, respectively, followed by recruitment of Spindly [60, 67,68,69,70,71]. Depletion of the Spindly or RZZ components introduces defects in SAC signaling and also prevents formation of the expanded crescent structure on outer kinetochores in nocodazole-treated cells [43, 44, 47], suggesting that these proteins are responsible for kinetochore expansion.

Given that RZZ dimer structure is similar to that of the membrane coat protein complex, which self-assembles to forms oligomers, RZZ has been suggested to be oligomerized [60]. However, purified recombinant human RZZ components did not form oligomers, in vitro [47, 60]. However, the RZZ complex in a mixture with farnesyled Spindly oligomerizes into filamentous structures [47], suggesting that Spindly interaction with RZZ is crucial for RZZ oligomerization (Fig. 2b). Farnesylation is necessary for kinetochore expansion [47]. Furthermore, Sacristan et al. [47] demonstrated that an auto-inhibitory role of the N-terminal region of Spindly in RZZ oligomerization (Fig. 2b). The Spindly N-terminal region is arranged in an intramolecular inhibitory conformation to prevent RZZ-Spindly oligomerization [47]. During mitosis, Spindly has to be released from the inhibitory configuration to form the expanded crescent. One of the key factors that regulate the process is a mitotic kinase, Mps1 that is necessary for outer kinetochore expansion. Spindly ∆N induces the kinetochore expansion under an Mps1-inhibited condition (Fig. 2b) [43, 47]. Although the molecular mechanisms underlying the prevention of Spindly autoinhibition enabled by Mps1 are yet to be studied, Rod is one of the Mps1 substrates observed to be involved in the expansion [43]. Rod has two Mps1 phosphorylation sites—Thr13 and Ser15 of human Rod [43]. While Rod with Alanine substitution in these sites forms the RZZ complex, localizes to the kinetochore, and recruits Spindly and other crescent proteins in human cells, the outer kinetochores do not undergo expansion in nocodazole-treated cells expressing the Rod phospho-dead mutant [43].

These findings indicate that post-translational modifications of the RZZ complex and Spindly play key roles in the kinetochore expansion. In addition to Spindly farnesylation and Mps1-mediated phosphorylation for the RZZ complex, Cdk1 activity is also involved in the expansion by tethering of RZZ onto outer kinetochore [44, 47].

As suggested by light and electron microscopy studies, the expanded outer kinetochores composed of Spindly-RZZ oligomers facilitate lateral attachment of microtubules to establish robust kinetochore–microtubule attachment [38, 47]. This is consistent with a prediction by a mathematical model [45]. The expanded Spindly-RZZ structure is also engaged in SAC signal maintenance, interacting with the Mad1–Mad2 complex [43, 68, 70, 72,73,74,75]. After end-on attachment is established, the expanded crescent structure is released from the kinetochores by poleward transport of dynein [12, 48,49,50,51]. Spindly has dynein–dynactin binding domains in its N-terminal region [5, 64, 66, 71]. Deletion or mutations in the domains prevent dynein recruitment to the kinetochores, even though the expanded crescent structures still form at the kinetochores, indicating that dynein is not necessary for outer kinetochore expansion [47, 64, 66, 71]. However, the expanded outer kinetochores, with the Spindly mutants that do not bind to dynein, are maintained even in metaphase. The cells with the persistent expanded crescent structures cause microtubule malattachments and subsequent chromosome mis-segregation, indicating that temporal morphological dynamics of the outer kinetochore in conjunction with mitotic progression is essential for correct microtubule–kinetochore attachment, ensuring faithful chromosome segregation [45, 47, 64].

Kinetochore stretches

Inter- and intra-kinetochore stretch

Once microtubules emanating from opposing spindle poles capture the kinetochores on sister chromatids, the sister kinetochore distance is increased [76]. This structural change is called centromere stretch or inter-kinetochore stretch (Fig. 3). The change in inter-kinetochore distance indicates tension applied on the kinetochores by microtubule-pulling force, as treatment of cells with Taxol, which stabilizes microtubules and inhibits microtubule dynamics, prevents the inter-kinetochore stretch even with kinetochore-attached microtubules [76, 77].

Fig. 3
figure 3

Kinetochore stretches. When the spindle microtubules attach to sister kinetochores, the distance between inner kinetochore proteins of a sister kinetochore is stretched out (centromere stretch or inter-kinetochore stretch: dinnner). The kinetochore itself is stretched by microtubule attachment (intra-kinetochore stretch). The intra-kinetochore distance is defined by half of the length difference between dinnner and douter (distance between the outer kinetochore proteins of a sister kinetochore)

Full size image

In addition to the increase in the distance between the sister kinetochores, the kinetochore itself stretches in length between inner and outer kinetochore proteins upon microtubule attachment (intra-kinetochore stretch) (Fig. 3) [77, 78]. The intra-kinetochore distance is measured by colocalization analysis of two-color fluorescence proteins (outer and inner kinetochore proteins labeled with different colors) in living and fixed cells (Fig. 3) [77, 78]. To calculate intra-kinetochore distance, the distance between inner or outer kinetochore proteins in the sister kinetochores are measured, respectively. The distance between the inner proteins is subtracted from that between the outer proteins, and half of this value is defined as the intra-kinetochore distance (Fig. 3). This method allows correction of errors due to chromatic aberration [77,78,79,80].

The bipolar attachment of microtubules causes both inter- and intra-kinetochore deformation. However, they do not appear to be a simple two-linked Hookean spring, as the weak treatment of cells with nocodazole, which inhibits microtubule polymerization, allows inter-kinetochore stretch, while it does not induce intra-kinetochore stretch [78]. Moreover, weak Taxol treatment leads to intra-kinetochore stretch, while it prevents inter-kinetochore stretch [77]. These findings suggest that the microtubule-pulling force might not simply induce the intra-kinetochore stretch, and there might be other mechanisms involved in the intra-kinetochore stretch, such as conformational changes in the kinetochore.

Kinetochore stretch and error correction

The deformation of kinetochore is proposed to regulate error correction of microtubule malattachments [14, 15]. Aurora B kinase (AurB) is an important kinase controlling the error correction machinery [29, 34, 81,82,83,84,85]. AurB is a catalytic subunit of chromosome passenger complex (CPC) along with other subunits (INCENP, Borealin and Survivin) [86,87,88,89,90,91,92,93,94,95]. CPC is involved in various regulatory processes in mitotic progression and correlated with its function, its localization alters dynamically during mitotic progression from chromosome arms to spindle midzone via centromeres.

CPC localizes to centromeres between prometaphase and metaphase and plays a key role in the error correction. The kinase activity of AurB in CPC forms a gradient centered at the inner centromere (Fig. 4) [14, 96,97,98,99,100]. In the absence of microtubule-binding or tension due to malattachment of microtubule–kinetochore, AurB substrates in the kinetochore are placed within or near to the AurB kinase gradient, leading to their phosphorylation. Among those substrates, KMN components, which provides a microtubule-binding surface of the kinetochore, are major AurB substrates in the outer kinetochore [100]. Ndc80 (also called Hec1 in vertebrates) is a critical substrate of AurB for error correction in KMN (Fig. 4). Ndc80 directly binds to microtubules through a globular domain in its N terminus [29, 33,34,35]. AurB phosphorylates the Ndc80 N-terminal tail preceding the globular domain, and the phosphorylated Ndc80 does not bind properly to microtubules [29, 34, 100,101,102]. When kinetochores attach incorrectly to microtubules, AurB phosphorylates Ndc80 and destabilizes the Ndc80–microtubule interaction [29, 33, 34, 84, 103,104,105]. Once the precise bi-polar attachment is established, AurB substrates in the outer kinetochore are physically separated from the AurB activity gradient and are dephosphorylated by protein phosphatase counteracting AurB, such as protein phosphatase 1 (PP1) (Fig. 4) [102, 105,106,107,108,109,110]. The dephosphorylated Ndc80 stabilizes the microtubule–kinetochore attachment to enable proper chromosome segregation (Fig. 4) [29, 34, 100,101,102].

Fig. 4
figure 4

Microtubule attachment and spindle assembly checkpoint. Aurora B kinase (AurB) localizes in inner centromere and forms a gradient of kinase activity centered at the centromere. AurB phosphorylates proteins in the kinetochore. Ndc80 is an AurB substrate that resides in the outer kinetochore. Without the correct bipolar microtubule attachment, Ndc80 remains in the AurB activity gradient and is phosphorylated, leading to unstable attachment to the microtubules. The unattached and malattached kinetochores activate spindle assembly checkpoint (SAC) signals. Correct and stable microtubule–kinetochore attachment silences SAC signals and further stabilizes the attachment by pulling out Ndc80 from the AurB kinase active zone through the intra-kinetochore stretch. Although the kinetochore stretch enables stable microtubule attachment, the stretch itself is not necessary; however, a stable microtubule attachment is sufficient for SAC silencing

Full size image

Distance of kinetochore substrates from the AurB active zone is a significant factor for their dephosphorylation. However, the removal of tension after dephosphorylation of substrates does not result in rephosphorylation [101], suggesting that additional mechanisms are necessary to sustain dephosphorylation.

SAC silencing and microtubule attachment

The bipolar attachment induces the kinetochore stretches, which further stabilize the microtubule–kinetochore attachment (Fig. 3). Since the inter-kinetochore stretch cab easily be observed under a light microscope, it is often used as an indicator of the bioriented stable microtubule–kinetochore attachment that is necessary for SAC silencing in anaphase onset. However, it was observed through detailed live-cell imaging analyses that the inter-kinetochore stretch itself is not necessary for SAC silencing; instead, the intra-kinetochore stretch is necessary for it [77, 78]. As mentioned above, it has been proposed that the inter- and intra-kinetochores are not two-linked Hookean springs. Inter- and intra-kinetochore stretches can be separated using microtubule-targeting drugs [77, 78]. The mitotic duration is much longer when the inter-, but not the intra-kinetochore stretch occurs, compared to that in an unperturbed control condition, suggesting that SAC silencing is ineffective without the intra-kinetochore stretch [78]. In contrast, during only intra-kinetochore stretch, without an increase in the inter-kinetochore distance in mitotic cells, the cells progress enter into anaphase with the same mitotic duration as the control cells, indicating that the SAC signal is silenced without the inter-kinetochore stretch [77]. These observations indicate that the intra-kinetochore stretch without inner-kinetochore stretch is sufficient for SAC inactivation.

Provided that the SAC monitors the bipolar microtubule–kinetochore attachment, SAC machinery has to sense some changes in the kinetochore upon the bipolar microtubule–kinetochore attachment for SAC silencing. As the intra-kinetochore stretch is involved in SAC silencing, the change in intra-kinetochore distance might directly trigger SAC silencing. However, this does not appear to be the case. The bipolar microtubule–kinetochore attachment stretches the kinetochore by microtubule-pulling force, separating outer kinetochore proteins form the centromere. This facilitates dephosphorylation of AurB substrates in the outer kinetochore. As mentioned above, dephosphorylated Ndc80 stabilizes microtubule–kinetochore attachment [29, 33, 34]. When an Ndc80 mutant, with all potential AurB phosphorylation sites substituted with alanine, was expressed in human cells, the cells silenced the SAC and exit from metaphase in the absence of bi-oriented sister and microtubule-pulling force [111, 112], suggesting that the intra-kinetochore stretch might not play a direct role in SAC silencing; yet, the stable microtubule–kinetochore interaction is sufficient for it (Fig. 4). Although the intra-kinetochore stretch is not a direct cue for SAC inactivation, this stretch is an event occurring upstream of SAC silencing, since stable microtubule–kinetochore attachment requires Ndc80 dephosphorylation induced by the intra-kinetochore stretch (Fig. 4). It is important to clarify next how the stable microtubule–kinetochore attachment induces changes in the kinetochore structure and how these changes lead to SAC silencing.

Dynamic interaction between CCAN and centromere

CCAN and centromere interaction interface

The kinetochore bridges centromeric chromatin with microtubules. One of the kinetochore components associated with the centromeric chromatin is CCAN, which is a 16-protein complex and constitutively localizes to the centromere during cell cycle progression (Fig. 1b) [8, 17]. It has been demonstrated that CCAN has three interfaces for the centromeric chromatin binding: CENP-T, CENP-N and CENP-C (Fig. 5). CENP-T has a histone-hold domain in its C terminus and forms a complex with three other histone-hold proteins (CENP-W, -S, and -X) to assemble a nucleosome-like structure (Fig. 5). This complex directly binds to centromeric DNA. However, the mechanism underlying binding of CENP-T complex to the centromeric DNA in cells remains to be determined, though DNA rounds up the nucleosome-like CENP-T/W/S/X complex in vitro, and it is suggested that CENP-T associates with H3 nucleosomes on centromeres [113, 114]. CENP-N and CENP-C bind to the CENP-A nucleosome directly in vitro, and it is proposed that these factors recognize centromere chromatin (Fig. 5) [115,116,117,118].

Fig. 5
figure 5

CCAN-centromere interaction interfaces. CENP-T interacts with centromere chromatin containing histone H3 nucleosomes. CENP-N and CENP-C bind to the CENP-A nucleosome. CENP-N interacts with CENP-A L1 loop (also known as RG loop because of the presence of conserved Arg and Gly residues). CENP-C binds to the CENP-A C-terminal tail and the histone H2A/B acidic patch in the CENP-A nucleosome

Full size image

Recent studies using Cryo-EM have determined the structures of the human CENP-A nucleosome with CENP-N [119,120,121,122]. The CENP-N N-terminal region binds to the CENP-A nucleosome surface as well as to the DNA. CENP-N interacts with CENP-A via L1-loop (also known as RG loop, because of the presence of conserved Arg and Gly residues) of CENP-A that is unique to CENP-A, though it is absent from the canonical histone H3 [123]. The RG loop interaction of CENP-N is necessary for CENP-N recruitment to the centromere in cells [120] (Fig. 5). In addition, basic residues in the N-terminal region of CENP-N interact with nucleosomal DNA and appear to stabilizes the DNA segments at the entrance and exit of the CENP-A nucleosome that have been shown to be flexible [123].

CENP-C forms another contact surface with the CENP-A nucleosome. CENP-C has the evolutionarily conserved motif called “CENP-C motif” in its C-terminal region (Fig. 5) [115, 118]. Human CENP-C has another contact sequence in the central region called “central domain”, which shows sequence similarity to the CENP-C motif [115, 118, 124]. These sequences directly bind to the CENP-A nucleosome [21, 115, 118, 124]. Structural analysis using a short peptide, including the CENP-C motif, revealed that CENP-C motif binds to the C-terminal tail of CENP-A and the acidic patch of histone H2A/2B in the CENP-A nucleosome [118].

Cell cycle-dependent interaction changes between CCAN-CENP-A interaction

Although the CENP-C motif stably binds the CENP-A nucleosome in vitro, recent studies have revealed that the interaction is dynamic during cell cycle progression in cells [21, 125]. The CENP-C C-terminal region, including the CENP-C motif, interacts with centromeric chromatin containing the CENP-A nucleosome only in M-phase, and not in interphase in human RPE-1 and chicken DT40 cells (Fig. 6) [21, 125]. This mitotic interaction requires phosphorylation at a CDK1 consensus site (Thr 651 and Thr734 for chicken and human CENP-C, respectively) adjacent to and upstream of the CENP-C motif in both human and chicken CENP-C (Fig. 6) [21]. The site is directly phosphorylated by CDK1 and the phosphorylation positively regulates the interaction between the CENP-C motif and CENP-A nucleosome in vitro [21]. The CDK1-mediated phospho-regulation of CENP-C could be significant for proper kinetochore assembly in M-phase [21]. Since the phosphorylation site does not include the CENP-C motif peptide used in current structural studies and modelling [118, 122], use of the longer CENP-C C-terminal region including the phospho-residue is crucial in future structural analyses for better understanding of the molecular basis of the CENP-C–CENP-A nucleosome interaction.

Fig. 6
figure 6

Cell cycle-dependent switching of interaction between CENP-A and CENP-C. As a CCAN component, CENP-C constitutively localizes to the centromere throughout the cell cycle. During interphase, the CENP-C localization relies on the CENP-H complex (CENP-H, -I, -K, and -M) and the CENP-L-N complex. During M-phase, in addition to the contacts, CENP-C binds directly to the CENP-A nucleosome. The M-phase interaction requires threonine phosphorylation by CDK1 (Thr651 in chicken CENP-C and Thr734 in human CENP-C)

Full size image

Localization hierarchy of CENP-C to centromeres changes between interphase and M-phase [126, 127]. In interphase, centromere localization of CENP-C requires the CENP-H complex, as centromeric CENP-C signals are not detected in CENP-H- or CENP-K (a member of the CENP-H complex)-depleted interphase cells. However, CENP-C localizes to mitotic kinetochore in the CENP-H- or CENP-K-depleted cells, suggesting that CENP-C directly binds to centromere during M-phase, while it localizes to centromeres via the CENP-H complex in interphase cells. This change is well explained by the CDK1-mediated CENP-C interaction with the CENP-A nucleosome during M-phase.

It has also been suggested that CENP-N interaction to the centromere (CENP-A nucleosome) is dynamic during cell cycle progression. CENP-N is loaded onto the centromere and its levels are increased during S-phase [128, 129]. The CENP-N levels at the centromeres peak prior to G2/M-phase and decrease in M-phase [122, 129]. Whereas structural change in centromere chromatin is one phenomenon that controls the CENP-N levels, regulation of CENP-N dynamics during cell cycle progression remains a matter of intrigue. Additionally, it is important to address the arrangement of CENP-C and CENP-N on the CENP-A nucleosome. Since CENP-C and CENP-N recognize different surfaces of the CENP-A nucleosome, current structural models suggest that CENP-N and CENP-C bind to a CENP-A nucleosome simultaneously [119, 120]. This idea is supported by a recent study that determined the structure of the CENP-A nucleosome bound with CENP-C and CENP-N using Cryo-EM [122]. However, since these studies used a short peptide for CENP-C, there is large scope for future efforts to understand how CENP-C and CENP-N are arranged on a CENP-A nucleosome in centromeric chromatin. Building a CENP-A nucleosome bound to a larger CENP-C fragment including the CENP-A-binding region and the post-translational modification is necessary to clarify how CCAN recognizes the CENP-A nucleosome.

Two CCAN pathways to recruit KMN onto the kinetochore: changes in KMN–CCAN interaction during M-phase progression

In contrast to CCAN, which constitutively associates with the centromere throughout the cell cycle, KMN starts localizing to the centromere at the onset of G2/M-phase transition and remains localized to the centromere during M-phase by binding to CCAN [2, 18, 130]. Since Ndc80C in KMN binds to microtubules along with other microtubule-binding proteins [19], KMN–CCAN interaction establishes a linkage between microtubules and the centromere.

Two CCAN components, CENP-C and CENP-T, provide KMN recruiting platforms in CCAN (Figs. 1, 7). As both CENP-C and CENP-T directly bind to centromeric chromatin [113, 118, 131, 132], a two-pathway model (CENP-C- and CENP-T-pathways) is proposed for bridging of centromeres and microtubules during mitosis (Fig. 7a) [1, 133, 134].

Fig. 7
figure 7

Two-pathway links between centromere chromatin and spindle microtubules. a Two-pathway: CENP-C and CENP-T pathways. CENP-C has a Mis12 complex-binding (Mis12C-binding) region in its N terminus and two CENP-A nucleosome-binding domains in its middle and C-terminal regions (central domain and CENP-C motif, respectively). CENP-T binds to the Ndc80 complex (Ndc80C) and Mis12C via its N-terminal region. The CENP-T C terminus has a histone fold domain, which forms a nucleosome-like structure with CENP-W, -S, and -X, binding to centromere chromatin. Both CENP-C and CENP-T recruit to Ndc80C, which binds to spindle microtubules directly or indirectly through Mis12C, and also interacts with the centromere chromatin, forming bridges between the centromere and the microtubules. b Dynamic interaction of the KMN network with the two pathways. Two CCAN components, CENP-C and CENP-T, interact with and recruit the KMN network to the centromere in M-phase. The multiple phosphoregulations by CDK1 and AurB induce changes in the interaction between CCAN and the KMN network during M-phase progression, making the CENP-T pathway the major load-bearing KMN scaffold

Full size image

CENP-C is conserved and essential for chromosome segregation in most model organisms [135,136,137]. In addition to the centromeric chromatin interacting motifs, CENP-C also has the Mis12C-binding domain in its N-terminal end [138,139,140,141], which directly binds to Mis12C in vitro and further recruits Ndc80C on to CENP-C [140, 141]. The CENP-C–Mis12C interaction is regulated by AurB-mediated phosphorylation of Mis12C [100, 141, 142] (Fig. 7B). Structural analyses of the Mis12C–CENP-C complex demonstrated that a Mis12C subunit, Dsn1, has a conserved basic domain which plays an auto-inhibitory role in the Mis12C–CENP-C interaction by masking the CENP-C-binding surface from Mis12C [141, 143]. AurB phosphorylates the basic domain and prevents autoinhibition of Mis12C–CENP-C interaction, leading to their stable binding and Mis12C localization to the centromere during mitosis [100, 141, 142].

CENP-T forms another platform to recruit KMN onto CCAN. In contrast to CENP-C, which binds to Mis12C to further recruit Ndc80C, the CENP-T N-terminal region directly interacts with Ndc80C (Fig. 7a) [131, 138, 144, 145]. Human CENP-T has two Ndc80C-binding sites, whereas chicken CENP-T and the budding yeast CENP-T homologue Cnn1 have one site [131, 145, 146]. In addition, the CENP-T N-terminal region directly binds to Mis12C, which recruits another Ndc80C (Fig. 7a) [134, 138, 144, 145]. Phosphorylation also plays a key role in the regulation of the CENP-T–KMN interaction. The Ndc80C-binding region of CENP-T interacts with the Spc24–Spc25 sub-complex in Ndc80C (Fig. 7b) [131]. This interaction is stabilized by phosphorylation at Cdk1 consensus motifs in the CENP-T N-terminal region; human CENP-T has two Cdk1 consensus motifs, whereas chicken CENP-T has one motif [131, 144, 145]. Phosphorylation at these sites is necessary for Ndc80C recruitment to the CENP-T-pathway in mitotic human cells [144, 145]. The CENP-T–Mis12C interaction is also regulated by Cdk1 [144, 145]. Cdk1 phosphorylates Thr195 and Ser201 in human CENP-T and facilitates CENP-T–Mis12C interaction [144, 145]. Besides the phosphoregulation, the CENP-T–Mis12C interaction requires the direct binding of Ndc80C to the CENP-T N-terminal region; deletion of the Ndc80-binding region of CENP-T or knock-down of Ndc80C components hinders the CENP-T–Mis12C interaction [134, 144]. Further studies are necessary to decipher the multi-layer regulation of the CENP-T–Mis12C interaction.

Evidences of the direct binding of KMN to CENP-C and CENP-T N-terminal regions are provided by the artificial tethering of either the CENP-C or CENP-T N-terminal fragment to a non-centromeric locus in human and chicken cells [134, 138, 144]. The artificially tethered CENP-C or CENP-T N-terminal fragment recruits KMN on a non-centromeric locus. Notably, neither the CENP-C nor the CENP-T N-terminal fragments recruit other CCAN components at the tethering locus. Even in the absence of other CCAN components, KMN recruited onto the non-centromeric site is sufficient for forming a functional kinetochore in chicken DT40 cells, which is able to ensure normal chromosome segregation after removal of the native centromere on chicken chromosome Z [134].

The native kinetochores have both CENP-C and CENP-T (Fig. 1). Given the stable interactions between KMN and CENP-C or CENP-T in vitro, they seem to firmly bind to KMN to form two stringent pathways in M-phase. However, in fact, our recent studies using chicken DT40 cells revealed that the KMN–CCAN interaction is dynamic during M-phase progression (Fig. 7b) [20].

Centromere localization of Mis12C is observed in a subset of interphase cells and all mitotic cells [130]. Our recent studies on Mis12C localization analyses using various DT40 mutant cell lines have revealed how Mis12C is recruited onto two different platforms in CCAN [20]. In late G2 phase and prophase, the Mis12C localization depends on CENP-C, and does not depend on CENP-T [20, 142]. However, in the subsequent prometaphase, Mis12C starts binding to CENP-T in addition to CENP-C, resulting in an increase in levels of Mis12C on centromeres [20]. Thereafter, the levels of CENP-T-binding Mis12C are constant in metaphase and anaphase, while those of CENP-C-binding Mis12C reduce [20]. On the other hand, the Ndc80C levels are undetectable in prophase and peak in prometaphase [20]. Notably, centromeric localization of a majority of Ndc80C molecules depend significantly on CENP-T rather than on CENP-C. Ndc80C remains attached to CENP-T from metaphase to anaphase, though its levels are reduced [20]. These dynamic changes in KMN–CCAN interaction results in binding of the majority of KMN to CENP-T in metaphase and anaphase (Fig. 7b), indicating that CENP-T is a major KMN-binding scaffold in the native kinetochores in chicken DT40 cells. Consistent with this observation, the KMN-binding of CENP-C is dispensable, while that of CENP-T is essential for DT40 cell viability [20].

These dynamic changes in KMN–CCAN interaction during mitosis are regulated by phosphorylation and dephosphorylation of KMN and CCAN components. AurB-mediated phosphorylation of the Dsn1 basic motif facilitates the CENP-C–Mis12C interaction, leading to CENP-C-dependent Mis12C localization in G2 phase and prophase, while it decreases in late mitosis [20, 100, 141, 142]. In correlation with high Cdk1 activity in prometaphase and metaphase, the CENP-T N-terminal region must be highly phosphorylated to induce the direct binding of Ndc80C and Mis12C to CENP-T. We have showed that Cdk1 also phosphorylates the Dsn1 C-terminal region, which binds to Ndc80C. However, unlike CENP-T phosphorylation, Dsn1 phosphorylation reduces the Ndc80C–Mis12C interaction [20]. These phosphorylation profiles account for the low levels of Ndc80C on CENP-C even when Mis12C is present (Fig. 7b) [100, 141, 142]. Once the bipolar-attachment is established, AurB substrates in the outer kinetochore are dephosphorylated in a PP1-dependent manner [102, 105,106,107,108,109,110]. Dsn1 basic domain could also be dephosphorylated at this stage, presumably leading to reduction of binding-affinity between CENP-C and Mis12C in metaphase and anaphase; the levels of CENP-C binding Mis12C reduce in metaphase and anaphase (Fig. 7b) [20]. On the other hand, Cdk1 substrates should remain phosphorylated until mitotic exit in general [147]. The Cdk1-dependent interaction between KMN and CENP-T can be maintained through M-phase, indicating that CENP-T functions as the major load-bearing pathway for faithful chromosome segregation (Fig. 7b).

The proposition that CENP-T forms the major link between centromeres and microtubules is further supported by studies using supper-resolution microscopy for analysing detailed localization of kinetochore proteins in mitotic human cells [148]. They demonstrate that the CENP-C C terminus is placed near the surface of the centromeric chromatin, as expected, due to its binding to CENP-A. The middle part of CENP-C stretches away from the centromere, while its N terminus turns back inside, close to the centromere surface [148]. In contrast, CENP-T forms an elongated structure in the kinetochore; the CENP-T C terminus, which binds to DNA, localizes close to the centromeric chromatin, whereas its KMN-binding region in the N terminus stretches outside of the kinetochore, where Ndc80C is placed [148]. This observation suggests that while the microtubule-pulling force is applied on CENP-T, it is applied insignificantly or not applied on CENP-C in human cells, consistent with the proposition that CENP-T functions as the major load-bearing pathway. A tension sensor probe also suggests that a major microtubule-pulling force is applied on CENP-T and less on CENP-C in chicken DT40 cells [20].

Conclusion

The kinetochore is a dynamic molecular complex observed during the cell cycle progression and the dynamic changes in the kinetochore structure appear to be critical for faithful chromosome segregation. The pioneering EM studies captured the expansion and compaction of the kinetochore structures that is further confirmed by fluorescent microscopy observations. Molecular genetics and biochemical reconstitution continue to reveal the elaborate protein interaction network of kinetochore proteins and their dynamic features in detail. Since the changes in the protein–protein interactions and the structures are significant for kinetochore functions, further studies on regulations of the kinetochore protein interaction network will provide a comprehensive understanding of regulatory mechanisms ensuring faithful chromosome segregation.