The whole is greater than the sum of its parts: At the intersection of order, disorder, and kinetochore function

Abstract

The kinetochore field has matured tremendously since Earnshaw first identified CENP-A, CENP-B, and CENP-C (1, 2). In the past 35 years, the accumulation of knowledge has included: defining the parts list, identifying epistatic networks of interdependence within the parts list, understanding the spatial organization of sub-complexes into a massive structure - hundreds of megadaltons in size, and dissecting the functions of the kinetochore in its entirety as well as of its individual parts. Like nearly all cell and molecular biology fields, the structure-function paradigm has been foundational to advances in the kinetochore field. A point nicely highlighted by the fact that we are at the precipice of the in vitro reconstitution of a functional kinetochore holo complex. Yet conventional notions of structure cannot provide a complete picture of the kinetochore especially since it contains an abundance of unstructured or intrinsically disordered constituents. The combination of structured and disordered proteins within a kinetochore results in an assembled system that is functionally greater than the sum of its parts.

INTRODUCTION

The kinetochore (KT) is a molecular superstructure that is hundreds of MDa in size in metazoans. The KT assembles at centromeres during cell division to fulfill two essential functions: 1) attachment of chromosomes to spindle microtubules (MTs) and 2) coordination of a signaling pathway that regulates the spindle assembly checkpoint (SAC). Both functions require intricate coordination of KT and centromere components (3). Attachment-mediation is a classic goldilocks scenario whereby kinetochore-microtubule (kMT) affinities must be “just right” – attachments that are either too weak or too tight can result in aneuploidy (4). Similarly, the SAC is exquisitely sensitive to a titratable signal, which is produced by misaligned and/or unattached KTs, that delays anaphase onset (5–7). When necessary, the anaphase-delaying signal emanating from KTs provides the cell with more time to achieve complete chromosome biorientation – the ideal configuration for accurate chromosome segregation whereby each sister KT is attached to MTs oriented towards opposite spindle poles. The SAC becomes satisfied once the titratable wait-anaphase signal is no longer produced and, as a result, anaphase ensues within minutes of the final chromosome biorientation event (8, 9).

Key molecular mediators of both the attachment and SAC-regulating functions are contained in the KMN network comprised of the protein KNL1 (K), the Mis12 (M) complex, and the Ndc80 (N) complex (10). The core attachment factor is the Ndc80 complex (11), which binds directly to MTs via a calponin homology (CH) domain assembled from the Ndc80 and Nuf2 subunits (12, 13). KNL1 is a large highly disordered protein that serves as a checkpoint signaling hub by directly binding many of the proteins that comprise the diffusible wait-anaphase signal. These checkpoint proteins associate via an array of internally distributed motifs known as MELTs (14–23). Interestingly, the disordered N-terminus of KNL1 possesses multiple protein-phosphatase 1 (PP1) binding motifs (24–26) as well as MT binding activity that contributes in a synergistic manner with the Ndc80 complex to the affinity of kMT attachments (10).

The attachment and SAC functions of the KT are highly phospho-regulated. The KT’s affinity for MTs is reduced by phosphorylation of the N-termini of both KNL1 and Ndc80 by Aurora B kinase (11, 27), which is highly enriched as part of the chromosomal passenger complex (CPC) at the region between sister KTs called the inner centromere. Consequently, Aurora B activity promotes unattached or weakly attached KTs. The checkpoint kinase Mps1 promotes the production of the wait-anaphase signal by increasing the affinity of KNL1 for checkpoint proteins via phosphorylation of the MELT motifs (16, 19). Mps1 binds directly to the CH domain in the Ndc80 complex and is competitively released upon MT binding (28–30). Since sustained Mps1 activity is required to maintain a mitotic delay, attachment of the Ndc80 complex to MTs contributes directly to extinguishing the SAC by reducing KT-levels of Mps1 (31, 32) although cytosolic Mps1 activity also contributes to production of the wait-anaphase signal (33). De-phosphorylation of KT substrates is largely mediated by two major KT-localized serine/threonine phosphatases: PP2A-B56 and PP1 (34, 35). The substrate specificity of PP2A-B56 and PP1 comes from the spatiotemporal recruitment pattern of each phosphatase to the KT throughout mitosis (36, 37). Understanding the independent functions of these phosphatases is challenging because they are known to be promiscuous and there is significant regulatory cross-talk between them. More remains to be teased apart about how PP1 and PP2A-B56 contribute to kMT attachment stability and SAC satisfaction.

The KT is arguably the most important force-transducing macromolecular complex in the cell. MT dynamics and motor proteins exert forces on KTs. In toto, these forces can be substantial as numerous studies have reported that metazoan KTs can be subjected to >100 pN of force (38–41). Much of the force exerted on bioriented KTs likely comes from depolymerizing MTs that can generate upwards of 30 pN of pulling force per MT (42–44) – forces that are entirely consistent with microneedle studies by Bruce Nicklas that yielded an estimate of 30 pN per MT being required to stabilize kMT attachments (45). While tension generated by MT-depolymerization forces is a central input into stabilizing kMT interactions, the molecular basis of tension dependent stabilization is still an area of active research. Although there is broad agreement that tension stabilizes kMT attachments, whether tension also directly regulates the SAC remains a long-standing debate in the field that will be discussed further below.

The KT is enriched for proteins that are predicted to be intrinsically disordered, which by our estimate comprise ~50% of the molecular mass of the D. melanogaster KT. Intrinsically disordered proteins (IDPs) by definition lack the fixed tertiary structure of a structured protein. IDPs tend to have low sequence complexity with a high proportion of charged and polar residues and a low proportion of hydrophobic amino acids (46). Therefore, disordered proteins typically lack the hydrophobic core that encourages the formation of globular structures. There are some disordered proteins that adopt a transient folded structure given specific environmental conditions or interactions with a binding partner (47). Additionally, many structured proteins have intrinsically disordered regions that may act as flexible linkers and are hotspots for phosphorylation (48, 49). IDPs often act as signaling hubs via recruitment of binding partners, but are not typically under tension in cells although focal adhesions and sarcomeres offer some noteworthy examples of force-sensing IDPs. For example, the focal adhesion protein p130Cas undergoes extension-mediated substrate priming whereby the application of force leads to its hyper-phosphorylation by Src kinase and enhancement of downstream signaling (50). Similarly, the giant sarcomere protein titin contains disordered “PEVK” modules that extend under tension in a manner that could modulate titin’s signaling capacity by affecting its association with SH3-containing binding partners (51–53). Given the abundance of IDPs at the KT, mechanistic insights into how tension impacts attachment stability and SAC signaling requires a more advanced understanding of the impact of forces on disordered KT proteins.

TENSION AND THE SPATIAL POSITIONING MODEL: ATTACHMENT STABILIZATION AND LINKS TO SAC REGULATION

The KT has spatially distinct domains (54) that were originally identified as separate electron-dense regions by electron microscopy (55, 56). The inner KT is the DNA-proximal region and is comprised of chromatin/DNA-binding components and the outer KT contains the KMN network (10, 11, 57). The outer KT moves away from the inner KT upon chromosome biorientation (58, 59). This structural change, which we deemed intrakinetochore (intraKT) stretch, underpins “spatial positioning” models of tension-based stabilization of kMT interactions and regulation of the SAC. Spatial positioning in this regard refers to the proximity of the centromere-enriched Aurora B to its outer KT substrates. The model posits that Aurora B efficiently phosphorylates KMN components when intraKT stretch is low while a tension-dependent increase in intraKT stretch results in reduced phosphorylation of the outer KT by Aurora B (60–63).

Indirect SAC regulation via attachment stabilization

There is broad agreement in the field on two points: 1) unattached KTs produce a wait-anaphase signal, and 2) tension stabilizes kMT attachments. The spatial positioning model provides an appealing explanation for how tension stabilizes kMT attachments: tension-based increase in intraKT stretch moves the Ndc80 complex and the N-terminal MT binding domain of KNL1 away from Aurora B. The increased distance between KNL1 and Ndc80 complex from Aurora B makes them less likely to be phosphorylated thereby increasing their affinities for MTs and stabilizing attachments. Conversely, reduction in intraKT stretch favors unattached KTs via Aurora B-dependent phosphorylation of Ndc80 and KNL1. In this case, lack of tension makes an indirect contribution to SAC regulation through the production of unattached KTs that produce a robust wait anaphase signal. However, there is ongoing debate as to whether kMT attachments are the only input required for SAC satisfaction or if tension directly contributes as well.

“Sort of direct” SAC regulation via PP1 recruitment to KNL1

The molecular identity of the wait-anaphase signal is a heterotetramer called the mitotic checkpoint complex (MCC) comprised of: Mad2, Bub3, BubR1, and Cdc20 (64). Enrichment of the MCC components at KTs catalyzes assembly of the complex and production of the wait-anaphase signal. As described earlier, KNL1 serves as a key location for enriching checkpoint components by providing tens of docking sites called MELT motifs distributed throughout a central disordered region. Checkpoint protein docking is phospho-regulated by Mps1 kinase, which is enriched at unattached KTs. In many organisms, the MELT motifs must be phosphorylated by Mps1 (MELTp) for Bub3 (and its binding partners Bub1 and BubR1) to bind to KNL1. The establishment of kMT attachments disrupts the interaction between Mps1 and the Ndc80 complex and thereby reduces the propensity of MELTs to be phosphorylated. However, this is only half of the phospho-regulation equation. The spatial positioning model may fill out the other half via regulation of PP1 phosphatase recruitment to KTs. The N-terminus of KNL1 contains multiple PP1 binding motifs that are negatively regulated by Aurora B-mediated phosphorylation (25). Thus, just as the tension-dependent increase in intraKT stretch is envisioned to increase affinity of the KT for MTs, it also increases the affinity of KNL1 for PP1. KNL1-bound PP1 promotes de-phosphorylation of the MELT motifs, which contributes to SAC satisfaction by reducing their affinity for the Bubs. Interestingly, recent work showing that PP1 does not directly dephosphorylate MELT motifs (65) indicates that MELTs are dephosphorylated by an alternative phosphatase. Nonetheless, the KNL1-bound population of PP1 is a critical input to extinguishing the wait-anaphase signal. It remains unclear whether the KNL1-PP1 population is also a major regulator of attachment stability. Increased PP1 recruitment to KNL1 through spatial positioning represents a more direct contribution of tension to SAC satisfaction than the attachment stabilization pathway especially if kMT attachments can be formed independent of the KNL1-bound pool of PP1. We lightheartedly describe this pathway as “sort of direct” in comparison to the models outlined below that evoke direct KT tension-sensor molecules.

Caveats to spatial positioning

As appealing and logical as the spatial positioning model has been in explaining tension dependent regulation of attachment stability and SAC signaling, it should be noted that the model is not ironclad (66). Non-centromeric populations of Aurora B have been observed at the KT (67–69) and attachment regulation can occur even in the absence of centromere-enriched CPC (70–75). Thus, the contributions of non-centromeric pools of Aurora B to attachment stability and SAC regulation (76) as well as if/how the activities of such pools are impacted by tension are important and unresolved questions.

TENSION AND DISORDERED PROTEINS: DIRECT REGULATORS OF SAC SIGNALING?

KMN binding to microtubules

It cannot be fully appreciated how the KMN network interacts with dynamic spindle MTs without also thinking about MT structure. The subunit of a MT is an obligatory heterodimer of α- and β-tubulin (77). Heterodimers are assembled end-to-end into a protofilament (PF), thirteen of which (in vivo) associate laterally to form a hollow tube (78). PFs can bend/curl at their plus-ends while further back on the lattice the PFs are straight. At depolymerizing plus-ends, PFs curl and eventually peel away from the MT. If the KT can manage to hold on to a shortening plus-end, it is the mechanical work done by bending PFs that results in the application of force to KTs (as discussed earlier). Importantly, the curling of a PF results in a conformational bend at the so-called “hinge point” within the interface of the α and β tubulin monomers that comprise the heterodimer (79).

The interaction between Ndc80 and dynamic MTs is mediated by a structured CH domain and an unstructured N-terminal tail (80). A so-called “toe” within the CH domain binds at the inter- and intra-dimer interfaces between each tubulin monomer (81). Interestingly, the Ndc80 toe print overlaps with the hinge point such that PF curling sterically hinders binding of the Ndc80 complex. Accordingly, the Ndc80 complex has a higher affinity for straight PFs than for curled ones. It is hypothesized that the Ndc80 toe acts as a MT conformation sensor preferentially binding to straightened stretches of the MT lattice rather than to bending PFs at the plus-ends. Of course, the KT should ideally be able to hang on to depolymerizing MTs. This critical function can be conferred to the Ndc80 complex by accessory factors that will not be discussed here. Using a streptavidin-based oligomerization system to control Ndc80 complex copy number in vitro, it was recently shown that ensembles of Ndc80 complexes attached to shortening MT plus-ends (82) in the absence of accessory proteins. This activity was dependent on the disordered and highly basic N-terminal tail of Ndc80 that electrostatically interacts with the acidic, disordered C-terminal tails of α and β tubulin. Thus, fully functional kMT interactions in cells require the structured CH domain of Ndc80 to bind the MT lattice and the unstructured tails of multiple Ndc80 complexes to efficiently harness the force of MT depolymerization (83).

While Ndc80 binding to MTs has been studied extensively, present knowledge of MT binding by KNL1 is limited due, in part, to its highly unstructured nature. Recent advances in understanding how the intrinsically disordered MT-associated protein Tau binds to MTs may provide insights into the nature of KNL1-MT interactions. Similar to Ndc80, Tau binds to the inter- and intra-dimer interfaces between tubulin monomers and electrostatic interactions contribute to the binding affinity (84). However, rather than associating with the hinge point, hydrophobic residues across multiple Tau repeats (the MT binding module) sit in a “greasy” pocket in α-tubulin and bind the so-called “anchor point” that maintains a constant structural conformation between curled and straight protofilaments (85). Furthermore, Tau’s unstructured nature may give it the molecular “slack” or flexibility for its anchor-point bound repeats to remain associated with highly curled tubulin oligomers. A recent characterization of Tau “condensates” indicated that Tau condensation preferentially occurred at regions of high curvature and only on GDP-MT lattices leading the authors to argue that the Tau condensation phenomenon responds to the spacing between tubulin dimers (86). Thus, Tau utilizes a combination of electrostatic and hydrophobic MT binding residues distributed through an intrinsically disordered region to bind straight and curled PFs and may preferentially condense on GDP-bound curling protofilaments.

KNL1 possesses a conserved basic patch at its immediate N-terminus that binds directly to MTs in vitro (24, 27). Deletion or mutation of this basic patch in C. elegans resulted in a checkpoint-dependent mitotic delay even in the presence of normal load-bearing kMT attachments indicating that the MT binding activity of KNL1 contributes to silencing the checkpoint (24). Importantly, this was not through the mis-localization of PP1 as the MT binding mutants were still able to bind PP1 in vitro. However, the basic MT-binding patch is not the entire picture as one of the N-terminal PP1 binding motifs was recently shown to also bind MTs in vitro (87). Interestingly, PP1 and MT binding were mutually exclusive in vitro - a finding that warrants further investigation since PP1 is highest at bioriented KTs when one would expect KNL1 to interact with kMTs considering that 1) the MT binding motif should no longer be phosphorylated by Aurora B (see the spatial positioning section above) and 2) MT binding by KNL1 contributes to timely checkpoint satisfaction. There are numerous small basic and hydrophobic patches distributed through the first several hundred amino acids of KNL1. Deletion of a hydrophobic region in the first ~240 aa of C. elegans KNL1, which contributes to oligomerization of the protein in vitro, phenocopied the basic patch MT binding mutants in C. elegans as both mutants exhibited a reproducible delay in anaphase onset (88). One possible explanation is that oligomerization increases KNL1’s affinity for MTs through promoting multivalent electrostatic interactions between the N-terminal basic patches and MTs. Alternatively, like Tau, KNL1 could utilize hydrophobic and electrostatic interactions distributed through an intrinsically disordered region to bind MTs (Figure 1). If KNL1 utilizes a Tau-like mode of interaction, then it may associate with polymerizing and depolymerizing kMTs. Given its long unstructured nature, it is appealing to us that KNL1 bound to curled kMT PFs could be a constituent of the conserved McIntosh fibrils (89, 90) that link the KT to flared PFs – the molecular identity of which remains unknown. Better understanding how KNL1 interacts with MTs will certainly provide further insight into KNL1’s function in regulating the SAC.

Figure 1. Tau as a model for KNL1-MT interactions.

The KMN network is made up of the largely disordered protein KNL1 (K), the Mis12 complex (M), and the Ndc80 complex (N). The Mis12 complex acts as a hub that binds both the Ndc80 complex and the C-terminus of KNL1 at a junction point. The Ndc80 (blue) toe has been proposed to act as a conformation sensor by preferentially binding straight protofilaments over curled protofilaments while its disordered tail contributes to the complex remaining associated with depolymerizing MTs. The N-terminus of KNL1 binds to MTs although the nature of this association is not well-characterized. In this speculative model we envision KNL1 as binding to MTs in a manner similar to that recently characterized for the MT-associated protein Tau. Specifically, MT interaction motifs distributed through a disordered N-terminal region of KNL1 (pink) span multiple tubulin monomers, possibly through interacting with anchor points (see main text), and its disordered nature provides flexibility for the protein to bind straight protofilaments (top) and remain associated with curling protofilaments on depolymerizing MTs (bottom). Upon MT shortening, the central disordered region of KNL1 would be extended due to the force generated by bending protofilaments. The model yields extended KNL1 molecules reminiscent of the fibrils that McIntosh and colleagues have observed extending from the KT to the flared protofilaments of kMTs (89, 90). The molecular identity of proteins comprising the McIntosh fibrils is unknown.

Tension-sensor molecules: How force-transducing kinetochore proteins could directly regulate the SAC

How could tension directly regulate the SAC? We hypothesize that a direct regulatory input would be mediated by KT-associated tension-sensor molecules that possess several characteristics. First, the sensor must be force-transducing such that the sum of the force-vectors acting on the molecule, most notably from pulling forces generated by depolymerizing MTs, would produce sufficient tension to alter the properties of the sensor. While it may be ideal for the sensor to directly bind and hang onto shortening MTs, this feature is not essential since non-MT binding molecules in the linkage between the DNA and depolymerizing MTs are subjected to tension (41). Second, a tension-sensor molecule must bind directly to checkpoint regulators. Finally, the application of force to the sensor must reduce its affinity for checkpoint proteins. Under low/no tension the sensor molecule would recruit checkpoint proteins to contribute to the production of a robust wait-anaphase signal. Under tension the affinity of the sensor for checkpoint proteins would be reduced resulting in a tension-dependent reduction of KT-associated checkpoint proteins. In this model, force (we favor pulling forces generated by depolymerizing kMTs) acts as a direct physical input into extinguishing production of a wait-anaphase signal.

Chromosome biorientation produces tension across both structured and unstructured KT components (41, 91), however, their response to physiological forces would be drastically different. Chromosome biorientation in metazoans may generate hundreds of pNs, but this force is distributed through hundreds of molecular linkages that are likely arranged in parallel between the DNA and the MT. We previously estimated that the average magnitude of force applied to each molecular linkage is in the low pN range (41). While this amount of force may orient/straighten the Ndc80 complex, which is quite “bendy” because it has a long alpha-helical coiled-coil that is interrupted briefly with an unstructured loop in Ndc80 (80, 92), it is insufficient to unfold well-structured complexes or protein domains. However, this magnitude of force could readily extend unstructured proteins tens of nanometers – lengths that are greater than the diameter of the average folded protein. Thus, intrinsically disordered proteins are well-situated to act as tension-sensing molecules and we envision various means by which the application of force to an unstructured protein could reduce its affinity for binding partners (Figure 2). It does not escape our attention that KNL1 could meet our requirements for a checkpoint tension sensor since it binds MTs and checkpoint proteins and it is well-positioned to be force-transducing. However, it is presently unclear if KNL1 can harness the force of depolymerizing MTs to become extended at metaphase. There are published data indicating that KNL1 may be relaxed at unattached KTs and extended at metaphase albeit suppression of MT dynamics with taxol did not result in a shortening of KNL1 in these studies (93–95). This is somewhat contradictory to a recent in-depth, image-based analysis of human KT organization that reported KNL1 is relaxed at unattached and metaphase KTs but extended at taxol-treated “tensionless” KTs (54). While the model outlined here broadly aligns with the authors’ proposal that KNL1 acts as a tension sensor by “unraveling” to relay tension, the data showing that KNL1 is extended in the tensionless state and relaxed at metaphase is the inverse of what our tension-sensor model envisions. Thus, further investigation of KNL1 (and other unstructured KT elements) as possible checkpoint tension-sensor molecules will be an exciting line of inquiry.

Figure 2. Models for how the application of force to disordered proteins could affect their affinity for binding partners.

(A) The top model envisions a structured binding partner with multiple interaction motifs (red) that interact with dispersed binding sites (blue) on the disordered protein. Under low force the disordered protein can accommodate a spatial organization that allows multiple binding sites (3 in this example) to be positioned in such a manner that a higher affinity interaction can occur. Application of increasing force to the disordered protein decreases the affinity for its binding partner by reducing the likelihood of multivalent interactions. This could result in a graded (pink) or switch-like (dashed line) change in affinity. Interestingly, multiple MELT motifs are required for efficient checkpoint signaling (and recruitment of Bubs) (23) and multivalent electrostatic interactions with a phospho-peptide binding pocket and basic patches on the surface of metazoan Bub3 homologues contribute to their affinity for KNL1 and to checkpoint signaling (19). (B) Molecular recognition features (MoRFs) are intrinsically disordered regions that undergo a transition to a structured state (such as an alpha-helical fold) upon binding to a structured partner (47). Accordingly, the affinity of the interaction for the ordered state is higher than for the disordered state. The bottom model envisions that under low/no forces transition to order would be favored while the application of force to the MoRF would drive the transition to a disordered state thereby lowering the binding affinity. It should be noted that even when localized to KTs, checkpoint proteins cannot have too high an affinity for their binding partners because they associate in a highly dynamic fashion and turnover on a seconds timescale (96, 97).

SUMMARY POINTS.

• The structure-function paradigm has underpinned advancements in our understanding of the kinetochore over the past 35 years.

• A significant fraction of the kinetochore’s molecular mass is comprised of proteins that are predicted to be unstructured.

• Understanding of the kinetochore cannot be complete without studying unstructured kinetochore proteins and their responses to force, which is difficult to do with conventional structure-function-based methodologies.

• Since the kinetochore is the most important force-bearing complex in dividing cells, studying how tension impacts unstructured kinetochore proteins and their functions offers exciting research opportunities moving forward.

ABBREVIATIONS

CH

Calponin homology

CPC

Chromosomal passenger complex

IDP

Intrinsically disordered protein

KMN

KNL1, Mis12 complex and Ndc80 complex

kMT

Kinetochore-MT

KT

Kinetochore

MT

Microtubule

MCC

Mitotic checkpoint complex

MoRFs

Molecular recognition features

PF

Protofilament

SAC

Spindle assembly checkpoint