Kinetochore attachment sensed by competitive Mps1 and microtubule binding to Ndc80C

See allHide authors and affiliations

Science  12 Jun 2015:
Vol. 348, Issue 6240, pp. 1260-1264
DOI: 10.1126/science.aaa4029

How cells sense connected chromosomes

Cells have a “checkpoint” that pauses cell division until all chromosomes are properly arranged on the mitotic spindle to allow precise distribution of one copy of each chromosome to each daughter cell. Hiruma et al. and Ji et al. explain the molecular mechanism by which cells sense that they are ready to divide. The protein kinase MPS1 associates with a protein complex at the kinetochore of the chromosome. Its activity produces signals that pause the cell cycle. When the chromosome becomes properly attached to the mitotic spindle, microtubules of the spindle physically compete for binding to the same site on the kinetochore where MPS1 is bound. Thus, once the kinetochore is properly attached, MPS1 dissociates, the inhibitory signal is lost, and cell division is allowed to proceed.

Science, this issue pp. 1264 and 1260


The spindle checkpoint of the cell division cycle senses kinetochores that are not attached to microtubules and prevents precocious onset of anaphase, which can lead to aneuploidy. The nuclear division cycle 80 complex (Ndc80C) is a major microtubule receptor at the kinetochore. Ndc80C also mediates the kinetochore recruitment of checkpoint proteins. We found that the checkpoint protein kinase monopolar spindle 1 (Mps1) directly bound to Ndc80C through two independent interactions. Both interactions involved the microtubule-binding surfaces of Ndc80C and were directly inhibited in the presence of microtubules. Elimination of one such interaction in human cells caused checkpoint defects expected from a failure to detect unattached kinetochores. Competition between Mps1 and microtubules for Ndc80C binding thus constitutes a direct mechanism for the detection of unattached kinetochores.

Unattached kinetochores (KTs) recruit and activate spindle checkpoint proteins to produce wait-anaphase signals, ensuring the fidelity of chromosome segregation. Microtubule (MT) attachment releases checkpoint proteins from KTs and attenuates checkpoint signaling. Whether KT signaling of any checkpoint proteins is directly inhibited by MT attachment remains to be established, however. The KMN network of KT proteins acts both as a critical MT receptor and a signaling platform for the spindle checkpoint (14). In humans, KMN contains the kinetochore null 1 protein (Knl1), the minichromosome instability 12 complex (Mis12C), and the nuclear division cycle 80 complex (Ndc80C). Ndc80C directly engages MTs through the N-terminal tail of highly expressed in cancer 1 (Hec1) and Calponin Homology (CH) domains of Hec1 and nuclear filament–containing protein 2 (Nuf2) (5). Ndc80C also mediates the KT localization of key checkpoint proteins, including monopolar spindle 1 (Mps1) and the mitosis arrest deficiency 1–2 complex (Mad1–Mad2) (6, 7). MT attachment releases both Mps1 and Mad1–Mad2 from KTs. In metazoans, removal of Mad1–Mad2 from KTs is, however, indirectly mediated through dynein-dependent poleward transport along MTs (8).

Mps1 phosphorylates Knl1 at multiple methionine-glutamate-leucine-threonine (MELT) motifs and creates docking sites for the budding uninhibited by benomyl 1–3 complex (Bub1–Bub3) (912). Mps1 further phosphorylates Bub1 and promotes its binding to Mad1–Mad2 (1315). Mps1 is thus an initiator of checkpoint signaling and a candidate for being a direct sensor of KT-MT attachment.

We first tested whether human Mps1 directly interacted with Ndc80C. Mps1 contains an N-terminal extension (NTE) critical for KT targeting (16), a tetratricopeptide repeat (TPR) domain, and a C-terminal kinase domain (Fig. 1A). Purified recombinant Mps1 and Ndc80C interacted with each other (fig. S1A). Furthermore, Mps1 NTE and a conserved middle region (MR) each independently bound to Ndc80C or Ndc80Cbonsai [miniaturized Ndc80C comprising mostly the head domains of its subunits (17)] (Fig. 1A and fig. S1, B to E).

Fig. 1 Direct interactions of Mps1 with Ndc80C through two distinct motifs.

(A) Domains and motifs of Mps1 and sequence alignment of the MR. Mps1 fragments that can or cannot bind to Ndc80C are depicted in red or black, respectively. H.s., Homo sapiens; M.m., Mus musculus; G.g., Gallus gallus; X.l., Xenopus laevis. The sequence of MR24 is boxed. Asterisk marks S281 phosphorylated in human Mps1. (B) Binding of Ndc80Cbonsai containing Hec1 WT, Δtail, or 9D (Spc24 and Spc25 blot) to pMR24-coupled beads. (C) Model explaining how Aurora B phosphorylation of the Hec1 tail stimulates MR binding to Ndc80C. (D) Binding of Mps11-200 (blotted with antibody to Mps1; top) to beads bound to GST-Hec1CH WT or the indicated mutants (Coomassie staining; bottom). The relative band intensities (as ratios to that of WT) were indicated below. (E) Binding of Ndc80Cbonsai containing Nuf2 WT or N126A (Spc24 and Spc25 blot) to pMR24 beads. Relative band intensities were indicated below. (F) Surface drawing of the CH domains of Hec1 and Nuf2. Hec1 and Nuf2 residues critical for Mps1 NTE and MR binding were labeled.

S281 of Mps1 MR is phosphorylated in mitotic human cells (18). The Mps1 MR24 and phospho-S281 pMR24 peptides bound to Ndc80Cbonsai with its Hec1 tail deleted (Δtail) with dissociation constants (Kd) of 8.4 and 2.0 μM, respectively, as measured with isothermal titration calorimetry (fig. S1F). Thus, S281 phosphorylation moderately enhanced binding of Mps1 MR to Ndc80C. Deletion of the Hec1 tail enhanced pMR24 binding to Ndc80Cbonsai (Fig. 1B), suggesting an autoinhibitory effect of this tail on binding of Mps1. The centromeric protein kinase Aurora B phosphorylates multiple residues in the Hec1 tail (17, 19, 20). Ndc80Cbonsai phosphorylated by Aurora B or with phosphomimicking mutations in Hec1 (9D) bound to pMR24 more strongly, indicating that Aurora B phosphorylation relieved the autoinhibition (Fig. 1, B and C, and fig. S1G). Thus, the Ndc80C–MR interaction appears to be regulated by mitotic phosphorylation of both partners.

Aurora B phosphorylation or deletion of the Hec1 tail had little effect on Mps1 NTE binding to Ndc80C (fig. S1H). The Kd of the NTE–Ndc80C interaction was 9.6 μM as determined by means of microscale thermophoresis (fig. S1J). Mps11-300 containing both motifs bound to Ndc80Cbonsai Δtail with a stoichiometry of 1:1 and a Kd of 8.7 μM, which was comparable with the affinity of either motif alone. Thus, Mps1 NTE and MR bind noncooperatively to Ndc80C.

We mapped the sites within Ndc80C bound by the two Mps1 motifs. Mps11-200 bound to the CH domain of Hec1 (Fig. 1D) (21). K89E and K166E mutations of Hec1 CH compromise binding of Ndc80C to MTs (17). (Single-letter abbreviations for the amino acid residues are as follows: A, Ala; C, Cys; D, Asp; E, Glu; F, Phe; G, Gly; H, His; I, Ile; K, Lys; L, Leu; M, Met; N, Asn; P, Pro; Q, Gln; R, Arg; S, Ser; T, Thr; V, Val; W, Trp; and Y, Tyr. In the mutants, other amino acids were substituted at certain locations; for example, K89E indicates that lysine at position 89 was replaced by glutamic acid.) These mutations also diminished binding of Mps11-200 to Ndc80C (Fig. 1D), indicating that Mps1 NTE bound to the MT-binding surface on Hec1. Deletion of Nuf2 CH, but not Hec1 CH, abolished binding of pMR24 to Ndc80C (fig. S2A). Several Nuf2 CH single mutants exhibited reduced binding to pMR24, with N126A being the most deficient (Fig. 1E and fig. S2, B and C). These mutated residues cluster around the MT-binding site of Nuf2 (Fig. 1F) (17). Thus, Mps1 MR binds to the MT-binding site of Nuf2.

We tested whether Mps1 NTE and MR competed with MTs for Ndc80Cbonsai binding. Taxol-stabilized MTs reduced binding of Ndc80C to both Mps11–200 (containing NTE) and pMR24 in a dose-dependent manner (Fig. 2, A and B, and fig. S3, A and B). Binding of Mps1 to Ndc80C 9D or Δtail (which are deficient in binding to MT) was less affected by MTs. Unpolymerized tubulin did not block the two Mps1–Ndc80C interactions (fig. S3C). Conversely, pMR24 at high concentrations reduced the MT-binding affinity of Ndc80C from 0.56 to 5.87 μM (Fig. 3C and fig. S3D). It did not affect binding between MT and the MR-binding–deficient Ndc80C N126A. Thus, both Mps1 NTE and MR compete with MTs for binding to Ndc80C.

Fig. 2 Competition of Mps1 and MTs for binding to Ndc80C.

(A and B) Release of Ndc80Cbonsai proteins bound to beads containing (A) GST-Mps11-200 or (B) pMR24 by taxol-stabilized MTs. Relative intensities of Ndc80C bound to beads were plotted against MT concentrations. AU, arbitrary unit. Error bars, SD (n = 3 independent experiments). (C) Inhibition of Ndc80Cbonsai (Nuf2 WT or N126A) binding to taxol-stabilized MTs by pMR24, as revealed by MT pelleting assay. Fractions of MT-associated Ndc80Cbonsai proteins were plotted against MT concentrations. Data were fitted with one-site binding curves. Error bars, SD (n = 3 independent experiments). (D) Localization of stably expressed GFP-Mps1 proteins in HeLa cells depleted of endogenous Mps1. ΔNTE, mutant lacking NTE; ΔMR, mutant lacking MR; ΔNM, mutant lacking NTE and MR; CREST, calcinosis, Raynaud phenomenon, esophageal dysmotility, sclerodactyly, telangiectasia; DAPI, 4',6-diamidino-2-phenylindole. Boxed regions were magnified and shown in insets. (E) Quantification of the KT enrichment of GFP signals in (D). Each dot represents one cell. Bars indicate the median. ****P < 0.0001; N.S., not significant; Student’s t test. (F) Immunofluorescence images of GFP-4MD–expressing HeLa cells. (G) Quantification of the percentages of mitotic cells with or without GFP-4MD expression that exhibited the indicated chromosome alignment phenotypes. Mean and range of two independent experiments (n > 100 cells each) are shown.

Fig. 3 Requirement of Mps1–Ndc80C interactions during checkpoint signaling.

(A and B) Mitotic indices of HeLa cells stably expressing the indicated (A) GFP-Mps1 or (B) Nuf2-Myc proteins transfected with the indicated small interfering RNAs (siRNAs) and treated with nocodazole (Noc) with or without ZM447439 (ZM). WT, wild type. Error bars, SD (n = 4 independent experiments). ***P < 0.001; ****P < 0.0001; Student’s t test. (C) Chromosome spreads of cells in (B) treated with nocodazole and MG132 were stained with DAPI and indicated antibodies. Boxed regions were magnified and shown in insets. (D) Quantification of the relative KT intensity of phospho-T875 Knl1 staining in (C). Each dot represents one KT pair. Bar indicates the median (n = 460 pairs of KTs). ****P < 0.0001; Student’s t test. (E) Model of MT-mediated inhibition of Mps1 signaling at KTs. Molecules and structures are not drawn to scale. Knl1 and Ndc80C are drawn as straight rods for simplicity.

Dynein-dependent poleward transport (8, 22) releases Mad1–Mad2 from attached KTs. This mechanism is blocked by the F258A mutation in Spindly, a KT receptor of dynein (23). Expression of Spindly F258A in HeLa cells depleted of endogenous Spindly blocked Mad1 release from attached KTs but did not block Mps1 release (fig. S4). Release of Mps1 from attached KTs is thus independent of dynein, likely through direct MT competition.

Mps1 ΔNTE failed to localize to unattached KTs, whereas Mps1 ΔMR localized normally to KTs (Fig. 2, D and E, and fig. S5A). Aurora B inhibition expectedly abolished the KT localization of both Mps1 wild type (WT) and ΔMR (fig. S5, B and C). Thus, the KT localization of Mps1 in vivo appears to depend mainly on the NTE–Ndc80C interaction. Competition of this interaction with MTs likely releases Mps1 from attached KTs. Four repeats of MR containing the phosphomimicking S281D mutation (4MD) fused to green fluorescent protein (GFP) localized to KTs (fig. S6). This localization was abolished by the Nuf2 N126A mutation, indicating that the MR–Nuf2 interaction could target Mps1 to KTs to some extent. GFP-4MD still localized to attached metaphase KTs and interfered with chromosome alignment (Fig. 2, F and G). The artificial GFP-4MD might outcompete MTs for binding to Ndc80C or displace the endogenous Mps1 bound to Nuf2.

HeLa cells depleted of endogenous Mps1 and expressing Mps1 ΔNTE failed to undergo mitotic arrest in nocodazole, demonstrating a requirement for NTE in checkpoint signaling (16) (Fig. 3A). Although Mps1 ΔMR was slightly deficient in supporting nocodazole-induced mitotic arrest, it was largely inactive when Aurora B was inhibited by ZM447439 (ZM). Bub1 KT localization was compromised in Mps1 ΔMR cells, even without Aurora B inhibition (fig. S7). Furthermore, the MR-binding–deficient Nuf2 N126A failed to restore mitotic arrest of cells depleted of endogenous Nuf2 and treated with ZM (Fig. 3B and fig. S8A). Nuf2 N126A cells had moderate checkpoint defects, even in the absence of ZM. Mps1-dependent events, including Knl1 phosphorylation and Bub1-related protein 1 (BubR1) KT localization, were defective in Nuf2 N126A cells (Fig. 3, C and D, and fig. S8, B to E). Thus, the Mps1 MR–Nuf2 interaction also contributes to checkpoint signaling. Disruption of this interaction in nocodazole-treated cells (with KTs unattached) renders the checkpoint dependent on Aurora B, similar to the checkpoint in taxol-treated cells (with most KTs attached) (24). These findings suggest that the Mps1 MR–Nuf2 interaction senses the status of KT attachment in human cells.

Our results establish a straightforward cellular mechanism for sensing KT–MT attachment (Fig. 3E). Because NTE and MR bind independently to Hec1 and Nuf2 heads, there might be two types of Mps1–Ndc80C interactions at KTs: a major one involving the NTE–Hec1 interface and a minor one involving the MR–Nuf2 interface. Binding of MT to Ndc80C releases both and inhibits Mps1 signaling. The weak, multisite Mps1–Ndc80C interactions explain the transient nature of Mps1 at KTs and the inability to detect these interactions in human cell lysates. These interactions might have evolved to balance Mps1 recruitment to unattached KTs and its release from attached ones.

The MR–Nuf2 interaction is required for Mps1-dependent phosphorylation of Knl1 and events downstream of this phosphorylation. Because Knl1 and Ndc80C are components of the same constitutive protein network, the weak MR–Nuf2 interaction is reminiscent of docking interactions between kinases and substrates. In the case of Mps1, MR does not dock on Knl1 itself but on Ndc80C, which indirectly associates with Knl1 through Mis12C. The two degrees of separation ensure that Mps1 preferably phosphorylates Knl1 in the intact KMN.

Our study further points to an intricate cross-talk among the mitotic kinases Mps1, Bub1, and Aurora B in sensing KT–MT attachment. Increasing phosphorylation of the Hec1 tail by Aurora B progressively weakens MT binding by Ndc80C (25). In contrast, Aurora B–dependent phosphorylation of the same Hec1 tail enhances binding of Mps1 MR to Nuf2, which promotes phosphorylation of Knl1 and KT targeting of Bub1. In a feed-forward mechanism, Bub1 can then phosphorylate histone H2A and further enrich Aurora B at centromeres (26, 27). Thus, Aurora B regulates the competing MT- and Mps1-binding activities of Ndc80C in opposite directions.

MT- and dynein-dependent release of Mad1–Mad2 from attached KTs indirectly contributes to KT–MT attachment sensing in metazoans but not in yeast (28). Our study establishes competition between Mps1 and MTs for binding to Ndc80C as a direct mechanism for monitoring KT attachment. Although the NTE and MR motifs of vertebrate Mps1 are not conserved in yeast, Mps1 interacts with Ndc80C in yeast (29). Thus, Mps1 may be a primordial sensor of KT–MT attachment.

Supplementary Materials

Materials and Methods

Figs. S1 to S8

References (3035)

References and Notes

  1. Acknowledgments: We thank C. Brautigam for assistance with isothermal titration calorimetry and microscale thermophoresis, H. Ball for peptide synthesis, and the Animal Resource Center on campus for antibody production. H.Y. is an investigator with the Howard Hughes Medical Institute. This work is supported by the Cancer Prevention and Research Institute of Texas (RP110465-P3 and RP120717-P2) and the Welch Foundation (I-1441).
View Abstract

Navigate This Article