Spindle checkpoint silencing at kinetochores with submaximal microtubule occupancy

Errors in chromosome segregation cause aneuploid karyotypes, which are devastating to embryonic development and are strongly associated with cancer (de Wolf and Kops, 2017; Duijf et al., 2013; Hanahan and Weinberg, 2011; Ricke and van Deursen, 2013). To ensure proper chromosome segregation, the spindle assembly checkpoint (SAC) prevents anaphase initiation until all chromosomes are stably attached to spindle microtubules. These attachments are powered by kinetochores, specialized structures assembled on centromeric chromatin (Musacchio and Desai, 2017). Microtubule binding by kinetochores is mediated predominantly by the NDC80 complex (Cheeseman et al., 2006; DeLuca and Musacchio, 2012; DeLuca et al., 2002; Tooley and Stukenberg, 2011). When unbound by microtubules, however, this complex recruits the MPS1 kinase (also known as TTK) to kinetochores (Hiruma et al., 2015; Ji et al., 2015; Liu and Winey, 2012), where it initiates a cascade of events that culminates in production of the anaphase inhibitor. The cascade involves phosphorylation of the short linear MELT sequences in the kinetochore protein KNL1 to form the binding sites for the BUB3-bound SAC proteins BUBR1 and BUB1 (Krenn et al., 2014; Overlack et al., 2015; Primorac et al., 2013; Vleugel et al., 2013; Zhang et al., 2014). MPS1 also ensures localization of the MAD1–MAD2 complex (also known as MAD1L1 and MAD2L1, respectively), at least in part by promoting BUB1–MAD1 interactions (Kim et al., 2012; London and Biggins, 2014; Silió et al., 2015). MAD1–MAD2 recruitment additionally requires the RZZ (ROD-ZW10-Zwilch) kinetochore complex but the mechanism of this has not been elucidated (Caldas et al., 2015; Matson and Stukenberg, 2014; Silió et al., 2015). Although poorly understood at the molecular level, a subset of these SAC proteins then form a multiprotein assembly with potent anaphase inhibitory activity (Chao et al., 2012; Herzog et al., 2009; Kulukian et al., 2009; Sudakin et al., 2001).

Whereas recruitment of SAC proteins to kinetochores is essential for proper SAC activation, their removal is crucial for efficient SAC silencing and timely anaphase onset (Ballister et al., 2014; Ito et al., 2012; Jelluma et al., 2010; Kuijt et al., 2014; Maldonado and Kapoor, 2011). Microtubule attachments disrupt SAC signalling from kinetochores by mediating poleward transport of SAC proteins by the dynein motor complex (a process referred to as ‘stripping’) (Howell et al., 2001), and by affecting the balance of SAC-regulating kinases and phosphatases (Etemad and Kops, 2016; Funabiki and Wynne, 2013; Saurin, 2018). For example, RZZ–MAD1 is a cargo of dynein via interactions with the kinetochore-specific dynactin adaptor Spindly (SPDL1) (Barisic et al., 2010; Caldas et al., 2015; Chan et al., 2009; Gassmann et al., 2008; Kops et al., 2005; Silió et al., 2015). By contrast, BUB protein removal is dependent on inhibition of local MPS1 activity and reversal of MELT phosphorylations by the PP1 phosphatase (Etemad and Kops, 2016; Hiruma et al., 2015; Ji et al., 2015; London et al., 2012; Meadows et al., 2011; Nijenhuis et al., 2014; Rosenberg et al., 2011; Zhang et al., 2014).

The subcellular architecture of kinetochores is substantially more complex than illustrated above. A single human kinetochores contains ∼240 NDC80 complexes, probably configured in a lawn-like macro-structure (Suzuki et al., 2015; Zaytsev et al., 2014). This lawn can bind up to 20 microtubules that together form a so-called kinetochore (k)-fibre (DeLuca et al., 2005; McEwen et al., 2001; Nixon et al., 2015; Wendell et al., 1993). Likewise, when unbound by microtubules, a single human kinetochore probably binds hundreds of SAC modules (Howell et al., 2004; Vleugel et al., 2015). This subcellular complexity of kinetochores raises numerous questions about the response dynamics of SAC modules to increasing amounts of bound microtubules. A current model of SAC signalling suggests that the SAC signal from kinetochores as a function of microtubule binding is not binary, but can exist in intermediate states (Collin et al., 2013). Whether SAC signalling is fully shut down only when kinetochores have acquired close to maximal microtubule occupancy is unknown (Burke and Stukenberg, 2008; Stukenberg and Burke, 2015). Two pieces of recent evidence support a model in which submaximal occupancy is sufficient for SAC silencing: MAD1 removal is initiated before a full occupancy state is reached (Kuhn and Dumont, 2017), and reduction of microtubule occupancy at kinetochores to ∼65% using a microtubule poison cannot prevent SAC silencing (Dudka et al., 2018). It is unclear, however, what occupancy state is sufficient for SAC silencing and how the different SAC modules respond leading up to these states. Here, we address these questions by quantitative correlation imaging of SAC protein levels and microtubule occupancy at single kinetochores, and by assessing SAC activity and SAC protein levels on kinetochores with experimentally manipulated average occupancy states. Our results allow a comprehensive view of the interaction between core kinetochore proteins, SAC signalling proteins and microtubules, and how they affect mitotic exit.

Kinetochores occupied by a full complement of microtubules have decreased or diminished levels of SAC proteins. The removal dynamics of key SAC proteins in response to increased microtubule occupancy is, however, not known. To address this, we wished to simultaneously quantify the relative amounts of SAC proteins and microtubules on individual kinetochores. We developed a method that allows accurate measurements of SAC protein and tubulin signal intensities on individual kinetochores with immature k-fibres (Fig. S1A–C). MG-132 and high doses of nocodazole were used to either allow full occupancy of kinetochores (‘FULL’ condition) or relieve all kinetochore–microtubule attachments (‘NULL’ condition), respectively. To create intermediate attachments [‘VAR’ (variable) condition], cells were fixed in prometaphase after release from a G2/M-boundary block (see Materials and Methods for details). The resulting population of kinetochores had a mixture of attachment states (Fig. 1A), including unattached and fully attached, as evident from comparisons with simultaneous imaging of kinetochores from the NULL and FULL conditions (Fig. 1A–H). We next quantified levels of six SAC proteins (MAD1, MAD2, ZW10, Spindly, BUBR1 and BUB1) and one SAC-regulating post-translation modification [KNL1-pT180, hereafter referred to as ‘pMELT’ (Vleugel et al., 2015)] in all three attachment conditions on individual kinetochores, to examine SAC protein behaviour in response to microtubule attachment (Fig. 1B–H). We found that the levels of all SAC proteins were substantially reduced on kinetochores when microtubule occupancy reached ∼30% or more (relative to mean tubulin intensity in the FULL condition, which has a normalized value of 1). However, whereas most kinetochores had no or barely detectable MAD1, MAD2, ZW10 and Spindly (‘RZZS-MAD1-MAD2’ group) at ∼50% of mean maximum occupancy, members of the ‘BUB1-BUBR1’ group (pMELT, BUBR1, BUB1) remained clearly detectable and showed retention of 29–53% of the median of the levels measured on unattached kinetochores (Fig. 1B–H; insets). This is in good agreement with previous observations of SAC protein behaviour on metaphase vs prometaphase kinetochores (Bomont et al., 2005; Hoffman et al., 2001; Howell et al., 2004; Martinez-Exposito et al., 1999; Skoufias et al., 2001). Detectability of the BUB1-BUBR1 group may be due to residual MPS1 on metaphase kinetochores (Hiruma et al., 2015; Howell et al., 2004), and/or the contribution of PLK1 to BUB protein recruitment (Espeut et al., 2015; von Schubert et al., 2015). To ensure that the observed differences between the two groups did not have a technical origin, we calculated signal-to-noise ratios for the NULL and FULL conditions and found that they are similar for all proteins in our panel, supporting a biological origin of this pattern (Fig. S1D,E). In addition, the results from quantitative immuno-imaging were verified using genome-edited cell lines that express N-terminal HA-mCherry-tagged versions of MAD2 or BUB1, as representative of their groups, from their endogenous locus, excluding differences between antibodies used and/or staining efficiency as a cause for differences between SAC protein behaviour at kinetochores (Fig. S2A–F). These observations support a model in which SAC proteins respond to intermediate attachments.

Although all SAC proteins showed the expected reduction on kinetochores in cells treated with MG-132 compared to those treated with nocodazole (Howell et al., 2004) (Fig. 1B–H, whisker plots), there was substantial variation in the amount of accumulated SAC proteins on unattached kinetochores (i.e. kinetochores with very high and very low SAC protein levels) (Fig. 1B–H, insets). Variation was also observed within single cells and did not correspond to possible inter-kinetochore variation of more stable kinetochore components such as CENP-C (Fig. S3A,B) or HEC1 (also known as NDC80) (Collin et al., 2013). Inter-kinetochore variation was maintained in metaphase for the group of proteins showing substantial retention at this stage (BUB1-BUBR1 group), but not for the RZZS-MAD1-MAD2 group (Fig. 1B–H; insets Fig. S3C, x-axis). The cause of this variation is unclear, but one can envision an inability of phosphatases to efficiently shut down MPS1 signalling or, for example, the existence of dynamic occupancy states at metaphase to which MPS1 kinase activity is highly sensitive. Pairwise comparison of the inter-kinetochore variation of all SAC proteins in the FULL condition resulted in clustering of proteins displaying high or low variation into two distinct groups that corresponded to the RZZS-MAD1-MAD2 and BUB1-BUBR1 groups (Fig. S3D). In addition, a fraction of attached kinetochores had accumulated as much BUBR1, BUB1 and pMELT as some of their unattached counterparts (Auckland et al., 2017), which was never observed for MAD2, MAD1 and Spindly, and only to a limited extent for ZW10 (Fig. S3E). Signal-to-noise ratios in the FULL condition were similar for proteins in the BUB1-BUBR1 group, supporting a biological source for our observations (Fig. S1E). Of note, variability between kinetochores was independent of the method used for measuring local protein levels (Fig. S1A–C), of differences between biological replicates (Fig. S3F), of different antibody penetration proficiencies per cell (Fig. S3G,H), or of kinetochore size (Fig. S3I).

To further explore the different behaviour of SAC proteins in response to intermediate attachments, we performed segmented linear regression on the data presented in Fig. 1 (Fig. 2A), and hierarchical clustering of various extracted mathematical features of their response to different microtubule occupancy states (features and clustering shown in Fig. 2B,C). We chose features that are biologically relevant, such as the difference between NULL and FULL (Null2Full decrease) and the level of tubulin intensity when the minimum levels of SAC protein are reached (Breakpoint) (Fig. 2B). Interestingly, occupancy response curves separated into two clusters after hierarchical clustering: one containing the profiles of MAD1, MAD2, Spindly and ZW10 and the other containing those of BUB1, BUBR1 and pMELT (Fig. 2C). This was consistent with the behaviour of these proteins with regards to full attachment (Fig. S3D) and suggests a mechanistic difference in their response to microtubules. Interestingly, MAD1 and BUB1 cluster as outliers in their groups, even though they are interaction partners of MAD2 and BUBR1, respectively. These differences in behaviour might be attributed to functions or localization mechanisms that do not rely on those interaction partners (Akera et al., 2015; Emre et al., 2011; Zhang and Nilsson, 2018; Zhang et al., 2016).

The clustering analysis predicted that members of the RZZS-MAD1-MAD2 module should behave differently in response to identical occupancy than members of BUB1-BUBR1 module when measured on the same kinetochore. Indeed, (partly) attached kinetochores with no detectable mCherry–MAD2 had a variety of BUB1 levels, and BUB1 displayed greater variability than MAD2 (Fig. 2D). In contrast, the signal intensities of BUB1 and BUBR1 on the same kinetochores of prometaphase cells were strongly correlated (Fig. 2E).

To examine whether response of the two SAC modules to intermediate attachment states indeed reflected underlying mechanistic differences, we inhibited MPS1 chemically in nocodazole-treated cells and analysed the levels of MAD1 and BUB1 as representatives of their groups. As expected, BUB1 levels declined to metaphase levels after MPS1 inhibition (Fig. 3A,B). MAD1 levels, however, remained largely unaffected, showing that MPS1 inhibition was insufficient to cause removal of MAD1 in the absence of microtubules, even though its recruitment has a substantial MPS1-dependent component (Hewitt et al., 2010; Ji et al., 2017; London et al., 2012). Removal of the two SAC modules from kinetochores is thus guided by different mechanisms.

To understand what extent of microtubule occupancy is sufficient to silence the SAC, we next wished to experimentally tweak mean maximal occupancy. HEC1 is the major microtubule-binding protein on the kinetochore (DeLuca et al., 2002, 2005; Liu et al., 2006; McCleland et al., 2003, 2004; Wigge and Kilmartin, 2001) and its affinity for microtubules is controlled by phosphorylation of its N-terminal tail (Cheeseman et al., 2006; Wei et al., 2007). Designed combinations of phospho-site substitutions to phosphomimetic or non-phosphorylatable amino acids (aspartic acid and alanine, respectively) generates HEC1 versions with a variety of microtubule-binding affinities, which in cells results in a controlled range of mean k-fibre intensities (Zaytsev et al., 2014). We constructed cell lines expressing mutant versions of HEC1 to achieve a range of occupancy states (Fig. 4A; Fig. S4A–D). Six of these mutants (HEC1-9A, -9D, -1D, -2D, -3D and -4D) have previously been characterized extensively with regards to their microtubule affinity (Zaytsev et al., 2014, 2015); others were generated and analysed by us. Considering that single microtubules can be bound by many HEC1 molecules, our approach enabled creation of uniform HEC1 lawns with specified microtubule-binding affinities, unlike, for example, diminishing the total amount of HEC1 on kinetochores (DeLuca et al., 2002, 2003), mixing high and low affinity HEC1 species or changing microtubule dynamics artificially, which might affect other processes as well (Dudka et al., 2018). Moreover, the HEC1 mutants simulate the phosphorylation states of kinetochores during unperturbed mitosis (Zaytsev et al., 2014), providing insight into the SAC response during k-fibre maturation.

Cells expressing the HEC1 variants were analysed using time lapse imaging for their ability to silence the SAC: occupancy states that cannot silence the SAC are predicted to delay mitosis indefinitely, while those that can should allow progression. As shown in Fig. 4B,C and reported before (Etemad et al., 2015; Guimaraes et al., 2008; Zaytsev et al., 2014), cells expressing wild-type HEC1 or HEC1-9A (high microtubule affinity) were able to silence the SAC, whereas those expressing HEC1-9D (low microtubule affinity) were not. While HEC1-5D was likewise unable to silence the SAC, HEC1-4D (intermediate microtubule affinity) was proficient in SAC silencing, albeit relatively inefficiently (Fig. 4B,C). As expected, HEC1-6D, -7D and -8D were unable to silence the SAC (data not shown). Quantitative immunofluorescence showed that HEC1-4D k-fibres were on average 45% of mean maximal intensity of wild-type HEC1 cells, in line with a previous report (Zaytsev et al., 2014), and those of HEC1-5D were ∼20% (Fig. 4D,E; Fig. S1). Simultaneous live imaging of mCherry–MAD2 and eGFP–HEC1-4D showed that HEC1-4D kinetochores had shed most or all of the MAD2 by 30 min following mitotic entry (Fig. 4F). Some kinetochores, however, had retained substantial MAD2 levels after 93 min, explaining why mitotic exit was relatively inefficient in these cells compared to controls. Quantitative immuno-imaging of single attached kinetochores showed that 82% of HEC1-4D kinetochores had MAD2 levels that were as low as those of wild-type HEC1 kinetochores (Fig. 4G,H). For comparison, this was true for 30% and 0% of HEC1-5D and HEC1-9D kinetochores, respectively. These data support the hypothesis that kinetochores can inactivate SAC signalling at intermediate (∼45%) occupancy states and that SAC silencing becomes more efficient with increasing occupancy.

Each human kinetochore consists of hundreds of microtubule-binding complexes that can each recruit SAC proteins. During metaphase these kinetochores are bound by ∼20 microtubules, which shuts down SAC signalling (DeLuca et al., 2005; Guimaraes et al., 2008; McEwen et al., 2001; Wendell et al., 1993). Kinetochores are unlikely to transition from zero to a full complement of microtubules in a single step, yet there is little knowledge about SAC responses to intermediate microtubule occupancies. We show here that key SAC proteins are substantially depleted from kinetochores at ∼30% occupancy and are nearly undetectable at ∼50% occupancy or above. Our quantitative immuno-imaging of SAC protein levels in relation to microtubule intensities on single kinetochores distinguished two response types. Levels of ZW10, Spindly, MAD1 and MAD2 inversely correlated to microtubule intensities and became not or barely detectable at ∼50% occupancy. BUBR1, BUB1 and pMELT, however, although also declining strongly at low occupancy, were not sensitive to further increases in occupancy and showed variable levels. The behaviours of the members of these two groups are consistent with their mutual physical interactions, and correlate with the distinct delocalization mechanisms that have been proposed for these groups. Removal of the RZZS-MAD1-MAD2 group occurs through dynein motor activity (Caldas et al., 2015; Kim et al., 2012; London and Biggins, 2014; Matson and Stukenberg, 2014; Silió et al., 2015) and molecular inhibition of this mechanism does not affect BUB removal (Gassmann et al., 2010). By contrast, removal of the BUB1-BUBR1 group requires dephosphorylation by PP1 and decreased localization and activity of MPS1 (Etemad and Kops, 2016; Hiruma et al., 2015; Ji et al., 2015; Nijenhuis et al., 2014). Here, we show that inhibition of MPS1 activity in cells with no microtubules decreases BUB1 levels on unattached kinetochores, while leaving MAD1 levels mostly unaffected, supporting probable independent removal mechanisms for these SAC signalling proteins. These different responses raise the question of what attachment features (i.e. microtubule stability, tension, lateral vs end-on interaction, etc.) are recognized by either signalling module. Interestingly, MAD1 also interacts with BUB1 in an MPS1-dependent manner (Ji et al., 2017; London and Biggins, 2014; Musacchio and Salmon, 2007), and our data suggest that this interaction is either not the main MAD1-localizing mechanism in prometaphase or is insensitive to MPS1-counteracting phosphatases. Finally, although MAD1 and MAD2 form a heterotetramer, their behaviour in our analyses is not entirely overlapping. The molecular basis for this is unknown, but MAD2-independent functions for MAD1 at kinetochores have been reported (Akera et al., 2015; Emre et al., 2011), and a pool of MAD2 depends on kinetochore-localized CDC20 (Zhang and Nilsson, 2018).

Cells in which kinetochores reach ∼45% occupancy on average (HEC1-4D) can silence the SAC and exit mitosis, while those with ∼20% occupancy (HEC1-5D) cannot. These data show that the full complement of microtubules seen on metaphase kinetochores is not strictly required for SAC silencing. The kinetochore, therefore, acts as a single unit with respect to SAC signalling; when a threshold of bound microtubules is reached, the entire unit switches off its signalling output. This has important implications for our understanding of the SAC as it suggests that the signal from hundreds of microtubule-binding complexes is quenched by only a few (∼7–10) microtubules. We envision several ways in which this can be achieved. Firstly, a few microtubules may be sufficient to pull a stiff kinetochore away from a SAC activating signal (e.g. Aurora B) originating from an inner centromere or kinetochore (Burke and Stukenberg, 2008; Santaguida and Musacchio, 2009; Saurin et al., 2011; Stukenberg and Burke, 2015). We do not favour this hypothesis, as we and others recently showed that distance between sister kinetochores or between inner and outer kinetochore is not required for SAC silencing. Secondly, a low number of microtubules may suffice to elicit a signal that sweeps the kinetochore. For example, phosphatases such as PP1 could be ‘unleashed’ from a site of recruitment or activation upon a threshold of microtubule binding. Concurrent with sufficient MPS1 displacement, this could switch the SAC signal to an ‘off’ state. It is unclear, however, how dynein-mediated removal of RZZS-MAD1-MAD2 proteins would occur in such a scenario. Thirdly, the kinetochore may be flexible, allowing only a few microtubules to engage the majority of microtubule-binding complexes and thus displace sufficient MPS1 molecules and recruit sufficient PP1 and dynein molecules to achieve substantial SAC protein delocalization. Transition to full occupancy may then be facilitated by kinetochore flexibility and many low-affinity microtubule interactions (Etemad and Kops, 2016; Hiruma et al., 2015). Fourthly, attachments may be highly dynamic, engaging and disengaging kinetochores frequently. This may allow most of the microtubule-binding complexes to briefly bind microtubules and shed SAC proteins. A sufficiently high frequency of these labile interactions could conceivably convert the kinetochore to a SAC-silenced state.

HeLa and HeLa FlpIn cells (gift from the Stephen Taylor lab, University of Manchester, UK) were grown in DMEM (Sigma-Aldrich; 4.5 g glucose/l) supplemented with 8% tetracycline-free FBS (Bodingo), penicillin/streptomycin (Sigma-Aldrich; 50 μg/ml), GlutaMAX (Gibco; 5 ml), and hygromycin (200 μg/ml) or puromycin (1.6 μg/ml). Cell lines were tested frequently for contaminations. Plasmids were transfected using Fugene HD (Roche) according to the manufacturer's instructions. To generate stably integrated constructs, HeLa FlpIn cell lines were transfected with pCDNA5-constructs and pOG44 recombinase simultaneously in a 1:9 ratio (Klebig et al., 2009). Constructs were expressed by addition of 1 μg/ml doxycycline for 24 h. siHEC1 (custom; Thermo Fisher Scientific; 5′-CCCUGGGUCGUGUCAGGAA-3′) and siGAPDH (Thermo Fisher Scientific; D-001830-01-50) was transfected using HiPerfect (Qiagen) according to manufacturer's instructions.

pCDNA5-pEGFP-HEC1 constructs and cloning strategies are described in Nijenhuis et al. (2013). Other constructs (the 1D, 2D, 3D, 4D and 5D mutants) were made using site-directed mutagenesis by PCR.

Inserting the gene for mCherry into the endogenous loci of MAD2 and BUB1 was performed using a self-cloning CRISPR strategy (Arbab et al., 2015). In brief: 3×FLAG-spCas9 was subcloned from spCas9-BLAST to pcDNA3-MCS-IRES-PURO using NdeI/EcoRI restriction digestion to allow selection for spCas9 expression in HeLa FLPin cells. To generate HA-mCherry, pcDNA3Zeo-CyclinB-mCherry (Kuijt et al., 2014) was used as template and amplified by PCR using forward (5′-AAGCTTTACCCGTACGACGTGCCAGATTACGCTGTGAGCAAGGGCGAGGAGG-3′) and reverse (5′-gcgccgTCTAGATCCGCAGCCACCGCCAGATCCGCCCTTGTACAGCTCGTCCATGC-3′) primers. The PCR fragment was digested with HindIII/XbaI and ligated into pcDNA3.0 via HindIII/XbaI.

To create the homology arms, three consecutive PCR-amplifications were done on HA-mCherry template to create MAD2 (120 bp 5′ of ATG and 120 bp 3′) or BUB1 (119 bp 5′ of ATG and 124 bp 3′). spCas9 was directed to MAD2 using sgRNA: 5′-GAGCTGCAGCGCCATGGCC-3′, or BUB1 using 5′-GTCCTCTGGCCATGGACACCC-3′. Both homology arms and sgRNA template were subcloned into pJet1.2 (Thermo Fisher Scientific) and sequence verified before PCR fragments were generated for transfection.

To generate HeLa FlpIn cells expressing endogenous tagged MAD2 or BUB1, cells were transfected with 1.5 µg spCas9-IRES-PURO, 1.5 µg sgPAL7-Hygro, 3 µg homology PCR template and 3 µg sgRNA PCR template at a ratio 1:3 DNA:Lipofectamine LTX (Thermo Fisher Scientific). 24 h after transfection, 1 µg/ml puromycin and 200 µg/ml hygromycin B were used for 48 h, after which cells were grown until confluence in a 10 cm Petri dish. HeLa FlpIn cells were subsequently FACS-sorted as single cells using a BD FACSAria FUSION (640 nm excitation laser, autofluorescence 670 nm/30 vs 651 nm excitation laser, 610 nm/20 mCherry channels, 100 µm nozzle, 2.0 flowrate). Clones were verified to have correct labelling of MAD2 or BUB1 by PCR on genomic DNA, western blotting and live-cell immunofluorescence microscopy.

To knock down and reconstitute HEC1 in HeLa FLpIn cell lines, cells were transfected with 40 nM siRNA against HEC1 or mock siRNA and arrested in early S phase for 24 h by addition of thymidine (2 mM). Cells were then allowed cell cycle re-entry by washing the cells once with appropriate media. 8 h after thymidine release, cells were treated with doxycycline (1 μg/ml), arrested again using thymidine and incubated with both reagents for 16 h after which they were released from thymidine and further processed. Cells processed for immunofluorescence imaging of SAC proteins and k-fibres were released from thymidine in CDK1 inhibitor RO-3306 (5 μM) and incubated for 8 h or more. Subsequently, cells were washed three times with warm media, incubated between each wash for 5 min at 37°C, and incubated for 120 min with nocodazole (3.3 μM, ‘NULL’ condition), or MG-132 (5 μM, ‘FULL’ condition). Next, cells were fixed and processed appropriately. To fix cells before all kinetochores had reached full occupancy (the ‘VAR’ condition), cells were fixed and processed 25 min after release from a G2/M-boundary block by RO-3306 (5 μM). For immunofluorescence imaging of cells expressing HEC1 variants, cells were treated with MG-132 (5 μM) for 120 min prior to fixation. To follow BUB1 and MAD1 levels in time after MPS1 inhibition, cells were synchronized with thymidine and blocked at the G2/M-boundary using RO-3306 (5 μM). Cells were then released from RO-3306 inhibition using nocodazole (3,3 μM). After 1 h nocodazole block, cells were treated with MG-132 for 15 min, after which Cpd5 (250 nM) was added to inhibit MPS1 activity.

For live-cell imaging experiments, cells were plated in 24-well plates (Greiner Bio-One), and subjected to DIC microscopy on a Nikon Ti-E motorized microscope equipped with a Zyla 4.2 Mpx sCMOS camera (Andor). A 20×0.45 NA objective lens (Nikon) was used. Cells were kept at 37°C and 5% CO₂ using a cage incubator and Boldline temperature/CO₂ controller (OKO-Lab). Images were acquired every 4 min at 2×2 binning and processed by Nikon Imaging Software (NIS). Analysis of live-cell imaging experiments was carried out with ImageJ software and time in mitosis was defined as the time between nuclear envelope breakdown and anaphase onset or cell flattening.

Live-cell imaging of mCherry-tagged MAD2 and BUB1 in single cells was performed on a Nikon Time-Lapse system (Applied Precision/GE Healthcare) equipped with a Coolsnap HQ2 CCD camera (Photometrics) and Insight solid-state illumination (Applied Precision/GE Healthcare). Cells were plated in 8-well plates (μ-Slide 8-well, Ibidi) and imaged in a heated chamber (37°C and 5% CO₂) using a 60×/1.42 NA or 100×/1.4 NA UPlanSApo objective (Olympus) at 2×2 binning. Images were acquired every 15 s (for the mCherry–MAD2 cells), or 1 min (for the mCherry–BUB1 cells), and deconvolved using standard settings in SoftWorx (Applied Precision/GE Healthcare) software. Multiple z-layers were acquired and projected to a single layer by maximum intensity projection. For simultaneous imaging of GFP (HEC1) and mCherry–MAD2, the same system was used. Cells were plated in 8-well plates (μ-Slide 8-well, Ibidi), treated with siRNA, thymidine and RO-3306 as described above. Images were acquired 30 and 60 min after mitotic entry, and then every three minutes.

For fixed-cell immunofluorescence microscopy, cells plated on round 12-mm coverslips (no. 1.5) were pre-extracted with 37°C 0.1% Triton X-100 in PEM (100 mM PIPES pH 6.8, 1 mM MgCl₂, 5 mM EGTA) for ∼45 s before fixation (with 4% paraformaldehyde) for 10 min. Coverslips were washed twice with cold PBS and blocked with 3% BSA in PBS for 16 h at 4°C, incubated with primary antibodies for 16 h at 4°C, washed four times with PBS containing 0.1% Triton X-100, and incubated with secondary antibodies for an additional hour at room temperature. Coverslips were then washed twice with PBS/0.1% Triton X-100, incubated with DAPI for 2 min, washed again twice with PBS and mounted using Prolong Gold antifade (Invitrogen). For cold-shock experiments, cells were placed on ice water in 500 μl media for 8 min prior to pre-extraction and fixation with the appropriate buffers.

All images were acquired on a deconvolution system (DeltaVision Elite; Applied Precision/GE Healthcare) with a 100×/1.40 NA UPlanSApo objective (Olympus) using SoftWorx 6.0 software (Applied Precision/GE Healthcare). Deconvolution is applied to all images and maximum projection is shown in figures, except for Fig. S1A, which is a sum-projection image, and Fig. 4F in which single planes are shown. For quantification of immunostaining, all images of simultaneously stained experiments were acquired with identical illumination settings. For analysis of the HEC1 mutant-expressing cell lines, cells expressing comparable levels of exogenous protein were selected for analysis and analysed using ImageJ. For measurement of protein levels and k-fibre intensities on single kinetochores, kinetochores were selected in maximum-projection images. The 7–8 slices that contained a single kinetochore and corresponding k-fibre were selected and sum-projection images were used for quantification. Line plots were used to determine the highest intensity at kinetochores and/or k-fibres and local background was subtracted from these values (Fig. S1). The same method was applied to determine protein levels, k-fibre intensity and CENP-C levels. K-fibre and protein measurements were normalized to CENP-C to correct for biological and technical variation between kinetochores. Further normalization steps included normalization of k-fibre levels to the median levels measured in FULL conditions, and normalization of protein levels to the levels measured for the median of the same protein in the NULL conditions.

Sample size for live imaging and immunofluorescence experiments was chosen based on common practice in the field

Data analysis was performed in R (3.3.2) using the pheatmap (1.0.8, CRAN), MarkdownReports (2.5, doi:10.5281/zenodo.594683) packages. Raw measurements per kinetochore were normalized as described in ‘Immunofluorescence and image quantification’.

To quantify features of individual occupancy response curves in transient (left) and steady (right) phase separately, piecewise linear regression was applied, where a breakpoint separates the two phases. Each feature is extracted from either the FULL, NULL or the VAR datasets as denoted. These features are: VarianceInPhase2 (variance in protein concentration in the steady phase, right of the split point), MedianPhase2 (median protein concentration in the steady phase), Null2Full decrease (relative protein decrease between the two conditions defined as the ratio of median protein levels NULL over FULL attachment, corresponds to data presented in insets in Fig. 1B–H), MedianFull (median protein levels in the full condition), Breakpoint (x- or tubulin-coordinate of the split point in the piecewise linear regression), and Slope (slope of the fitted line in the transient phase, left of the split point).

To investigate variance across proteins at full attachment, measured values were tested for normality. Based on these results, Levene's test was used to compare variances.

Cells were treated as described above and entered mitosis in the presence of nocodazole. Cells were collected and lysed in Laemmli lysis buffer (4% SDS, 120 mM Tris pH 6.8, 20% glycerol). Lysates were processed for SDS-PAGE and transferred to nitrocellulose membranes for immunoblotting. Immunoblotting was performed using standard protocols. Visualization of signals was performed on a scanner (Amersham Imager 600) using enhanced chemiluminescense.

The following primary antibodies were used for immunofluorescence imaging: CENP-C (guinea pig polyclonal, 1:2000; Sigma-Aldrich, PD 030), α-tubulin (mouse monoclonal, 1:10,000; Sigma-Aldrich, T5168), HEC1 (mouse monoclonal 9G3, 1:2000; Abcam, Ab-3613), GFP (custom rabbit polyclonal raised against full-length GFP as antigen, 1:10,000; Jelluma et al., 2008), GFP (mouse monoclonal, 1:1000; Roche, 12-814-460-001), MAD2 (custom rabbit polyclonal raised against full-length 6×His-tagged MAD2 as antigen, 1:2000; Sliedrecht et al., 2010), BubR1 (rabbit polyclonal, 1:1000; Bethyl, A300-386 A), BUB1 (rabbit polyclonal, 1:1000; Bethyl, A300-373 A-1), Spindly (rabbit polyclonal, 1:1000; Bethyl, A301-354A), ZW10 (rabbit polyclonal, 1:1000; Abcam, ab21582), MAD1 (rabbit polyclonal, 1:1000; Santa Cruz, cat. no. sc-67337), RFP (rat monoclonal, 1:1000; Chromotek, 5F8) GFP-Booster (Atto 488, 1:500; Chromotek, gba488). Secondary antibodies (Invitrogen Molecular Probes, all used at 1:600) were highly cross-absorbed goat anti-guinea pig Alexa Fluor 647 (A21245), anti-rat Alexa Fluor 568 (A11077), goat anti-rabbit Alexa Fluor 488 (A11034) and 568 (A11036), and anti-mouse Alexa Fluor 488 (A11029) and 568 (A11031).

The following primary antibodies were used for immunoblotting: HEC1 (mouse monoclonal, 1:1000; Abcam, ab3613), α-tubulin (mouse monoclonal, 1:10.000; Sigma, T5168), mCherry (rabbit polyclonal, 1:1000; Abcam, ab167453), BUB1 (rabbit polyclonal, 1:1000; Bethyl, A300-373 A-1), MAD2 (custom rabbit polyclonal raised against full-length 6xHis-tagged MAD2 as antigen, 1:2000; Sliedrecht et al., 2010). Secondary antibodies were: goat anti-rabbit IgG, HRP-conjugate (1:10.000; Bioke, #7074) and goat anti-mouse IgG, HRP-conjugate (1:10.000; Bio-Rad, #1706516).

We thank Jennifer DeLuca (Colorado State University, Fort Collins, CO) for reagents, members of the Kops lab for discussions, and Sophie Dumont and Jonathan Kuhn (University of California, San Fransisco, CA) for sharing unpublished data.