Low kinetochore-microtubule occupancy leaves a mitotic memory by delaying checkpoint silencing

Abstract

The spindle assembly checkpoint (SAC) promotes faithful chromosome segregation by delaying mitosis until all kinetochores attach to spindle microtubules. However, unusually prolonged mitoses block daughter cell proliferation through a p53-dependent memory mechanism—the “mitotic stopwatch”—suggesting a selective pressure to avoid significant mitotic delays. Here, we show that microtubule occupancy at kinetochores is a cornerstone linking SAC silencing with mitotic duration and memory. By combining live-cell with super-resolution microscopy, photobleaching and laser microsurgery in Indian muntjac fibroblasts, we demonstrate that SAC silencing is gradual and confined to microtubule attachments within kinetochores. Augmin promotes timely SAC silencing with high microtubule occupancy at kinetochores, whereas MPS1/CDK1 inhibition silences the SAC irrespective of microtubule occupancy. Conversely, low microtubule occupancy delays SAC silencing, increases segregation errors and blocks daughter cell proliferation due to mitotic stopwatch surveillance. Thus, timely SAC silencing with high microtubule occupancy avoids “bad memories” of mitosis to allow daughter cell proliferation.

How do cells faithfully segregate chromosomes within an optimal time? Using naturally “super-resolved” kinetochores, the authors show that maximal microtubule occupancy prevents bad mitotic memories to ensure proliferation of the next generations.

Introduction

To promote faithful chromosome segregation during mitosis, the spindle assembly checkpoint (SAC) delays anaphase until all kinetochores attach to mitotic spindle microtubules¹. At the root of the SAC signaling cascade, a MAD1/MAD2 complex recruited to unattached kinetochores initiates a catalytic process that produces a diffusible Mitotic Checkpoint Complex (MCC) and inhibits the Anaphase Promoting Complex/Cyclosome (APC/C), an E3-ubiquitin ligase that targets Cyclin B1 and Securin for degradation by the proteasome. This releases the cysteine-protease Separase to cleave the Cohesin molecules that hold sister-chromatids together, thereby triggering their initial separation in anaphase (reviewed in refs. ^2–4).

While there is a large consensus that SAC silencing can occur with low microtubule occupancy at kinetochores^5–10, whether and how SAC signaling along vertebrate kinetochores responds locally or globally to microtubule attachments has remained a central open question. Current models suggest that mammalian kinetochores respond as a single unit that globally halts SAC signaling upon the attachment of just a few microtubules^7–9. As so, low microtubule occupancy would cause kinetochores to respond as a whole and predicts a uniform loss of MAD1 along the kinetochores. These switch-like models imply the existence of an elusive microtubule “counting” mechanism that determines SAC silencing when a given occupancy at each individual kinetochore is reached. Moreover, they remain difficult to reconcile with the finding that the rate of MAD1/MAD2 loss from kinetochores increases with microtubule occupancy, with SAC silencing under conditions of low microtubule occupancy resulting in extensive mitotic delays^7,8,10. Importantly, even ~70% microtubule occupancy at metaphase kinetochores, which is sufficient to silence the SAC with only minor mitotic delays, increases the incidence of lagging chromosomes during anaphase⁶. Thus, although cells appear to be able to silence the SAC with low microtubule occupancy, there may be selective pressure for high microtubule occupancy at kinetochores before initiating anaphase, ensuring proper chromosome segregation and avoiding significant mitotic delays^6,11. This hypothesis gains weight in light of the recently uncovered “mitotic stopwatch” complex formed by 53BP1 and the deubiquitinase USP28 that stabilizes p53 in response to even moderate delays in mitosis, leading to subsequent cell cycle arrest or death in G1, independently of DNA damage^12–19. However, how SAC silencing is coordinated with the mitotic stopwatch to control daughter cell proliferation remains unknown. Here we investigate how the SAC responds to microtubule attachments within kinetochores, while testing whether the extent of microtubule occupancy underlying SAC silencing impacts daughter cell proliferation.

Results

SAC silencing at Indian muntjac kinetochores is gradual

According to switch-like SAC-silencing models, kinetochores respond as a single unit to globally silence the SAC after the establishment of just a few microtubule attachments^7,8. However, because human kinetochores cannot be resolved by conventional light microscopy, this postulate has so far been impossible to test experimentally. To overcome this limitation, we took advantage of the unique cytological features of female Indian muntjac (Muntiacus muntjak) fibroblasts, which carry only six chromosomes with naturally “super-resolved” kinetochores that arose by tandem and centric fusions of smaller ancestral chromosomes^20–26. As a first step towards investigating SAC silencing in this system, we determined the time between MAD1 disappearance from kinetochores and anaphase onset by tracking individual kinetochores in live, hTERT-immortalized, female Indian muntjac fibroblasts stably expressing mScarlet-CENP-A (a constitutive kinetochore marker) and Venus-MAD1²⁷. We found that MAD1 signal at kinetochores became undetectable ~20 min before anaphase onset (Supplementary Fig. 1a, b), in agreement with previous reports in other vertebrate cell lines^28,29 and consistent with the temporal framework of SAC silencing upon laser-mediated ablation of the last unattached kinetochore¹.

To determine the kinetics of SAC silencing in response to microtubule attachments at individual kinetochores in live Indian muntjac fibroblasts, we combined stable expression of fluorescent Venus-MAD1 with carefully titrated SiR-tubulin to avoid mitotic delays resulting from spindle defects²⁴, under slight cell compression to improve visualization²⁵. As a control, cytoplasmic Venus-MAD1 signal was simultaneously measured, and photobleaching was found to be negligible throughout the course of the experiment (Supplementary Fig. 1c). MAD1 decayed gradually at individual kinetochores as chromosomes bi-oriented and aligned at the equator during prometaphase (Fig. 1a, c and Supplementary Movie 1), with a rate of decay that was well fit to a single exponential (Fig. 1e and Supplementary Fig. 2a), indicating that SAC silencing in this system faithfully replicates established kinetics in other vertebrate systems⁹. Noteworthy, pole-proximal chromosomes aligned last and did so tangentially along the curvature of the spindle edge. Moreover, their leading kinetochores were enriched with MAD1 from the onset of congression, with the signal gradually decreasing as chromosomes approached the equator (Fig. 1a-a” and Supplementary Movie 1). Importantly, MAD1 typically decayed first on trailing kinetochores, likely reflecting differences in the timing or efficiency in the formation of stable end-on microtubule attachments compared with leading kinetochores. (Fig. 1a-a” and Supplementary Movie 1). These results are fully consistent with congression models where pole-proximal, mono-oriented chromosomes use plus-end-directed motors on the leading kinetochore to initiate lateral gliding along pre-existing spindle microtubules towards the equator^9,30–32, while at odds with models where stabilization of end-on attachments and bi-orientation take place before pole-proximal chromosomes initiate congression³³.

Fig. 1. MAD1 is gradually removed from the kinetochore in a microtubule- and mitotic kinases-dependent manner.

a Representative live-cell recordings (n = 7 cells, 5 independent experiments) of Indian muntjac fibroblasts. Venus-MAD1 (orange) and SiR-tubulin labeled microtubules (gray). Scale bar: 5 µm. Insets: early congressing kinetochore pair (dashed square) and polar kinetochores/chromosomes (dashed circles), scale bar: 2 µm. a’ Insets of 1 half-spindle highlight the congression of polar chromosomes (dashed circles) (from 1 cell in (a)). Scale bar: 5 µm. a” Kymographs depicting Venus-MAD1 signal of the 2 polar chromosomes from (a). The closest pole (dashed line) and the leading (black asterisk) and trailing kinetochores (white asterisk) are indicated. Scale bar: 5 µm. b Representative live-cell recordings of Indian muntjac fibroblasts treated with nocodazole (n = 8 cells, 5 independent experiments), MPS1 (n = 12 cells, 5 independent experiments) and CDK1 (n = 10 cells, 6 independent experiments) inhibitors. Venus-MAD1 (orange) and SiR-tubulin labeled microtubules (gray). Scale bar: 5 µm. Insets: kinetochore pair, scale bar: 2 µm. c Normalized Venus-MAD1 fluorescence decay in Indian muntjac cells control or treated with nocodazole, MPS1 and CDK1 inhibitors. t = 0 s corresponds to the peak of MAD1 fluorescence before it starts to decay in control cells; for drug-treated cells, t = 0 s corresponds to the average first 10 frames after nocodazole or the first frame after MPS1i/CDK1i treatment. Data are mean ±95% confidence interval (CI; Control n = 53 kinetochores, 7 cells; nocodazole n = 31 kinetochores, 8 cells; MPS1i n = 63 kinetochores, 12 cells; CDK1i n = 18 kinetochores, 10 cells; minimum five independent experiments per condition). d Normalized Venus-MAD1 fluorescence decay in large or small kinetochores from control cells. Data are mean ±95% CI (Large KTs n = 19 kinetochores, 7 cells, 5 independent experiments; small KTs n = 34 kinetochores, 5 cells, 4 independent experiments). e Parameters extracted from a single exponential fitting of Venus-MAD1 fluorescence decay across the condition shown in (c) and (d). Control large vs small kinetochores, p = 8.7 × 10⁻⁵; control vs MPS1i, p < 1 × 10⁻¹⁵; control vs CDK1i, p = 0.0018, extra sum-of-squares F-test.

To test whether gradual MAD1 decay at kinetochores during mitosis depends on microtubule occupancy, we monitored MAD1 levels in the presence of 1 µM nocodazole, which depolymerized all spindle microtubules. As expected, under these conditions, MAD1 localized on every kinetochore without any measurable decay for over 20 min (Fig. 1b, c, Supplementary Fig. 2a and Supplementary Movie 2). These data indicate that gradual SAC silencing during mitosis depends on the establishment of microtubule attachments at kinetochores.

Microtubule-independent SAC silencing upon MPS1/CDK1 inhibition

MPS1 and CDK1 kinases are critical to maintain SAC signaling by counteracting the activities of PP1 and PP2A phosphatases required for mitotic exit^34–39. Therefore, we next investigated the kinetics of MAD1 decay at kinetochores after acute MPS1 or CDK1 inactivation with MPS1-IN-1⁴⁰ or RO 3306⁴¹, respectively. We found that MAD1 decayed more abruptly from kinetochores after MPS1 or CDK1 inhibition, regardless of microtubule occupancy, reaching 50% of the initial levels 3.8× or 1.5× faster than during normal mitosis, respectively (Fig. 1b, c, e, Supplementary Fig. 2a and Supplementary Movies 3, 4). These data indicate that inactivation of MPS1 or CDK1 overrides the requirement of a given microtubule occupancy at kinetochores to silence the SAC and is consistent with recent findings suggesting that substantial loss of MPS1 activity correlates with SAC silencing⁴².

SAC silencing depends on kinetochore size

A prediction from switch-like models that favor independent microtubule binding at kinetochores and where a given percentage of occupancy is key to uniformly silence the SAC^7,8 is that small or large kinetochores would simultaneously reach the same percentage of occupancy (and, consequently, have similar MAD1 half reduction times), despite binding a different total number of microtubules. However, despite ample evidence of kinetochore size differences in humans^43–51, this prediction is currently impossible to test experimentally in human cells for the same reasons evoked before. In contrast, the large kinetochore from chromosome 3 + X in female Indian muntjac cells can be unequivocally distinguished from those on the smaller chromosomes 1 and 2, allowing us to track MAD1 decay as a function of kinetochore size^21,23,24. Surprisingly, we found that MAD1 signal on the large kinetochores of chromosome 3 + X requires ~38% more time to decay, when compared with smaller kinetochores on chromosomes 1 and 2 (Fig. 1d, e, and Supplementary Fig. 2a). These findings suggest that switch-like models do not fully account for how the SAC is normally silenced.

The stable MAD1 pool does not diffuse within the kinetochore

In order to obtain a deeper understanding of the mechanisms underlying SAC silencing we investigated MAD1 dynamics within kinetochores. Two possible scenarios could be envisioned: 1) if MAD1 diffuses within kinetochores, few microtubule attachments might be sufficient to “extinguish” all MAD1, for example by Dynein-mediated stripping along microtubules^52,53 (Fig. 2a); 2) if MAD1 does not diffuse within kinetochores, Dynein-mediated stripping along microtubules would be confined to the sites or domains where microtubules are attached (Fig. 2a). To distinguish between these two scenarios, we photobleached Venus-MAD1 on a fraction (up to 50%) of the large kinetochore from chromosome 3 + X in the absence of microtubules. Clear predictions about MAD1 diffusion within kinetochores could be made, depending on the Fluorescence Recovery After Photobleaching (FRAP) profile. If MAD1 diffuses within kinetochores, the photobleached fraction of the kinetochore is expected to recover due to migration of fluorescent Venus-MAD1 molecules from the unbleached fraction, resulting in the convergence of signals to an intermediate and uniform level along the entire kinetochore (Fig. 2a’). In contrast, if MAD1 does not diffuse within kinetochores, the photobleached fraction of the kinetochore would be expected to only recover fluorescent Venus-MAD1 molecules that are exchanging with the cytoplasmic pool, while the unbleached fraction would remain stable over time (Fig. 2a’). We were successful in partially photobleaching Venus-MAD1 in a fraction of the kinetochore, with minimal unintended photobleaching of Venus-MAD1 on the remaining fraction (Fig. 2b-b” and Supplementary Movie 5). Importantly, after photobleaching of the intended fraction of the kinetochore, Venus-MAD1 fluorescence recovered up to ~50%, with a half recovery time of 1–7 s, while remaining stable (or increasing slightly; see ahead) over time in the unbleached fraction (Fig. 2b-b”, d, e and Supplementary Movie 5). As a control, we photobleached Venus-MAD1 on entire kinetochores, which resulted in identical recovery parameters when compared to partial photobleaching, and in line with previous measurements in other mammalian systems^28,29 (Fig. 2c–e and Supplementary Movie 6). Because fluorescence recovery of the mobile fraction of Venus-MAD1 after photobleaching entire kinetochores is exclusively due to exchange with the cytoplasmic pool^28,29, we concluded that this pool also accounts for the observed fluorescence recovery of Venus-MAD1 on partially photobleached kinetochores. Taken together, these data indicate that the stable pool of MAD1 does not diffuse within kinetochores, suggesting that it can only be removed from spatially confined regions upon the establishment of microtubule attachments or by a microtubule-independent SAC-silencing mechanism, such as the one involving CDK1 or MPS1 inactivation.

Fig. 2. MAD1 does not diffuse within kinetochores.

a Schematic representation of Dynein-mediated MAD1/MAD2 complex removal from kinetochores. a’ Predicted curves of MAD1 fluorescence recovery after partial kinetochore bleaching assuming that MAD1 diffuses (left) or not (right) within the kinetochore. Lines represent the fluorescence levels from the region of the kinetochore that was bleached (light blue) or left unperturbed (dark blue). Representative images of Indian muntjac cells pre- and post-bleach of part (n = 7 cells, 3 independent experiment) (b) or an entire kinetochore (n = 8 cells, 3 independent experiment) (c). Insets: kinetochore pair that was perturbed. Dashed line boxes indicate the ablated kinetochore (partial bleach KT: non-bleached region - dark blue; bleached region – blue; total bleach KT - yellow). Scale bars: 2 µm. (b’), c’ Venus-Mad1 signal intensity profile along the kinetochore pre-bleach and at selected times post-bleach for kinetochores shown in (b) and (c), respectively. b”, c” Normalized Venus-Mad1 fluorescence recovery curves for cells shown in (b) and (c), respectively (dark blue line non-bleached region; light blue bleached region; yellow line totally bleached kinetochore; black and dark gray lines unperturbed kinetochores from the bleach or a control kinetochore pair, respectively; light gray cytoplasm; fit curves are shown by the semi-translucid line). d Kinetochore Venus-MAD1 fluorescence recovery after partial or total kinetochore bleach (Partial KT bleach n = 7 kinetochores from 7 cells; total KT bleach n = 8 kinetochores from 8 cells from 3 independent experiments). Dots show individual kinetochores, boxes show interquartile ranges, horizontal lines show median and error bars show min max. P = 0.2319, Mann–Whitney test (two-sided). e Half recovery times of kinetochore Venus-MAD1 fluorescence after partial or total kinetochore bleach. Dots show individual kinetochores, boxes show interquartile ranges, horizontal lines show median and error bars show min max. Samples as in (d). P = 0.5358, Mann–Whitney test (two-sided).

SAC silencing along kinetochores is confined to attachment sites

The conclusion from our FRAP experiments implies that, in a normal mitosis, there may be low microtubule occupancy states in which MAD1 removal is limited to regions where microtubule attachments are established. To test this prediction, we monitored Venus-MAD1 decay along the entire kinetochore length as microtubules attach during mitosis. Near-simultaneous tracking of individual kinetochores, MAD1 and microtubules in living cells revealed a non-uniform decay of MAD1 along kinetochores (Supplementary Fig. 2b, b’). To further dissect the relationship between non-uniform MAD1 decay and microtubule occupancy along kinetochores, we used super-resolution CH-STED microscopy⁵⁴ and 3D image rendering to inspect (fixed) late prometaphase cells with only partial MAD1 signal along kinetochores (Fig. 3a, b, Supplementary Fig. 3a, b and Supplementary Movie 7). We found clear examples where fractions of the kinetochore had end-on attached microtubules with no detectable MAD1, while other fractions were devoid of microtubules and clearly enriched for MAD1 (Fig. 3a, b, Supplementary Fig. 3a, b and Supplementary Movie 7). A line scanning profile along the entire length of several such individual kinetochores confirmed a negative correlation between MAD1 signal and microtubule density in 15/21 (71.4%) of the cases (Fig. 3c, d and Supplementary Fig. 3a’, b’).

Fig. 3. MAD1 removal along individual kinetochores correlates with local microtubule density.

a Representative CH-STED image (n = 17 cells, 7 independent experiments) of an Indian muntjac fibroblasts with partially attached kinetochores. Microtubules (gray; STED), centromeres (ACA; blue; STED) and Venus-MAD1 (orange; confocal). Insets: partially attached kinetochore. Scale bars: 2 µm. b Three-dimensional reconstruction of the cell shown in (a) (different perspective). Inset: partially attached kinetochore. The kinetochore pair of interest is traced by a dashed line. Microtubules (gray), centromeres (ACA; blue) and Venus-MAD1 (orange). Scale bars, left: 2 µm; right: 1 µm. c Max normalized Venus-Mad1 (orange) and tubulin (microtubules; gray) signal intensity profile along the longitudinal axis of the kinetochore highlighted in (a) (arrowhead). d Pearson correlation coefficient between MAD1 and microtubules signal intensity along the kinetochore. Bars show individual partially attached kinetochores (n = 21 kinetochores, 17 cells, 7 independent experiments, see also Supplementary Fig. 3c). e Illustrative/simulated distribution of kinetochore sub-regions showing positive, negative and no correlation between MAD1 (y-axis) and microtubules (x-axis) intensity z-scores. Dots represent distinct positions along the “kinetochore” shown below the graphs. NE and SW quadrants contribute positively, while NW and SE quadrants contribute negatively to the Pearson correlation coefficient r. f Correlation map of MAD1 and microtubule intensity z-scores for the kinetochores shown in (d). The underlying scatterplot contains data from multiple kinetochores (see also Supplementary Fig. 3c), with each dot representing a sub-kinetochore region. For visualization, the scatterplot was converted into a regional density map. P-value was calculated with Student’s t test for zero correlation (two-sided).

To improve the resolution of our analyses we divided the longitudinal axis of partially attached kinetochores into multiple subsections and quantified MAD1 and microtubule intensity at each subsection (Fig. 3e). We found a weak, yet highly significant, negative correlation between MAD1 and microtubule signal intensity along kinetochores (Fig. 3f and Supplementary Fig. 3c). Of note, the resulting L-shape curve indicates that while high microtubule signal is a good predictor of the absence of MAD1 at kinetochores, the opposite is not the case, suggesting that there are regions at kinetochores without microtubules that have no detectable MAD1, consistent with models evoking cooperative microtubule binding at kinetochores^55–57. Overall, these data show that SAC silencing at kinetochores is non-uniform and confined to microtubule attachment sites.

Localized SAC response to microtubule detachments

To directly test how the SAC responds to microtubule attachments/detachments, we used laser microsurgery to partially ablate individual metaphase kinetochores (and, consequently, microtubules that were attached to the ablated region) upon SAC silencing in live Indian muntjac fibroblasts (Fig. 4a). These cells stably expressed GFP-CENTRIN-1 to mark spindle pole positions, 2xGFP-CENP-A to define the kinetochore target, and mScarlet-MAD1 to monitor SAC status before and after surgery (Fig. 4b and Supplementary Movie 8). We found that, within few minutes after partial kinetochore ablation, the unperturbed sister kinetochore typically bent towards the attached spindle pole without any detectable MAD1 (Fig. 4a–c and Supplementary Movie 8). However, ~5 min after bending, MAD1 was often recruited to the tip of the bent unperturbed sister kinetochore, presumably due to local microtubule detachments (Fig. 4a–d and Supplementary Movie 8). In other cases, MAD1 was recruited to a more central region of the bent unperturbed sister kinetochore (Supplementary Fig. 4a, b). Strikingly, in all cases MAD1 accumulation was confined and distinct from the uniform accumulation of MAD1 on fully unattached kinetochores resulting from nocodazole treatment (Fig. 4b, d, e and Supplementary Fig. 4b). Therefore, local kinetochore perturbations result in a local SAC response.

Fig. 4. Localized SAC response inversely correlates with microtubule detachment.

a Schematic of kinetochore response to partial laser ablation. b Representative live-recordings of Indian muntjac cells upon partial kinetochore surgery or nocodazole treatment. 2x-GFP-CENP-A and GFP-CENTRIN-1 (both in blue) and mScarlet-MAD1 (orange/ gray). Insets: perturbed kinetochore pair pre- and post-surgery (solid and hollow arrow, respectively). Arrowheads show MAD1 recruitment. Scale bars: 2 µm. b’ Max normalized profile of mScarlet-MAD1/2x-GFP-CENP-A signal ratio along the longitudinal axis of the kinetochores highlighted in (a). c Time from surgery to sister kinetochore deformation (n = 12 kinetochores) and MAD1 signal re-appearance (n = 14 kinetochores) in cells with SAC reactivation. Dots represent individual kinetochores from different cells, (13 independent experiments). P < 0.0015, Wilcoxon matched-pairs signed rank test (two-sided). d Max normalized mScarlet-MAD1/2x-GFP-CENP-A ratio profile along longitudinal axis of kinetochores after partial surgery of sisters or upon nocodazole treatment (NOC). Cohort of cells recruiting MAD1 to kinetochore tip rather than center (n = 11/16 MAD1 recruiting kinetochores; see also Supplementary Fig. 4). Thin lines show individual kinetochores and thick lines show average (surgery n = 11 kinetochores, 11 cells, 10 independent experiments; nocodazole n = 19 kinetochores, 19 cells, 2 independent experiments). e Coefficient of variation of mScarlet-MAD1/2x-GFP-CENP-A signal ratio along kinetochores after partial surgery of sisters. Dots represent individual kinetochores pooled from all MAD1 recruiting cells(surgery n = 16 cells, 14 independent experiments; nocodazole n = 19 cells, 2 independent experiments). Boxes show interquartile ranges, horizontal lines show median and error bars show min max. P = 7 × 10⁻⁶, Mann–Whitney test (two-sided). Correlative live cell spinning disk (f) and fixed CH-STED microscopy (f ’) (n = 1 experiment) of cell expressing 2x-GFP-CENP-A and GFP-CENTRIN-1 (both in blue) and mScarlet-MAD1 (orange/gray) and stained microtubules (gray) after partial kinetochore surgery. Cell recruiting MAD1 to kinetochore tip rather than center. Insets: perturbed kinetochore pair pre- and post-surgery (solid and hollow arrow, respectively). Arrowheads show MAD1 recruitment. Scale bars: 2 µm. f ” Max normalized mScarlet-MAD1/2x-GFP-CENP-A (live cell data just before fixation, orange) and microtubules/2x-GFP-CENP-A (fixed cell data, gray) ratio profile along longitudinal axis of kinetochore highlighted in (f) and (f ’).

To assist us in the interpretation of the laser microsurgery data and determine whether the SAC was responding to microtubule detachments in the bent unperturbed sister kinetochore, we performed correlative super-resolution CH-STED microscopy after fixation of cells that showed MAD1 recruitment upon partial sister kinetochore laser ablation by live imaging (Fig. 4f, f’ and Supplementary Fig. 4c, c’). We found that MAD1 accumulation, either at the tip or at a more central region of the unperturbed sister kinetochore, negatively correlated with microtubule density (Fig. 4f’, f” and Supplementary Fig. 4c’, c”). We reasoned that MAD1 accumulation near the tip of the unperturbed kinetochore may result from local microtubule detachments due to lack of tension in the partially ablated sister. In other cases, localized forces applied to both ends of the unperturbed kinetochore may generate enough tension along its length to allow a few microtubules to remain stably attached near the tip, resulting in more central accumulation of MAD1. Taken together, these experiments demonstrate that, once silenced, the SAC remains responsive to local microtubule detachments from kinetochores, even when more than 50% occupancy persists.

Low microtubule occupancy at kinetochores delays SAC silencing

Next, we sought to investigate the cellular consequences of low microtubule occupancy at kinetochores. To do so we monitored SAC silencing after RNAi-mediated depletion of HAUS6, a subunit of the Augmin complex required for K-fiber maturation^26,58. We found that MAD1 decay from kinetochores was slowed-down in HAUS6-depleted cells, as revealed by ~50% increase in the half reduction time of MAD1 signal at kinetochores relative to mock-depleted controls (Fig. 5a–c, Supplementary Fig. 2a and Supplementary Movies 9, 10). Moreover, even when HAUS6-depleted cells were treated with the proteasome inhibitor MG132 for 1 h to provide more time for K-fiber maturation, many more (66% vs. 12%) cells had at least one MAD1-positive kinetochore and overall higher MAD1 levels at kinetochores when compared with control metaphase cells (but lower than prometaphase levels) (Fig. 5d, e). Closer inspection of mock- and HAUS6-depleted fixed cells with CH-STED super-resolution microscopy revealed that, while kinetochores in control mock-depleted metaphase cells treated with MG132 were abundantly occupied with microtubules and without detectable MAD1 signal, in HAUS6-depleted cells treated with MG132 MAD1 signal could still be detected along the kinetochores, despite interspersed end-on microtubule attachments (Fig. 5f). However, unlike in control cells with clear partially attached kinetochores that showed a negative correlation between MAD1 and microtubule signal (r = −0.29, p < 1e−10; Supplementary Fig. 5e), in HAUS6-depleted cells this correlation was lost (r = −0.06, p = 0.14; Supplementary Figs. 5d, e and 6c), suggesting that the resulting perturbations of microtubule occupancy were beyond our resolution capacity (Supplementary Figs. 5a–c, 6a, b’). Overall, these data are in line with previous findings in human cells^7,8 indicating that low microtubule occupancy at kinetochores delays MAD1 removal from kinetochores and consequently SAC silencing.

Fig. 5. Partial microtubule occupancy at kinetochores after Augmin depletion delays MAD1 removal and SAC silencing.

a Representative live-cell recording of Indian muntjac fibroblasts mock or HAUS6 siRNA treated. Venus-MAD1 (orange) and SiR-tubulin labeled microtubules (gray). Insets: kinetochore pair. Scale bars: 2 µm. b Normalized Venus-MAD1 fluorescence decay from (a). t = 0 s corresponds to the peak of MAD1 fluorescence before it starts to decay. Data are mean ±95% CI (Mock n = 48 kinetochores, 10 cells, gray; siHAUS6 n = 47 kinetochores, 10 cells, blue; minimum 5 independent experiments). c Parameters extracted from a single exponential fitting of Venus-MAD1 fluorescence decay across the conditions shown in (b). Mock vs siHAUS6, p = 1.4 × 10⁻⁵,extra sum-of-squares F-test. d Percentage of cells with at least one MAD1 positive kinetochore in mock (gray) or siHAUS6 (blue) treated cells after 1 h of MG132 arrest. Bars show mean, error bars show s.d., circles show experiments (n ≥ 293 cells per condition, N = 3 independent experiments). Prometaphase mock vs siHAUS6, p = 9.4 × 10⁻⁵; Metaphase mock vs siHAUS6, p = 0.0008; unpaired t test (two-sided). e Ratio between kinetochore and cytoplasmatic Venus-MAD1 signal intensity in mock (gray) or siHAUS6 (blue) treated cells fixed after 1 h of MG132 arrest. Dots show individual kinetochores, boxes show interquartile ranges, horizontal lines show median and error bars show min max from 3 independent experiments (Prometaphase mock n = 320 kinetochore, 17 cells; prometaphase siHAUS6 n = 309 kinetochores, 17 cells; metaphase mock n = 418, 25 cells; metaphase siHAUS6 n = 581, 34 cells). Prometaphase mock vs siHAUS6, p = 5.5 × 10⁻⁷; Metaphase mock vs siHAUS6, p < 1 × 10⁻¹⁵, Mann–Whitney test (two-sided). f Representative CH-STED image of Indian muntjac control (4 cells, 2 independent experiments) or HAUS6 depleted cells (9 cells, 2 independent experiments) treated with MG132 for 1 h. Microtubules (gray; STED), centromeres (ACA; blue; STED) and Venus-MAD1 (orange; confocal). Inset: full (control) or partially (siHAUS6) attached kinetochore pair. Scale bars: 2 µm.

Low microtubule occupancy leaves a mitotic memory

To further characterize the cellular consequences of low microtubule occupancy at kinetochores, we tracked HAUS6-depleted cells over several days by live-cell microscopy and determined their fates (Fig. 6a). Mitoses in HAUS6-depleted cells were significantly delayed and error-prone, proportional to the extent of HAUS6 depletion over time (Fig. 6b, c”). When inspecting the outcomes resulting from mitotic delays in HAUS6-depleted cells more closely, we found three regimens. The majority of cells with mitotic durations below 200 min divided without segregation errors or spindle defects (Fig. 6c). However, after a delay of ~200 min, virtually all cells divided with mitotic defects (Fig. 6c), pointing towards persistently immature K-fibers at anaphase onset⁶. Strikingly, most cells that were stuck in mitosis for longer than 600 min ultimately died in mitosis (Fig. 6c).

Fig. 6. Delayed SAC silencing due to partial microtubule occupancy impacts daughter cell proliferation.

a Schematics of experimental design. b Mitotic duration of Indian muntjac H2B-GFP⁺ cells with decreasing HAUS6 levels (time of recording after control or HAUS6 siRNA treatment). siCtrl n = 823 cells, siHAUS6 n = 1134 cells, 5 independent experiments). Circles and error bars denote median and interquartile range in each 12 h time bin. Note: after 84 h many cells remain in mitosis beyond the movie ending, making measurements inaccurate. c Mitotic cell fate as a function of mitotic duration in cells treated with control or HAUS6 siRNA grouped into 24 h time bins. Bars show individual cells (siCtrl 12–36 h n = 202 cells, 36–60 h n = 204 cells, 60–84 h n = 243 cells, siHAUS6 12–36 h n = 316 cells, 36–60 h n = 309 cells, 60–84 h n = 322 cells, 5 independent experiments). c’ Bars show the percentage of mitoses in (c) that divide error-free, with errors, or die in mitosis. c” Defect categories found in (c). Bars show percentage of total analyzed mitoses. d Data from (c) filtered for cells that divide defect-free between 12 h to 36 h (Gen 1 cohort). Dots show mitotic duration of individual cells (siCtrl n = 140 cells, siHAUS6 = 207 cells, 5 independent experiments). Boxes show interquartile ranges, horizontal lines show median and error bars show min max. P < 1 × 10⁻¹⁵, Mann–Whitney test (two-sided). e Bars show the percentage of mother cells from the selected cohort shown in (d) that has 0, 1 or 2 daughter cells dividing at least once, as a function of mother cells time in mitosis in siCtrl (left) or siHAUS6 (right). Data are time-binned (siCtrl 30–60 n = 84 cells, 60–90 n = 49 cells, 90–120 n = 6 cells, >120 n = 1 cell, siHAUS6 30–60 n = 36 cells, 60–90 n = 90 cells, 90–120 n = 59 cells, >120 n = 21 cells, 5 independent experiments). siCtrl 30–60 vs 90–120, p = 0.049; 30–60 vs 60–90, p = 0.35; 60–90 vs 90–120, p = 0.21. siHAUS6 30–60 vs 60–90, p = 0.04; 30–60 vs 90–120, p = 0.0007; 30–60 vs >120, p = 8.2 × 10⁻⁵; 60–90 vs 90–120, p = 0.007; 60–90 vs >120, p = 0.004, chi-square test with Benjamini–Hochberg multiple comparison correction (two-sided).

To determine the fate of daughter cells from mothers that did divide despite immature K-fibers, we defined a cohort of cells that underwent mitosis within 12 and 36 h of recording and followed their offspring for 60 h (Fig. 6a). Since prolonged mitotic duration under low microtubule occupancy was tightly associated with mitotic errors (Fig. 6c, c’) that may independently influence daughter cell fate^59–62, we sought to disentangle the two events. To this end, the low chromosome number of Indian muntjac fibroblasts allowed us to unequivocally exclude cells dividing with errors from our cell fate analysis (Fig. 6d, e’). Importantly, although the remaining cohort only underwent a short Augmin depletion period (max. 42 h) and divided without errors, siHAUS6-treated cells showed significantly prolonged mitotic times, presumably due to persistent low-microtubule occupancy (Fig. 6d). Of note, within the control population, the progeny of cells that sporadically took longer to divide was also more likely to arrest proliferation, when compared with the one of cells that divided within the expected duration (15% [13/84] of cells with 30–60 min mitoses vs 67% [4/6] with 90–120 min mitoses have both daughters arresting; Fig. 6e). However, the frequency of these events was significantly higher upon Augmin depletion (Fig. 6d) where mother cells with increasing mitotic duration were increasingly more likely to give rise to daughter cells that stopped proliferating (3% [1/36] of cells with 30–60 min mitoses vs 39% [23/59] of cells with 90–120 min mitoses have both daughters arresting; Fig. 6e). Consistently, we observed a similar gradual trend in cells released from up to 6 h of mitotic arrest mediated by the Eg5/kinesin-5-inhibitor monastrol⁶³ (Supplementary Fig. 7a–d). Here, mother cells that experienced mitotic times longer than 100 min gave rise to daughters that showed increased cell cycle lengths or stopped proliferating (Supplementary Fig. 7b), even following defect-free mitoses (5% [1/22] of cells with 30–60 min mitoses vs 35% [16/46] of cells with >120 min mitoses have both daughters arresting; Supplementary Fig. 7e). Together, these results hinted at a conserved “mitotic stopwatch” surveillance of mitotic duration, in line with previous findings in human cells^12,14,15,17.

Such a “memory” of mitotic duration was shown to result from the formation of a protein complex containing 53BP1, USP28 and p53^14,15,17. To test whether this is conserved in Indian muntjac cells, we probed for mitotic stopwatch complex formation in cells arrested for up to 6 h in the presence of monastrol (Fig. 7a). Indeed, we were able to detect a significant accumulation of p53 and USP28 bound to 53BP1 in the mitotic population, but not in cells that remained adherent after drug treatment (LO population; Fig. 7b), suggesting that the interaction specifically resulted from increased time in mitosis.

Fig. 7. A conserved stopwatch mechanism stops proliferation of cells that delay mitosis due to low-microtubule occupancy.

a Schematics of co-immunoprecipitation assay experimental design. b Analysis of 53BP1 or IgG control immunoprecipitates (IP) in asynchronous (“A”), leftovers (“LO”) and 6 h mitotic arrested (“M”) H2B-GFP⁺ Indian muntjac cells (2 independent experiments). Inputs correspond to whole cell lysates. GAPDH as loading control. c Schematics of live cell imaging experimental design. d Mitotic duration of Indian muntjac cells expressing H2B-GFP treated with HAUS6 in combination with control or USP28 siRNAs. Dots show mitotic duration of individual cells from a select cohort - divide defect-free between 12 h to 36 h (Gen 1 cohort, siCtrl n = 96 cells, siHAUS6+siCtrl = 153 cells, siHAUS6+siUSP28 n = 199 cells, 3 independent experiments). Boxes show interquartile ranges, horizontal lines show median and error bars show min max. siCtrl vs siHAUS6+siCtrl, p = 1.6 × 10⁻¹¹; siCtrl vs siHAUS6+siUSP28, p = 1.6 × 10⁻⁹, siHAUS6+siCtrl vs siHAUS6+siUSP28, p = 0.72, Kruskal-Wallis test with Dunn’s multiple comparisons test. e Bars show the percentage of mother cells from the selected cohort that has 0, 1 or 2 daughter cells dividing at least once, as a function of mother cells time in mitosis in siCtrl (left), siHAUS6+siCtrl (middle) or siHAUS6+siUSP28 (right). Data are time-binned (siCtrl 30–60 n = 50 cells, 60–90 n = 34 cells, 90–120 n = 11 cells, >120 n = 1 cell, siHAUS6+siCtrl 30–60 n = 20 cells, 60–90 n = 69 cells, 90–120 n = 49 cells, >120 n = 14 cells, siHAUS6+siUSP28 30–60 n = 41 cells, 60–90 n = 84 cells, 90–120 n = 58 cells, >120 n = 16 cells, 3 independent experiments). siCtrl p = 0.12, chi-square test; siHAUS6+siCtrl 30–60 vs 60–90, p = 0.22; 30–60 vs 90–120, p = 0.07; 30–60 vs >120, p = 0.002; 60–90 vs 90–120, p = 0.07; 60–90 vs >120, p = 0.002; 90–120 vs >120, p = 0.04, chi-square test with Benjamini–Hochberg multiple comparison correction (two-sided); siHAUS6+siUSP28 p = 0.24, chi-square test.

To test whether prolonged mitoses due to low microtubule occupancy at kinetochores were indeed registered and bookmarked by mitotic stopwatch complex formation, we co-depleted USP28 together with HAUS6 and again determined daughter fates (Fig. 7c). Importantly, Augmin depleted cells took longer to divide then controls, irrespective of USP28 presence or absence (Fig. 7d). Consistent with reports on USP28-deficient RPE-1 cells that were released from prolonged mitotic arrest⁶⁴, we found that siUSP28-treated cells lost stopwatch sensitivity and continued proliferating even after considerable mitotic arrest due to HAUS6 deficiency (2% [1/41] of cells with 30–60 min mitoses vs 2% [1/58] of cells with 90–120 min mitoses have both daughters arresting; Fig. 7e). In conclusion, combining long-term cell fate analysis and biochemistry allowed us to pinpoint a conserved mitotic stopwatch mechanism in Indian muntjac cells, controlling daughter proliferation when SAC silencing is delayed due to low microtubule occupancy at kinetochores.

Discussion

In stark contrast with the predictions from switch-like SAC-silencing models, here we found that SAC silencing at Indian muntjac kinetochores is gradual, non-uniform, correlates with microtubule attachments and kinetochore size, and depends on Augmin-mediated maturation of K-fibers. However, these apparently conflicting models can be reconciled by evoking a two-tier SAC silencing mechanism, in which cells first attempt to maximize microtubule occupancy at kinetochores to ensure gradual SAC silencing and avoid significant mitotic delays, but may eventually silence the SAC in a switch-like manner if this premise cannot be fulfilled. In line with a model in which the number of unattached kinetochores ultimately determines the rate of MCC production that proportionally inhibits the APC/C^27,65, low microtubule occupancy at kinetochores may result in less MCC production than fully unattached kinetochores¹⁰. Consequently, the APC/C would become sufficiently active and allow for slower, yet steady, Cyclin B1 degradation (and CDK1 inactivation), ultimately silencing the SAC after an extensive mitotic delay¹⁰. Whether kinetochores eventually reach normal microtubule occupancy by the time the SAC is silenced after a delay remains unclear (Fig. 8), but it should be noted that even cells that are unable to silence the SAC after extremely prolonged mitotic delays due to the complete absence of microtubules degrade Cyclin B1 and eventually exit mitosis due to residual APC/C activity^66,67. Our finding that CDK1 inhibition overrides SAC silencing irrespective of microtubule occupancy further supports this interpretation (Fig. 8).

Fig. 8. Delayed SAC silencing due to low microtubule occupancy at kinetochores triggers the mitotic stopwatch to control daughter proliferation.

(Top) During normal mitosis as K-fibers mature by Augmin activity, MAD1/MAD2 molecules are gradually and locally removed at the sites of microtubule attachment, resulting in timely SAC silencing. In this way, mitotic timing is optimal, fidelity is maintained, and daughter cells continue proliferating. (Bottom) When K-fiber maturation is impaired (for example through disruption of Augmin-mediated microtubule amplification), MAD1/MAD2 removal and consequently Cyclin B/CDK1 degradation, is slower, extending the duration of mitosis. This provides time for mitotic stopwatch complex accumulation and p53 stabilization. Eventually, the cell divides after a delay (either with high or low microtubule occupancy), and the mitotic stopwatch bookmarks daughter cells to halt proliferation.

Although low microtubule occupancy at kinetochores may eventually silence the SAC, there is an unavoidable impact on the respective mitotic duration. This has important implications in light of the recently-uncovered “mitotic stopwatch” mechanism, where even mild mitotic delays elicit a subsequent USP28-53BP1-p53-dependent cell cycle arrest in G1^{12,14,15,18,68}, and the observation that SAC silencing with low microtubule occupancy increases the incidence of lagging chromosomes during anaphase, potentially leading to aneuploidy⁶. As so, the mechanism underlying SAC silencing at individual kinetochores is likely to have evolved to maximize microtubule occupancy within a strict temporal window to avoid “bad memories” of mitosis, and we show here that the Augmin complex, through its role in K-fiber maturation^26,58, is instrumental in this process (Fig. 8). Likewise, a recent CRISPR-Cas9 screen in human RPE1 cells independently uncovered the Augmin complex as an important factor controlling mitotic duration, perturbation of which triggered a mitotic stopwatch response⁶⁹. Importantly, Augmin expression was found to correlate with p53 signaling in certain cancers⁷⁰, whereas Augmin deficiency in neural stem cells was shown to cause p53-dependent apoptosis and compromise brain development in mice⁷¹. Overall, these results hint at functional conservation of the underlying molecular network linking timely SAC silencing with mitotic memory, with important implications for tissue architecture and homeostasis.

One may still argue that the compound architecture of Indian muntjac kinetochores does not reflect what is normally found in other vertebrate species, including humans. However, we have no evidence that kinetochore architecture in Indian muntjac reflects a significant deviation at the level of the core microtubule-binding moduli or MAD1 decay during SAC silencing, when compared with other vertebrates^9,24,26,72. Noteworthy, recent works have revealed that many human, mouse and chicken centromeres/kinetochores also have a compound, often bipartite, architecture^49,51,73,74. In particular, the same human kinetochore could be found with one sub-domain end-on attached, and the other laterally attached to microtubules via the fibrous corona⁷³. This is consistent with our finding of non-uniform, highly localized, MAD1 removal in response to microtubule occupancy in Indian muntjac kinetochores. Importantly, while in these cases MAD1 and microtubule density often negatively correlated along two apparently insulated adjacent domains on the same large kinetochore of chromosome 3 + X (simply because these extreme cases are easier to detect), there were other cases (particularly evident after Augmin perturbation) where this negative correlation was observed along three or more sub-domains, without following a strict spatial pattern (also evident in the distinct responses after laser microsurgery experiments). Lastly, our finding that the stable pool of MAD1 does not diffuse within unattached kinetochores further indicates that it can only be locally removed upon the establishment of microtubule attachments or, eventually, by a microtubule-independent SAC-silencing mechanism involving MPS1 or CDK1 inactivation.

Despite their unique cytological features, not even Indian muntjac kinetochores allow us to resolve individual microtubule attachments by light microscopy. However, our quantitative fluorescence analyses of MAD1 within Indian muntjac kinetochores indicated that while high microtubule signal is a good predictor for the absence of MAD1 at kinetochores, there may be sites without microtubule attachments that also lack MAD1. This suggests that SAC silencing within the kinetochore is confined to points of individual microtubule attachments and possibly their immediate vicinity, in agreement with cooperative models in which a single microtubule dynamically binds approximately a dozen of NDC80 complexes at the kinetochore^55–57. Given that there is one MAD1 dimer for every five to six NDC80 complexes^57,75,76, this would be consistent with few MAD1 molecules leaving the kinetochore upon each individual microtubule attachment.

Taken together, our model reconciles gradual SAC silencing with increasing microtubule occupancy at kinetochores, without the need for a dedicated microtubule “counting” system that triggers SAC inactivation. Most relevant, if cells ultimately silence the SAC after experiencing significant mitotic delays due to low microtubule occupancy at kinetochores, the resulting daughter cells are bookmarked by the mitotic stopwatch to prevent their proliferation and the potential propagation of segregation errors. Thus, cells can neither rush, nor significantly delay mitosis, and microtubule occupancy at kinetochores is a cornerstone defining this balance. These findings open the exciting possibility of modestly hampering microtubule occupancy at kinetochores for therapeutic purposes, in order to delay mitosis and block daughter cell proliferation through activation of the mitotic stopwatch.

Methods

Cell lines and culture conditions

Indian muntjac cell lines were grown in Minimum Essential Media (MEM; Corning; Cat. No. 15313531), supplemented with 10% FBS (GIBCO, Life Technologies; Cat. No. A5256801). All cells were kept at 37 °C in humidified conditions with 5% CO₂. Indian muntjac hTERT-immortalized fibroblasts were a gift from Shay⁷⁷. Indian muntjac fibroblasts stably expressing Venus-MAD1 were generated by lipofection of pVenus-MAD1 (kind gift from Jakob Nilsson⁷⁸). Transfection was performed by incubating the cells with 1:400 Lipofectamine 2000 (Invitrogen; Cat. No. 11668027) in Opti-MEM (GIBCO, Life Technologies; Cat. No. 31985047) for 6 h, and stable and uniformly expressing cells were selected by FACS sorting (as detailed in ref. ²⁵). For the production of Venus-MAD1 and mScarlet-CENP-A co-expressing cell line, Venus-MAD1 cells were transduced with pLVx-mScarlet-CENP-A lentiviral plasmid (generated in ref. ²⁶). Indian muntjac cells co-expressing 2x-GFP-CENP-A, GFP-CENTRIN-1 and mScarlet-MAD1 were produced by lentiviral transduction of pRRL-2x-EGFP-CENP-A (generated in ref. ²⁶), pLVX-EGFP-C1-CENTRIN-1 (kind gift from Manuel Thery; Addgene plasmid # 73331) and pInducer-mScarlet-MAD1 (generated in this paper). Cells expressing GFP-H2B were produced by lentiviral transduction of LV-GFP (kind gift from Elaine Fuchs; Addgene plasmid # 25999). Lentiviral transduction was performed as detailed in ref. ²⁵. Cells were incubated with lentivirus particles in the presence of 1:2000 Polybrene (Sigma-Aldrich; TR-1003) for 24 h before supplied with fresh complete media. Stable lines with sufficient fluorescence intensity were selected by FACS sorting.

Plasmid design

pInducer-mScarlet-MAD1 was produced by cloning mScarlet (from pRRL-mScarlet-CENP-A) and MAD1 (from pVenus-MAD1) into the pInducer-20 backbone under a Tet-inducible promotor. To construct the empty backbone, KASH-mCherry was removed by SalI + XhoI restriction digestion of pinducer 20 DN-KASH (kind gift from Daniel Conway; (Addgene plasmid # 125554)). Restriction sites were amplified from the plasmid by PCR (custom designed primers, fwd: 5′-TTAAAGGAACCAATTCAGGCTAGCACGCGTATATCTAGACCCAGCTTTCT-3′; rev: 5′-TAAGCGTAGTCTGGGACG-3′) and added to the emptied vector through Gibson assembly. mScarlet and MAD1 were cloned into the NheI linearized pInducer-20 backbone through Gibson assembly.

siRNA experiments

Protein knockdown was performed as detailed in ref. ²⁵. Briefly, Indian muntjac fibroblasts seeded at 70% confluency were starved with MEM supplemented with 5% FBS for 30 min before transfection. siRNA transfection was performed by adding Lipofectamine RNAi Max (1:400; Invitrogen; Cat. No. 13778075) and 75 nM, 50 nM or 25 nM of siRNA in serum free-medium (Opti-MEM; Gibco; Cat. No. 31985047) for 6 h before replacing the solution with fresh complete media. Cells were analyzed 72 h after depletion, unless specified otherwise. For the depletion of HAUS6, the following sequence was used: 5′-GGUUGGUCCUAAGUUUAUU[dT][dT]-3′²⁶ at 50 nM. For the depletion of USP28, the following sequence was used: 5′- GCTTCCGGACATGTTGAAATA[dT][dT]-3′ at 25 nM. Cells mock transfected (lipofectamine only) or transfected with luciferase-targeting siRNA (5′-CAUUCUAUCCUCUAGAGGA[dT][dT]-3′) at 75 nM, 50 nM or 25 nM were used as controls.

Drug treatments

For acute inhibition of Mps1 and CDK1 kinases, 4 µM MPS1-IN-1 (kind gift from Gray⁴⁰) and 10 µM RO 3306 (ChemCruz; CAS 872573-93-8), respectively, were added shortly after Nuclear Envelop Breakdown (NEB). To arrest cells in G2, these were treated with 10 µM RO 3306 for 14 h. Microtubule depolymerization was triggered using 1 µM of nocodazole (Sigma-Aldrich; CAS 31430-18-9) for 30 min before the analysis. To increase mitotic duration cells were arrested in prometaphase with 75 or 100 µM of the EG5 inhibitor, Monastrol (Tocris Bioscience, Cat. No.1305) for 6 h. Metaphase arrest was obtained using 5 µM MG132 (Sigma-Aldrich; CAS 133407-82-6). Fixed cell analysis using MG132 was performed in the first 1.5 h after drug addition to avoid cohesion fatigue. SiR-tubulin (Spirochrome; SC002)⁷⁹ was used to visualize microtubules, at 50 nM concentration incubated for 1 h prior to live-cell imaging.

Immunofluorescence

Indian muntjac fibroblasts were seeded on fibronectin coated coverslips (#1.5 thickness) 24 h before the experiment. Depending on the microscopy routine, different fixation protocols were used: for widefield imaging cells were incubated with 4% paraformaldehyde (Delta Microscopies; GF750170-1010) in PBS for 10 min at room temperature; for STED microscopy PFA 4% supplemented with 0.1–0.2% Glutaraldehyde (Delta Microscopies; GF132574-1010)) in Cytoskeleten Buffer (CB - 274 mM NaCl, 10 mM KCl, 2.2 mM Na₂HPO₄, 0.8 mM KH₂PO₄, 4 mM EGTA, 4 mM MgCl₂, 10 mM Pipes, 10 mM glucose, pH 6.1) for 10 min at room temperature was used; for same cell correlative live confocal and super-resolution STED microscopy pre-warmed 2× concentrated PFA + Glutaraldehyde diluted in imaging media to a final concentration of 4% and 0.2%, respectively, was added to the imaging chamber. After a 5 min incubation at the microscope (37 °C), the fixation solution was replaced with PFA 4% supplemented with 0.2% Glutaraldehyde in CB for an additional 5 min at room temperature. Autofluorescence was quenched by a 0.1% sodium borohydride solution (Sigma-Aldrich; 71320) for 7 min. Cells were permeabilized with CB-0.5% Triton for 20–30 min and blocked with CB-0.05% Tween 20 with 10% FBS (blocking buffer) for 1 h at RT. Samples were incubated with primary antibodies diluted in blocking buffer over-night at 4 °C. The following primary antibodies were used: human anti-centromere antiserum (ACA, Fitzgerald; 90C-CS1058) 1:2000/1:200 (Widefield/STED); anti-tyrosinated tubulin (Bio-Rad; MCA77G) 1:2000/1:100 (Widefield/ STED); anti α-tubulin (clone B-5-1-2, Sigma-Aldrich; T5168) 1:100 (correlative live-cell and fixed STED microscopy). Cells were washed with PBS-0.05% Tween before incubation with the corresponding secondary antibodies for 1 h at RT – Alexa Fluor 568 and 647 (Thermo Fisher Scientific) 1:1000; or abberior STAR 580 (Abberior Instruments; ST580) and abberior STAR-Red (Abberior Instruments; STRED) 1:200 for STED microscopy. DNA was labeled with a quick incubation in the presence of 1 µg/mL 4’6’-Diamidino-2-phenylindole (DAPI; Sigma-Aldrich; D9564) in PBS-0.05% Tween. The samples were washed in PBS and mounted on glass slides with a mounting solution (20 mM Tris pH8, 0.5 N-propyl gallate, 90% glycerol).

Monitoring kinetochore MAD1 levels in live recordings

Indian muntjac fibroblasts stably expressing Venus-MAD1 or co-expressing Venus-MAD1 and mScarlet-CENP-A were plated on fibronectin coated glass bottom 35 mm FluoroDish (World Precision Instruments; FD35-100), 36–48 h before imaging. Before imaging, normal culture media was replaced with Leibovitz’s L15 medium (GIBCO, Life Technologies; Cat. No. 21083027) supplemented with 10% FBS and the cell-permeable microtubule dye, SiR-tubulin (Spirochrome AG; SC002). Live-cell imaging was performed on a temperature-controlled (37 °C) Nikon-Ti microscope equipped at the camera port with a Yokogawa CSU-X1 spinning-disc head with Borealis and an iXon+ DU-897 EM-CCD (Andor). Cells were imaged with an oil-immersion 60× 1.4 NA Plan-Apo DIC CFI objective (Nikon, lambda series), yielding a 176 nm/pixel sampling, or a 100× 1.4 NA Plan-Apo DIC CFI objective (Nikon, VC series), yielding a 106 nm/pixel. Images were acquired using 488 nm, 561 nm and 647 nm (Coherent) laser lines and acquisition was controlled by NIS Elements AR software (3.2). To ensure that kinetochore could be tracked overtime with a reduced number of optical sections, cells were imaged under modest cell confinement. An image stack was acquired: every 6 s, 9 planes separated by 0.5 µm (Figs. 1 and 5a, b and Supplementary Fig. 1c); 1 min, 11 planes separated by 0.5 µm, Supplementary Fig. 1a, b); 10 seconds, 9 planes separated by 0.5 µm (Supplementary Fig. 2b, b’). Fluorescence intensity of kinetochore Venus-MAD1 was manually tracked throughout mitosis using the image analysis software Fiji (2.16.0)⁸⁰. The recorded image stacks were sum projected and a region was drawn encircling the kinetochore and three different cytoplasmic background sites at each timepoint. MAD1 kinetochore fluorescence intensity (S_KT) was calculated by subtracting the averaged cytoplasmatic signal (S_BG), represented in the following equation: S_{KT, corrected} = S_KT – S_{BG, averaged}$\times$ A_KT/A_BG S: raw integrated density; A: Area. The calculated signal intensity was normalized to the followings: the clear last peak of the signal decay curve in untreated/control, mock treated and siHAUS6 conditions; the first frame after adding the inhibitor in MPS1i and CDK1i treatments; the average of the first 10 timepoints in nocodazole treated cells. A single-phase exponential decay function was fitted globally across curves polled from all kinetochores in each experimental condition using the equation: Y = (Y₀ − Plateau) × (exp^(−Kx)) + Plateau, in GraphPad Prism 9 (Boston, Massachusetts USA). The half decay time (t_1/2) for each condition was derived from the fit according to t_1/2 = ln(2)/K. To test for differences in the rate of MAD1 decay (K, derived from the global fit) in untreated vs treated cells or large vs small kinetochores an extra sum-of-squares F-test was used. Quantification of the signal intensity ratio between kinetochore and cytoplasm (Supplementary Fig. 1b), was performed manually using Fiji (2.16.0). Three regions were drawn encircling the kinetochore (KT), the cytoplasm (cyto) and an area outside of the cell (background signal, BG) at define time points (anaphase onset, 10, 20, 30 and 40 min before anaphase). The kinetochore and cytoplasmic fluorescence intensities were calculated by subtracting background signal, represented in the following equations: S_{KT, corrected} = S_KT/A_KT − S_BG/A_BG; S_{cyto, corrected}= S_cyto/A_cyto − S_BG/A_BG; S: signal; A: Area. The ratio of the corrected kinetochore over cytoplasmatic signals was then calculated. To observe how Mad1 signal spatial distribution within the kinetochore changed throughout time kymographs were produced from live recordings of Indian muntjac cells. Image stacks were sum projected using Fiji (2.16.0) and kymographs were generated with a custom-written routine in MATLAB (R2023a; Mathworks, Natick, MA) that compensates for spindle rotation and translations previously described in ref. ⁸¹.

Cell confinement

For cell confinement, we adapted a cell confiner as previously described²⁵ using a custom-designed polydimethylsiloxane (PDMS, RTV615, GE) layout to fit a 35 mm diameter fluorodish. A suction cup was custom-made with a 10:1 mixture (w/w PDMS crosslinker A/ B) and baked on an 80 °C hot plate for 30 min and left to dry over-night before unmolding. A PDMS confinement slide was then prepared with a 10:1 mixture (w/w PDMS crosslinker A/B) molded into micropillars with a height of 8 μm (using a SU-8 wafer mold) polymerized onto a 10 mm round coverslip and baked for 95 °C for 15 min. The PDMS confinement slide was attached to the PDMS suction cup and connected to a vacuum generator (AF1-dual, Elveflow).

Quantification of MAD1 intra-kinetochore mobility by FRAP

For fluorescence recovery after photobleaching (FRAP) experiments, Indian muntjac cells stably expressing Venus-MAD1 were plated on fibronectin coated 35 mm FluoroDish (World Precision Instruments, FD35-100) 36–48 h before imaging. To elicit uniform and maximal kineotchore-MAD1 occupancy, cells were treated with nocodazole for 30 min before the experiment. To ensure that the bleached kinetochore could be tracked overtime with a reduced number of optical sections, an agar overlay was placed on top of the cells decreasing the cell volume, as described in ref. ⁸². Briefly, a 170 μm thick layer of 2% low-melting-point agarose in L15 medium supplemented with 10% FBS was prepared in advance and incubated for 30 min with pre-warmed L15 + 10% FBS containing nocodazole. Immediately before imaging, a small piece of the agar layer was gently placed over the cells that were maintained in a minimal amount of medium to prevent cell dehydration while ensuring compression. FRAP experiments were conducted on a Leica Scanning Confocal STELLARIS 8 FALCON (Leica Microsystems, Germany), equipped with a temperature control chamber and a power HyD S detector. Cells were imaged with a water-immersion 86× 1.20 NA HC PL APO STED white objective (Leica Microsystems, Germany), plus a 4.5× zoom, yielding an 86 nm/pixel sampling and acquisition was controlled by LAS X software (4.90.30221). The 488 nm excitation wavelength of a white Light Laser (WLL) was used to bleach MAD1 signal from either part of or an entire kinetochore with 2 pulses of 0.3 seconds each. An image stack of 3 z-planes at a 0.5 μm interval was acquired with the following routine: 3 pre-bleach frames, every 1.3 s; kinetochore bleaching; post-bleach 5 frames every 1.3 s, 10 frames every 2 s and then every 3 s for up to 55 s. The selected representative images show sum-intensity projections of the image stacks. Venus-MAD1 fluorescence at the kinetochore was manually tracked on the sum projected image stacks using Fiji (2.16.0) and a custom-designed script in Python (3.13). Four rectangular ROIS were defined corresponding to: (i) the bleached region of the experiment kinetochore; (ii) the non-bleached region of the experiment kinetochore; (iii) the unperturbed kinetochore from the pair (sister KT) and (iv) a kinetochore from an unperturbed pair (control KT). Additionally, two square ROIs with 0.86 µm sides were drawn over the cytoplasm to measure background fluorescence (BG). A custom-designed script was used to draw the kinetochore ROIs with a constant width of 0.6 µm, and length defined by the distance of the kinetochore extremities in the pre-bleach frame (manually marked based on MAD1 signal using Fiji (2.16.0)). In the case of the partially ablated kinetochore the length of the non-bleached region was defined at the first frame post-bleaching, and the length of the bleached region was defined as the difference between the total kinetochore length and the length of the non-bleached region. The long axis of the rectangle was defined based on the vector described by the points marking the extremities of the non-bleached signal and was adjusted at every time point to track the kinetochore motion. MAD1 kinetochore fluorescence intensity (S_{KT, corrected}) was corrected for fluctuations in the cytoplasmatic fluorescent signal as described in the following equation: S_{KT, corrected} = S_KT/(a × time + b); S_KT: kinetochore MAD1 raw integrated density; a, b: constants provided by a linear fit of the cytoplasmatic MAD1 raw integrated density. The corrected kinetochore signal was then normalized to the average of the 3 pre-bleach values (S_{KT, normalized}) and fit to a single exponential association function: Y = Y₀ + (Plateau − Y₀) × (1−(exp^(−Kx))); Y0 value was constrained to be the S_{KT, corrected} at the first frame post-bleaching. Half recovery time (t_1/2) was derived from the fit according to t_1/2 = ln(2)/K. Fluorescence recovery percentage (Recovery) was calculated according to Recovery = (Plateau − Y₀)/(1 − Y₀). Fitting and downstream statistical analysis (Mann–Whitney test) was performed using GraphPad Prism 9 (Boston, Massachusetts USA). Cells where bleaching efficiency was <50%; focal plane changes were >1 µm; or where the normalized MAD1 fluorescence could not be fitted to a single exponential (if the 95 % confidence interval for any of the parameters could not be determined) were excluded from the analysis. Line intensity profiles showing Venus-MAD1 signal along the kinetochore were measured from a ROI covering the entire kinetochore length and a width of 0.250 µm at selected times.

Quantification of kinetochore MAD1 signal distribution with STED Microscopy

Indian muntjac fibroblasts stably expressing Venus-MAD1 were seeded on glass coverslips and processed as described in the “Immunofluorescence” section. 3D image stacks were acquired using an Abberior Instruments ‘Expert Line’ gated-STED coupled to a Nikon Ti microscope and equipped with avalanche photodiode detectors. Acquisition was controlled by Abberior Instruments iMSPECTOR software (16.3). Nikon Lambda Plan-Apo 1.4NA 60x objective lens was used. CH-STED mode was implemented as described before⁵⁴. Images were acquired using excitation wavelengths at 488 nm, 561 nm, and 640 nm and a single doughnut-shape depletion beam at 775 nm, a 0.8 Airy unit pinhole and the STED channel had a time-gate threshold of 500 ps. Pixel size was set to 30 nm in xy and 200 nm in z and image stacks covering near full spindle volume (~8 µm) were acquired. The selected representative images show sum-intensity projections of a fraction of the Z-slices (5–10) and in images that show merged channels, histograms were cropped to allow visualization of all structures. Venus-MAD1 and tubulin signal distribution was analyzed along individual kinetochores in sum projected 3D stacks (5–10 slices) using Fiji (2.16.0). A line of 0.6 µm width was drawn along the longitudinal axis of the kinetochore (defined by the ACA signal). Intensity line profiles were extracted along the kinetochore for both MAD1 and tubulin channels and background subtracted (average cytoplasmatic intensity value). Corrected signal intensity was normalized to the maximum intensity value within each kinetochore. To visualize and measure correlations between MAD1 and tubulin signals, we calculated the intensity z-scores, standardized internally (for each kinetochore). Each position along the kinetochore is represented as a point in a 2D space where the z-scores are the x-y axes and an ensemble of kinetochores was either overlaid in the same scatter plot or displayed and calculated for isolated kinetochores. The product zMAD1.ztubulin at each data point is the local contribution to an overall correlation. Averaging over all data points yields the covariance normalized to the standard deviation product, known as the Pearson correlation coefficient. The Matlab (R2023a) function corrcoef was used, which calculates Pearson’s correlations and p-values with a two-tailed Student’s t test for zero correlation. 3D reconstructions of the image stacks were obtained with Imaris (9.3.1) (Oxford Instruments).

Laser microsurgery and correlative live-fixed cell microscopy

Indian muntjac cells expressing 2x-GFP-CENP-A and GFP-CENTRIN-1 and mScarlet-MAD1 were seeded on fibronectin coated 25 mm, no. 1.5, round coverslips (Fisher scientific; CAT 10593054,) 48 h prior to the experiment. To induce mScarlet-MAD1 expression, cells were incubated with 1 μg/mL of doxycycline (Sigma-Aldrich; Cat. No. D9891) for 24 h. Before imaging the coverslips were mounted on a 35 mm Attofluor™ Cell Chamber (Invitrogen; CAT A7816) and the media replaced with in Leibovitz’s L15 medium (GIBCO, Life Technologies; Cat. No. 21083027) supplemented with 10% FBS. Cells were imaged using a temperature-controlled (37 °C) Nikon Eclipse TE2000-U spinning-disk confocal microscope equipped with a Nikon plan-apochromatic DIC 1.4 NA 100× objective, and an iXonEM + EM-CCD camera (Andor Technology), yielding a 113 nm/pixel sampling. Acquisition was controlled by NIS Elements AR software (3.2). Cells in metaphase were identified by having MAD1 negative kinetochores aligned at the equator and increased inter-kinetochore and pole-pole distance (assessed by 2x-GFP-CENP-A and GFP-CENTRIN-1 signals). After recording a pre-surgery image stack, one of the large kinetochores (from the Indian muntjac chromosome 3 + X) was partially ablated with a 532 nm laser, controlled by a custom routine. A detailed description of the microsurgery setup can be found in ref. ⁸². 2–5 consecutive pulses with a 0.35 μm step and a 12 Hz repetition rate were applied. The pulse width was 10 ns and the pulse energy was 3.9–4.4 μJ. After surgery an image stack (5 planes, 0.75 µm z-step) was recorded every minute until MAD1 signal was detected at the kinetochore, for up to 30 min; after MAD1 detection the acquisition routine was changed: 7 planes, 0.5 µm z-step every 20 s for 5 min. As a control for uniform MAD1 signal distribution, non-ablated cells were imaged upon acute nocodazole treatment (5–60 min incubation). For correlative live-cell confocal and fixed CH-STED microscopy analysis, once MAD1 signal was detected at the kinetochore a fixation solution was added to the imaging chamber. The location of the cell of interest was marked with a laser-engraved pattern on the glass coverslip. The sample was processed for staining as described in the “Immunofluorescence” section. After locating the surgery cell with a 10× objective, a super-resolved image stack of the spindle was acquired with CH-STED as described in the “Quantification of kinetochore MAD1 signal distribution with STED Microscopy” section (pixel size 35 nm in xy and 150 nm in z). Representative images show a sum (live-cell data) or max-intensity (fixed-cell data) projection of all z-slices that contain the kinetochore of interest and histograms were adjusted per frame to allow better visualization of the representative phenotypes. To measure MAD1 recruitment after laser microsurgery, kinetochores were manually tracked in 4D using the GFP-tagged CENP-A as a reference in Fiji (2.16.0). Kinetochore deformation was defined as the first time point when an increase in the angle between the line connecting the ends of the surgery kinetochore and the line connecting the ends of the unperturbed kinetochore from the pair (sister kinetochore) was observed. MAD1 recruitment was defined as the first time point when mScarlet-MAD1 signal was visible at the sister kinetochore. MAD1 signal distribution along the sister kinetochore was measured at the time point when signal intensity appeared highest, based on visual assessment. On the cells where correlative live-cell spinning-disk confocal and CH-STED microscopy was performed, tubulin signal distribution was measured in the fixed sample. Measurements were performed on sum projected stacks containing the kinetochore pair of interest. A segmented line of 0.8 µm width (for MAD1, live imaging) or 0.9 µm width (for tubulin, fixed cell imaging) was drawn to span the full length of the kinetochore, as defined by the CENP-A signal, starting from the end that retained its counterpart (or randomly in nocodazole treated cells). Intensity line profiles were extracted along the kinetochore for all channels and background subtracted (cytoplasmatic signal near the kinetochore). The corrected intensity profile was normalized to the maximum within each kinetochore, as exemplified for CENP-A: S_{CENP-A, norm} = (S_CENP-A − S_{BG CENP-A})/max (S_CENP-A − S_{BG CENP-A}), S: mean pixel intensity. To correct for intensity variations caused by kinetochore deformation or movement, intensity profiles are shown as a ratio of MAD1 or tubulin over CENP-A. Kinetochore length was normalized to a relative scale ranging from 0 (start of the line profile) to 1 (end of the line profile). The average MAD1/CENP-A ratio intensity profile for each condition (Nocodazole vs Surgery), was calculated using the ‘Multiple average curves’ function with linear interpolation on OriginPro (2022; OriginLab Corporation, Northampton, MA, USA). The coefficient of variation of the MAD1/CENP-A ratio was calculated as the standard deviation of the MAD1/CENP-A intensity values along each kinetochore, divided by their mean. Cells that failed to recruit MAD1 within the mean recruitment time +1 s.d. (~12 min) or exhibited only very faint MAD1 signal were excluded from the MAD1 distribution analysis. Statistical analysis (Wilcoxon matched-pairs signed rank test; coefficient of variation and Mann–Whitney test) and plots were done using GraphPad Prism 9 (Boston, Massachusetts USA).

Quantification of kinetochore MAD1 levels in prometaphase and metaphase HAUS6 depleted cells

Control and HAUS6 depleted cells expressing Venus-MAD1 were treated with MG132 for 1 h and then fixed and stained as detailed in the “Immunofluorescence” section. Samples were imaged on Zeiss AxioImager Z1 equipped with a CCD camera (ORCA-R2, Hamamatsu) operated by Zen software (2.3; Carl Zeiss, Inc.). 3D image stacks covering the entire cell volume (0.23 μm z-step) were acquired using a 63× Plan-Apochromatic oil differential interference contrast objective lens, 1.4 NA (Carl Zeiss Microimaging Inc.) and excitation wavelengths at 488 nm, 561 nm, and 640 nm. The number of cells with MAD1 positive kinetochores was visually assessed for each condition. Kinetochore Venus-MAD1 signal intensity was measured on a single focal plane using Fiji (2.16.0). Cytoplasmatic Venus-MAD1 signal was averaged from 3 regions outside the spindle area. The ration between kinetochore and cytoplasm signal was calculated according to the equation: S_KT/S_cytoplasm; S: mean pixel intensity. Statistical analysis (unpaired t test and Mann–Whitney test) and plots were done using GraphPad Prism 9 (Boston, Massachusetts USA).

Cell tracking and daughter fate analysis in live recordings

Indian muntjac fibroblasts stably expressing GFP-H2B were plated on fibronectin coated µ-Slide 4 Well Glass Bottom (Ibidi GmbH; 80427) with Ham’s F-10 media (GIBCO, Life Technologies; Cat. No. 11550043) supplemented with 20% FBS 48 h before imaging. Cells were recorded using the Zeiss LSM 980 Airyscan confocal microscope built on an Axio Observer 7 inverted stand with a motorized piezo stage and active focus stabilization. A controlled atmosphere chamber (5% CO₂ and 37 °C) was used. Objective lens was a dry 0.3 NA 10× Plan Neofluar. Detection was performed with two GaAsP and a transmitted light (PMT) detectors. Image acquisition was carried out with ZEN (3.11) software. 3D image stacks (pixel size 830 nm in xy, 4 planes separated by 6.7 µm, pinhole Airy unit of 1.23, scan speed of 7, no averaging) were recorded every 7 min using both transmitted light and an excitation wavelength at 488 nm. In experiments where mitotic delay was induced through monastrol treatment, cells were imaged in the presence of the drug for 6 h before its wash out into fresh media. The fate of cell families was tracked for 72 h. In experiments where USP28 and/or HAUS6 was depleted, cells were treated with control, USP28 or HAUS6 siRNAs 6 h before imaging. Cell families were then tracked for 96 h. Cells were tracked manually using TrackMate (8.1.6)⁸³. Single-cell lineages and cell cycle durations (frame of anaphase mother cell to frame anaphase daughter cell) were analyzed by modifying previously published⁸⁴ Jupyter notebooks (7.0.8). We defined mitotic duration as the frame of mitotic onset (condensing chromatin) until the first frame of anaphase (clearly separated daughter chromatin). For each division event, we determined mitotic defects and errors, and the fate of the daughters. For daughter fate analysis tracks where daughter cells migrated out of the field were excluded. For binned data, bins with n < 3 cells were grayed out or not shown. Data analysis, filtering and plotting was integrated using Jupyter notebooks (Python 3.12) and the pandas (2.2.2)⁸⁵ and Altair (5.4.1)⁸⁶ libraries. Fiji/ImageJ⁸⁰ macros. Statistical analysis (chi-squared test) was performed using the SciPy.stats library (1.17.0)⁸⁷.

Co-immunoprecipitation

Indian muntjac fibroblasts stably expressing GFP-H2B were grown on with Ham’s F-10 media supplemented with 20% FBS. To obtain a mitotic enriched population, cells were synchronized in G2 with RO-3306 for 14 h and then released into fresh media supplemented with monastrol for 6 h. 10 million mitotic cells were collected by shake-off. The adherent population after shake-off (leftovers, LO) and untreated (asynchronous) cells were collected by cell scrapping. Cells were pelleted at 300 × g for 5 min, washed with cold PBS and cell pellets were snap frozen in liquid nitrogen. A co-immunoprecipitation assay was optimized from the protocol described in ref. ¹². Cell pellets were resuspended in lysis buffer (20 mM Tris/HCl, pH 7.5; 50 mM NaCl; 0.5% Triton X-100; 5 mM EGTA; 1 mM dithiothreitol; 2 mM MgCl₂; Pierce Protease and Phosphatase inhibitor mini tablets, EDTA-free [Thermo Fisher Scientific; A32961]) and incubated for 45 min at 4 °C with agitation. Cell lysates were centrifuged at 21,000 × g for 15 min at 4 °C to collect the supernatant. Protein concentration was quantified and a lysate solution containing 1.5 mg of protein in 1 mL of lysis buffer was prepared for each condition. Lysates were incubated with 3 mg of anti-53BP1 antibody (Novus; NB100-304) or control rabbit IgG (Antibodies Incorporated; 43-630-0010) for 2 h at 4 °C with agitation and subsequently with 20 mL of Pierce™ Protein A Magnetic Beads (Thermo Fisher Scientific; 88845), pre-washed 3 times with cold lysis buffer, for 1 h at 4 °C with agitation. The non-binding supernatant fraction was collected and the beads were washed 4 times with cold lysis buffer. The bound fraction was eluted from the beads with 30 μL of Laemmli sample buffer (100 mM Tris-HCl, pH 6.8, 4% SDS, 20% glycerol, 10% β-mercaptoethanol and 0.01% bromophenol blue) for 5 min. After separated from the beads, the eluate was denatured at 95 °C for 5 min and the samples were analyzed through western blot.

Western blotting

Whole protein samples (input) containing 40 μg in Laemmli sample buffer (50 mM Tris-HCl, pH 6.8, 2% SDS, 10% glycerol, 5% β-mercaptoethanol and 0.005% bromophenol blue) or co-IP eluate samples were loaded on a 4–15% Mini-PROTEAN® TGX™ Precast Protein Gel (Bio-Rad, 4561084), mounted on a Mini-PROTEAN vertical electrophoresis apparatus (Bio-Rad). NZYColour Protein Marker II (NZYTech, MB09002) was used as a ruler. Blotting was performed with a Transfer-Blot Turbo transfer system (Bio-Rad) onto 0.2 µm PVDF membranes (Bio-Rad; 1704156). Membranes were blocked in blocking buffer (5% powder milk in PBS Tween 0.1%) for 1 h. Anti-53BP1 mouse 1:1000 (Millipore; 05-725), anti-USP28 rabbit 1:1000 (Abcam; ab126604), anti-p53 mouse 1:100 (Thermo Fisher Scientific; MA1-12549), anti-Cyclin B1 mouse 1:1000 (Cell Signaling Technology; 4135) and anti-GAPDH mouse 1:10 000 (Proteintech; 60004-1) primary antibodies were diluted in blocking buffer and incubated 1 h at room temperature with agitation. Anti-rabbit HRP (Jackson ImmunoResearch; 111-035-003), anti-mouse HRP (Jackson ImmunoResearch; 115-005-003) and anti-mouse light chain specific HRP (Jackson ImmunoResearch; 115-035-174) secondary antibodies were diluted 1:5000 in blocking buffer and incubated 1 h room temperature with agitation. Membranes were developed with Clarity Western ECL Blotting Substrate (Bio-Rad; 1705061) and detected in the Chemidoc XRS system and Image Lab software (4.1) (Bio-Rad).

Statistical analysis

In each figure legend, the number of independent experiments, analyzed cells, and kinetochores measured (n) are detailed. No statistical methods were used to pre-determine sample sizes. No randomization was performed, as experiments were conducted with independent cell culture populations. Data collection and analysis were not performed blind to the conditions of the experiments. Exclusion criteria for each analysis is detailed in the corresponding method section. Data distribution was tested for normality. Data analysis was performed using Jupyter notebooks (Python 3.12), pandas (2.2.2), Altair (5.4.1) and scipy.stats library (1.17.0), Matlab (R2023a), OriginPro (2022) or GraphPad Prism (9) as detailed for each analysis in the corresponding method section. The statistical tests applied and calculated p-values for each analysis are detailed in the corresponding methods section and figure legends. Intervals were calculated with 95% confidence.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

Conceptualization: H.M.; Methodology: N.O., J.S.-O., T.K., A.J.P., and E.W.; Investigation: N.O., J.S.-O., and T.K.; Visualization: N.O., J.S.-O., T.K., and A.J.P.; Funding acquisition: H.M.; Project administration: H.M.; Supervision: H.M. and A.J.P.; Writing – original draft: N.O. and H.M.; Writing – review & editing: N.O., J.S.-O., T.K., A.J.P., E.W., and H.M.

Peer review

Peer review information

Nature Communications thanks the anonymous reviewers for their contribution to the peer review of this work. A peer review file is available.

Funding

Fundação para a Ciência e a Tecnologia of Portugal grants SFRH/BPD/118126/2016, SFRH/BD/07730/2021 and 2022.13008.BD (N.O., J.S.-O., and E.W.). European Research Council Advanced grant KAREVO grant 101140624 (H.M.).

Data availability

All data are available in the main text or the supplementary materials. Source data are provided with this paper.

Code availability

Analysis code is available as public repositories at https://github.com/CID-LAB-i3S/KinoFRAP⁸⁸ and https://github.com/CID-LAB-i3S/MitoticMemory⁸⁹ and archived on Zenodo.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41467-026-76654-6.