Chromosome size-dependent polar ejection force impairs mammalian mitotic error correction

The kinetochore is the multivalent interface that connects chromosomes to spindle microtubules at cell division. Accurate chromosome segregation requires biorientation, a state in which sister kinetochores attach to opposite spindle poles. To preserve genome integrity, each kinetochore must monitor chromosomes’ attachment status and signal to the cell whether or not it is ready to enter anaphase (Musacchio and Desai, 2017). Correct, bioriented attachments form alongside incorrect attachments, which must be detected and corrected through a process called error correction. Both physical and biochemical features differ between correct and incorrect attachments, and understanding how these cues govern error correction and how their detection varies across chromosomes is central to understanding mitotic fidelity (Lampson and Grishchuk, 2017).

Current models for error correction propose that incorrect kinetochore–microtubule attachments are molecularly destabilized to promote detachment, while correct attachments are stabilized (Funabiki, 2019; Sarangapani and Asbury, 2014). A key regulator in this process is Aurora B kinase, which phosphorylates the kinetochore’s primary loadbearing complex Ndc80C (Powers et al., 2009) to reduce its affinity for microtubules (Biggins and Murray, 2001; Cheeseman et al., 2006; DeLuca et al., 2011; Zaytsev et al., 2015). This phosphorylation decreases with kinetochore tension and is required for error correction to occur (DeLuca et al., 2006; Lampson et al., 2004). Moreover, applying ectopic tension to incorrect attachments is sufficient to delay their correction indefinitely in grasshopper spermatocytes (Nicklas and Koch, 1969), indicating a causative role for tension in distinguishing between correct and incorrect attachments in some cell types. However, how the mammalian error correction machinery responds to acute changes in tension and what cellular sources of said tension are relevant remain poorly understood.

While tension at bioriented kinetochores is generated primarily by kinetochore-bound microtubules (k-fibers), exerting opposing poleward pulling force on sisters, non-kinetochore microtubules also exert force in the spindle that influences kinetochore behavior. Polar ejection force arises from non-kinetochore microtubules growing outward from spindle poles generating outward force on chromosome arms (Rieder et al., 1986). This polar ejection force, mediated both by microtubule polymerization and by two chromokinesins in mammalian cells (Kif22/Kid and Kif4a), contributes to chromosome congression and regulates the amplitude of kinetochore oscillations at metaphase (Barisic et al., 2014; Iemura and Tanaka, 2015; Ke et al., 2009; Levesque and Compton, 2001; Wandke et al., 2012). Overexpression of the Drosophila ortholog of Kif22/Kid induces attachment errors, which elude error correction (Cane et al., 2013b), suggesting that polar ejection forces can contribute to attachment stabilization, possibly by increasing kinetochore tension. In mammalian cells, tension perturbations are needed to understand how the error correction machinery responds to tension changes and how the magnitude of polar ejection force—which in principle can vary throughout mitosis (Ke et al., 2009; Thompson et al., 2022) and across different chromosomes—impacts kinetochore tension and attachment stability.

Segregation error rates vary not only across cell types but also across chromosomes within the same cell type. Indeed, long chromosomes have been shown to missegregate more frequently than short chromosomes in human cells (Klaasen et al., 2022; Worrall et al., 2018). Many mechanisms could underlie this difference including differential nuclear positioning in interphase (Klaasen et al., 2022; Tovini and McClelland, 2019), distinct positions of long and short chromosomes along the metaphase plate (Rieder and Salmon, 1994; Wan et al., 2012), or variable biophysical features between long and short chromosomes. Indeed, given the biophysical nature of error detection models, error correction cues might be read differently based on the physical features of the chromosome or kinetochore in question, complicating our understanding of how tension impacts kinetochore–microtubule attachment stability. However, studying segregation outcomes is not sufficient to understand the dynamic formation of correct attachments. Instead, probing how efficiently chromosomes of different sizes become correctly bioriented sheds light not only on the cues driving error correction but also on the mechanisms underlying chromosome-based differences in mitotic trajectories and segregation outcomes.

Here, we provide direct evidence for the tension model for error correction in mammalian cells and show that differential tension at chromosomes of different sizes impacts both their biorientation and error correction efficiency. In mammalian rat kangaroo (PtK2) cells, we show that long chromosomes become bioriented later and experience higher pushing force than short chromosomes in the same cell. By enriching for errors and letting them correct in live imaging of drug washouts, we show that long chromosomes are less efficient at correcting errors, therefore delaying their biorientation. Increasing size-based force via chromokinesin Kif22/Kid overexpression increases the probability of chromosomes of any size becoming persistently stuck at spindle poles. Finally, with laser ablation, we cut chromosome arms and found that chromosome size, rather than simply identity, sets biorientation efficiency or success rate. Here, biorientation efficiency does not set a strict order of events in chromosome alignment but rather refers to the probability of successfully becoming bioriented. We propose a model in which elevated force on long chromosomes increases kinetochore tension, stabilizing kinetochore–microtubule attachments, thus prolonging the lifetime of erroneous attachments at long chromosomes. These findings provide a framework for understanding not just mechanisms of error detection, which may vary in response to chromosome-specific differences, but also aneuploidy and karyotype evolution.

To determine which physical features promote correct kinetochore–microtubule attachment in mitosis, and whether they vary across chromosomes, we assessed whether dividing rat kangaroo (PtK2) cells exhibit a bias in chromosome alignment efficiency. With just 14 chromosomes, widely ranging in size, PtK2 cells are particularly well suited to this question. We classified chromosomes by size into two, roughly even groups for easy identification: long chromosomes (≥7 μm along the longest dimension by phase microscopy) and short ones (<7 μm) (Fig. 1, A and B). We imaged live spindle assembly in PtK2 cells expressing eYFP-Cdc20 to mark kinetochores and scored the frequency of early and late aligning chromosomes belonging to either the long or short chromosome group. Here, we define early aligning chromosomes as the first three chromosomes to begin oscillating at the metaphase plate, a marker of biorientation, and late aligning chromosomes as the last three chromosomes to align (Fig. 1 C). While long and short chromosomes were evenly represented in the early aligning group, long chromosomes were significantly over-represented among late aligning chromosomes (Fig. 1, D and E; and Video 1; Fisher’s exact test P = 0.0041). Thus, on average, long chromosomes biorient and form correct attachments later in spindle assembly than short chromosomes in the same cell, suggesting less efficient or delayed attachment formation. Such a delay could be due to chromosome position differences in the spindle, intrinsic differences in chromosome or kinetochore identity, and/or differences in the formation or correction of incorrect attachments that must be resolved prior to alignment.

A longstanding error correction model posits high tension across sister kinetochores as the primary cue that promotes formation of correct attachments while disfavoring incorrect ones (Funabiki, 2019; Sarangapani and Asbury, 2014). Thus, we sought to test whether different-sized chromosomes are subject to differing forces. At correct attachments, tension is generated by the opposing force from k-fibers pulling sister kinetochores toward opposite spindle poles; in contrast, incorrect, syntelic attachments (with both sister kinetochores attached to the same spindle pole) may experience kinetochore tension when poleward pulling is counteracted by anti-poleward force along chromosome arms, and this may, in turn, affect attachment stability (Fig. 1 F). To test how polar ejection forces affect long and short chromosomes, we generated monopolar spindles by treating cells with Eg5 inhibitor S-trityl-L-cysteine (STLC) and tracked kinetochore movements with respect to the pole. Assuming roughly equal poleward pulling force at all chromosomes due to standard k-fiber size (McEwen et al., 1997, 1998), the distance from poles will be proportional to the strength of the outward polar ejection force. In live monopolar spindles, the kinetochores on long chromosomes were significantly farther from poles on average than short chromosomes (Fig. 1, G and H; and Video 2; unpaired t test P = 0.0005). Thus, long chromosomes biorient less efficiently but also experience higher pushing force than their shorter counterparts.

We hypothesized that increased pushing force at long chromosomes may generate enough tension to stabilize incorrect attachments, prolonging their lifetime and delaying congression. To test this, we enriched for errors in attachment using the Eg5 inhibitor STLC to generate monopolar spindles (Kapoor et al., 2000; Wu et al., 2018) and then washed out the drug, allowing spindle bipolarization and error correction to occur (Lampson et al., 2004). Following drug washout, as spindle poles separate, monotelic attachments (one attached sister kinetochore, one unattached), incorrect syntelic attachments (both sister kinetochores attached to the same spindle pole), and other complex attachment errors convert to bioriented attachments, which oscillate at the new metaphase plate (Fig. 2 A). We measured the frequency of these attachment types after STLC washout by immunofluorescence and found that while ∼80% of monopolar spindles exhibit 1–2 unambiguously monotelic kinetochores (Fig. S1), consistent with previous work (Kapoor et al., 2000), only ∼20% of bipoles contain monotelic attachments (Fig. S1), suggesting most chromosome alignment events following drug washout represent error correction events, though we suspect these errors transit between attachment states from syntelic to monotelic and back. We measured the time needed for chromosomes to start oscillating after drug washout and used it as a proxy for biorientation and error correction (Fig. 2, B and C; and Video 3). Alignment times were normalized to measure rank order and account for high variance in time required to achieve bipolarity between different cells (Fig. S2). We found that long chromosomes become bioriented later than short chromosomes following drug washout (Fig. 2, C and D; Mann–Whitney test P = 0.0003). In principle, this could either reflect that long chromosomes have more trouble moving within the spindle than short ones or that long chromosomes correct errors less efficiently (i.e., have a lower success rate converting errors to bioriented attachments). To determine whether drag force limits chromosome movement within the spindle, we measured the velocity of kinetochore movement on long and short chromosomes in monopolar spindles (a prolonged prometaphase-like state) to avoid differences in kinetochore maturation that could affect kinetochore grip and speed of movement in bipoles. We did not see a significant difference in kinetochore speed between groups (Fig. 2 E; P = 0.318), consistent with previous findings in grasshopper cells (Nicklas, 1965). Thus, chromosome movement velocity does not impact biorientation. Instead, these data indicate that long chromosomes take longer to correct erroneous attachments. Indeed, we find clear instances of syntelic attachments in fixed, regularly cycling cells (Fig. S1), and together with our previous findings (Fig. 2 D), this suggests that the observed alignment delay might be due to the higher forces long chromosomes experience.

To test if elevated force on long chromosomes is responsible for their delayed error correction, we globally increased polar ejection force and imaged spindle assembly. If elevated force on long chromosomes does not meaningfully contribute to kinetochore tension or attachment stability, increasing anti-poleward force should speed chromosome alignment and reduce dwell time of chromosomes near poles. Indeed, chromokinesins are known to promote chromosome alignment by pushing chromosomes toward the metaphase plate, and their depletion slows the progression to metaphase in human cells (Wandke et al., 2012). Alternately, if polar ejection force on long chromosomes can stabilize incorrect attachments, increasing this force globally may result in more delayed chromosome alignment. To discriminate between these possibilities, we perturbed the force balance between poleward pulling at the kinetochore and anti-poleward pushing along chromosome arms by overexpressing the rat kangaroo chromokinesin Kif22/Kid tagged with a HaloTag (Fig. 3 A). We confirmed chromokinesin overexpression (Kid OE) and proper localization by immunofluorescence, which showed significantly more total Kid on chromosomes in overexpressing cells compared with control (Fig. 3 B). Additionally, we measured the distance of kinetochores on long and short chromosomes from the spindle pole in monopoles and found that short chromosomes are significantly further from the pole in Kid OE than in control (Fig. S3 and Video 4), indicating an increase in ejection force on these short chromosomes.

We imaged spindle assembly and quantified the number of polar chromosomes that persisted at poles for longer than 15 min (Fig. 3, C and D; and Videos 5 and 6). Here, we define polar chromosomes as those that did not move to the spindle center but rather oriented sister kinetochores toward the same spindle pole and took on a characteristic V shape of chromosomes under polar ejection force (Fig. 3 D). Among control cells, only 53% of cells displayed persistently stuck polar chromosomes compared with 82% of Kid OE cells (Fig. 3 E), and on average, Kid OE cells had significantly more chromosomes stuck at poles per cell (Fig. 3 F; 0.9 ± 1.1 versus 2.8 chromosomes ± 2.8, mean ± SD; P = 0.0202), consistent with observations in Drosophila S2 cells (Cane et al., 2013b). This effect was a specific result of increasing polar ejection force along chromosome arms as overexpressing KidΔC, lacking 89 amino acids at the C terminus responsible for DNA binding, was indistinguishable from control cells (Fig. 3, C and F; and Video 5). Moreover, when categorized by size, instance of short chromosomes stuck at poles increased significantly with Kid OE (Fig. 3 F; 0.2 chromosomes versus 1.2 chromosomes; P = 0.0009), while long chromosomes in persistent polar attachments increased more modestly with no significant difference (0.7 chromosomes versus 1.7 chromosomes; P = 0.07). This distribution suggests a shift in the size-effect delaying chromosome alignment, where specifically increasing polar ejection force on short chromosomes (Fig. S3 and Video 4) pushes them over a tension threshold and leads to chromosomes of all sizes getting stuck in polar attachments.

Additionally, consistent with impaired error correction of chromosomes under elevated force, we observed polar chromosome attachments that do not resolve in the duration of imaging (Fig. 3 G) and cells entering anaphase with chromosomes stuck at poles as well as lagging chromosomes (Fig. 3 H). This indicates the force balance between poleward pulling at the kinetochore and anti-poleward pushing by chromokinesins impacts the ability of cells to detect attachment errors. Despite this, errors that are detected seem to reach alignment on similar timescales to controls (Fig. 3 G). Together, we find that globally increasing polar ejection force leads to an enrichment of chromosomes stuck near poles, consistent with a model in which polar ejection force contributes to kinetochore attachment stability even at polar, non-bioriented attachments.

To directly test the role of chromosome size and size-based force in chromosome alignment efficiency, we sought to acutely change chromosome size. To do so, we used laser ablation to cut chromosome arms and measured alignment efficiency. If chromosome size, either directly or indirectly, leads to differential alignment efficiency, shortening chromosome arms should promote earlier chromosome alignment with respect to other long chromosomes in the same cell (model 1; Fig. 4 A). If, instead, intrinsic chromosome-specific differences, like kinetochore composition or size (Drpic et al., 2018), dictate chromosome alignment order, ablating chromosome arms should not affect alignment efficiency (model 2; Fig. 4 A). We ablated long chromosomes in early prometaphase in cells with many unaligned long chromosomes (Fig. 4, B and C; and Video 7). Ablated chromosomes congressed to the metaphase plate before other long, unablated chromosomes in the same cell did (Fig. 4 D; Mann–Whitney test P = 0.002), indicating that chromosome size, not simply identity, sets alignment efficiency (model 1). This effect was specific to chromosome arm shortening and not due to damaging the arms as control ablations, which nick chromosomes at their ends or along their arms without shortening them (Fig. 4, E and F), did not change alignment order (Fig. 4 G). Given the indistinguishable speeds of kinetochores on long and short chromosomes that exclude any drag effect (Fig. 2 E), we conclude that the expedited alignment of ablated chromosomes is the result of reducing polar ejection forces. We propose a model wherein increased polar ejection force on long chromosomes prematurely stabilizes incorrect or incomplete attachments, causing a delay in congression and biorientation (Fig. 5).

Accurate chromosome segregation requires cells to be able to distinguish between correct and incorrect attachments, global features of the spindle, despite having only local cues. Here, we ask how the ability both to form bioriented attachments and to detect incorrect attachments varies across chromosomes in the same cell. We demonstrate that in regularly cycling PtK2 cells, with just 14 chromosomes widely ranging in size, the formation of correct, bioriented attachments is biased by chromosome size such that long chromosomes tend to be the last to become correctly attached (Fig. 1). Error correction models have long posited that the kinetochore relies on physical differences—specifically variable tension on kinetochores leading to differential kinetochore phosphorylation—to determine which attachments to reinforce (bioriented) and which to destabilize (syntelic) (Funabiki, 2019; Sarangapani and Asbury, 2014). By enriching for attachment errors, we find that mammalian error correction efficiency varies based on the physical features of the chromosome in question (Fig. 2), and specifically perturbing chromosome size-based force reveals that is the result of the kinetochore sensing and responding to differences in tension (Fig. 3 and Fig. 4). Together, these findings suggest that while all chromosomes are competent to form both correct and incorrect attachments, long chromosomes more readily stabilize incorrect attachments, delaying their alignment and biorientation (Fig. 5). In addition to defining biophysical models for error correction, this work defines new questions about protective mechanisms for not just minimizing segregation error rates but also coordinating the threshold for error detection across different chromosomes in the cell.

Overall, this work provides insight into the biophysical mechanism of error correction and reveals that the dynamics of chromosome biorientation are not chromosome agnostic. Indeed, long chromosomes exhibit distinct behaviors in spindle assembly, more frequently getting stuck at poles (Fig. 3), congressing later in mitosis (Fig. 1), and oscillating with lower amplitude than their short counterparts (Ke et al., 2009). Chromosome arm ablation speeds alignment (Fig. 4 D) indicating these are effects of chromosome size and not simply identity. Still, long and short chromosomes move at similar speeds (Fig. 2 E). While chromosome identity, kinetochore position, or early spindle position could affect the propensity to form errors and possibly contribute to the rate of their correction, our data show that the main factor regulating error correction efficiency or success rate is chromosome size. As such, we propose that differences in alignment order and tendency to dwell at spindle poles stem primarily from chromosome size-based force specifically stabilizing attachments at long polar chromosomes rather than drag force or steric effects. Consistent with this, increasing polar ejection force increases the number of stuck polar chromosomes and reduces the effect of size on the tendency to become persistently stuck at poles (Fig. 3, E and F), indicating chromosome size and force influence the resolution of incorrect attachments but not necessarily the formation of correct ones. Regardless, stuck chromosomes that do align do so on a similar timescale to those in control cells (Fig. 3 G), suggesting that beyond a certain tension threshold, additional tension will not slow alignment further but may impact the probability of error detection. This is consistent with work that shows that the tension state impacts kinetochore detachment probability (Chen et al., 2021; De Regt et al., 2022; Miller et al., 2016, 2019; Parmar et al., 2023; Sarangapani et al., 2013) and intra-kinetochore stretch, in particular, is the relevant feature for attachment stability (Drpic et al., 2015; Etemad et al., 2015; O’Connell et al., 2008; Tauchman et al., 2015). Additionally, attachment stability and error correction efficiency may also be linked to chromosome position in the spindle, leading to differential access to known regulators like Aurora A kinase (Eibes et al., 2023; Ye et al., 2015). Indeed, increasing dwell time of chromosomes near spindle poles may tip the balance to favor Aurora A–mediated error correction such that proximity to poles not only stabilizes errors on long chromosomes temporarily but also activates the pathway required for their correction after some delay. Together, these findings support a model in which high force near poles increases the propensity to stabilize polar chromosome attachments (Fig. 5). In healthy cells, elevated polar ejection force at long chromosomes might come from more numerous chromokinesins, non-kinetochore microtubules polymerizing against a larger chromatin surface and generating force, or both (Schneider et al., 2022). Regardless, this size-dependent mechanism for error detection is especially interesting considering recent work showing higher rates of missegregation of large chromosomes in human cells (Klaasen et al., 2022; Worrall et al., 2018) and formation of complex attachment errors in polar chromosomes (Tovini and McClelland, 2019; Vukušić and Tolić, 2022).

Our model, that error correction dynamics depend on chromosome size and resultant kinetochore tension, may have wide-ranging functional implications for segregation outcomes, and thus propensity toward aneuploidy. How this differential tension and attachment stability across chromosomes of different sizes alters molecular maturation of the kinetochore in Aurora B regulated processes such as Ndc80 phosphorylation (Sarangapani et al., 2021), Ska (Cheerambathur et al., 2017; Hanisch et al., 2006), and SKAP (Fang et al., 2009; Schmidt et al., 2010) recruitment, and checkpoint satisfaction (Etemad et al., 2015; Tauchman et al., 2015) remains an exciting open question. Relatedly, while we propose elevated polar ejection force opposes error correction at long chromosomes, it is also clear that polar ejection force is required for chromosome alignment (Iemura and Tanaka, 2015; Wandke et al., 2012). We see these two opposing roles as part of an important balancing act that drives evolution of kinetochore tension sensitivity. It is worth noting that missegregation rates were found to be similar across all chromosomes in PtK1 cells by fixed cell imaging (Torosantucci et al., 2009), yet these results are not inconsistent with a size-based model for error correction efficiency. Chromosome segregation errors rely on failure of both the error correction machinery and the spindle assembly checkpoint (SAC), but how tightly these pathways are coupled may vary across systems. Indeed, observations of SAC satisfaction dynamics have been variable. While Drosophila S2 cells show partial SAC satisfaction at persistently incorrect attachments (Cane et al., 2013a), live monitoring of the SAC in PtK2 cells revealed switch-like loss of Mad1 from the kinetochore (Kuhn and Dumont, 2017). Thus, how and whether persistently stuck polar chromosomes signal the checkpoint in mammalian cells remains unclear. Our work demonstrates that however many errors do or do not slip through at anaphase, the dynamics of error correction do, indeed, vary in PtK2 cells with chromosome size (Fig. 2). Future work will be required to untangle the crosstalk between the SAC and error correction.

Characterizing not just the propensity of individual chromosomes for missegregation but also the features that tune the propensity up or down will inform our understanding of different organisms’ or cell types’ oncogenic capacity and tendency toward aneuploidy. Interesting evolutionary questions arise from non-random missegregation rates. Is there evolutionary pressure to maintain oncogenes on short chromosomes, which may be less likely to missegregate? Or instead, in organisms with long chromosomes carrying oncogenes, is there selective pressure to either homogenize chromosome size over evolutionary time or develop compensatory molecular mechanisms that will equalize missegregation rates? The size dependence revealed here is likely just one of many features that impact error correction efficiency. Indeed, in different species, chromosome size (Klaasen et al., 2022; Tovini and McClelland, 2019; Worrall et al., 2018), kinetochore position (metacentric, telocentric, or holocentric) (Dumont et al., 2020), material properties (Cojoc et al., 2016), and size (Drpic et al., 2018), as well as chromatin material properties (Schneider et al., 2022), all have the potential to impact segregation outcomes. Future work on these topics will shed light on how organisms employ either a low threshold for sensing incorrectness, compensatory mechanisms that tune error detection or correction capacity on a per chromosome basis, or both to avoid catastrophic chromosomal instability. Ultimately, the findings herein provide insight into the biophysical mechanism of error detection at the kinetochore and provide a framework through which to think more broadly about how biophysical processes must shape molecular mechanisms across evolution.

PtK2 cells expressing human eYFP-Cdc20 (gift from Jagesh Shah, Harvard Medical School, Boston, MA, USA) and wildtype PtK2 cells (ATCC) were cultured at 37°C and 5% CO₂ in MEM (Invitrogen) supplemented with sodium pyruvate (Invitrogen), non-essential amino acids (Invitrogen), penicillin/streptomycin, and 10% qualified and heat-inactivated fetal bovine serum. Karyotyping and cytogenic analysis were performed on G-banded metaphase cells from the eYFP-Cdc20 PtK2 line by Cell Line Genetics. Cells were transfected using Viafect (Promega) and imaged 72 h after transfection with mCherry-α-tubulin (gift from M. Davidson, Florida State University, Tallahassee, FL, USA). Transfection reactions were prepared in 100 μl reactions using a 1:6 ratio of DNA to Viafect in OptiMEM media. Overexpression experiments were done with transient lentiviral infection. The coding sequence for the rat kangaroo chromokinesin Kif22/Kid was obtained from the PtK transcriptome (Udy et al., 2015) and cloned into a puromycin-resistant lentiviral backbone. Rat kangaroo KidΔC (89 amino acids deleted at the C terminus) was designed based on predicted secondary structure of Kid (Paysan-Lafosse et al., 2023) and homology of rat kangaroo Kid to human Kid at a helix-hairpin-helix domain known to be involved in non-specific DNA binding (Tokai et al., 1996). Lentivirus was generated in Hek293T cells, and PtK2 cells were infected 48–72 h prior to imaging spindle assembly.

Live imaging was done using an inverted (Eclipse Ti-E; Nikon) spinning-disk confocal microscope (CSU-X1; Yokogawa Electric Corporation) with Di01-T405/488/568/647 dichroic head (Semrock), 100x 1.45 Ph3 oil objective, 405 nm (100 mW), 488 nm (120 mW), 561 nm (150 mW), 642 nm (100 mW) diode lasers, emission filters (ET455/50M, ET525/50M, ET630/75M, and ET690/50M; Chroma Technology Corp.), and either an iXon3 camera (Andor Technology) (Fig. 1 G; Fig. 2 B; Fig. 3 B–H; Fig. 4, B–E; Fig. S1, A, B, and D; and Fig. S3 A) or a Zyla 4.2 sCMOS camera (Andor Technology) (Fig. 1, B and D). Cells were plated for imaging on 35-mm dishes with #1.5 poly-D-lysine–coated coverslips (MatTek Corporation). For all live experiments except drug washouts, images were collected at bin = 1, while drug washout and monopole movies were collected at bin = 2 on MetaMorph (7.8; MDS Analytical Technologies) or Micro-Manager (2.0.0). Cells were imaged in phase contrast and fluorescence in a single z-plane (Fig. 3, Fig. S3, and Fig. 4), six z-planes spaced 0.35 µm apart (Fig. 1 G and Fig. 2), or three z-planes spaced 0.35 µm apart (Fig. 1, B and D). Cells were imaged in a humidified stage-top incubation chamber (Tokai Hit) at 37°C (or 30°C for monopole and STLC washout experiments [Fig. 1 G and Fig. 2]) with 5% CO₂. Cells expressing HaloTag constructs were labeled with 200 nM JF549 or JF646 (Promega) for 30 min prior to imaging.

For all immunofluorescence experiments described below, cells were seeded on poly-L-lysine–coated coverslips. For immunofluorescence to validate chromokinesin overexpression, cells were transiently infected with lentivirus driving expression of PtK Kid-HaloTag for 48 h. Cells were then labeled with HaloTag dye JF549 and fixed in 99.8% methanol for 3 min at −20°C and permeabilized in TBS1x (Tris-buffered saline) + 2% BSA + 0.1% Triton for 30 min (IF Buffer hereafter). For immunofluorescence after STLC washout experiments, cells were treated with 7.5 μM STLC for 1 h, then washed out of solution (5× media exchanges) and incubated at 37°C for 55–85 min prior to fixation in 99.8% methanol for 3 min at −20°C and permeabilization in IF Buffer. For immunofluorescence of regularly cycling cells, these were cold-treated by incubating in 4°C media for 5 min prior to the same fixation and permeabilization conditions. For all above experiments, primary antibodies were incubated overnight at 4°C (chromokinesin overexpression) or for 1 h at room temperature (STLC washout and regularly cycling cells) at the following concentrations in IF Buffer: mouse anti-α-tubulin (DM1Α, 1:1,000, T6199; Sigma-Aldrich), rabbit anti-rat kangaroo-Kid (3 μg/ml, Genscript), human anti-CREST (1:25, 15-234-0001; Antibodies Incorporated), rat anti-tubulin (1:2,000, MCA77G; RRID: AB325003; Bio-Rad), and rabbit anti-centrin2 (1:1,000, ABE480; Millipore). The anti-rat kangaroo Kid antibody was raised in rabbits against the full Kid protein translated from the coding sequence obtained from the rat kangaroo transcriptome (Udy et al., 2015). Coverslips were washed three times in IF Buffer (5 min each) before incubating with secondary antibodies (1:500 in IF Buffer, 30 min at room temperature): goat anti-rabbit IgG Alexa Fluor 488 (A11008; Invitrogen) for Kid OE, and goat anti-rabbit IgG Alexa Fluor 647 (A21245; Invitrogen), goat anti-human IgG Alexa Fluor 561 (A21090; Invitrogen), goat anti-mouse 488 (A11001; Invitrogen), and goat anti-rat 488 (A11006; Invitrogen) for STLC washout and regularly cycling cells. Samples were washed three times in IF Buffer, incubated with Hoechst 33342 (1 μg/ml, H3570; Invitrogen) for 1 min, and washed in IF Buffer once more prior to mounting on slides with ProLong Gold Antifade Mountant (p36934; Thermo Fisher Scientific).

Laser ablation experiments were done using 514 nm ns-pulsed laser light and a galvo-controlled MicroPoint Laser System (Andor, Oxford Instruments) operated through Micro-Manager. Spindle assembly was imaged with single z-planes in both phase contrast and 488 nm (to visualize kinetochores expressing eYFP-Cdc20) to identify unaligned long chromosomes (by phase contrast). Chromosomes were selected to ablate if they were relatively isolated (a stretch of chromosome not overlapping with others), not at the spindle center, and if most of the chromosome arms were in focus at a single z-plane (little to no tilt). Ablations were performed by firing the laser at 4–8 discrete points across the width of chromosome arms (40–80 pulses of 3 ns at 20 Hz), and successful chromosome arm ablations were verified by a visible continuous gap between chromosome segments and mechanical uncoupling of ablated arms and the kinetochore-containing fragment (i.e., moving in different directions). Similarly, control ablations were considered successful if a noticeable change could be seen at the site of laser targeting (a loss of roundness at chromosome ends or visible damage along chromosome arms).

To generate monopolar spindles, Eg5 motor activity was inhibited by addition of STLC (7.5 μM). Monopoles were imaged and kinetochores were tracked for no more than 45 min. Error correction was assessed by STLC washout, imaging every 4 min for the first 40–60 min until spindles appeared roughly bipolar but had not reached metaphase, then every 20 s thereafter.

Kinetochores were manually tracked using the MTrackJ plugin on FIJI (Schindelin et al., 2012) with a combination of eYFP-Cdc20 and phase contrast to follow individual kinetochores and measure the size of chromosomes. For kinetochores that left the imaging plane for a short period during a movie, a rough position (polar or at the metaphase plate) was recorded using the phase contrast channel. Chromosome size was measured in two dimensions using the segmented line tool on FIJI on single z-plane phase contrast images to measure the longest dimension of a chromosome. In a crowded spindle, kinetochore tracks were used to find the clearest frame for size measurement. Long chromosomes were considered those ≥7 µm and short chromosomes were those <7 µm in length.

For alignment order (Fig. 1 D) and alignment time (Fig. 2, C and D; Fig. 3, and E; and Fig. 4, D and G), the metaphase plate was defined as the region between poles in which oscillating chromosomes resided. Oscillating chromosomes were defined as chromosomes with an inter-kinetochore (K-K) distance of ≥1.8 μm moving periodically toward one spindle pole and then the other. Distance to pole (Fig. 1 H) was calculated in monopolar spindles by finding the shortest distance between kinetochore position and pole position (found by MTrackJ using mCherry-α-tubulin fluorescence) and speed of kinetochore movement (Fig. 4 E) was determined by measuring the change in kinetochore-to-pole distance over the change in time.

For immunofluorescence, brightness and contrast for each channel were scaled identically within each experiment. A threshold mask (Yen method) was applied using the Hoechst signal (for chromosome selection) to determine the area in which we would measure fluorescence intensity of our protein of interest. Fluorescence intensity was measured of both the chromokinesin Kid and the reference (Hoechst) and normalized by the intensity of the reference (Fig. 3 B). Three independent experiments were performed and quantified separately, yielding comparable results.

Statistical analyses were performed in GraphPad Prism 9. For parametric datasets (Fig. 1 H and Fig. S3 B), the two-tailed student’s t test was used. For non-parametric datasets (Fig. 2, D and E; Fig. 3, B, F, and G; and Fig. 4, D and G), the two-tailed Mann–Whitney was used. Fisher’s exact test was used in Fig. 1D.

Fig. S1 shows that the chromosomes experiencing delayed biorientation may be delayed as a result of syntelic attachment in both STLC washout and regularly cycling cells. Fig. S2 shows that the time to biorientation after STLC washout varies widely between cells. Fig. S3 shows that the chromokinesin overexpression increases polar ejection force experienced by short chromosomes. Video 1 shows that long chromosomes align later than short chromosomes during spindle assembly, related to Fig. 1. Video 2 shows that the polar ejection forces push long chromosomes further than short ones from the pole of monopolar spindles, related to Fig. 1. Video 3 shows that the monopolar spindles regain bipolarity and correct errors following STLC drug washout, related to Fig. 2. Video 4 shows that the chromokinesin Kid overexpression increases polar ejection force on short chromosomes, related to Fig. S3. Video 5 shows the effects of Kid overexpression on spindle assembly depend on chromosomal localization, related to Fig. 3. Video 6 shows that chromosomes of all sizes become persistently stuck at poles under high polar ejection force induced by Kid overexpression, related to Fig. 3. Video 7 shows that the reducing chromosome arm size by laser ablation speeds biorientation during spindle assembly, related to Fig. 4.

All the data generated in this paper have been deposited at Mendeley Data and are publicly available as of the date of this publication (https://doi.org/10.17632/v2d59hzyj6.2).

We thank Zachary Mullin-Bernstein for his help with karyotyping. We thank Jagesh Shah for the gift of the PtK2 eYFP-Cdc20 cell line. We thank Hironori Funabiki, Charles Asbury, Sue Biggins, Michael Lampson, Wallace Marshall, Geeta Narlikar, Orion Weiner, and members of the S. Dumont laboratory for helpful discussions.

This work was supported by the National Institutes of Health (NIH) R35GM136420, National Science Foundation Center for Cellular Construction DBI-1548297, and NIH F31CA265136 and NIH T32EB009383 (M.K. Chong). S. Dumont is a Chan Zuckerberg Biohub investigator.

Author contributions: M.K. Chong: Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Project administration, Software, Supervision, Validation, Visualization, Writing—original draft, Writing—review and editing; M. Rosas-Salvans: Conceptualization, Methodology, Writing—review and editing; V. Tran: Investigation, Methodology, Writing—review and editing; S. Dumont: Conceptualization, Funding acquisition, Project administration, Supervision, Writing—review and editing.