Untangling the contribution of Haspin and Bub1 to Aurora B function during mitosis

To maintain genomic integrity during mitosis, the duplicated chromosomes need to be correctly distributed over the two daughter cells. This requires that sister chromatids become connected to microtubules emanating from opposing poles of the mitotic spindle (amphitelic attachment). Microtubules attach to chromosomes via specialized protein structures called kinetochores, which assemble on centromeres (Musacchio and Desai, 2017). Formation of correct, amphitelic attachments of kinetochore microtubules (kMTs) is facilitated by a dynamic kinetochore–microtubule interface (KT–MT) that allows the detachment of improper connections such as syntelic attachments (both kinetochores attached to microtubules from the same mitotic spindle pole) or merotelic attachments (one kinetochore attached to microtubules from both sides of the mitotic spindle), and the stabilization of amphitelic attachments. A key player in this “error correction” process is the chromosomal passenger complex (CPC), consisting of Aurora B kinase, INCENP, Survivin, and Borealin. Aurora B destabilizes KT–MT attachments by phosphorylating several outer kinetochore proteins that directly bind microtubules, including components of the Knl1/Mis12 complex/Ndc80 complex (KMN) network (Cheeseman et al., 2006; Cimini et al., 2006; DeLuca et al., 2006; Tanaka et al., 2002; Welburn et al., 2010). Destabilization of KT–MT attachments transiently generates unattached kinetochores, which provide the sister chromatids with another opportunity to be captured by microtubules. Additionally, unattached kinetochores activate the mitotic checkpoint, a surveillance mechanism that prevents the onset of anaphase until all kinetochores have become attached to microtubules of the mitotic spindle (Foley and Kapoor, 2013; Lampson and Cheeseman, 2011). Aurora B also feeds into the mitotic checkpoint in a more direct way by facilitating the rapid recruitment of the essential checkpoint kinase Mps1 to kinetochores (Santaguida et al., 2011; Saurin et al., 2011) and by phosphorylating the kinetochore protein Knl1. Phosphorylation of Knl1 prevents the binding of PP1y, the phosphatase that counteracts Mps1-dependent phosphorylation of Knl1 (Liu et al., 2010; Nijenhuis et al., 2014). Thus, Aurora B contributes to faithful chromosome segregation by facilitating error correction and mitotic checkpoint maintenance.

During the early stages of mitosis, Aurora B is predominantly observed at the inner centromere, a specialized region on the chromatin that lies at the intersection of the inter-kinetochore axis and the inter-sister chromatid axis (Hindriksen et al., 2017a; Yamagishi et al., 2010). The typical inner centromere localization of Aurora B is considered important for its activity toward substrates at the outer kinetochore: it concentrates Aurora B kinase in proximity of these substrates, while at the same time allowing spatial regulation of kinetochore substrate phosphorylation (Andrews et al., 2004; Krenn and Musacchio, 2015; Liu et al., 2009; Tanaka et al., 2002; Wang et al., 2011; Welburn et al., 2010).

Two evolutionarily conserved kinases, Haspin and Bub1, direct the docking of the CPC to the inner centromere. The cohesin-associated kinase Haspin phosphorylates histone H3 on threonine 3 (H3T3ph), and H3T3ph directly interacts with the CPC via Survivin (Dai et al., 2005; Du et al., 2012; Jeyaprakash et al., 2011; Kelly et al., 2010; Niedzialkowska et al., 2012; Wang et al., 2010; Yamagishi et al., 2010). The kinetochore-localized kinase Bub1 phosphorylates (centromeric) histone H2A on threonine 120 (H2AT120ph). H2AT120ph recruits the paralogs Shugoshin 1 and Shugoshin 2 (Sgo1/2), which in turn bind to the CPC subunit Borealin (Kawashima et al., 2007, 2010; Liu et al., 2015; Tsukahara et al., 2010; Yamagishi et al., 2010). The prevailing model is that the CPC is recruited to the chromatin region where H3T3ph and H2AT120ph overlap (Yamagishi et al., 2010), implying that the CPC inner centromere confinement may be defined by simultaneous interactions of the CPC with H3T3ph and Sgo1/2 that localize to H2AT120ph (Krenn and Musacchio, 2015; Trivedi and Stukenberg, 2016). However, the two histone marks do not evidently overlap: H3T3ph appears as a single dot at the inner centromere, while H2AT120ph can be observed as two foci that are in closer proximity to the kinetochore (Liu et al., 2013; Yamagishi et al., 2010). This led us to hypothesize that instead of converging to accumulate the CPC precisely at the inner centromere, Haspin and Bub1 might each recruit a separate functional CPC pool to the centromere (Fig. 1 A). Indeed, we demonstrate that Haspin or Bub1 kinase activity can individually recruit Aurora B to an ectopic chromosomal locus. Furthermore, a detailed analysis of HCT116 Haspin CRISPR mutant (CM) cells revealed a small, Bub1-dependent Aurora B pool that was sufficient for faithful chromosome segregation in otherwise unchallenged cells. Inhibition of Bub1 in Haspin CM cells abolished accumulation of Aurora B in the centromere region, and while this increased the frequency of anaphases with lagging chromosomes, it surprisingly did not compromise the mitotic checkpoint, or the phosphorylation of the Aurora B kinetochore substrates Dsn1, Hec1, and Knl1. Our findings suggest that the correction of erroneous KT–MT attachments is more sensitive to reductions in centromeric Aurora B levels than activation of the mitotic checkpoint and the phosphorylation of the tested kinetochore substrates.

We hypothesized that Haspin and Bub1 might each localize a functional CPC pool near the centromere, instead of collectively creating a local environment where the CPC makes simultaneous interactions with H3T3ph and H2AT120ph, in case of the latter most likely via Sgo1/2 (Fig. 1 A; Tsukahara et al., 2010; Yamagishi et al., 2010). To discriminate between these possibilities, while excluding effects of cross-talk between the different proteins, we made use of U-2 OS-LacO cells, which harbor an array of lac operator repeats on chromosome 1 (Janicki et al., 2004). Expression of Haspin or Bub1 fused to LacI-GFP allowed us to assess CPC recruitment to the ectopic LacO locus by each individual kinase (Fig. 1 B). As readout for CPC recruitment, we measured Aurora B levels using quantitative immunofluorescence (IF) microscopy. Expression of LacI-GFP-Haspin in U-2 OS-LacO cells resulted in local phosphorylation of H3T3 and in the recruitment of Aurora B, in contrast to cells expressing LacI-GFP (Fig. 1, C–E; and Fig. S1 A). Aurora B recruitment required Haspin kinase activity, because expression of a kinase dead mutant of Haspin did not result in H3T3ph signal at the LacO locus or in Aurora B recruitment (Fig. 1, C–E; and Fig. S1 A).

To study the capacity of Bub1 to recruit the CPC, we used a LacI-GFP-Bub1 fusion protein harboring a mutation in its GLEBS domain (E252K, hereafter called LacI-GFP-Bub1^E252K; Fig. S1, B and C; Krenn et al., 2012; Overlack et al., 2015; Ricke et al., 2012). Bub1 recruitment to the kinetochore protein Knl1 depends on the interaction of its GLEBS domain with Bub3 (Larsen et al., 2007; Primorac et al., 2013; Taylor et al., 1998). The use of Bub1^E252K excluded recruitment of kinetochore proteins to the LacO locus, which we observed when expressing WT LacI-GFP-Bub1 (Fig. S1, D–H). In contrast to LacI-GFP, ectopic localization of LacI-GFP-Bub1^E252K resulted in local phosphorylation of H2AT120 and recruitment of Sgo1, Sgo2, and Aurora B, which was not observed upon ectopic targeting of a kinase dead mutant of Bub1^E252K (Fig. 1, F–K; and Fig. S1, B and C). Since Bub1 is thought to recruit the CPC via Sgo1/2 (Kawashima et al., 2007, 2010; Tsukahara et al., 2010), we bypassed Bub1 and expressed either Sgo1-LacI-GFP or LacI-GFP-Sgo2. Indeed, both LacI fusion proteins recruited Aurora B to the LacO array (Fig. S1, I–K).

To exclude the possibility that endogenous Bub1 kinase activity contributed to Aurora B recruitment by ectopically localized Haspin, we analyzed Aurora B recruitment by LacI-GFP-Haspin in cells depleted of Bub1 using siRNA (Fig. S1, L–O). Transfection of siBub1 abolished chromatin-associated H2AT120ph, indicating that Bub1 depletion was successful (Fig. S1, L and N). In the absence of phosphorylated H2AT120, LacI-GFP-Haspin could still recruit Aurora B to the LacO array (Fig. S1, L–O). To analyze if Aurora B recruitment by Bub1 could occur independently of Haspin activity, we mutated the endogenous Haspin gene in U-2 OS-LacO cells using CRISPR/Cas9 (U-2 OS-LacO Haspin CM; Table S1). While H3T3 was no longer phosphorylated in these cells, expression of LacI-GFP-Bub1^E252K was still capable of recruiting Aurora B to the LacO array (Fig. 1, L–N; and Fig. S1, P and Q). These data show that either Haspin or Bub1 activity is sufficient to recruit Aurora B to an ectopic locus.

Next, we performed a detailed analysis of Aurora B localization in cells in the absence of Haspin and/or Bub1 kinase activity, to investigate if Haspin and Bub1 could also recruit separate pools of the CPC to centromeres. We generated two Haspin mutants in HCT116 cells using CRISPR/Cas9, each generated with a unique single guide RNA (sgRNA) targeting GSG2 (the gene encoding Haspin). The first gRNA targets GSG2 just before the kinase domain at position bp 1306–1308, yielding Haspin CM1, while the second gRNA targets the start of GSG2, at position bp 39–41, hereafter called Haspin CM2. In each cell line, both Haspin alleles were mutated, resulting in premature stop codons (Table S1). Loss of Haspin activity was further confirmed by the lack of H3T3ph detected by Western blot and by IF imaging (Fig. S2, A, B, and D). Compared with WT HCT116 cells, centromeric levels of Aurora B were reduced by ∼50% in the Haspin CM cell lines (Fig. 2, A and C; and Fig. S2, A and C). Interestingly, chromosome spreads consistently revealed two kinetochore-proximal foci of Aurora B in the Haspin CM cells, similar to previous observations for Borealin in cells treated with the Haspin inhibitor 5-ITu (Bekier et al., 2015; Fig. 2, D and F). Because Haspin also helps to maintain centromeric cohesin (Dai et al., 2006, 2009; Liang et al., 2018; Zhou et al., 2017), it was unclear if the two Aurora B foci represented an alternative CPC pool, or reflected separated sister centromeres caused by weakened cohesion. Measurements of CENP-C/CENP-C distances in these chromosome spreads showed that on average, inter-kinetochore distances were increased in the Haspin CM cells as compared with WT cells (Fig. 2, D and E). However, even on chromosomes with CENP-C/CENP-C distances comparable to WT cells (552 nm), Aurora B was observed as two kinetochore-proximal foci in the Haspin CM cells (556 nm, Fig. 2, D-F). In WT HCT116 cells, Aurora B was mainly localized to the inner centromere, even on chromosomes with larger inter-kinetochore distances (Fig. 2, D and F). Interestingly, in WT cells, a small kinetochore-proximal pool of Aurora B was also observed, in addition to the main inner centromere pool (Fig. 2, D and F). These observations implied that the kinetochore-proximal Aurora B localization in Haspin CM cells most likely represents an alternative pool of the CPC, distinct from that at the inner centromere. Because of its localization, proximal to the kinetochore, we hypothesized this pool of Aurora B may depend on Bub1 activity. Indeed, inhibition of Bub1 kinase activity using the small molecule inhibitor BAY-320 (Baron et al., 2016) abolished this pool of Aurora B in Haspin CM cells (Fig. 2, A–D). This suggests that Bub1 activity can recruit a small, kinetochore-proximal pool of Aurora B to centromeres independently of Haspin.

Of note, we chose to inhibit Bub1 kinase activity instead of depleting Bub1, because Bub1 also has nonenzymatic functions, including the recruitment of mitotic checkpoint proteins to the kinetochore, a function that is maintained after inhibition of its kinase activity (Baron et al., 2016; Meraldi and Sorger, 2005; Raaijmakers et al., 2018; Rodriguez-Rodriguez et al., 2018; Zhang et al., 2019). As previously reported, treatment of cells with the Bub1 inhibitor BAY-320 markedly reduced H2AT120ph and Sgo1 levels at the centromere (Fig. 2, A and B; and Fig. S2 E; Baron et al., 2016). Centromeric levels of Aurora B were reduced by ∼40% in cells treated with BAY-320 (Fig. 2, A and C). Inspection of chromosome spreads showed that in Bub1-inhibited cells, Aurora B localized along the inter-sister chromatid axis, although some enrichment in the inner centromere region was still visible (Fig. 2 D). While we cannot fully exclude that this is caused by residual Bub1 activity, the observed localization is similar to what has been reported for Bub1 knockdown in cell lines and for mouse embryonic fibroblasts derived from Bub1 kinase-dead mutant mice (Ricke et al., 2012; Yamagishi et al., 2010). The inter-sister chromatid and inner centromere pool of Aurora B, observed upon Bub1 inhibition in WT cells, is not observed when Bub1 is inhibited in Haspin CM cells, strongly suggesting that Haspin is responsible for the recruitment of this pool of Aurora B (Fig. 2, A and D). Bub1 inhibition in Haspin CM cells resulted in low levels of residual Aurora B, dispersed over the chromatin, and quantifications at the centromeres revealed a reduction in Aurora B levels of ∼70% (Fig. 2, A and C). Collectively, our data support the idea that Haspin and Bub1 can recruit separate pools of Aurora B to the inter-sister chromatid region and the kinetochore-proximal centromere, respectively, in early mitosis.

To study the function of the two observed pools of the CPC during mitosis, we followed mitotic progression by live cell imaging using WT HCT116 or Haspin CM cells stably expressing H2B-mCherry. WT HCT116 and Haspin CM cells aligned their chromosomes on the metaphase plate with similar timing, and the incidence of chromosome segregation errors in anaphase in Haspin CM cells was comparable to WT cells (Fig. 3, A–C). Bub1 inhibition delayed anaphase onset by ∼10 min, due to a delay in chromosome congression (Fig. 3, B and C). Bub1 inhibition had only a minor effect on chromosome segregation fidelity (Fig. 3 A). This suggests that both the Aurora B pool controlled by Haspin and the pool controlled by Bub1 are largely capable of supporting faithful chromosome segregation. However, precluding centromeric accumulation of Aurora B, treatment of Haspin CM cells with BAY-320 caused a substantial increase in the fraction of cells with anaphase lagging chromosomes (Fig. 3 A). Importantly, the mitotic defects observed in Bub1-inhibited Haspin CM cells differed from the defects induced by Aurora B kinase inhibition. Inhibition of Aurora B with the small molecule inhibitor ZM447439 in WT HCT116 or Haspin CM cells severely disturbed chromosome alignment, and cells exited mitosis without discernable anaphases (Fig. 3, A and B), similar to previously reported effects of Aurora B inhibition in other cell lines (Cimini et al., 2006; Ditchfield et al., 2003; Girdler et al., 2008; Harrington et al., 2004; Hauf et al., 2003; Tao et al., 2008). This suggests that despite the lack of centromere-concentrated Aurora B in Bub1-inhibited Haspin CM cells, residual kinase activity limits the severity of chromosome bi-orientation and segregation errors.

Our data suggest that the two Aurora B pools, controlled by either Haspin or Bub1, are largely redundant for faithful chromosome segregation during an otherwise unperturbed mitosis. However, error correction becomes more challenging when many faulty attachments are established early in mitosis (Ganem et al., 2009). Under these conditions, both Haspin and Bub1 might be required to accumulate sufficient amounts of active Aurora B at centromeres. To increase the frequency of erroneous KT–MT attachments, we inhibited Eg5 with monastrol to accumulate cells in mitosis with monopolar spindles (Kapoor et al., 2000; Khodjakov et al., 2003; Mayer et al., 1999). We then let cells progress through mitosis for a range of time points (from 45 to 180 min) after release from the monastrol block in the presence of the proteasome inhibitor MG132 to prevent anaphase onset (Lampson et al., 2004). The fraction of cells with fully aligned and misaligned chromosomes was scored and used as a proxy for error correction (Fig. 4, A and B). After 45 min, ±80% of WT cells had achieved full alignment, compared with ±49% for the Haspin CM cell lines (Fig. 4 B). After 90 min, WT cells reached maximal alignment (±91%). Importantly, Haspin CM cells largely recovered with ±81% and ±76% of cells showing complete alignment for Haspin CM1 and CM2, respectively. This suggests that Haspin CM cells are capable of correcting erroneous KT–MT attachments when given sufficient time. However, Haspin CM cells displayed a dramatic drop in the number of correct metaphases after 180 min with only ±25% of cells showing full chromosome alignment. Closer inspection of these cells revealed mainly chromosome scattering, as opposed to chromosome misalignment (Fig. 4 C). Since Haspin also contributes to maintenance of centromeric cohesion by locally blocking the activity of the cohesin removal factor WAPL (Dai et al., 2006; Liang et al., 2018; Zhou et al., 2017), we tested if depletion of WAPL could rescue the observed alignment defect in Haspin CM cells. Indeed, depletion of WAPL by siRNA transfection rescued chromosome alignment after 180 min but not after 60 min (Fig. 4, C–F), suggesting that the alignment defect observed early after the monastrol release is not a consequence of cohesin loss but due to impaired error correction.

We next studied the consequences of Bub1 inhibition on chromosome alignment in WT and Haspin CM cells, 60 min after monastrol release. The addition of BAY-320 did not affect chromosome alignment in WT cells; however, Bub1 inhibition exacerbated the alignment defect in Haspin CM cells, with only ±35% of cells reaching full alignment, compared with ±60% in Haspin CM cells. Moreover, depletion of WAPL was unable to rescue alignment (Fig. 4, E and F). To test if the congression defect observed in the Haspin CM cells plus Bub1 inhibition could recover when given more time to align, we also measured alignment after 3 h (Fig. S3 A). When Bub1 was inhibited in Haspin CM cells, it rendered the cells largely resistant against rescue by WAPL depletion, suggesting that the observed alignment defects were due to perturbed error correction and not loss of cohesion.

Taken together, our data show that Haspin loss only delays error correction, while the combined loss of Haspin and Bub1 activity severely compromises error correction. To test if the observed alignment defects after monastrol release were caused by the reduced centromeric Aurora B levels in Bub1-inhibited Haspin CM cells (Fig. 2, A and C), we expressed CENP-B (CB)–mCherry or CB-INCENP-mCherry in a doxycycline-inducible manner in HCT116 WT and Haspin CM cells (Huang et al., 2018; Liu et al., 2009; van der Waal et al., 2012). Addition of doxycycline restored centromeric Aurora B levels in CB-INCENP–expressing cells, in contrast to cells expressing CB-mCherry (Fig. S3, B and C). Expression of CB-INCENP but not CB-mCherry resulted in a near complete rescue of alignment in Bub-inhibited Haspin CM cells, with ±81% of cells reaching full alignment (Fig. 4 G). This suggests that the centromeric levels of Aurora B become the limiting factor for efficient error correction upon the concurrent loss of Haspin and Bub1 activity.

While our data support the idea that Haspin and Bub1 recruit separate functional pools of Aurora B to centromeres, the mitotic defects observed in Bub1-inhibited Haspin CM cells were less severe than the defects induced by Aurora B kinase inhibition, suggesting residual Aurora B kinase activity in the absence of centromere concentrated Aurora B. Aurora B phosphorylates multiple kinetochore substrates in the KMN network, resulting in a dynamic KT–MT interface that facilitates correction of faulty microtubule attachments (Bakhoum et al., 2009b; DeLuca et al., 2011; Welburn et al., 2010; Zaytsev et al., 2016). We therefore performed quantitative IF for phosphorylated Hec1 (S44) and Dsn1 (S109), two Aurora B substrates of the KMN network (Kim and Yu, 2015; Welburn et al., 2010; Yang et al., 2008). Cells were blocked in mitosis after release from a G2 arrest into nocodazole in the presence or absence of BAY-320 (Fig. 5, A–D). Remarkably, phosphorylation of both Hec1 at S44 and Dsn1 at S109 were not affected by Haspin knockout, Bub1 inhibition, or the combined loss of Haspin and Bub1 activity (Fig. 5, A–D). This is in line with the notion that loss of centromeric Aurora B still permits residual Aurora B activity (Fig. 3).

Since the net phosphorylation status of a kinase substrate depends on the balance between kinase and phosphatase activity, we tested if perturbation of Haspin and Bub1 activity might reduce the activity of local phosphatases that counteract Dsn1 S109 and Hec1 S44 phosphorylation. Addition of the small molecule Aurora B inhibitor ZM447439 to WT- or Bub1-inhibited Haspin CM cells (Ditchfield et al., 2003) resulted in a comparable decrease in Dsn1 and Hec1 phosphorylation, suggesting that the activity of Aurora B counteracting phosphatases is largely unaffected in the absence of Haspin and Bub1 activity (Fig. S4, A–D). In fact, when titrating ZM447439 into Bub1-inhibited Haspin CM cells, these cells lost Dsn1 S109 phosphorylation at an even lower concentration of ZM447439 than WT cells (Fig. S4 G).

We then reasoned that while the steady-state levels of Dsn1 S109 and Hec1 S44 phosphorylation remain unchanged in mitotic cells with no visible enrichment of centromeric Aurora B, perhaps the kinetics of kinetochore substrate phosphorylation was compromised: in the absence of centromere-concentrated Aurora B, cells may require more time to phosphorylate outer-KT substrates. To test this, Dsn1 S109 and Hec1 S44 phosphorylation were followed during mitotic progression by quantitative IF. We released WT or Haspin CM cells from an RO3306-induced G2 arrest in the presence or absence of BAY-320 and discriminated early phases of mitosis by nuclear envelope breakdown. Based on DAPI (chromatin morphology) and Lamin B (nuclear envelope) staining, we identified four categories of early mitotic cells (Fig. 5 E). Phosphorylation of Dsn1 S109 was already high in the category 1 cells and only slightly increased until cells reached category 3, and no clear differences could be observed between WT- and Bub1-inhibited Haspin CM cells (Fig. 5, E and F). In WT cells, phosphorylation of Hec1 S44 increased while cells progressed through mitosis, peaking in category 3 and decreasing in metaphase (category 5), in agreement with previous observations (Fig. 5, E and G; DeLuca et al., 2006, 2011; Welburn et al., 2010). We observed a similar pattern of Hec1 S44 phosphorylation in Bub1-inhibited Haspin CM cells. However, the peak of Hec1 phosphorylation appeared to shift from category 3 to category 2 cells, indicating that the kinetics of Hec1 phosphorylation might be slightly accelerated (Fig. 5, E and G). Moreover, Hec1 S44 phosphorylation levels in metaphase appeared similar in all conditions, suggesting that dephosphorylation of Hec1 S44 in metaphase is not affected by the loss of Haspin and Bub1 activity (Fig. 5, E and G).

Finally, we also tested if the observed kinetochore phosphorylation was caused by a recently described microtubule-dependent pool of the CPC (Trivedi et al., 2019). We blocked cells in mitosis in the presence of increasing concentration of nocodazole and again measured kinetochore phosphorylation using quantitative IF. However, we saw no difference in Hec1 S44 phosphorylation using a nocodazole concentration described to fully prevent microtubule formation (3.3 µM; Fig. S4, H and I).

Taken together, our data suggest that phosphorylation of the tested outer kinetochore substrates does not require centromeric accumulation of Aurora B, and that the limited error correction efficiency observed in Bub1-inhibited Haspin CM cells cannot be explained by reduced Hec1 phosphorylation.

We next considered reductions in the levels of centromeric MCAK (mitotic centromere-associated kinesin; KIF2C) as a possible cause of the limited error correction in Haspin- and Bub1-deficient cells. MCAK is a microtubule depolymerase that controls mitotic spindle assembly and faithful chromosome segregation (Bakhoum et al., 2009a,b; Kline-Smith et al., 2004; Maney et al., 1998; Walczak et al., 2002). Recruitment of MCAK to centromeres depends on Bub1 and Aurora B kinase activities. Specifically, Bub1 phosphorylates histone H2A T120, resulting in recruitment of Sgo2 to centromeres. Sgo2 then recruits MCAK in an Aurora B kinase–dependent manner (Huang et al., 2007; Kawashima et al., 2010; Kitajima et al., 2005; Tanno et al., 2010). MCAK recruitment was shown to be strongly reduced after loss of Haspin or Haspin activity (De Antoni et al., 2012; Wang et al., 2010, 2012), and we reasoned this might explain the delayed error correction observed in the Haspin CM cells (Fig. 4, A and B). Surprisingly, quantitative IF did not reveal an obvious decrease in centromeric MCAK levels in Haspin CM cells compared with WT cells blocked in mitosis with nocodazole (Fig. 6, A and B; and Fig. S5, A and B). siRNA-mediated knockdown of Haspin or Haspin inhibition by 5-ITu resulted in a small decrease in centromeric MCAK levels in WT HCT116 cells, but not to the extent previously reported (Fig. S5, C–H; De Antoni et al., 2012; Wang et al., 2010, 2012). Instead, we observed a redistribution of MCAK from the inner to the KT proximal centromere in Haspin CM and Haspin knockdown cells. In addition, we also observed a more continuous distribution of MCAK between the two CENP-C foci (Fig. 6 A). The displacement of MCAK followed Sgo2, in line with Sgo2 being the main MCAK receptor at centromeres (Fig. 6, C and D). After inhibition of Bub1 with BAY-320 in WT and Haspin CM cells, both MCAK and Sgo2 levels were strongly reduced (Fig. 6, A–D), supporting the idea that Bub1 activity is the main driver of MCAK centromere localization via Sgo2 (Fig. 6, A–D; Baron et al., 2016; Huang et al., 2007; Kawashima et al., 2010; Kitajima et al., 2005; Tang et al., 2004; Tanno et al., 2010; Williams et al., 2017). Indeed, both LacI-Bub1 and LacI-Sgo2 could recruit MCAK to ectopic LacO arrays both in WT and in Haspin CM cells, while Sgo1 and LacI-Haspin did not (Fig. 6, E–G; and Fig. S5, F–H).

Taken together, these data suggest that Haspin is not directly involved in the recruitment of Sgo2 and MCAK toward centromeres. However, Haspin seems to promote the inner centromere positioning of both proteins, potentially via regulation of local Aurora B activity. It remains unclear if the alternative localization of MCAK in Haspin knockout cells could explain the observed delay in chromosome alignment after monastrol wash-out (Fig. 4). The inner centromere pool of Aurora B has been suggested to inhibit the inner centromere pool of MCAK during early mitosis, with MCAK activity rising later in mitosis when MCAK translocates toward the kinetochore, away from the inner centromere pool of Aurora B, and resolving merotelic attachments during metaphase (Andrews et al., 2004; Bakhoum et al., 2009b; Gómez et al., 2007; Knowlton et al., 2006; Lan et al., 2004; Manning et al., 2007; McHugh et al., 2019). In Haspin CM cells, Aurora B displays a similar redistribution as Sgo2 and MCAK, and may as such inhibit the activity of the alternatively localized MCAK. However, the fact that we do not observe an increase in anaphase errors in Haspin CM cells argues against this scenario (Fig. 3).

Finally, apart from regulating the stability of KT–MT attachments, Aurora B activity is also required for a robust mitotic checkpoint response (Ditchfield et al., 2003; Santaguida et al., 2011; Saurin et al., 2011; Vader et al., 2007). To examine the contribution of centromere accumulation of Aurora B in sustaining the mitotic checkpoint, we measured the time spent in mitosis in the presence of unattached kinetochores by depolymerizing spindle microtubules with a high dose of nocodazole (3.3 µM; Fig. 7). Aurora B inhibition by ZM447439 accelerated exit from mitosis in the presence of nocodazole, supporting the notion that a robust mitotic checkpoint response requires Aurora B activity (Fig. 7 C; Ditchfield et al., 2003; Santaguida et al., 2011; Saurin et al., 2011; Vader et al., 2007). However, in Bub1-inhibited Haspin CM cells, the mitotic checkpoint was strong enough to prevent mitotic exit (Fig. 7 B). To determine if Haspin CM in combination with Bub1 inhibition may have a more subtle effect on the mitotic checkpoint, we measured the extent of the mitotic delay in cells treated with nocodazole together with a low dose of reversine to partially inhibit the mitotic checkpoint kinase Mps1 (Santaguida et al., 2010). In the presence of nocodazole and reversine, WT cells gradually exited mitosis, and mitotic exit was accelerated by inhibition of Aurora B (Fig. 7, D and F), again confirming the contribution of Aurora B kinase activity to a functional mitotic checkpoint. However, even in this sensitized condition, Haspin CM, Bub1 inhibition, or a combination of these did not reduce the time spent in mitosis (Fig. 7 E), suggesting that centromeric recruitment of Aurora B by either Haspin or Bub1 is dispensable for a functional mitotic checkpoint and that the residual Aurora B kinase activity that phosphorylates the kinetochore substrates Hec1 and Dsn1 (Fig. 5) also supports a robust mitotic checkpoint response (Fig. 7). In line with this notion, we found that S60 in the RVSF motif of Knl1 is phosphorylated in nocodazole-treated Bub1-inhibited Haspin CM cells (Fig. 7, G and H). Importantly, Knl1 S60 phosphorylation was reduced to a similar extent after addition of ZM447439 in both WT and Haspin CM cells inhibited with BAY-320, suggesting that the RVSF motif counteracting phosphatases is not perturbed by loss of Haspin and Bub1 activity (Fig. 7 H). Phosphorylation of the RVSF motif in Knl1 by Aurora B interferes with Knl1 binding to PP1y, the phosphatase that counteracts Mps1-dependent phosphorylation of Knl1 (Liu et al., 2010; Nijenhuis et al., 2014). As such, the residual Aurora B activity may support the mitotic checkpoint.

The combined kinase activities of Haspin and Bub1 drive the precise accumulation of Aurora B at the inner centromere and as such are thought to jointly control Aurora B activity toward kinetochore substrates involved in error correction and mitotic checkpoint activity (Hindriksen et al., 2017b; Kawashima et al., 2007, 2010; Kelly et al., 2010; Santaguida et al., 2011; Saurin et al., 2011; Tsukahara et al., 2010; Wang et al., 2010; Yamagishi et al., 2010). Here, we demonstrate that either Haspin or Bub1 kinase activity is sufficient to recruit Aurora B to an ectopic chromatin locus, and in line with this, we identified two centromere pools of Aurora B. First, inhibition of Bub1 activity revealed a Haspin-dependent pool of Aurora B that accumulated along the inter sister chromatid axis and at the inner centromere, as previously observed by others (Baron et al., 2016; Boyarchuk et al., 2007; Kawashima et al., 2007; Vanoosthuyse et al., 2007; Yamagishi et al., 2010). Second, loss of Haspin activity in Haspin CM cells exposed a small Bub1-dependent kinetochore-proximal pool of Aurora B in Haspin CM cells. This kinetochore-proximal pool of Aurora B was also infrequently observed in WT HCT116 cells, and we therefore propose that this Bub1-dependent kinetochore-proximal pool of Aurora B is present, but masked by the much higher levels of Aurora B recruited to the inner centromere by Haspin.

Recruitment of the CPC to the inner centromere is thought to serve several purposes. First, it locally concentrates and clusters the CPC, which is required to fully activate Aurora B (Kelly et al., 2007, 2010; Krenn and Musacchio, 2015; Wang et al., 2011; Zaytsev et al., 2016). Second, it is thought to confer Aurora B substrate specificity as it places the kinase in proximity to its substrates at the kinetochore (Fu et al., 2009; Hans et al., 2009; van der Horst and Lens, 2014). Third, CPC inner centromere localization would support tension-dependent phosphoregulation of Aurora B kinetochore substrates: In the presence of inter-kinetochore tension, Aurora B confined to the inner centromere is no longer be able to reach its substrates at the kinetochore, and hence only bi-oriented KT–MT attachments would be stabilized (Andrews et al., 2004; Liu et al., 2009; Tanaka et al., 2002; Welburn et al., 2010). However, we find that either the Haspin or the Bub1 kinase-dependent pools of Aurora B can largely support faithful chromosome segregation in otherwise unchallenged cells. These findings challenge the spatial separation model that is based on the assumption that there is a single pool of Aurora B at the inner centromere, and kinetochore substrates are phosphorylated depending on their distance from this pool (Andrews et al., 2004; Liu et al., 2009; Tanaka et al., 2002; Welburn et al., 2010). However, our findings are in line with several recent studies concluding that the precise inner centromere localization of Aurora B is not a prerequisite for chromosome bi-orientation. Removal of the N-terminal centromere targeting domain of INCENP in budding yeast disturbed Aurora B recruitment to centromeres, but did not affect its capacity to bi-orient KT–MT attachments (Campbell and Desai, 2013). Moreover, placing Aurora B in closer proximity to the kinetochore in mammalian cells did not impair the stabilization of bi-oriented KT–MT attachments (Hengeveld et al., 2017).

Inhibition of Bub1 kinase activity in Haspin CM cells abolished any detectable enrichment of Aurora B at centromeres. While this impaired KT–MT error correction, it did not alter the phosphorylation of the Aurora B kinetochore substrates Dsn1, Hec1, and Knl1, nor did it weaken the mitotic checkpoint. This implies that Aurora B–dependent kinetochore substrate phosphorylation and mitotic checkpoint function do not necessarily require (inner) centromere accumulation of Aurora B. Because activation of Aurora B depends on clustering of the CPC (Kelly et al., 2007, 2010; Wang et al., 2011), clustering and activation of Aurora B most likely takes place at an alternative location, in the absence of Haspin and Bub1 activity. This could simply be chromatin, as we observed low levels of Aurora B dispersed over the chromatin in Bub1-inhibited Haspin CM cells, and in Xenopus laevis egg extracts this was shown to be sufficient for the activation of Aurora B kinase activity (Haase et al., 2017; Kelly et al., 2007). Alternatively, recent work proposed a role for microtubules that traverse the centromere in depositing active Aurora B from the centromere at the kinetochore (Trivedi et al., 2019). A Borealin mutant incapable of binding microtubules in vitro was unable to rescue KT substrate phosphorylation, despite the presence of active Aurora B at centromeres. We deem it unlikely that a microtubule-associated pool of the CPC is responsible for the KT substrate phosphorylation in Bub1-inhibited Haspin CM cells because we observed no difference in Hec1 phosphorylation in either low (0.33 µM) or high (3.3 µM) concentrations of nocodazole. However, we cannot exclude that the CPC core complex, consisting of Borealin, Survivin, and the N terminus of INCENP, might play a role in Aurora B–dependent KT substrate phosphorylation independently of microtubule binding and inner centromere clustering of Aurora B (Haase et al., 2017). Finally, a potential third, transient kinetochore-associated pool of Aurora B, one that does not depend on Haspin and Bub1 kinase activity, may phosphorylate the KMN network (Caldas et al., 2013; DeLuca et al., 2011; Fischböck-Halwachs et al., 2019; García-Rodríguez et al., 2019; Broad et al., 2020).

Irrespective of the source of Aurora B activity that can phosphorylate substrates within the microtubule binding KMN network, in cells with combined Haspin and Bub1 inhibition, impaired error correction does not correlate with a reduction in Hec1 and Dsn1 phosphorylation. This seems to be in line with a recent study by Yoo et al. (2018), who used a fluorescence lifetime imaging microscopy–Förster resonance energy transfer–based approach to estimate the fraction of Hec1/Ndc80 complexes bound to kMTs. While inhibition of Haspin by the small molecule inhibitor 5-ITu reduced the accumulation of Aurora B at centromeres and increased kMT binding to Hec1/Ndc80 in the absence of tension, it did not reduce phosphorylation of an Aurora B Förster resonance energy transfer probe that was localized at the outer-KT (Yoo et al., 2018). This suggests that KT–MT dynamics and hence error correction efficiency are more sensitive to reductions in Aurora B centromeric levels than the phosphorylation of KMN substrates and mitotic checkpoint activity. In fact, we deem it likely that the observed phosphorylation of KMN substrates in Bub1-inhibited Haspin CM cells could explain their capacity to sustain the mitotic checkpoint in nocodazole with partial inhibition of Mps1. Phosphorylation of these outer kinetochore substrates by Aurora B, including the RVSF motif in Knl1, would prevent the recruitment of PP1y, thereby precluding PP1y-mediated dephosphorylation of Mps1-phosphorylated MELT motifs in Knl1, a prerequisite for mitotic checkpoint silencing (Hengeveld et al., 2017; Liu et al., 2010; London et al., 2012; Meadows et al., 2011; Nijenhuis et al., 2014; Rosenberg et al., 2011; Shepperd et al., 2012; Yamagishi et al., 2012). Of note, based on experiments using the Haspin inhibitors 5-ITu and LDN-211898, it was concluded that Haspin is important for mitotic checkpoint signaling, either directly or indirectly through localization of Aurora B (De Antoni et al., 2012; Wang et al., 2012). In contrast, we find that Haspin CM cells sustain the mitotic checkpoint in nocodazole even when the checkpoint kinase Mps1 was partially inhibited (Fig. 7, D and E). Therefore, our data argue against a role for Haspin in the mitotic checkpoint.

Our findings raise the question how the loss of the Aurora B centromere pools impairs KT–MT error correction, when Hec1 can still be phosphorylated. The Bub1-dependent pool of Aurora B in Haspin CM cells supports chromosome alignment in otherwise unchallenged cells, but is less efficient in bi-orienting chromosomes when the frequency of erroneous KT–MT attachments is increased. Further reducing centromeric levels of Aurora B by inhibition of Bub1 exacerbates the alignment defect in Haspin CM cells. One could argue the latter might be caused by a loss of the microtubule-destabilizing kinesin MCAK from centromeres, which is present at very low levels in Bub1-inhibited Haspin CM cells. However, the levels of centromeric MCAK are already very low in Bub1-inhibited WT cells that do align efficiently when released from a mitotic arrest induced by Eg5 inhibition. Moreover, the observation that expression of CB-INCENP can rescue alignment in Bub1-inhibited Haspin CM cells argues against a role for centromeric MCAK levels, but again suggests that the total level of Aurora B at centromeres is an important determinant for efficient resolution of erroneous attachments. Additionally, it shows that the bi-orientation defects observed upon loss of Haspin and Bub1 activity can be attributed to loss of centromeric Aurora B and not to loss of Aurora B–independent functions of Haspin or Bub1. The Aurora B substrates involved remain to be identified, but we expect these to be specifically phosphorylated by the centromere-associated pools of Aurora B.

In conclusion, we have uncovered at least three different functional pools of Aurora B: a Haspin-dependent inner centromere pool, a Bub1-dependent kinetochore proximal pool, and a Haspin- and Bub1-independent pool, responsible for phosphorylation of kinetochore substrates. The Haspin- and Bub1-dependent Aurora B pools are largely redundant for KT–MT error correction. We propose that through phosphorylation of proteins at the kinetochore, the third pool contributes to mitotic checkpoint signaling but also regulates KT–MT attachments.

Together, these three pools of Aurora B ensure that chromosome segregation occurs with high fidelity even in challenging conditions.

U-2 OS-LacO cells (Janicki et al., 2004) were cultured in DMEM (Sigma-Aldrich) supplemented with 10% FBS (Bodinco BV), 2 mM UltraGlutamine (Lonza), 100 U/ml penicillin, and 100 µg/ml streptomycin (Lonza). HCT116 cells (a kind gift from O. Kranenburg, University Medical Center Utrecht, Utrecht, Netherlands) were cultured in DMEM F12 (Lonza) supplemented with 10% FBS, 2 mM UltraGlutamine, 100 U/ml penicillin, 100 µg/ml streptomycin, and 25 mM Hepes (Lonza). Cell lines were maintained at 37°C and 5% CO_2. HCT116 cells stably expressing H2B-mCherry were generated by lentiviral transduction as described before (Hengeveld et al., 2017). Briefly, HEK293T cells (ATCC, CRL-3216) were cotransfected with pWPT-H2B-mCherry, pRSV, pMD2-G, and pMDLG-I using X-tremeGENE (Roche) according to the manufacturer’s instructions. After 48 h, viruses were harvested and used for viral transduction of HCT116 cells. HCT116 cells expressing doxycycline-inducible CB-INCENPdCEN-mCherry and CB-mCherry were generated by lentiviral transduction as described above and selected on G418. Sf9 cells (ATCC, CRL-1711) were cultured in Insect-EXPRESS (Lonza) supplemented with 5% heat-inactivated FBS, 100 U/ml penicillin, and 100 µg/ml streptomycin. Cells were maintained at 27°C in suspension flasks.

sgRNAs targeting GSG2 are 5′-GCA​ATG​ATT​TTT​ATG​GCT​AC-3′ (U-2 OS-LacO Haspin CM), 5′-CAA​GTG​GTG​CTC​CGT​CCT​CT-3′ (HCT116 Haspin CM1), and 5′-CGT​CCG​CAG​CCC​CAT​ATG​TG-3′ (HCT116 Haspin CM2). sgRNAs were cloned into pAceBac-Cas9 (Hindriksen et al., 2017a). Bacmids were generated using the Bac-to-Bac system in conjunction with EMBacY cells (Berger et al., 2004; Bieniossek et al., 2012).

LacI-GFP, LacI-GFP-Haspin, LacI-GFP-Haspin kinase dead (K511A), LacI-GFP-Bub1, LacI-GFP-Bub1^E252K, LacI-GFP-Bub1^E252K kinase dead (D946A), Sgo1-LacI-GFP, or LacI-GFP-Sgo2 was cloned into pAceBac1-CMV or pAceBac1-Ubc (van der Horst et al., 2015). Bacmids were generated using the Bac-to-Bac system in conjunction with EMBacY cells (Berger et al., 2004; Bieniossek et al., 2012). For generating doxycycline-inducible HCT116 cells, CB-INCENPdCEN-mCherry and CB-mCherry were cloned into pInducer20.

Baculovirus was produced by transfection of bacmids into Sf9 cells as described previously (Hindriksen et al., 2017a). For baculoviral transduction, cells were trypsinized and taken up in DMEM F12 supplemented with 10% heat-inactivated FBS, 2 mM UltraGlutamine, 100 U/ml penicillin, and 100 µg/ml streptomycin. Cells were plated, and baculovirus was added immediately.

U-2 OS LacO and HCT116 Haspin knockout cells were generated using CRISPR/Cas9. Briefly, cells were transduced with recombinant baculovirus expressing Cas9-2a-GFP and the gRNA sequences indicated above. After 48 h, cells were seeded at low density in 96-well plates. Cells from wells harboring single colonies were selected for further analysis.

Genomic DNA was isolated using the DNeasy Blood and Tissue kit (Qiagen), and PCR was performed using the following primers: Haspin-Fw: 5′-AAA​CAC​CTC​CTC​AGA​GTG​ATC​CCA​GGA​GAT​CTC​TGC-3′; and Haspin-Rv: 5′-TTA​CTT​AAA​CAG​ACT​GTG​CTG​GCA​GAG​CAA​GTC​AGT​G-3′. PCR products were subjected to Sanger sequencing.

The following siRNAs were used in this study: Luciferase (siLuc, 20 or 100 nM, Luciferase GL2 duplex; Dharmacon/D-001 100–01-20), Bub1 (siBub1, 100 nM, 5′-GAAUGUAAGCGUUCACGAADTDT-3′; Dharmacon), WAPL (siWAPL, 40 nM, 5′-GAGAGAUGUUUACGAGUUUDTDT-3′; Dharmacon), and Haspin (siGSG2, 40 nM, Ambion AM51331, ID1093). Cells were plated and transfected immediately using HiPerfect (Qiagen) or DharmaFECT 2 (Dharmacon), according to the manufacturers’ instructions.

Cells were cultured on 12-mm glass coverslips and synchronized in mitosis by the addition of either S-trityl-L-cysteine (STLC; Tocris Bioscience; 20 µM) or nocodazole (Sigma-Aldrich; 0.83 µM), as indicated in the figure legends. Alternatively, cells were blocked in G2 by using the CDK1 inhibitor RO-3036 (Calbiochem; 7.5 µM) overnight. The next morning, RO-3036 was washed away in the presence of nocodazole (0.83 µM), leaving the cells another 1–4 h in the presence or absence of BAY-320 (10 µM), ZM447493 (2 µM), 5-ITu (2 µM), or DMSO as a control. For the LMNB1 experiments, the cells were released after overnight RO-3306 block for 30 min.

Cells were fixed using PFA, permeabilizing cells either before or after fixation. In the first case, cells were treated with 0.2% Triton X-100 (vol/vol) in PEM buffer (100 mM Pipes KOH, pH 6.8, 5 mM EGTA, and 1 mM MgCl₂) for ±60 s. Next, an equal volume of 4% PFA in PBS was added (final concentration, 2% PFA). 4 min later, this was replaced with 4% PFA in PBS and incubated for another 4 min. For IF staining of Hec1 S44ph, Knl1 S60pH and Dsn1 S109pH cells were first permeabilized in 0.25% Saponin (wt/vol) in PHEM buffer (60 mM Pipes, 20 mM Hepes KOH, pH 6.9, 10 mM EGTA, and 4 mM MgSO₄) for ±3 min, followed by fixation of cells in 4% PFA in PHEM buffer for ±10 min. For the permeabilization after fixation, cells were treated with 0.5% Triton X-100 in PBS for 1 min. For Histon3 T3ph, Histon2A T120pH and MCAK cells were fixed with 4% PFA in PBS for 10 min. After two washes with PBS, the cells were incubated in ice-cold methanol for a minimum of 20 min. Cells were washed with PBS and then blocked with 3% BSA in PBS with 0.01% Tween 20 (PBST0.01).

Cells were incubated with primary antibodies diluted in 3% BSA in PBST0.01 for 2 h. After washing with PBST0.01, cells were incubated with secondary antibodies with or without GFP-Booster (Chromotek, cat. no. ABIN509419, 1:5,000) diluted in 3% BSA in PBST0.01 for 1 h. After washing with PBST0.01, cells were treated with 500 ng/ml DAPI in PBST0.01 for a few minutes. After a final wash in PBS, the coverslips were mounted onto glass slides using Prolong Diamond Antifade Mountant (Thermo Fisher Scientific).

Images were acquired on a DeltaVision imaging system (GE Healthcare), upgraded with a seven-color InsightSSI Module & TruLight Illumination System Module using a UPlanSApo 100×/1.40 objective and a CoolSnap HQ2 camera (Photometrics). Presented images are deconvolved maximum intensity projections, processed using Softworx v6. Quantifications were performed using an in-house-developed macro in ImageJ that automatically selects kinetochores. This selection was enlarged with 3 pixels (px), and this region of interest (ROI) was then used to measure fluorescence intensities in different channels. For the Knl1 S60ph and Hec1 S44ph, the ROI of the centrosomes staining was removed for the images. For background subtraction, an area surrounding the DAPI signal was selected. This area was enlarged with 4 px (ROI-A) and with 6 px (ROI-B). ROI-A was subtracted from ROI-B, and this selected region was used to determine the background fluorescence.

Primary antibodies used were anti-Aurora B (Millipore, cat. no. EP1009Y, rabbit monoclonal, 1:1,000), anti-Aurora B (BD Transductions, cat. no. 611083, mouse monoclonal, 1:500), anti-CENP-C (MBL International, cat. no. PD030, guinea pig polyclonal, 1:1,000), anti-CENP-T (MBL International, cat. no. D286-3, rat monoclonal, 1:500, discontinued) anti-phospho-Histone H3 (Thr3; Merck Millipore, cat. no. 07–424, rabbit polyclonal, 1:2,000), anti-phospho-Histone H3 (Thr3; Southern Biotech, cat no. 13500–01, mouse monoclonal, 1:4,000–1:8,000), anti-Sgo1 (Abnova, cat. no. h00151648, mouse monoclonal, 1:1,000), anti-Sgo2 (Novus, cat. no. NB100-60454, rabbit polyclonal, 1:1,000), anti-MCAK (Abcam, cat. no. ab187652, rabbit polyclonal, 1:500), anti-H2AT120ph (Active Motif, cat. no. 39391, rabbit polyclonal, 1:2,000), anti-LMNB1 (Atlas antibodies, cat. no AMAB-91251, monoclonal, 1:1,000), and anti-Dsn1 (GeneTex, GTX120402, rabbit polyclonal, 1:1,000). Anti-Knl1 (rabbit, polyclonal, 1:500) was a kind gift from G. Kops, Hubrecht Institute–KNAW, University Medical Center Utrecht, Utrecht, Netherlands. Anti-phospho-Dsn1 (S109; rabbit polyclonal, 1:1,000) and phosphor-Knl1 (S60; rabbit polyclonal, 1:2,000) were kind gifts from I. Cheeseman, Whitehead Institute for Biomedical Research, Massachusetts Institute of Technology, Cambridge, MA (Welburn et al., 2010). Anti-phospho-Hec1 (S44; rabbit polyclonal, 1:2,000) was a kind gift from J. DeLuca, Colorado State University, Fort Collins, CO (DeLuca et al., 2011).

Secondary antibodies used were goat anti-mouse IgG Alexa Fluor 488 conjugate (Thermo Fisher Scientific, cat. no. A-11029, 1:500), goat anti-mouse IgG Alexa Fluor 568 conjugate (Thermo Fisher Scientific, cat. no. A-11031, 1:500), goat anti-mouse IgG Alexa Fluor 647 conjugate (Thermo Fisher Scientific, cat. no. A-1103121236, 1:500), goat anti-rabbit IgG Alexa Fluor 488 conjugate (Thermo Fisher Scientific, cat. no. A-11034, 1:500), goat anti-rabbit IgG Alexa Fluor 568 conjugate (Thermo Fisher Scientific, cat. no. A-11036, 1:500), goat anti-rabbit IgG Alexa Fluor 647 conjugate (Thermo Fisher Scientific, cat. no. A-21245, 1:500), and goat anti-guinea pig IgG Alexa Fluor 647 conjugate (Thermo Fisher Scientific, cat. no. A-21450, 1:500).

To assay chromosome alignment, cells were incubated with 5 µM monastrol for ±2 h. Cells were then washed three times with medium, after which medium containing 5 µM MG-132 (Merck Millipore) with or without 10 µM BAY-320 was added for various times (45–180 min) before fixing the cells as described above. For the inducible HCT116 cells, induction with 1 µg/ml doxycycline was started 6 h before fixing the cells.

For the preparation of chromosome spreads, HCT116 cells were treated with 0.83 µM nocodazole for 3 h before harvesting by mitotic shake off. Cells were then incubated in 75 mM KCl for 10 min (room temperature) and spun onto coverslips for 4 min at 800 rpm using a Shandon Cytospin 4 (Thermo Fisher Scientific). Cells were then fixed and stained as described above. CENP-C/CENP-C distances were calculated using the x, y, and z coordinates of the centroids of CENP-C kinetochore pairs, obtained from 3D z-stacks using the 3D objects counter in FIJI. Line plots were generated with the Plot Profile tool in FIJI.

Cells were synchronized for 6 h with 0.83 µM nocodazole and harvested. Protein concentrations were determined using a Lowry assay to ensure equal sample loading. Proteins were separated by SDS-PAGE using a Bolt 4–12% Bis-Tris Plus gel (Thermo Fisher Scientific, cat. no. NW0412BOX). Proteins were then transferred to nitrocellulose membranes using a Trans-Blot Turbo Transfer system (Bio-Rad, cat. no. 170–4155). Membranes were blocked in PBST containing 3% BSA. Primary antibody incubation was performed in PBST including 3% BSA for ∼18 h at 4°C. Membranes were washed three times with PBST before incubation with the secondary antibody diluted in PBST. Membranes were washed three more times followed by signal detection using ECL.

Primary antibodies used were anti-Aurora B (ARK-2; Santa Cruz, sc-25426, rabbit polyclonal, 1:1,000), anti-phospho-Histone H3 (Thr3; Southern Biotech, 13500-01, mouse monoclonal, 1:1,000), and anti-α-tubulin (Sigma-Aldrich, cat. no. T5168, mouse monoclonal, 1:10,000). Secondary antibodies used were swine anti-rabbit-HRP (DAKO, P0217, 1:2,500) and rabbit anti-mouse-HRP (DAKO, P0161, 1:2,500).

Cells were seeded in eight-well imaging chambers (μ-Slide, Ibidi) and blocked in G2 by overnight incubation in 7.5 µM RO-3306. The cells were released from the RO-3306 block by washing three times with warm medium, followed directly by filming. Film experiments were performed on a DeltaVision imaging system with an imaging chamber maintained at 37°C and 5% CO₂ (GE Healthcare). The system was upgraded with a seven-color InsightSSI Module and TruLight Illumination System Module and made use of an Olympus UApo N340 40×/1.35 objective and a CoolSnap HQ2 camera (Photometrics).

For assessing the duration of the mitotic arrest in nocodazole by live cell imaging, cells were first seeded in eight-well imaging chambers (μ-Slide, Ibidi) and treated with 2.5 mM thymidine (Sigma-Aldrich) for 24 h. Cells were washed three times with medium, after which culture medium was changed to Leibovitz’s L-15 medium (Gibco), supplemented with 10% FBS, 2 mM UltraGlutamine (Lonza), 100 U/ml penicillin, and 100 µg/ml streptomycin (Lonza), containing 3.3 µM nocodazole. ZM447439 (2 µM; Tocris Bioscience), reversine (200 nM; Cayman Chemical), and/or BAY-320 (10 µM; gift from Bayer Pharma AG; Baron et al., 2016) were added as indicated. Cells were imaged on an Olympus Cell M system with a UPlanFL N 20×/0.5 objective and analyzed using ImageJ software.

Fig. S1 shows the quantified levels of GFP for the various LacI-GFP fusions expressed in Figs. 1 and 2 (A–H and Q), shows recruitment of Aurora B by both Sgo1-LacI-GFP and LacI-GFP-Sgo2 (I–K), shows Aurora B recruitment by LacI-GFP-Haspin upon depletion of Bub1 (L–O), and shows a Western blot showing loss of H3T3 phosphorylation in U-2 OS LacO Haspin CM cells (P). Fig. S2 shows IF images and quantifications highlighting loss of H3T3 phosphorylation in HCT116 Haspin CM cell lines (A–C), shows a Western blot depicting loss of H3T3 phosphorylation in HCT116 Haspin CM cell lines (D), and shows a quantification of an IF experiment determining centromeric Sgo1 levels in the presence or absence of BAY-320 in HCT116 WT and Haspin CM cells (E). Fig. S3 shows a quantification of a chromosome alignment assay of HCT116 WT and Haspin CM cells with or without Bub1 inhibition upon depletion of WAPL (A) and shows IF images and quantifications of HCT116 WT and Haspin CM cells expressing CB-INCENP, which rescues centromeric levels of Aurora B (B and C). Fig. S4 shows IF images and quantifications of HCT116 WT and Haspin CM for Dsn1 S109, Hec1 S44, and Knl1 S60 phosphorylation with or without Aurora B inhibition (A–F), shows IF quantification of Dsn1 S109 phosphorylation at various concentrations of ZM447439 in HCT116 WT and Haspin CM cells (G), and shows IF quantification of Hec1 S44 phosphorylation and tubulin in HCT116 WT and Haspin CM cells at various concentrations of nocodazole (H–I). Fig. S5 shows IF images and quantifications of HCT116 cells with or without depletion of MCAK using two different siRNAs to validate our MCAK antibody (A and B), shows IF images and quantifications of MCAK levels in HCT116 cells with depletion (C–E) or inhibition (F–H) of Haspin (activity), and shows IF images and quantifications of U-2 OS LacO WT and Haspin CM cells expressing various LacI-GFP fusion proteins and their ability to recruit MCAK (I–K). Table S1 summarizes the alterations induced by Cas9 at the GSG2 locus in the cell lines used in this study, both at the genetic level and the resultant protein level.

We thank Dr. I. Cheeseman and Dr. G. Kops for their generous gift of antibodies and Dr. G. Siemeister from Bayer AG, Berlin, Germany, for the generous gift of BAY-320.

This work is part of the Oncode Institute, which is partly financed by the Dutch Cancer Society and was funded by a grant from the Dutch Cancer Society (KWF research project 10366).

The authors declare no competing financial interests.

Author contributions: M.A. Hadders and S.M.A. Lens conceived the project. M.A. Hadders, S. Hindriksen, M.A. Truong, A.N. Mhaskar, J.P. Wopken, and M.J.M. Vromans performed experiments under supervision of M.A. Hadders and S.M.A. Lens. M.A. Hadders, S. Hindriksen, and S.M.A. Lens wrote the manuscript.