α-Tubulin detyrosination impairs mitotic error correction by suppressing MCAK centromeric activity

Successful chromosome segregation during mitosis requires that each sister kinetochore (KT) is stably attached to microtubules (MTs) oriented to opposite spindle poles (amphitelic attachments). However, due to stochastic interactions between KTs and spindle MTs early in mitosis, many chromosomes establish erroneous attachments that have been implicated in chromosomal instability, a hallmark of human cancers (Bakhoum and Cantley, 2018; Cimini et al., 2003). To prevent this, cells rely on error correction mechanisms that regulate MT dynamics at the KT (Bakhoum et al., 2009). These mechanisms may act globally through the regulation of Cdk1 activity during early mitosis (Kabeche and Compton, 2013) or, more locally, by promoting MT detachment from KTs in response to low centromeric tension (Cimini et al., 2006; Liu et al., 2009). At the heart of this local error correction mechanism, the Aurora B kinase regulates the recruitment and/or activity of several centromeric/KT proteins, including the kinesin-13 MCAK (Andrews et al., 2004; Bakhoum et al., 2009; Knowlton et al., 2006; Lan et al., 2004), a potent MT depolymerase present in animal cells. Therefore, mitotic error correction is currently viewed as a “blind” process that results from the nondiscriminatory renewal of MTs at the KT interface, regardless of their attachment status (i.e., correct or incorrect).

How MCAK mediates mitotic error correction has been difficult to determine, mostly due to its independent localization at centromeres/KTs and MT plus ends, and its global impact on spindle MT dynamics (Bakhoum et al., 2009; Domnitz et al., 2012; Huang et al., 2007; Kline-Smith et al., 2004; Rizk et al., 2009; Wordeman et al., 2007). Interestingly, the contribution of MCAK for KT MT (kMT) turnover appears to occur primarily in metaphase (Bakhoum et al., 2009), when most KT-MT attachments are amphitelic and stabilized, and it does so without any measurable impact on MT polymerization dynamics associated with MT poleward flux (Ganem et al., 2005). This apparent paradox led us to hypothesize that centromeric MCAK is able to discriminate correct and incorrect KT-MT attachments independently of its global effect on spindle MT dynamics. In support of this hypothesis, active MCAK is enriched at centromeres/KTs of misaligned chromosomes (Andrews et al., 2004; Lan et al., 2004), as well as in aligned chromosomes with erroneous merotelic attachments (Knowlton et al., 2006).

In vitro reconstitution experiments have shown that MCAK activity is significantly suppressed by α-tubulin detyrosination (Peris et al., 2009; Sirajuddin et al., 2014), a posttranslational modification that accumulates on long-lived MTs (Nieuwenhuis and Brummelkamp, 2019). α-tubulin detyrosination has been recently implicated in mitosis and meiosis, neuronal processes and cognitive brain function, heart and skeletal muscle contraction, and cancer (Akera et al., 2017; Barisic et al., 2015; Chen et al., 2018; Erck et al., 2005; Kerr et al., 2015; Lafanechère et al., 1998; Liao et al., 2019; Pagnamenta et al., 2019; Robison et al., 2016). The detyrosination/tyrosination cycle involves the catalytic removal of the C-terminal tyrosine of most mammalian α-tubulin isoforms by tubulin carboxypeptidases, such as the recently identified Vasohibin (VASH) 1/VASH2-SVBP complexes (Aillaud et al., 2017; Nieuwenhuis et al., 2017), followed by retyrosination of soluble α-tubulin by tubulin tyrosine ligase (TTL; Ersfeld et al., 1993).

Here we combined powerful gene manipulation tools (including RNAi, small-molecule inhibition, protein overexpression, and CRISPR-Cas9 gene editing) with state-of-the-art microscopy, including a novel super-resolution microscopy technique (Pereira et al., 2019), to investigate whether MCAK activity required for mitotic error correction is regulated by α-tubulin detyrosination. Our findings support a physiological role for α-tubulin detyrosination in the discrimination between correct and incorrect KT-MT attachments, establishing a new paradigm in the control of mitotic fidelity.

To investigate whether the mitotic error correction machinery “reads” the tyrosination/detyrosination state of α-tubulin on kMTs, we started by quantifying the ratio of detyrosinated/tyrosinated α-tubulin immediately adjacent to the KT in distinct experimental conditions that favor a particular attachment configuration. These included short-lived syntelic (when both KTs from a pair are attached to MTs from the same pole) or monotelic (when only one KT from the pair is attached to MTs) attachments, typically observed in monopolar spindles after kinesin-5 inhibition with monastrol (Kapoor et al., 2000); more stable amphitelic attachments that form during normal metaphase; and finally, hyperstabilized amphitelic attachments induced by treatment with 10 nM taxol for 1 h. We found that correct amphitelic attachments show higher α-tubulin detyrosination near KTs when compared with incorrect/incomplete syntelic/monotelic attachments, and this tendency correlated with increased MT stability (Fig. 1, A–F). Thus, α-tubulin detyrosination near KTs increases with the establishment of correct, more stable, MT attachments.

To investigate whether α-tubulin detyrosination impacts mitotic error correction, we inactivated TTL either by RNAi or CRISPR/Cas9-mediated gene knockout (KO) in human U2OS cells. Both conditions caused a significant increase in the overall α-tubulin detyrosination levels, with a proportional reduction in the tyrosinated α-tubulin and without detectable changes in the total tubulin pool (Figs. S1 A and S2 A). This was also the case in mitotic spindles, in which detyrosinated α-tubulin was increased along KT fibers, including in the immediate vicinity of the KT, and astral MTs after TTL inactivation (Fig. 2, A–C; Fig. S1, B and C; and Fig. S2 B). Live-cell imaging 72 h after TTL RNAi in cells stably expressing H2B-histone-GFP/mCherry-α-tubulin revealed a consistent and significant increase in the frequency of anaphase cells with lagging chromosomes (Fig. 2, D–F). Interestingly, this increase in chromosome missegregation gradually attenuated to control levels after chronic TTL inactivation (TTL KO), despite the continuous increase in detyrosinated α-tubulin over time (Fig. S2, C–F). This suggests the existence of compensatory or cellular adaptation mechanisms to the chronic loss of TTL, as shown previously for TTL KO mice (Erck et al., 2005).

To test whether the observed mitotic errors after TTL RNAi were a consequence of constitutively high α-tubulin detyrosination and not due to other putative roles of TTL, we performed rescue experiments with RNAi-resistant wild-type human TTL fused with YFP (TTL-YFP^R) or a corresponding catalytic dead version (TTL[CD]-YFP^R) in which a glutamic acid at position 331 was replaced by glutamine (E331Q), rendering TTL defective in α-tubulin retyrosination, but not tubulin binding (Prota et al., 2013; Szyk et al., 2011). Western blot and confocal fluorescence microscopy analyses confirmed the RNAi resistance of both TTL-YFP^R and TTL[CD]-YFP^R compared with conventional TTL-YFP, as well as the recovery of nearly normal α-tubulin detyrosination levels (total and in mitotic spindles) in the TTL-YFP^R, but not TTL[CD]-YFP^R rescue conditions (Fig. S3, A and B). Surprisingly, we noticed that ectopic expression of any version of TTL-YFP resulted in the overexpression of endogenous TTL (Fig. S3 A). To exclude potential deleterious effects associated with the overexpression of ectopic and endogenous TTL (Szyk et al., 2011), we considered only cells expressing low-to-mild levels of TTL-YFP^R or TTL[CD]-YFP^R that retained an apparently normal spindle architecture in our quantification of mitotic errors. Accordingly, we found that TTL-YFP^R, but not TTL[CD]-YFP^R, was able to rescue the increased frequency of anaphase cells with lagging chromosomes observed upon TTL RNAi (Fig. S3 C). These results not only demonstrate the specificity of the TTL RNAi phenotype, but also directly indicate that the observed mitotic errors were dependent on the catalytic activity of TTL required for α-tubulin retyrosination.

Last, experimental increase of α-tubulin detyrosination levels independently of TTL by overexpression of the tubulin carboxypeptidase VASH1-SVBP led to an equivalent increase in anaphase cells with chromosome segregation errors when compared with TTL inactivation by RNAi (Fig. 3, A–C, F, and G). Conversely, RNAi-mediated depletion of VASH1 and VASH2 (with or without SVBP depletion) reduced chromosome missegregation events below control levels (Fig. 3, D–G). Overall, these data indicate that experimental modulation of α-tubulin detyrosination levels on mitotic spindles, including in the vicinity of KTs, impacts chromosome segregation fidelity.

Persistent merotelic attachments in which a single KT is attached to MTs from both spindle poles are the main source of anaphase lagging chromosomes in mammalian cells (Cimini et al., 2003). To investigate whether the lagging chromosomes observed after experimental increase in α-tubulin detyrosination result from the formation of merotelic attachments, we used super-resolution coherent hybrid stimulated emission depletion (CH-STED) microscopy (Pereira et al., 2019) in fixed cells after TTL RNAi. Relative to conventional 2D-STED, this novel subdiffraction technique improves contrast in highly complex environments, such as the mitotic spindle. Accordingly, we observed a significant increase in lagging chromosomes with single stretched KTs attached to MTs from both poles, confirming the formation of merotelic attachments (Fig. 2, G and H). Of note, TTL-depleted cells also showed a significant increase in acentric (i.e., without KTs) chromosome fragments, probably resulting from the sheer stress undergone by the centromere during correction of merotelic attachments during anaphase (Cimini et al., 2003; Guerrero et al., 2010) or due to chromosome shattering of missegregated chromosomes in the previous division cycle (Crasta et al., 2012). Taken together, these data indicate that a significant fraction of the lagging chromosomes observed after experimental increase of α-tubulin detyrosination derive from erroneous merotelic attachments.

To distinguish whether increased α-tubulin detyrosination increases error formation and/or prevents efficient error correction, we induced the assembly of monopolar spindles with monastrol, which favors the formation of merotelic attachments during spindle bipolarization after monastrol washout (Cimini et al., 2003; Kapoor et al., 2000). This treatment led to an equivalent increase in lagging chromosomes in both control and TTL-depleted anaphase cells (Fig. 2 I), suggesting that the increased frequency of mitotic errors observed after experimental increase of α-tubulin detyrosination were not due to an increased error formation rate. This conclusion was substantiated by the observation that the timing of centrosome separation at nuclear envelope breakdown, an important variable that has been linked to increased error formation during mitosis (Silkworth et al., 2012), was indistinguishable between control and TTL-depleted cells (Fig. S4, A–C). Thus, increased α-tubulin detyrosination impairs efficient error correction during mitosis.

Given that α-tubulin detyrosination correlates with MT stability, which has been proposed as the main cause of persistent merotelic attachments behind chromosomal instability (Bakhoum et al., 2009; Cimini et al., 2003, 2006), we investigated whether constitutive amplification of α-tubulin detyrosination globally affects kMT stability. To do so, we used fluorescence dissipation after photoactivation (FDAPA) of GFP-α-tubulin (Fig. 4 A) to measure kMT turnover in control and TTL-depleted cells, as well as in MCAK-depleted cells used as positive control (Bakhoum et al., 2009; Ferreira et al., 2018). The resulting decay curves for each individual cell fit a double exponential (R² > 0.99), reflecting the presence of two MT populations: a fast fluorescence decay population that has been attributed to non-kMTs, and a slow fluorescence decay population thought to correspond to more stable kMTs (Zhai et al., 1995). From each fit we calculated and plotted the mean ± SD of the fluorescence intensity for each time point in the different conditions (Fig. 4 B), and extracted the relative percentages of stable (kMTs) vs. unstable (non-kMTs) MTs, as well as their respective t_1/2. We found that only MCAK depletion resulted in a statistically significant decrease in the proportion of kMTs in both prometaphase and metaphase (Fig. 4 C). KMT turnover also increased from prometaphase to metaphase in all conditions, but it was indistinguishable between control and TTL-depleted cells at each given stage (Fig. 4 D). In line with these data, measurement of inter-KT distances as a proxy for tension on amphitelic attachments was also indistinguishable between control and TTL-depleted cells (Fig. S1 D). In contrast, MCAK-depleted cells had slightly more stable kMTs specifically during metaphase (Fig. 4 D), as shown previously (Bakhoum et al., 2009). Altogether, these data are consistent with previous measurements of MT t_1/2 after experimental increase of α-tubulin detyrosination in interphase (Webster et al., 1990), and demonstrate that, contrary to MCAK, α-tubulin detyrosination does not globally interfere with kMT dynamics.

Interestingly, while TTL or MCAK depletion significantly increased the frequency of anaphase cells with lagging chromosomes, TTL-depleted cells showed a clear enrichment of detyrosinated α-tubulin along kMTs, whereas MCAK-depleted cells strongly accumulated detyrosinated α-tubulin preferentially on astral MTs (Fig. 5, A–D). These striking differences in the distribution of detyrosinated α-tubulin imply that global perturbation of the two proteins impacts kMT attachments through different mechanisms. Indeed, MCAK was proposed to control the formation of robust KT-MT attachments both by regulating the length of non-kMTs through a role at MT plus ends (Domnitz et al., 2012), and by controlling chromosome directional switching by a specific role at centromeres (Kline-Smith et al., 2004; Wordeman et al., 2007). Additionally, α-tubulin detyrosination was recently shown to regulate MCAK activity required for normal astral MT length (Liao et al., 2019).

To address whether and how MT detyrosination impairs MCAK activity required for mitotic error correction, we set up rescue experiments either by overexpressing exogenous full-length EGFP-MCAK (MCAKFL) or by enhancing endogenous MCAK activity with the small-molecule agonist UMK57 (Orr et al., 2016) in TTL-deficient cells. We found that both treatments partially rescued the frequency of anaphase cells with lagging chromosomes observed after TTL depletion (Fig. 6, A–D). Similar findings were obtained after overexpression of exogenous Kif2b, another kinesin-13 with MT depolymerizing activity that localizes to prometaphase KTs (Manning et al., 2007), but not overexpression of exogenous Kif2a, a third kinesin-13 associated with MT minus ends at spindle poles (Ganem and Compton, 2004; Fig. S5, A and B), as predicted by models in which global destabilization of KT-MT attachments is able to rescue mitotic errors (Bakhoum et al., 2009). However, because our measurements of kMT t_1/2 after TTL depletion indicated that the mitotic errors caused by increased α-tubulin detyrosination do not result from an overall increase in kMT stability, our results are better explained by an alternative, nonmutually exclusive model, in which α-tubulin detyrosination compromises the activity of a specific pool of MCAK, without globally changing kMT dynamics.

To distinguish between both models, we performed additional rescue experiments with MCAK mutant constructs that compromise MT depolymerizing activity (MCAKhypir) or centromeric localization (ΔNMCAK; Domnitz et al., 2012; Wordeman et al., 2007). These experiments suggested that α-tubulin detyrosination impairs error correction by specifically inhibiting MCAK activity at centromeres (Fig. 6, A–C and E). To further test this hypothesis, we constitutively targeted a GFP-tagged minimal motor domain of MCAK to the centromere fused with the centromere binding domain of CENP-B, with (GCPBM) or without (GCPBMhypir) MT depolymerizing activity (Wordeman et al., 2007). We found that overexpression of either GCPBM or GCPBMhypir did not cause any measurable differences in mitotic spindle and astral MT length (Fig. 6, A–C, F, and G). Yet GCPBM, but not its inactive counterpart, was sufficient to significantly rescue the abnormally high frequency of anaphase cells with lagging chromosomes observed in TTL-depleted cells (Fig. 6 E), suggesting that, in addition to affecting MCAK’s role on astral MTs (Liao et al., 2019), α-tubulin detyrosination also inhibits a restricted pool of centromeric MCAK required for error correction. To directly test this, we used FDAPA in U2OS cells stably expressing photoconvertible mEos-α-tubulin to measure global kMT t_1/2 after overexpression of GCPBM. We found that overexpression of GCPBM abolished the stabilization of kMTs typically observed from prometaphase to metaphase, without significantly affecting overall kMT t_1/2 (Fig. 7, A–C). Thus, MCAK depolymerizing activity at centromeres is required and sufficient to correct mitotic errors caused by a constitutive increase of α-tubulin detyrosination, independently from its global role in the regulation of spindle MT dynamics.

The prevalent view is that MTs are “passive” elements during mitotic error correction. The results reported in this work support a model in which differential α-tubulin tyrosination/detyrosination on kMTs allows MT depolymerizing enzymes localized at centromeres/KTs to discriminate between correct and incorrect kMT attachments. According to this model (Fig. 7 D), initial attachments (including incorrect ones) established during early mitosis are essentially formed by tyrosinated MTs, and thus are more permissive to depolymerize and detach. As chromosomes bi-orient and centromeric tension develops, amphitelic attachments become more stabilized and consequently more detyrosinated, leading to their further stabilization. This implies that the kinetics of α-tubulin detyrosination is sufficiently fast to accumulate near the MT plus ends as they become stably attached at KTs. Indeed, TTL was shown to only retyrosinate free tubulin dimers (Arce et al., 1975; Raybin and Flavin, 1975), and recent in vitro essays revealed that α-tubulin detyrosination by VASH1/VASH2-SVBP on polymerized MTs occurs within seconds and reaches steady-state in ∼10 min (Aillaud et al., 2017; van der Laan et al., 2019). Importantly, VASH1/SVBP also showed substantial detyrosinating activity toward the free tubulin dimer, whereas VASH2/SVBP appeared to specifically modify polymerized MTs, and it did so with a much higher overall detyrosinating activity (van der Laan et al., 2019). If a similar behavior is present in cells, it is conceivable that a fraction of α-tubulin detyrosination occurs even before the free dimer is incorporated into polymerized MTs, increasing exponentially as correctly attached MTs become stabilized. While allowing the selective preservation of correct amphitelic attachments, the proposed MT-based positive feedback mechanism explains how de novo merotelic attachments on already bi-oriented chromosomes can be surgically prevented/corrected, given that the low α-tubulin detyrosination of newly attached MTs would favor their selective destabilization. Thus, spindle MTs are not just passive elements during mitotic error correction, and their detyrosination/tyrosination works as an active signaling mechanism that discriminates mitotic errors. Such mechanism might have evolved to promote mitotic fidelity in vertebrates, since known TTL and/or VASH-SVBP orthologues are absent from fungi, Caenorhabditis elegans, and Drosophila melanogaster (Prota et al., 2013; Sanchez-Pulido and Ponting, 2016). In these systems, mitotic fidelity might be ensured by alternative mechanisms, including the closed or semi-open nature of mitosis, or functionally distinct KTs (Drechsler and McAinsh, 2012). Last, our findings have implications for the use of MT stabilizing drugs, such as taxol and its derivatives, in the treatment of human cancers, many of which acquire resistance by modulating MCAK activity (Ganguly et al., 2011a; Ganguly et al., 2011b).

All cell lines were cultured at 37°C in 5% CO₂ atmosphere in DMEM (Gibco, Thermo Fisher Scientific) containing 10% FBS (Gibco, Thermo Fisher Scientific). Parental, mEos-α-tubulin and H2B-GFP/mCherry-α-tubulin U2OS cells were kindly provided by S. Geley (Innsbruck Medical University, Innsbruck, Austria), whereas the PA-GFP-α-tubulin/mCherry-α-tubulin U2OS cells were kindly provided by R. Medema (Netherlands Cancer Institute, Amsterdam, The Netherlands). Stable H2B-GFP/mRFP-α-tubulin U2OS cells were generated in this study by lentiviral transduction.

To express fluorescently tagged proteins, cells were transfected using Lipofectamine 2000 (Thermo Fisher Scientific) with 0.25–2 µg of each construct. MCAK constructs (kindly provided by L. Wordeman, University of Washington, Seattle, WA): EGFP-MCAK full-length, EGFP-MCAKhypir, EGFP-ΔN-MCAK, GCPBM, and GCPBMhypir (Domnitz et al., 2012; Wordeman et al., 2007); human VASH1 (NM_014909.5), and SVBP (NM_199342.4) encoding gateway-compatible vectors (kind gift from J. Nilsson, Center for Protein Research, Copenhagen, Denmark) were used to create expression plasmids with C-terminal EGFP and 3X-myc tags by recombination between attL and attR sites (Invitrogen) according to the manufacturer’s instructions. Plasmid-encoding EGFP alone was used as control. Expression of VASH1 and SVBP was performed as described previously (Liao et al., 2019). Briefly, cells were transfected with 1 µg of the two plasmids using GeneJuice (Merck) transfection reagent according to the manufacturer’s protocol. To perform RNAi-mediated protein depletion, cells were plated at 30–50% confluence onto 22 × 22-mm no. 1.5 glass coverslips and cultured for 1–2 h in DMEM supplemented with 5% of FBS before transfection. RNAi transfection was performed using Lipofectamine RNAiMAX reagent (Thermo Fisher Scientific) with 50 nM of validated siRNA oligonucleotides against human TTL 5′-GUG​CAC​GUG​AUC​CAG​AAA​U-3′, MCAK 5′-GAU​CCA​ACG​CAG​UAA​UGG​U-3′, VASH1 5′-CCA​GAC​UAG​GAU​GCU​UCU​G-3′, VASH2 5′-GGA​UAA​CCG​UAA​CUG​AAG​U-3′, SVBP 5′-GCA​GAG​AUC​UAU​GCC​CUC​A-3′, and control (scramble) siRNA 5′-UGG​UUU​ACA​UGU​CGA​CUA​A-3′, diluted in serum-free medium (Opti-MEM, Thermo Fisher Scientific). Mock transfected or scramble siRNA results were indistinguishable. Validation of TTL depletion by RNAi was obtained by expressing exogenous human TTL-YFP (kindly provided by C. Janke, Institut Curie, Orsay, France) resistant to siRNA depletions. An siRNA-resistant TTL-YFP (TTL-YFP^R) was obtained by site-directed mutagenesis of five nucleotides (indicated with asterisks) at the target site using the following complementary primers: siTTLresist.Fw 5′-CAA​CCA​GGG​CCA​AGT​C*CAT*GTC*ATT*CAA*AAA​TAT​CTT​GAG​CAC​CCT​C-3′ and siTTLresist.Rev 5′-GAG​GGT​GCT​CAA​GAT​ATT​TT*TGA*ATG*ACA*TGG*ACT​TGG​CCC​TGG​TTG-3′. The same protocol was used to introduce siRNA resistance in a plasmid encoding for the catalytic dead version (E331Q; Prota et al., 2013; Szyk et al., 2011) of human TTL-YFP (TTL[CD]-YFP^R), kindly provided by C. Janke. The TTL knockout cells were generated by CRISPR/Cas9-mediated genome editing, using a lentiviral backbone containing both the Streptococcus pyogenes Cas9 (spCas9) nuclease and the single guide RNA scaffold (lentiCRISPRv2). Two 20-bp single guide RNA (sgTTL, 5′-CAC​CGA​ACA​GCA​GCG​TCT​ACG​CCG-3′ and sgCtrl, 5′-AAA​CCG​GCG​TAG​ACG​CTG​CTG​TTC-3′; Sigma-Aldrich) were designed to target the TTL gene and cloned into the lentiviral vector pLenti-CRISPR-v2 (no. 52961, addgene). All lentivirus was produced by cotransfection of HEK293T cells with the lentiviral vectors, psPAX2 (Gag, Pol, Rev, and Tat expressing packaging vector) and pMD2.G (VSV-G expressing envelope vector). The 72 h supernatant containing DNA particles was used to transduce the lentivirus to the target cells, as previously described (Ferreira et al., 2018). TTL knockout cells were selected by their resistance to puromycin (2 µg/ml, Calbiochem) and confirmed by Western blot analysis.

To induce lagging chromosomes by inhibition of kinesin-5, we added 100 µM Monastrol (Tocris, Bioscience) 12–16 h before washout. MT stabilization was obtained using 10 nM Taxol for 1 h (Sigma-Aldrich). Mitotic arrest at metaphase was obtained using 10 µM MG132 (Calbiochem). Live-cell analysis using MG132 was performed in less than 2 h in the presence of the drug to avoid cohesion fatigue. MCAK activity was enhanced using 250 µM of UMK57 (Orr et al., 2016) for 12–16 h before fixation. An equivalent volume of DMSO was used as control for each drug treatment.

U2OS cells were fixed with cold methanol (–20°C) for 4 min or 0.1–0.25% gluteraldehyde + 4% paraformaldehyde (Electron Microscopy Sciences) for 10 min. Autofluorescence was quenched by 0.1% sodium borohydride (Sigma-Aldrich) after aldehyde fixation and cells permeabilized with 0.5% Triton X-100 (Sigma-Aldrich) for another 10 min. Tyrosinated α-tubulin and detyrosinated α-tubulin were immunostained using rat monoclonal anti-tyrosinated α-tubulin clone YL1/2 (1:100–500, Bio-Rad) and rabbit polyclonal anti-detyrosinated α-tubulin (1:250–2,000; Liao et al., 2019), respectively. Other primary antibodies used were human anti-centromere antibodies (ACA, 1:100–2,000, kind gift from B. Earnshaw, Welcome Trust Centre for Cell Biology, University of Edinburgh, Edinburgh, UK, or S. Geley). GFP-tagged constructs were visualized by means of direct fluorescence. Alexa Fluor 488 (1:200–2,000, Themofisher), STAR-580, and STAR-RED (1:200, Abberior) were used as secondary antibodies, and DNA was counterstained with 1 µg/ml DAPI (Sigma-Aldrich).

3D wide-field images were acquired using an AxioImager Z1 (100× Plan-Apochromatic oil differential interference contrast objective lens, 1.46 NA, Carl Zeiss Microimaging Inc.) equipped with a CCD camera (ORCA-R2, Hamamatsu) operated by Zen software (Carl Zeiss, Inc.). Blind deconvolution of 3D image datasets was performed using Autoquant X software (Media Cybernetics). Confocal and super-resolution CH-STED images were acquired using Abberior Instruments “Expert Line” gated-STED coupled to a Nikon Ti microscope. An oil-immersion 60× 1.4 NA Plan-Apo objective (Nikon, Lambda Series) and pinhole size of 0.8 Airy units were used in all confocal acquisitions. Super-resolution images were acquired using a CH-STED beam (Pereira et al., 2019). A second LSM800 confocal microscope (Carl Zeiss Microimaging Inc.) mounted on a Zeiss-Axio imager Z1 equipped with plan-apochromat 63×/1.40 oil differential interference contrast (DIC) M27 objective (Carl Zeiss, Inc.) was used and operated by Zen 2008 software (Carl Zeiss, Inc.).

Detyrosination/tyrosination profile along astral MTs and KT fibers (including in a sub-region 1 µm distant from the KT) was performed using ImageJ by drawing a line segment along the region of interest (ROI) followed by plotting the fluorescence intensity values (command plot profile at each channel) as a function of distance from the KT centroid determined by ACA signal. Background was quantified in a region outside the spindle and subtracted from each channel.

The fluorescence intensity signal of tubulin detyrosination was measured directly at KTs in a 9 × 9-pixel-square ROI using LAPSO (MATLAB) and normalized to the intensity signal of ACA in the same ROI. Background fluorescence was measured outside the ROI and subtracted from each KT. The mean values of all KTs quantified in one cell were plotted in a scattered dot plot.

The spindle length was calculated as the distance between the two centrosome position adjusted to the focal plane according to the equation: spindle length = √[d(centrosome1 − centrosome2, xy)² + d(centrosome1 − centrosome2, z)², d = distance. Astral MT length was measured as the distance between the centrosome and the MT distal tip using maximum-projection images.

Inter-KT distances were measured using a custom program written in MATLAB 8.1 (QUANTA), which determines the 3D distance between the centroid of the sister KTs.

For phenotypic analysis of TTL depletion (siTTL), U2OS H2B-GFP/mCherry-α-tubulin cells were cultured in glass coverslips and assembled into 35-mm magnetic chambers (14 mm, no. 1.5, MatTek Corporation). Cell culture medium was replaced with phenol-red-free DMEM CO₂-independent medium (Invitrogen) supplemented with 10% FBS. Time-lapse imaging was performed in a heated chamber (37°C) using a 100× 1.4 NA Plan-Apochromatic differential interference contrast objective mounted on an inverted microscope (TE2000U; Nikon) equipped with a CSU-X1 spinning-disk confocal head (Yokogawa Corporation of America) and with two laser lines (488 nm and 561 nm). Images were acquired with an iXon+ EM-CCD camera (Andor Technology). Eleven 1-µm-separated z-planes covering the entire volume of the mitotic spindle were collected every 2 min. The phenotypic analysis of U2OS H2B-GFP/mCherry-α-tubulin cells overexpressing GFP and VASH1-GFP+SVBP-Myc, as well as the analysis of VASH1/2, VASH1/2+SVBP, and additional TTL depletion (used as positive control) in the same experimental dataset, were performed using a Plan-Apochromat DIC 63×/1.4 NA oil objective mounted on an inverted Zeiss Axio Observer Z1 microscope (Marianas Imaging Workstation from Intelligent Imaging and Innovations Inc. [3i]), equipped with a CSU-X1 spinning-disk confocal head (Yokogawa Corporation of America), iXon Ultra 888 EM-CCD camera (Andor Technology), and four laser lines (405 nm, 488 nm, 561 nm, and 640 nm). Fifteen 1-µm-thick z-planes were collected every 2 min for 2–3 h. Image processing was performed in ImageJ. All displayed images represent maximum-intensity projections of z-stacks. Additional phenotypic live-imaging analysis of TTL KO H2B-GFP/mCherry-α-tubulin cells and respective controls was performed in an IN Cell Analyzer 2000 microscope (GE Healthcare), equipped with temperature and CO₂ controller, using a Nikon 40×/0.95 NA Plan Fluor objective and a large chip CCD Camera (CoolSNAP K4) with a pixel array of 2,048 × 2,048. Single planes were acquired every 2 min.

All U2OS cell lines used in this assay were seeded onto 22 × 22-mm no. 1.5 glass converslips and cultured at 37°C in 5% CO₂ atmosphere in DMEM (Gibco, Thermo Fisher Scientific) supplemented with 10% FBS (Gibco, Thermo Fisher Scientific) for 72 h. 100 µM of Monastrol (Tocris, Bioscience) was added 12–16 h before release, which was performed by three consecutive washes with warm PBS followed by incubation with DMEM + 10% FBS for 40–50 min. Cells were fixed and immunostained for tyrosinated α-tubulin. DNA was counterstained with DAPI. Only chromosomes spatially separated from the main mass of chromosomes were accounted as laggards.

Quantification of lagging chromosomes was performed using 50–500 images of contiguous fields acquired in an IN Cell Analyzer 2000 microscope (GE Healthcare) with a Nikon 40×/0.95 NA Plan Fluor objective (binning 2 × 2), using a large chip CCD Camera (CoolSNAP K4) with a pixel array of 1,024 × 1,024 (2.7027 pixel/µm resolution). All early to late anaphase figures were classified regarding the presence or absence of lagging chromosomes. Any DAPI-positive material between the two chromosome masses, but distinguishably separated from them, was counted as lagging chromosomes. DNA bridges were excluded from this analysis.

For PA and PC assays, both U2OS cells stably expressing PA-GFP-α-tubulin/mCherry-α-tubulin or mEos-α-tubulin, respectively, were cultured on glass coverslips as described. Mitotic cells were identified by Differential Interference Contrast (DIC) microscopy, and imaging was performed using a Plan-Apo 100× NA 1.40 DIC objective on a Nikon TE2000U inverted microscope equipped with a Yokogawa CSU-X1 spinning-disc confocal head containing two laser lines (488 nm and 561 nm) and a Mosaic (Andor) photoactivation system (405 nm). Photoactivation and photoconversion were performed in late prometaphase or metaphase cells, the latter treated with 10 µM MG132 for less than 2 h, identified by mCherry-α-tubulin or mEos-α-tubulin signal, respectively. A region of interest for PA/PC was selected using a line segment placed perpendicular to the main axis in one of the sides adjacent to the metaphase plate. Seven 1-µm separated z-planes centered in the middle of the mitotic spindle were captured every 10–15 s for 4.5 min (a prePA/PC image was always acquired) using 561-nm and 488-nm lasers and an iXon+ EM-CCD camera. MT t_1/2 was calculated using a MATLAB algorithm for kymograph generation and analysis. Briefly, generated kymographs were collapsed, and the fluorescence intensity curve was created. Fluorescence intensity peaks of each time frame were defined by manually tracking regions of the highest fluorescence intensities followed by automated curve fitting and normalized intensities to the first time-point after PA/PC following background subtraction (background values obtained from quantifying the respective nonactivated half-spindle). Values were corrected for photobleaching by normalizing to the values obtained from the quantification of fluorescence loss of whole-cell sum projected images. MT turnover was calculated based on a fitted curve of the normalized intensities at each time point (corrected for photobleaching) to a double exponential curve A1*exp(–k1*t) + A2*exp(–k2*t) using MATLAB (MathWorks), in which t is time, A1 represents the less stable (non-kMTs) population, and A2 the more stable (kMTs) population with decay rates of k1 and k2, respectively (cells displaying an R² value <0.99 were excluded from quantification). From the curve, the rate constants and the percentage of MTs were extracted for the fast (typically interpreted as the fraction corresponding to non-kMTs) and the slow (typically interpreted as the fraction corresponding to kMTs) processes. The t_1/2 was calculated as ln2/k for each MT population.

Detailed quantitative analysis of centrosome location and nuclear envelope topology was performed using custom made MATLAB scripts MathWorks, R2018a). Tracking of centrosome position/trajectories was performed in 3D space using image stacks with a pixel size of 0.190 µm and z-step of 0.7 µm. Nuclear shape was reconstructed in 3D using H2B-GFP as a marker. From the centrosome locations and nuclear envelope reconstruction, it was possible to calculate the angle between the centrosome–centrosome axis and the nucleus major axis at nuclear envelope breakdown (NEB).

10-µm-width line micropatterns to control individual cell shape and adhesion pattern were produced as follows: glass coverslips (22 × 22 mm, no. 1.5, Van Waters and Rogers) were activated with plasma (Zepto Plasma System, Diener Electronic) for 1 min and incubated with 0.1 mg/ml of PLL(20)-g[3,5]-PEG(2) (SuSoS) in 10 mM Hepes at pH 7.4 for 1 h at room temperature. After rinsing and air-drying, the coverslips were placed on a synthetic quartz photomask (Delta Mask), previously activated with deep-UV light (PSD-UV, Novascan Technologies) for 5 min. 3 µl of MiliQ water was used to seal each coverslip to the mask. The coverslips were then irradiated through the photomask with the UV lamp for 5 min. Afterward, coverslips were incubated with 25 µg/ml of fibronectin (Sigma-Aldrich) and 5 µg/ml of Alexa 647–conjugated fibrinogen (Thermo Fisher Scientific) in 100 mM NaHCO₃ at pH 8.6 for 1 h at room temperature. Cells were seeded at a density of 50,000 cells/coverslip and allowed to spread for ~10–15 h before imaging. Nonattached cells were removed by changing the medium ~2–5 h after seeding.

Cells were grown until 90% confluence in complete growth medium and harvested by centrifugation at 1,200 rpm for 5 min. Cell pellets were washed once with warm PBS and resuspended in ice-cold lysis buffer (50 mM Tris HCl, pH 7.4, 150 mM NaCl, 1 mM EDTA, 1 mM EGTA, 0.5% NP-40, and 0.5% Triton X-100) supplemented with a cocktail of protease inhibitors (Roche). Protein samples were denatured in Laemmli buffer at 95°C for 5 min, and 50 µg of total protein were separated by 10% (vol/vol) SDS-PAGE electrophoresis. Proteins were transferred to a nitrocellulose membrane using an IBlot Dry Blotting System (Invitrogen). TTL and tubulins were probed using the following antibodies: rabbit polyclonal anti-TTL (1:2,000, ProteinTech), rat monoclonal anti-tyrosinated α-tubulin clone YL1/2 (1:2,000, Bio-Rad), rabbit polyclonal anti-detyrosinated α-tubulin (Liao et al., 2019), mouse monoclonal anti-polyglutamilated tubulin (1:1,000, Adipogen), mouse monoclonal anti-α-tubulin B-5-1-2 clone (1:10,000; Sigma-Aldrich), and mouse monoclonal anti-β-tubulin clone T5201 (1:2,000, Sigma-Aldrich). Other proteins were immunodetected using mouse monoclonal anti-Cas9 (1:2,500, Merck), anti-mCherry (1:2,500, kind gift from I. Cheeseman, Massachusetts Institute of Technology, Cambridge, MA), rabbit or goat polyclonal anti-GFP (1:2,500, made by our in-house facility and 1:1,000, Rockland Immunochemicals Inc., respectively), rabbit polyclonal anti-MCAK (1:5,000, kind gift from D. Compton, Geisel School of Medicine, Dartmouth College, Hanover, NH), mouse monoclonal anti-Myc (1:10,000; Cell Signaling Technology), and mouse monoclonal anti-GAPDH (1:40,000; Proteintech). HRP-conjugated secondary antibodies (1:5,000–10,000, Jackson Immunoresearch) were visualized using an ECL system (Bio-Rad). Protein levels were quantified by chemiluminescence using a ChemicDoc XRS+ system (Bio-Rad).

Logistic regression was performed to evaluate the impact of multiple treatments and the occurrence of lagging chromosomes during anaphase, adjusting the model for the different experiments. One of the advantages of logistic regression is that it takes into account the effect of a risk predictor (coefficient size) that evaluates the enhancement (positive coefficient) or reversal (negative coefficient) of a particular variable/phenotype. Data variation was also taken into account when the logistic regression model was adjusted for this variable. Significance was set at α = 0.05, and these analyses were performed using SPSS, version 24.0. An unpaired two-tailed Student’s t test was used to determine the significance of differences between two groups (fluorescence intensity, spindle length, inter-KT distance, and MT t_1/2). The Student’s t test was adjusted to the significance of the variance between experiments. Comparison between multiple conditions was performed using one-way ANOVA. The non-parametric Kolmogorov–Smirnov test was used to determine the significance of differences in centrosome separation between conditions. Significance was set at α = 0.05, and these analyses were performed using SPSS, version 24.0, and Prism, version 7.0a.

Fig. S1 shows that TTL depletion increases α-tubulin detyrosination on astral MTs and in the vicinity of KTs without affecting inter-KT distance. Fig. S2 shows evidence for cellular adaptation to the chronic loss of TTL using CRISPR-Cas9 KO cells. Fig. S3 shows that faithful chromosome segregation requires α-tubulin retyrosination and relies on the catalytic activity of TTL. Fig. S4 shows that chromosome missegregation observed after TTL depletion is not due to altered centrosome separation at nuclear envelope breakdown. Fig. S5 shows that overexpression of a second kinesin-13 (Kif2b) also rescues chromosome missegregation due to increased α-tubulin detyrosination.

We would like to thank Linda Wordeman for kindly sharing MCAK constructs, Ben Kwok (University of Montreal, Montreal, Quebec, Canada) and Duane Compton (Geisel School of Medicine, Dartmouth, NH) for the UMK57, Stephan Geley and René Medema for the U2OS lines, Jakob Nilsson (Novo Nordisk Foundation Center, University of Copenhagen, Copenhagen, Denmark) for gateway plasmids, Carsen Janke for the TTL-YFP plasmids, Martina Barisic (Danish Cancer Society Research Center, Copenhagen, Denmark) for technical support, and the Biosciences Screening Scientific platform at i3S.

This work was funded by the European Research Council under the European Union’s Horizon 2020 research and innovation program (grant agreement no. 681443) and FLAD Life Science 2020 (Proj. 261/2014 to H. Maiato), by grants from the Danish Cancer Society Scientific Committee (KBVU; R146-A9322) and the Lundbeck Foundation (R215-2015-4081 to M. Barisic), and grants from FEDER - Fundo Europeu de Desenvolvimento Regional funds through the COMPETE 2020 - Operacional Programme for Competitiveness and Internationalization (POCI) Portugal 2020 and by Portuguese funds through Fundação para a Ciência e a Tecnologia/Ministério da Ciência, Tecnologia e Ensino Superior in the framework of the project PTDC/BEX-BCM/1758/2014 (POCI-01-0145-FEDER-016589) to J.G. Ferreira and PPBI-POCI-01-0145-FEDER-022122 to the Biosciences Screening Scientific platform at i3S. L.T. Ferreira was supported in part by a studentship from Fundação para a Ciência e a Tecnologia (SFRH/BD/79174/2011).

The authors declare no competing financial interests.

Author contributions: Investigation, formal analysis, and validation (L.T. Ferreira, B. Orr, G. Rajendraprasad, C.G. Boldú, J.T. Lima); conceptualization, supervision, and funding acquisition (H. Maiato, M. Barisic, J.G. Ferreira); resources, methodology, and software (A.J. Pereira); validation and formal analysis (C. Lemos); writing - original draft (L.T. Ferreira, H. Maiato); writing – review and editing (L.T. Ferreira, G. Rajendraprasad, C.G. Boldú, J.T. Lima, H. Maiato, M. Barisic); project administration (H. Maiato).