Potent microtubule-depolymerizing activity of a mitotic Kif18b–MCAK–EB network

Within all eukaryotes, length control of microtubule polymers is essential. Microtubule length is controlled through the regulation of microtubule dynamics by motors and microtubule-associated proteins (Mitchison and Kirschner, 1984). At the onset of mitosis, the frequency of microtubule catastrophe increases dramatically, leading to disassembly of the interphase microtubule network (Belmont et al., 1990; Piehl and Cassimeris, 2003). These dynamic properties of microtubules are essential to enable the dynamic assembly and remodelling of the mitotic spindle throughout mitosis and promote chromosome alignment and segregation. In most eukaryotes, ranging from yeast to parasites and humans, the increase in catastrophe frequency is driven by the upregulation of microtubule-depolymerizing kinesin motors. These motors shorten microtubules by promoting their catastrophe (reviewed in Walczak et al., 2013). There are two major families of depolymerizing kinesins: kinesin-8 and kinesin-13. Generally, the absence of these motors is associated with defects in cell division (Gupta et al., 2006; Maney et al., 1998; Walczak et al., 1996). Yeast only have kinesin-8 motors: Kip3 and KlpB from budding yeast and Aspergillus nidulans, respectively, both essential for spindle positioning (Cottingham and Hoyt, 1997; Miller et al., 1998; Rischitor et al., 2004; Walczak et al., 1996). Metazoan species such as Drosophila melanogaster have both the kinesin-8 (Klp67) and kinesin-13 (Klp10, Klp59D) microtubule depolymerases with key roles in spindle assembly and positioning (Savoian and Glover, 2010; Goshima et al., 2005).

In mammalian cells, the kinesin-13 family consists of three members: Kif2a, Kif2b and MCAK (also known as Kif2c). Kif2a is present at centrosomes and weakly at kinetochores (Cameron et al., 2006), and Kif2b is absent from most cells and appears as a weak microtubule depolymerase in human cells (Tanenbaum et al., 2009; Welburn and Cheeseman, 2012). MCAK is the most potent and best-characterized microtubule-depolymerizing kinesin (Ems-McClung and Walczak, 2010; Friel and Welburn, 2018; Walczak et al., 2013). MCAK is strongly localized to centromeres in mitosis (Walczak et al., 1996; Wordeman et al., 1999). The endogenous localization of MCAK was reported to be cytoplasmic in interphase and at microtubule ends in mitosis (Moore et al., 2005; Tanenbaum et al., 2011; Wordeman et al., 1999). Most studies use transiently transfected MCAK tagged with a genetically encoded fluorophore to image MCAK plus-end targeting to microtubules (Moore et al., 2005). In vitro, rather than walking on microtubules, MCAK diffuses short distances on the microtubule lattice (Helenius et al., 2006), raising the question of how MCAK reaches the ends of crowded microtubules in the dense cytoplasm in cells. Also in vitro, this atypical motor associates with microtubule end-binding (EB) proteins to enhance its localization to microtubule plus ends, utilizing an SxIP-like motif, whereas in cells, disruption of this motif prevents the ability of GFP–MCAK to accumulate at microtubule ends (Honnappa et al., 2009; Montenegro Gouveia et al., 2010; Moore et al., 2005). EB3 (also known as MAPRE3)-dependent accumulation of MCAK at microtubule plus ends increases microtubule catastrophes (Montenegro Gouveia et al., 2010). However, the affinity of EB proteins for MCAK is low, around 10 µM (Buey et al., 2012).

The human kinesin-8 family has three members that regulate microtubule length: Kif19 at cilia, Kif18a on kinetochore fibres and Kif18b on astral microtubules (reviewed in Shrestha et al., 2018). The kinesin-8 motor Kif18b shortens astral microtubules in mitosis when it is released into the cytoplasm from the nucleus (Lee et al., 2010; McHugh et al., 2018; Tanenbaum et al., 2011; Walczak et al., 2016). It requires both EB proteins and its C-terminal microtubule-binding tail to accumulate at microtubule ends and depolymerize them (McHugh et al., 2018; Stout et al., 2011; Tanenbaum et al., 2011). The absence of Kif18b or MCAK leads to aberrant microtubule growth, defects in spindle assembly and positioning and lagging chromosomes (Domnitz et al., 2012; Huang et al., 2007; McHugh et al., 2018; Rankin and Wordeman, 2010). Although Kif18b shortens microtubule length through its depolymerase activity, this activity in vitro remains modest compared to kinesin-13 motors, and does not explain the strong depolymerization associated with Kif18b function in vivo. Here, using purified proteins, we show that Kif18b uses its motile properties to move MCAK and EB3 and promote their accumulation to microtubule plus ends. We also observed that MCAK and EB3 influence the motor properties of Kif18b. We reconstituted microtubule dynamics in the presence of Kif18b, EB3 and MCAK, allowing us to dissect the contributions of each factor to the activity of the network on microtubules and at microtubule plus ends. When all three components of the network were present at microtubule plus ends, we observed a large increase in catastrophe frequency, which resulted in a significant decrease in microtubule length. Overall, it is the collective effect of Kif18b, EB3 and MCAK on microtubules that enables robust change in microtubule dynamics and reduces microtubule length in mitosis.

First, we investigated the localization of endogenous MCAK at microtubule plus ends during mitosis. In HeLa cells, we observed co-localization of MCAK with endogenous EB1 (also known as MAPRE1) that marks the plus ends of astral microtubules. We then examined the localization of MCAK in a Kif18b-knockout cell line that we had previously generated using Cas9-induced engineering, in which the long microtubule phenotype could be reversed by re-expressing Kif18b (McHugh et al., 2018). In cells lacking Kif18b (Fig. S1A), MCAK was significantly reduced at microtubule plus ends (Fig. 1A,B). These data, alongside previous work that showed microtubule plus-end reduction of MCAK after siRNA-mediated depletion of Kif18b (Tanenbaum et al., 2011), indicate that the localization of MCAK to microtubule plus ends was compromised in mitosis when Kif18b was absent both in the long and short term (Fig. 1A,B).

Although MCAK requires Kif18b for plus-end localization, the molecular mechanism underlying this dependency is not known, despite a weak interaction reported (Tanenbaum et al., 2011). The high-resolution molecular structure of dimeric MCAK is unknown. MCAK has a complex conformation in solution that is stabilized through intramolecular interactions, which could reduce the interaction with Kif18b (Ems-McClung et al., 2013; Maney et al., 2001; McHugh et al., 2019; Zong et al., 2016). We therefore examined the Alphafold2-predicted model structure for monomeric MCAK, which revealed that a large part of the motor outside the catalytic domain is unstructured (Fig. S1B) (Jumper et al., 2021). This model correctly predicts that the final 20 residues of the C-terminus form a small helix and interact with the motor through conserved residues, which was previously shown in both Xenopus and human MCAK (Talapatra et al., 2015; Zong et al., 2016). We then used Colabfold, a Google Colab platform powered by AlphaFold2, to predict the structure of dimeric MCAK; however, we were not successful due to the large low-complexity regions (Mirdita et al., 2022 preprint). Similarly, much of Kif18b, apart from the catalytic domain and the dimerization domain, consists of low-complexity regions (Fig. S1C). These unstructured regions make it difficult to identify regions of interactions between MCAK and Kif18b.

To probe for the interaction between MCAK and Kif18b, we incubated Kif18b–His bound to Ni²⁺-agarose beads with purified SNAP–MCAK_1–177 or SNAP–MCAK_584–725. We found that MCAK did not bind to Kif18b in solution (Fig. S1D,E). Thus, we hypothesized that Kif18b might interact dynamically with MCAK only in the context of the microtubules, facilitated by EB proteins that bind both Kif18b and MCAK (Honnappa et al., 2009; Montenegro Gouveia et al., 2010; Stout et al., 2011). We first tested whether Kif18b interacts with EB proteins using guanylyl-(α,β)-methylene-diphosphonate (GMPCPP)-stabilized microtubules. EB3 alone does not strongly bind to GMPCPP-stabilized microtubules (Maurer et al., 2011; Roth et al., 2018). Using in vitro reconstitution and total internal reflection fluorescence (TIRF) microscopy imaging, we imaged microtubules (Rhodamine, HiLyte Fluor 647), Kif18b (GFP, mRuby3), EB3 (SNAP₆₄₇–EB3) and MCAK (GFP) in the presence of 2 mM ATP. The polarity of the microtubules could be inferred because Kif18b is a microtubule plus-end-directed motor that accumulates at microtubule plus ends (McHugh et al., 2018). The amount of Kif18b accumulation at microtubule plus ends was length dependent (Fig. S2A). Recombinant SNAP₆₄₇–EB3 alone uniformly and weakly decorated the GMPCPP-stabilized microtubule lattice (Fig. 1C,D), and soluble SNAP₆₄₇–EB3 provided some background fluorescence. Upon addition of 12.5 nM Kif18b, we observed that SNAP₆₄₇–EB3 strongly colocalized with Kif18b at the plus ends of microtubules (Fig. 1C,D; Fig. S2B). Thus, on GMPCPP-stabilized microtubules, Kif18b binds EB3 and enhances the localization of EB3 to plus ends.

Based on the Kif18b-dependent localization of MCAK to microtubule plus ends in HeLa cells (Fig. 1A,B), we tested whether Kif18b influences the localization of MCAK on microtubules in vitro. Purified 12.5 nM MCAK–GFP bound diffusely along the length of double-stabilized (taxol and GMPCPP) microtubules (Fig. 1E,F). With the addition of 12.5 nM Kif18b–mRuby3, the localization of MCAK shifted towards the plus end of the microtubules, although MCAK still decorated the lattice (Fig. 1E,F). The localization of Kif18b–mRuby3 to the microtubule plus end was slightly reduced by the addition of MCAK and EB3 (Fig. S2C). This could be due to the Kif18b C-terminus engaging with EB3 and MCAK rather than the microtubule lattice. Indeed, we previously showed that the last 90 amino acids of Kif18b that contain two SxIP motifs increase the residence time of Kif18b at microtubule plus ends by 3-fold (McHugh et al., 2018). Thus, in our assay, Kif18b promotes the accumulation of both MCAK and EB3 proteins individually to the plus ends of microtubules. In the presence of both 50 nM SNAP₆₄₇–EB3 and 12.5 nM Kif18b–GFP, MCAK plus-end accumulation was further enhanced (Fig. 1E,F), indicating that, although MCAK interacts weakly with EB proteins (Montenegro Gouveia et al., 2010) and Kif18b (Fig. 1E,F), multivalent weak interactions of MCAK with both EB3 and Kif18b increase its accumulation to microtubule ends. The distribution of MCAK at the plus ends of microtubules varied upon addition of Kif18b or Kif18b and EB3, possibly reflecting different degrees of taper at the plus end of the taxol- and GMPCCP-stabilized microtubules that is induced by depolymerases (MCAK–Kif18b motor domain) and microtubule-stabilizing factors (Kif18b C-terminus–EB3) (McHugh et al., 2018; Montenegro Gouveia et al., 2010).

To test whether the Kif18b-dependent localization of MCAK to microtubule plus ends was specific to MCAK, we mixed the microtubule crosslinker PRC1–GFP (Subramanian et al., 2010) and Kif18b and incubated them with stabilized microtubules that were pre-immobilized on coverslips (Fig. S2D,E). No accumulation of PRC1 at the ends of single microtubules was observed, confirming that the Kif18b-dependent delivery of MCAK and EB3 was specific. The presence of Kif18b at the ends of microtubules resulted in a small decrease in PRC1 intensity (Fig. S2E), perhaps due to competition for space on the lattice in this region. Our data indicate that Kif18b promotes EB3 and MCAK plus-end targeting and that Kif18b and EB3 proteins function cooperatively to facilitate microtubule plus-end accumulation of MCAK.

Previous work showed that in cells, Kif18b does not interact with a phosphomimetic mutant of MCAK, in which residues phosphorylated by Aurora kinases in the N-terminus of MCAK were mutated to glutamates (Tanenbaum et al., 2011). Thus, we hypothesized that MCAK interaction with Kif18b might be mediated by its N-terminus. To test this, we expressed the N-terminus of MCAK (MCAK_1–177) and its C-terminus (MCAK_584–725) fused to SNAP labelled with Alexa Fluor 546 (SNAP₅₄₆). Alone, SNAP₅₄₆–MCAK_1–177 did not bind to the microtubules, nor did it bind Kif18b in solution (Fig. 2A–C; Fig. S1E). In the presence of 12.5 nM Kif18b, we observed an enrichment of SNAP₅₄₆–MCAK_1–177 at the plus ends of GMPCPP-stabilized microtubules, where Kif18b accumulated (Fig. 2A–C). This was further enhanced by the addition of EB3. In contrast, the C-terminus of MCAK (MCAK_584–725) uniformly decorated the microtubule lattice both in the presence and absence of Kif18b (Fig. 2D–F). These results indicate that microtubule-bound Kif18b interacts specifically with the N-terminus of MCAK.

We next tested whether Kif18b can transport MCAK along microtubules using SNAP₅₄₆–MCAK_1–177. We observed SNAP₅₄₆–MCAK_1–177 moving with Kif18b–GFP along GMPCPP-stabilized microtubules in the absence and presence of unlabelled EB3 by two-colour fluorescence imaging (Fig. 2G). A fraction of Kif18b motors displayed co-localization with the N-terminal fragments of MCAK. This was increased in the presence of EB3, which interacts with both MCAK and Kif18b. In the presence of EB3, most Kif18b motors co-localized with the N-terminal fragments of MCAK (Fig. 2H). Kif18b displayed a modest increase in diffusive behaviour in the presence of EB3 and/or MCAK N-terminus, indicating that the binding of cargos impacts the behaviour of Kif18b on the lattice (Fig. 2I). Additionally, the velocity of Kif18b decreased when bound to the N-terminus of MCAK and/or EB3 (Fig. 2J). Kif18b–GFP remained dimeric on microtubules when alone or in the presence of EB3 (Fig. 2K; Fig. S3A,B). However, there was an increase in the fluorescence intensity of Kif18b–GFP moving on microtubules in the presence of the MCAK N-terminus, indicating potential oligomerization (Fig. 2K). Together, our work indicates that Kif18b interacts with MCAK through its N-terminus and that this interaction is further stabilized by EB3, which binds both Kif18b and MCAK. Kif18b transports the N-terminus of MCAK towards the plus end of the microtubule lattice, while MCAK also influences the motile properties of Kif18b.

To test whether Kif18b can promote transport of full-length MCAK, we then imaged low concentrations of labelled MCAK–GFP on stable microtubules to record single-molecule events using fast imaging (10 frames per second) (Fig. S3C). MCAK alone diffused on microtubules – we measured a diffusion coefficient of 5510±734 nm²/s (estimated by the least square fit; indicated as gradient±95% c.i.) (Fig. S3D) – and remained bound to the lattice for 1.3±0.5 s on average (Fig. S3E). In the presence of 10 nM Kif18b, MCAK lattice residency increased to 2.4±0.2 s and MCAK was more diffusive with a 4-fold increase in diffusion coefficient to 24,082±1478 nm²/s (Fig. S3C–E). At 10 nM Kif18b and the sub-nanomolar concentrations of MCAK–GFP that are required for single-molecule imaging, we saw MCAK accumulating at the microtubule ends. Although we could not distinguish the polarity of the stable microtubules, we could infer that MCAK was accumulating at microtubule plus ends, where the Kif18b concentration would have been at its highest, given the plus-end accumulation of Kif18b (Fig. S3C) (McHugh et al., 2018). We also observed events of directional movement of MCAK, often at rates of around 300 nm/s, corresponding to those seen for single Kif18b motors (McHugh et al., 2018) (Fig. S3C). This, when combined with our data for SNAP₅₄₆–MCAK_1–177 (Fig. 2G), indicates that the accumulation of MCAK at the plus ends of microtubules in the presence of Kif18b is due, at least in part, to direct plus-end-directed transport of MCAK as a cargo of Kif18b. Multivalent interactions between Kif18b, EB3 and MCAK proteins in the context of the microtubule create an Kif18b–MCAK–EB3 plus-end-localized network.

We previously showed that the C-terminus of Kif18b binds to microtubules and enables Kif18b to remain bound to the plus ends of both stable and dynamic microtubules (McHugh et al., 2018). It also opposes the weak depolymerase activity of Kif18b, similarly to budding yeast kinesin-8 Kip3 (Su et al., 2011). We therefore tested whether Kif18b interfered with MCAK-induced depolymerization of stable microtubules through its microtubule-binding C-terminus. It is important to note that full-length Kif18b–GFP accumulated at microtubule plus ends but did not depolymerize GMPCPP-stabilized microtubules, and that MCAK caused slow microtubule depolymerization of GMPCPP-stabilized microtubules from both ends rather than inducing catastrophe (Fig. 3A). This slow depolymerization of microtubules allowed us to probe how MCAK and Kif18b can work together at microtubule ends (Fig. 3A–E) and image single molecules of MCAK on microtubules (Fig. 3F). In the presence of 25 nM MCAK–GFP, the addition of Kif18b caused a decrease in depolymerization of microtubule ends that was dependent on Kif18b concentration and specific to the plus end where Kif18b–GFP accumulated (Fig. 3A,B). This led to an asymmetry in plus- and minus-end depolymerization rates, proportional to the concentration of Kif18b (Fig. 3C). In the presence of 25 nM MCAK and 25 nM full-length Kif18b, the rate of the microtubule plus-end depolymerization was severely reduced to 0.036 µm/min, whereas the free minus-end depolymerization rate was similar to that with MCAK alone (0.390 µm/min) (Fig. 3D; Table S1). We found that in the presence of MCAK and a C-terminally truncated Kif18b₁₋₅₉₀–GFP, the microtubule depolymerization rate was similar to the rate in the presence of MCAK alone, indicating that the plus-end localization of the C-terminus of Kif18b interfered with MCAK-mediated depolymerization of stable microtubules. We then used a high concentration of Kif18b truncation mutants to compensate for their inability to accumulate at microtubule plus ends (Fig. 3E; Table S2). The addition of 250 nM Kif18b_1–590 displayed strong depolymerase activity, as previously shown (McHugh et al., 2018). The addition of 500 nM Kif18b_591–852 (which is monomeric and corresponds to the concentration of molecules for 250 nM dimeric Kif18b) significantly reduced the rate of MCAK-induced depolymerization, without affecting the landing rate of MCAK, defined as the number of MCAK motors binding to the lattice (µm⁻¹/s) (Fig. 3F; Table S3). This indicates that the C-terminus of Kif18b further stabilizes the GMPCCP-stabilized microtubule lattice against MCAK depolymerase activity, rather than preventing MCAK association with the microtubules.

On dynamic microtubules, addition of 500 nM of the C-terminus of Kif18b increased the growth rate of microtubules (Fig. S4A), consistent with a role in microtubule lattice stabilization. However, it did not impact the catastrophe frequency or overall microtubule length (Fig. S4B,C). In the presence of 12.5 nM MCAK and 500 nM Kif18b_591–852, the growth rate of microtubules remained elevated (Fig. S4D), and the catastrophe frequency and overall microtubule length were largely unchanged (Fig. S4E,F). Thus, in the context of dynamic microtubules, the C-terminus of Kif18b alone did not stabilize microtubules against MCAK activity. Overall, stabilized microtubules allowed the accumulation of Kif18b and MCAK at microtubule plus ends where Kif18b mediated a stabilizing effect. In contrast, on dynamic microtubules, catastrophe led to loss of motors associated with the depolymerizing end, lessening any stabilizing role the C-terminus of Kif18b might have had. Most kinesin-8 proteins appear to use their C-terminal domains to counteract the activity of their microtubule-depolymerizing motor domains, but most importantly to facilitate their plus-end targeting (Stumpff et al., 2011; Su et al., 2013; Su et al., 2011).

We then sought to define the properties of the Kif18b–MCAK–EB3 network on dynamic microtubules. We first analysed the effects of MCAK alone on dynamic microtubules using GMPCPP-stabilized seeds with microtubules growing from them. We observed that MCAK acts as a strong catastrophe factor (Fig. 4A–C; Table S4). Increasing MCAK concentration from 0 to 25 nM led to high catastrophe frequencies, ranging from 0.34 min⁻¹ to 1.1 min⁻¹, respectively (Fig. 4A,B; Table S4), as previously reported (Gardner et al., 2011). The extension lengths of microtubules were significantly shorter upon increasing concentrations of MCAK, corresponding to 2.68±0.21 and 1.51±0.14 µm for 12.5 and 25 nM MCAK, respectively, whereas the microtubule extension lengths alone were 3.85±0.69 µm (indicated as median±95% c.i.) (Fig. 4C; Table S4).

We next analysed whether microtubule dynamics in the presence of MCAK were altered by the addition of EB3 (Fig. 4D–F). In the presence of 12.5 nM MCAK, catastrophe frequency increased significantly from 1.02 to 1.16 min⁻¹ upon the addition of 25 nM EB3 (Fig. 4E). However, there were no major changes in microtubule lengths between conditions in which MCAK or MCAK and EB3 were used (Fig. 4F; Table S5). In addition, we could see little enrichment of MCAK at microtubule plus ends under these conditions (Fig. 4D). This can be explained by the dissociation constant of EB proteins with MCAK being in the micromolar range of 10 µM (Buey et al., 2012). It was nevertheless possible to see robust EB3-dependent MCAK tracking the growing ends of microtubules when higher concentrations of EB3 and MCAK were used. However, the microtubule seeds had to be additionally stabilized with taxol to counteract MCAK depolymerase activity as seen in Montenegro Gouveia et al. (2010) and Fig. S5A. Thus, we did not quantify microtubule dynamics under these conditions. In total, these data indicate that at low nanomolar concentrations, MCAK depolymerizes microtubules independently of EB3.

We next tested how Kif18b and MCAK influence microtubule dynamics when combined. We have previously shown that Kif18b only has a modest effect on microtubule dynamics, increasing catastrophe frequency and growth rate, thereby leading to slightly reduced microtubule extension lengths (McHugh et al., 2018). Using 12.5 nM of each dimeric motor, we compared the effect of the motors alone and in combination (Fig. 4G). The addition of 12.5 nM MCAK increased catastrophe frequency (Fig. 4H; Table S6), and this occurred to a greater extent than for Kif18b (Fig. 4G,I). In the presence of both MCAK and Kif18b, catastrophe frequency further increased (Fig. 4I), whereas microtubule length was reduced when both Kif18b and MCAK were present (Fig. 4G,I).

We next defined the properties of the Kif18b–MCAK–EB3 network on dynamic microtubules. When using 12.5 nM MCAK, 12.5 nM Kif18b and 25 nM EB3, no dynamic microtubule extensions were seen (Fig. S5B), indicating that the combination of Kif18b, EB3 and MCAK had a dramatic effect on microtubules, which could not equate to the sum of their individual activities. To enable quantification of this effect, we decreased the amount of MCAK in our assays to 5 nM, while keeping 12.5 nM Kif18b and 25 nM EB3, and imaged the microtubule extensions grown from GMPCPP-stabilized seeds in the presence of 12 µM tubulin. A significant effect was seen in the presence of all three proteins on both catastrophe frequency and microtubule extension lengths. Catastrophe frequency was found to be the highest and, conversely, the dynamic microtubule extensions were the shortest under these conditions (Fig. 5A,B; Table S7). There was a 2-fold increase in catastrophe frequency from 0.34 min⁻¹ for treatment with MCAK and EB3 to 0.74 min⁻¹ for treatment with MCAK, EB3 and Kif18b (Fig. 5A). Overall, this resulted in a 1.8-fold reduction in the length of dynamic extensions when Kif18b was added to EB3 and MCAK, with the microtubules remaining dynamic and short (Fig. 5B,C; Table S7). We then imaged the three proteins simultaneously using microtubule seeds that were doubly stabilized with taxol and GMPCPP, prepared as in Montenegro Gouveia et al. (2010) (Fig. 5D). As expected, MCAK mostly diffused on the lattice, whereas EB3 tracked the growing ends of microtubules, and Kif18b exhibited processive behaviour walking towards microtubule plus ends. In the presence of Kif18b, there was a slight enrichment of MCAK towards the growing plus end of microtubules (white arrows, Fig. 5D) and plus-end-directed processive events could be seen with MCAK–GFP moving at 296 nm/s, similar to the results we previously reported for Kif18b alone (Fig. S5C,D) (McHugh et al., 2018). When all three proteins, Kif18b, MCAK and EB3, were present, MCAK tracked the plus ends of growing microtubules along with Kif18b and EB3 (Fig. 5D). The addition of EB3 increased the proportion of processive plus-end-directed MCAK–GFP motors, although the majority (>90%) of MCAK–microtubule-binding events were diffusive (Fig. S5E). When the protein concentration was slightly higher, Kif18b-mediated transport events of EB3 and MCAK towards the microtubule ends were more readily apparent (Fig. 5E; Movie 1). These complexes formed bright clusters of multiple proteins, indicating assembly of the Kif18b–MCAK–EB3 network, and were transported to the plus ends by Kif18b. Together, the Kif18b–MCAK–EB3 network cooperates in an integrated manner to dramatically increase microtubule catastrophe in mitosis.

Microtubule dynamics dramatically increase at mitotic onset when microtubules shorten to remodel the microtubule cytoskeleton and assemble a bipolar spindle (Belmont et al., 1990; Piehl and Cassimeris, 2003). This change in microtubule length is attributed mainly to an increase in the catastrophe frequency (Belmont et al., 1990). Although Kif18b alone is a weak microtubule depolymerase in vitro, it is a major microtubule depolymerization factor in mitosis once released from the nucleus (Lee et al., 2010). In the absence of Kif18b or MCAK, mitotic microtubules remain long, leading to spindle assembly and positioning defects in humans (McHugh et al., 2018; Rankin and Wordeman, 2010; Stout et al., 2011; Tanenbaum et al., 2011; van Heesbeen et al., 2016). Here, we show mechanistically how MCAK, Kif18b and EB3 cooperate to increase the catastrophe frequency of microtubules. Using single-molecule imaging and in vitro reconstitution, we find that Kif18b can transport MCAK via interactions with both the MCAK N-terminus and EB3 to microtubule plus ends, where they promote microtubule catastrophe. It is clear from our data that missing any one component of the Kif18b–MCAK–EB3 network results in reduced capability for limiting microtubule length.

Our data challenge previous models that postulate that MCAK uses diffusion alone to reach microtubule ends (Cooper et al., 2010; Helenius et al., 2006). Although one-dimensional diffusion on the microtubule lattice is an efficient way to reach a microtubule end over short distances, diffusion is not necessarily an efficient way to reach specifically the plus end of a microtubule in a cell. The diffusion rate of MCAK was calculated to be 5510 nm²/s with a residence time of 1.3 s. This would mean that a single motor on average explores just 84 nm of the microtubule lattice with each encounter. In the presence of Kif18b, both MCAK diffusion and residence time increase: MCAK would diffuse an average of 240 nm, thereby increasing the probability MCAK would reach or get close to a microtubule plus end. However, diffusion alone does not explain how MCAK localizes primarily to microtubule plus ends on crowded and long microtubules in cells.

The molecular structure of the Kif18b–MCAK–EB3 network remains poorly understood. MCAK binding to both Kif18b and EB proteins has been shown to be weak (Fig. S1E) (Buey et al., 2012). Given that Kif18b and MCAK have large regions of low complexity (Fig. S1B,C), it is likely that they assemble as higher oligomers through weak multivalent interactions within their unstructured domains on the microtubules (Fig. 5E), offering many binding sites to EB3.

The localization of endogenous MCAK at microtubule plus ends in interphase cells is barely detectable using immunofluorescence, prompting most studies to use fluorescently labelled MCAK (Moore et al., 2005). Endogenous MCAK has only been observed in mitosis and MCAK localization is dependent on the Kif18b and EB proteins (Fig. 1) (Tanenbaum et al., 2011), suggesting that less MCAK is present at the ends of interphase microtubules. Indeed, similarly to Kif18b, the most striking effects of the kinesin-13 proteins are on mitotic microtubules. Our work demonstrates that Kif18b, EB3 and MCAK collectively form a protein network at dynamic microtubule plus ends at very low concentrations of each component. At these plus ends, Kif18b plays a critical role as an integration platform to modulate the depolymerase activity of the network. Kif18b enhances the accumulation of MCAK on the lattice and at microtubule ends and delivers MCAK and EB3 proteins to the microtubule ends (Fig. 6), where they work together in an integrated manner to increase microtubule catastrophe. The unique and potent properties of the network cannot be imparted to the sum of the individual components (Fig. 5).

Two distinct populations of Kif18b decorating astral microtubule plus ends have been reported; one population of Kif18b colocalizes with EB1, whereas the other population precedes EB1 at the plus end of the microtubules (Shin et al., 2015). Kif18b is a fast motor, with a speed of 350 nm/s in vitro and up to 700 nm/s in vivo (McHugh et al., 2018; Tanenbaum et al., 2014), faster than microtubule growth speed that is around 200–250 nm/s (Sironi et al., 2011). Thus, Kif18b can reach the growing end of a microtubule and accumulate there using its C-terminal microtubule-binding tail independently of EB proteins (McHugh et al., 2018). Indeed, small regions at the growing ends of microtubules were found to be EB-free (Maurer et al., 2014) and lacked lateral stabilization of protofilaments (McIntosh et al., 2018). It is possible that Kif18b brings MCAK and EB3 to the terminal tubulin dimers of microtubules where tubulin is assembled into single protofilaments but is not yet closed, to promote depolymerization. There, the Kif18b–MCAK–EB3 complex could efficiently depolymerize microtubule growing ends before they have been stabilized through lateral contacts of protofilaments, to avoid an additional energy barrier to depolymerization.

Importantly, each interaction within this Kif18b–MCAK–EB3 network is individually regulated by mitotic phosphorylation, a rapid and reversible post-translational modification (Tanenbaum et al., 2011; McHugh et al., 2018). Additionally, Kif18b is confined to the nucleus before mitosis and irreversibly degraded at the end of mitosis. Thus, the integrity and activity of this depolymerizing network is tightly regulated to allow for spatial, local and temporal regulation of microtubule dynamics during mitosis. It would be interesting to define whether Kif18b also delivers other microtubule-associated proteins with low-complexity domains to plus ends through interactions with its unstructured C-terminus. Finally, our work highlights the importance of studying motors not only in isolation but examining their collective behaviour and the underlying emergent function. How cargos influence motor activity remains a poorly understood phenomenon. Rather than having a microtubule motor that provides transport to passive cargos, we reveal here how the three components of the mitosis-specific Kif18b–MCAK–EB3 network, two of which are motors, act in an integrated manner to provide increased depolymerization of microtubule plus ends with temporal and spatial layers of control.

Plasmids made for this study are summarized in Table S8. The Sf9 insect cell expression constructs for producing full-length human Kif18b–GFP, Kif18b_1–590–GFP and MCAK–GFP with a C-terminal 6×His have been described elsewhere (McHugh et al., 2018; Talapatra et al., 2015). For Kif18b–mRuby3–His, Kif18b was cloned into the pFL–mRuby3–His vector [constructed by inserting mRuby3–His6 into pFL (Addgene #110739)] using restriction enzymes. Kif18b_591–852 was cloned into pET3aTr (Addgene #64008) with an N-terminal 6×His. MCAK_1–177 and MCAK_584–725 were cloned into the pST50Trc1-HISNDHFR vector (Addgene #63940). EB3 was cloned into the pET3aTr vector containing a His, or the pST50Trc1-HISNDHFR vector containing His–TEV–SNAP or His–GFP tags. The pFL-His–TEV–GFP–PRC1 plasmid was a gift from Tarun Kapoor, Rockefeller University, NY and is described in Subramanian et al. (2010). mRuby3 was synthesized after codon optimization for expression in insect cells by GeneArt, Thermo Fisher Scientific. The DNA sequence for mRuby3 is as follows:

5′-ATGGTGTCTAAGGGCGAAGAGCTGATCAAGGAAAATATGCGTATGAAGGTGGTCATGGAAGGTTCGGTCAACGGCCACCAATTCAAATGCACAGGTGAAGGAGAAGGCAGACCGTACGAGGGAGTGCAAACCATGAGGATCAAAGTCATCGAGGGAGGACCCCTGCCATTTGCCTTTGACATTCTTGCCACGTCGTTCATGTATGGCAGCCGTACCTTTATCAAGTACCCGGCCGACATCCCTGATTTCTTTAAACAGTCCTTTCCTGAGGGTTTTACTTGGGAAAGAGTTACGAGATACGAAGATGGTGGAGTCGTCACCGTCACGCAGGACACCAGCCTTGAGGATGGCGAGCTCGTCTACAACGTCAAGGTCAGAGGGGTAAACTTTCCCTCCAATGGTCCCGTGATGCAGAAGAAGACCAAGGGTTGGGAGCCTAATACAGAGATGATGTATCCAGCAGATGGTGGTCTGAGAGGATACACTGACATCGCACTGAAAGTTGATGGTGGTGGCCATCTGCACTGCAACTTCGTGACAACTTACAGGTCAAAAAAGACCGTCGGGAACATCAAGATGCCCGGTGTCCATGCCGTTGATCACCGCCTGGAAAGGATCGAGGAGAGTGACAATGAAACCTACGTAGTGCAAAGAGAAGTGGCAGTTGCCAAATACAGCAACCTTGGTGGTGGCATGGACGAGCTGTACAAG-3′.

HeLa cells (93021013, Sigma-Aldrich, USA) were used and maintained in Dulbecco's Modified Eagle Medium (DMEM, Lonza) supplemented with fetal bovine serum (FBS; Thermo Fisher Scientific), glutamine and penicillin-streptomycin (15140122; Thermo Fisher Scientific) at 37°C in a humidified atmosphere with 5% CO₂. The generation of the Kif18b-knockout cell line is described in McHugh et al. (2018). Cells were purchased from American Type Culture Collection (ATCC) and were checked monthly for mycoplasma contamination (MycoAlert detection kit, Lonza).

Cells were washed in PBS and fixed in 3.8% formaldehyde in PHEM buffer (60 nM PIPES, 25 mM HEPES, 10 mM EGTA, 2 mM MgSO₄, pH 7.0) for 10 min. Immunofluorescence in human cells was performed using an antibody against mouse EB1 (610535; BD transduction laboratories, USA; 1:400) and a custom rabbit MCAK antibody (1:500 or 1:1000). The MCAK antibody, raised against GFP–MCAK, was generated by Eurogentec, Belgium, using the Speedy program. Over time, the antibody became rapidly unstable and non-specific. Images were acquired using a DeltaVision core microscope (bought from Applied Precision, Imsol, UK) with a 100× lens (NA: 1.4) equipped with a CoolSnap HQ2 CCD camera (Teledyne Photometrics, USA). 10–20 z-sections were acquired at 0.2–0.5 µm. Comet intensities were quantified as previously described (McHugh et al., 2018). Experiments were repeated three times independently.

Kif18b and MCAK proteins were expressed using a baculovirus expression system; Sf9 cells were infected with the virus for each construct for 60–72 h. His–Kif18b_591–852, His–SNAP–MCAK_1–177, His–SNAP–MCAK_584–725, His–EB3 and His–GFP–PRC1 proteins with the His₆ tag cleavable with the 3C protease were expressed in BL21 CodonPlus Escherichia coli cells overnight at 18°C. His-tagged Kif18b and MCAK proteins were purified as previously described (Talapatra et al., 2015). His–GFP–EB3, His–SNAP–EB3 and His–GFP–PRC1 proteins were purified using a HisTrap HP column and a gradient elution. His–SNAP–EB3, His–SNAP–MCAK_1–177 and His–SNAP–MCAK_584–725 were then coupled to SNAP-Cell 647-SiR or SNAP-Cell 546-SiR dyes (S9136S and S9132S, respectively; New England BioLabs) by incubation at 37°C for 30 min or overnight at 4°C. Labelling of His–SNAP–EB3 was around 30% of the total protein, whereas MCAK labelling was around 80–90%, as assessed by following the manufacturer's guidelines. All proteins were further purified using S200 Increase 10/300 gel filtration chromatography (GE Healthcare), in which the columns were pre-equilibrated in gel filtration buffer (25 mM PIPES, pH 6.9, 150 mM NaCl, 300 mM KCl, 5 mM β-mercaptoethanol, 1 mM MgCl₂, 1 mM Na-EGTA, and 1 mM ATP). Proteins were frozen in liquid nitrogen within 24 h of cell lysis and stored for up to 3 months at −70°C (Kif18b, MCAK) or 6 months (EB3, PRC1). Protein concentration is reported for its oligomeric state in solution. The activity of frozen proteins was found to be consistent with results from freshly purified proteins within these timescales. Porcine brain tubulin was purified as described (Castoldi and Popov, 2003) and stored in liquid nitrogen long term.

Samples were prepared as detailed in McHugh et al. (2018). An ATP concentration of 2 mM and a temperature of 30°C were used for all experiments. The stated concentrations of protein (or an equivalent volume of gel filtration buffer, up to a maximum of 3.5% total volume) were added to the final assay buffer {BRB80 buffer (80 mM K-PIPES, 1 mM MgCl₂, 1 mM EGTA, pH 6.8) with 2 mM ATP, 0.5 mg/ml casein and an oxygen-scavenging system [0.2 mg/ml glucose oxidase (Sigma-Aldrich, UK), 0.035 mg/ml catalase (C40; Sigma-Aldrich, USA), 4.5 mg/ml glucose (G2133; Sigma-Aldrich, USA) and 140 mM β-mercaptoethanol]}, slides were sealed with Valap [made of equal parts by weight of Vaseline (Unilever, UK and Ireland), Lanolin (NOW Solutions, USA) and uncoloured candle wax] and imaged immediately. For dynamic microtubule experiments, the assay buffer also contained 7–12 µM tubulin (6% Rhodamine-labelled), 1 mg/ml casein and 2 mM GTP. For dynamic microtubule experiments where seeds were double-stabilized with taxol (Sigma-Aldrich, UK) and GMPCPP (NU405S; Jena Biosciences, Germany) (Fig. S5A; Fig. 5D,E), the flow cell was washed with at least ten cell volumes of 1 mg/ml casein (with the exception of data in Fig. 5E, where to provide additional protection against catastrophe, only five cell volumes were used) and then incubated for 5 min before being flowed through with six flow-cell volumes of the assay buffer. Experiments that are compared (in the same graph or table) in this paper were performed in parallel using the same ‘master mix’ assay buffer with at least three independent repeats, with the exception of data in Fig. S4A–C where two repeats were performed for each condition. For photobleaching analysis, 0.25 nM Kif18b–GFP in BRB80 buffer was introduced to the flow cell and allowed to non-specifically adsorb to the surface. After 3 min, BRB80 with the oxygen-scavenger mix was used to wash away non-adsorbed motors. The sample chamber was imaged using the same conditions as described for single-molecule assays.

Imaging was performed on a Zeiss Axio Observer Z1 TIRF microscope using a 100× NA 1.46 objective and either a Photometrics Evolve Delta electron-multiplying charge-coupled device camera or a Photometrics Prime 95B sCMOS camera (Teledyne Photometrics, USA) controlled by Zeiss Zen Blue software. For single-molecule experiments in Fig. 1, Fig. S2 and Fig. S3C–E, an Optosplit III beam-splitter (Andor, Oxford Instruments, UK) in bypass mode inserted before the camera provided a further 2× magnification. Depolymerization assays were performed over 10 min at 1 fps or 0.5 fps for two colour imaging; this had no effect on depolymerization rates. Microtubule dynamics assays were performed over 15 min in either one, two or three colours at 0.3 fps. Two-colour imaging of Kif18b–GFP and SNAP₅₄₆–MCAK_1–177 was performed over 5 min at 0.5 fps. MCAK single-molecule imaging was performed in the GFP channel only at 10 fps (stable microtubules) or 6 fps (dynamic microtubules).

Kymographs were produced using ImageJ (National Institutes of Health), and MCAK single-molecule behaviour, catastrophe length and catastrophe time were manually measured from these kymographs. When analysing microtubule dynamics kymographs, we were unable to consistently detect microtubule growth events that occurred for less than 25 s, so these data were removed from our analysis. Line-scan intensity measurements along microtubules were taken using a fixed threshold in the microtubule fluorescence channel [Rhodamine or HiLyte 647 (TL590M-B and TL670M, respectively; Cytoskeleton, Inc., USA)] to specify the microtubule ends. Intensity measurements for Kif18b motors were taken from kymographs and represent the background-subtracted maximum intensity in the GFP channel along the trajectory of the motor. Graphpad Prism 9 was used for all statistical analysis.

Statistical analyses were performed using Prism 9 (GraphPad Software). For the calculation of the error on the median, we report the upper 95% confidence interval (c.i.). No statistical method was used to predetermine sample size. No samples were excluded from the analyses. The investigators were not blinded to allocation during experiments and outcome assessment. All experiments were performed and quantified from at least three independent experiments (unless specified otherwise), and representative data are shown.

We thank the Welburn lab and colleagues who took time to read the manuscript. We thank Agata Gluszek for the His–SNAP–EB3 plasmid and providing immunofluorescence images of the Kif18b-knockout cells, and Dave Kelly for Centre Optical Instrumentation Laboratory (COIL) support. The Wellcome Trust Centre for Cell Biology is supported by core funding from the Wellcome Trust (203149). We are grateful to Tarun Kapoor for the His–TEV–GFP–PRC1 plasmid.

The peer review history is available online at https://journals.biologists.com/jcs/article-lookup/doi/10.1242/jcs.260144.