During mitosis, microtubules in the spindle turn over continuously. At spindle poles, where microtubule minus ends are concentrated, microtubule nucleation and depolymerization, the latter required for poleward microtubule flux, happen side by side. How these seemingly antagonistic processes of nucleation and depolymerization are coordinated is not understood. Here, we reconstitute this coordination in vitro combining different pole-localized activities. We find that the spindle pole–localized kinesin-13 KIF2A is a microtubule minus-end depolymerase, in contrast to its paralog MCAK. Due to its asymmetric activity, KIF2A still allows microtubule nucleation from the γ-tubulin ring complex (γTuRC), which serves as a protective cap shielding the minus end against KIF2A binding. Efficient γTuRC uncapping requires the combined action of KIF2A and a microtubule severing enzyme, leading to treadmilling of the uncapped microtubule driven by KIF2A. Together, these results provide insight into the molecular mechanisms by which a minimal protein module coordinates microtubule nucleation and depolymerization at spindle poles consistent with their role in poleward microtubule flux.

The microtubule cytoskeleton is essential for a multitude of processes in eukaryotic cells, such as intracellular organization and trafficking, cell division, and differentiation. Microtubules are structurally polar filaments with two biochemically distinct ends with characteristic differences in their dynamic properties that are critical for microtubule function (Akhmanova and Steinmetz, 2015; Gudimchuk and McIntosh, 2021). Both ends of microtubules can switch stochastically between episodes of growth and shrinkage, a property called dynamic instability, which is ultimately a consequence of GTP hydrolysis at the microtubule end (Desai and Mitchison, 1997; Mitchison and Kirschner, 1984). The dynamic properties of microtubule plus ends and their regulation have been extensively studied, both in cells and also in vitro with purified proteins, but much less is known about the regulation of microtubule minus-end dynamics (Akhmanova and Steinmetz, 2019).

Across eukaryotic cells, major microtubule nucleation pathways require the γ-tubulin ring complex (γTuRC), a large multisubunit protein complex that serves as a template for microtubule formation (Kollman et al., 2011; Tovey and Conduit, 2018). Nucleation from purified γTuRC in vitro results in microtubules with stably γTuRC-capped minus ends and dynamic plus ends (Berman et al., 2023; Consolati et al., 2020; Rai et al., 2022,Preprint; Thawani et al., 2020; Wieczorek et al., 2021). To what extent γTuRC-nucleated microtubule minus ends remain capped in cells is not clear, and recent electron microscopy data suggests there is a mixture of capped and uncapped minus ends in cells (Kiewisz et al., 2022; Laguillo-Diego et al., 2023).

In mitotic and meiotic spindles during cell division, microtubule minus ends at the spindle poles appear to slowly depolymerize while dynamic microtubule plus ends in the spindle center display net polymerization, leading to a constant flux of microtubule mass from the spindle center to its poles (Barisic et al., 2021; Cameron et al., 2006; Ferenz and Wadsworth, 2007; Ganem et al., 2005; Mitchison, 1989; Risteski et al., 2022; Rogers et al., 2005; Steblyanko et al., 2020). How microtubule minus-end dynamics at spindle poles are regulated during poleward flux remains a major open question. On one hand, γTuRC is enriched at spindle poles, in agreement with the poles’ microtubule-nucleating properties. On the other hand, a number of microtubule-destabilizing proteins involved in regulating microtubule flux are also known to accumulate at spindle poles, such as the microtubule depolymerase KIF2A, a kinesin-13 (Gaetz and Kapoor, 2004; Ganem and Compton, 2004; Welburn and Cheeseman, 2012), and the microtubule severases, like katanin, fidgetin, and spastin (Jiang et al., 2017; McNally et al., 1996; Zhang et al., 2007). The molecular mechanism by which this mixture of seemingly antagonistic microtubule nucleating and destabilizing activities at the spindle pole can control microtubule minus-end dynamics to ensure correct microtubule flux and hence correct spindle function is unknown.

The kinesin-13s are not directionally motile, but instead bind microtubules in a diffusive manner and depolymerize them from their ends. This function requires ATPase activity as well as the positively-charged and class-specific “neck,” which is directly N-terminal to the conserved kinesin motor domain (Desai et al., 1999; Ems-McClung and Walczak, 2010; Friel and Welburn, 2018; Moores and Milligan, 2006). The N- and C-termini of the kinesin-13s drive cellular localization and homodimerization, respectively (Maney et al., 2001; Welburn and Cheeseman, 2012). In the ATP state, both the neck and motor domains bind two longitudinally bound tubulin dimers in a bent conformation, which suggests an “unpeeling” mechanism to remove tubulin from microtubule ends (Benoit et al., 2018; Trofimova et al., 2018; Wang et al., 2017).

The majority of kinesin-13 studies have focused on MCAK (KIF2C), which preferentially localizes to kinetochores in cells and is involved in the correct chromosome–spindle attachment (Ganem et al., 2005; Manning et al., 2007; Welburn and Cheeseman, 2012; Wordeman et al., 2007). In vitro, purified MCAK slowly depolymerizes both ends of artificially stabilized microtubules at similar speeds and promotes the transition from growth to fast shrinkage, called catastrophe, of dynamic microtubules (Desai et al., 1999; Gardner et al., 2011; Montenegro Gouveia et al., 2010). The paralog KIF2A, on the other hand, localizes mainly at spindle poles, and is important for correct spindle architecture and length, promoting poleward flux (Gaetz and Kapoor, 2004; Ganem and Compton, 2004; Steblyanko et al., 2020; Uehara et al., 2013; Wilbur and Heald, 2013). Although KIF2A has been studied less in vitro (Desai et al., 1999; Wilbur and Heald, 2013), it is thought that KIF2A and MCAK share identical enzymatic activities, with their functional differences in cells resulting from different localizations mediated by specific binding partners. It is however still unknown how KIF2A contributes to the control of microtubule minus-end dynamics at spindle poles, especially given that most minus ends at spindle poles are expected to be capped after having been nucleated by γTuRCs.

Microtubule uncapping by microtubule severases could allow KIF2A to depolymerize minus ends at spindle poles (Sharp and Ross, 2012; Zhang et al., 2007), and indeed severases promote poleward flux (Guerreiro et al., 2021; Huang et al., 2021). They are hexameric AAA proteins that pull on the negatively charged C-terminal tails of tubulin on the microtubule lattice, inducing damage that can lead to microtubule severing, exposing new plus and minus ends (Roll-Mecak and Vale, 2008; Zehr et al., 2020). This process has however been shown to be inefficient in the presence of soluble tubulin due to spontaneous reincorporation of tubulin at the damage sites (Jiang et al., 2017; Vemu et al., 2018), raising the question of how efficient minus-end uncapping by severases could be at the spindle poles. The molecular mechanisms of how γTuRC, KIF2A, and severases in combination control microtubule minus-end dynamics to ensure correct spindle function remain largely unknown.

Here, we demonstrate that KIF2A is an inherently asymmetric microtubule depolymerase. It preferentially binds to and depolymerizes the minus ends of microtubules, in agreement with its function in the spindle, and is distinct from the symmetric behavior of its paralog MCAK. KIF2A alone is sufficient to induce microtubule treadmilling that recapitulates poleward flux by slowly depolymerizing the minus ends of dynamic microtubules while permitting plus-end growth. We find that KIF2A still allows microtubule nucleation by γTuRC and that γTuRC capping of minus ends interferes with KIF2A-mediated minus end depolymerization. The severase spastin and KIF2A are together required to create fresh minus ends via active γTuRC-uncapping and to induce KIF2A-mediated microtubule treadmilling. Altogether, our results provide new insight into the molecular mechanism of the control of microtubule minus-end dynamics by the γTuRC/KIF2A/severase module, consistent with their role at spindle poles during cell division.

We expressed and purified the human kinesin-13 KIF2A fused to the fluorescent protein mScarlet (mScarlet-KIF2A; Fig. 1 A and Fig. S1 A) and studied its depolymerization activity on microtubules using a total internal reflection fluorescence (TIRF) microscopy–based in vitro assay (Fig. 1 A). We were surprised to observe strongly asymmetric depolymerization of glass-immobilized microtubules that were stabilized using the slowly hydrolyzable GTP analog GMPCPP (Video 1). One microtubule end depolymerized considerably faster than the other (Fig. 1 B), and mScarlet-KIF2A strongly accumulated at the more-quickly shrinking end (Fig. S2 A). This is in contrast to the activity of other kinesin-13s such as MCAK, which display rather symmetric depolymerization activity at both ends of stabilized microtubules (Desai et al., 1999; Helenius et al., 2006; Hunter et al., 2003).

Polarity-labeling of GMPCPP-microtubules showed that mScarlet-KIF2A preferentially bound to and depolymerized microtubule minus ends; plus ends depolymerized between 5 and 10 times more slowly (Fig. 1 C, left, Fig. 1 D). A purified KIF2A construct without a fluorescent tag (Fig. S1 B) showed similar asymmetric depolymerase behavior on polarity-marked GMPCPP microtubules (Fig. S2, B and C). To be able to directly compare the activities of KIF2A and MCAK under the same assay conditions, we expressed and purified MCAK fused to mCherry (mCherry-MCAK; Fig. 1 A and Fig. S1 C) and confirmed that MCAK does not have a minus end preference, neither for accumulation nor for depolymerization of GMPCPP microtubules (Fig. 1 C, right; Fig. 1 D).

Varying the mScarlet-KIF2A concentration, we observed that the depolymerization speed of both the minus and the plus end of non-polarity-marked microtubules (distinguished by considering that the minus end depolymerizes faster) increased with the KIF2A concentration, maintaining a strongly asymmetric depolymerization character at all the KIF2A concentrations tested (Fig. 1 E). All in all, this demonstrates that KIF2A is inherently different from MCAK, preferentially depolymerizing minus ends, consistent with its proposed role at spindle poles (Cameron et al., 2006; Gaetz and Kapoor, 2004; Ganem and Compton, 2004).

Kinesin-13s may recognize a particular lattice curvature at microtubule ends induced by ATP binding of the motor domain (Benoit et al., 2018; Desai et al., 1999; Hunter et al., 2003; Trofimova et al., 2018; Wang et al., 2017). Therefore, we asked if also the preferential accumulation of KIF2A at minus ends was dependent on its nucleotide state. We found that mScarlet-KIF2A accumulated at both ends of microtubules in the presence of the non-hydrolyzable ATP analog AMPPNP (Fig. 1 F, left), as observed previously with kinesin-13s (Desai et al., 1999; McHugh et al., 2019; Moores et al., 2002). However, we were surprised to find that in the presence of ADP, KIF2A accumulated only at the minus and not at the plus ends (Fig. 1 F, right). We quantified the extent of end versus lattice accumulation by measuring fluorescence intensity ratios. In the presence of ATP, KIF2A not only showed a marked binding preference for the minus end but also accumulated mildly at the plus end (Fig. 1 G, left, ATP), consistent with the observed difference in depolymerization rates at the two ends. In the presence of AMPPNP, KIF2A accumulated equivalently mildly at both plus and minus ends, remarkably without a minus-end preference (Fig. 1 G, left, AMPPNP), whereas with ADP, KIF2A accumulation only showed a minus-end preference (Fig. 1 G, left, ADP). Because ADP also increased KIF2A’s binding to the microtubule lattice (Fig. 1 H), the ratio of end-to-lattice intensity was lower for ADP than ATP, but the amount of KIF2A motors bound to minus ends was similar in the presence of ATP and ADP (Fig. 1 H). Taken together, these results indicate that the selective microtubule minus end accumulation by KIF2A is distinct from the indiscriminate end accumulation of MCAK and that there is a structure at minus ends, and not at plus ends, that KIF2A can recognize independent of ATP binding. In contrast, MCAK did not show any end accumulation in the presence of ADP, whereas it accumulated at both ends equally strongly in ATP and equally mildly in AMPPNP (Fig. 1 G, right), as observed previously (Desai et al., 1999; Helenius et al., 2006; Hertzer et al., 2006; McHugh et al., 2019).

To further investigate the mechanism of minus end recognition by KIF2A, we next purified the microtubule end capping protein iE5 (Fig. S1 D) and DARPin (D1)2 (Fig. S1 E), which specifically cap either the minus or the plus ends of microtubules, respectively (Campanacci et al., 2019; Pecqueur et al., 2012). Using polarity-marked microtubules, we observed that while (D1)2 had no influence on KIF2A’s capacity to accumulate at and depolymerize microtubule minus ends, as one would expect, iE5 effectively hindered both minus end accumulation and depolymerization by KIF2A (Fig. 1, I–K). These results suggest that KIF2A might interact with the exposed surface of the α-tubulins that iE5 binds to, suggesting a new model for microtubule minus end recognition by KIF2A.

Having established a robust minus-end preference for KIF2A using stabilized microtubules, we next investigated if KIF2A also acts asymmetrically on dynamic microtubules growing from immobilized GMPCPP microtubule seeds in the presence of soluble GTP-tubulin (Fig. 2 A and Video 2). Dynamic microtubule extensions could be distinguished from seeds by including a higher ratio of fluorescent tubulin in the seed (Fig. 2 B). In the absence of mScarlet-KIF2A, the faster-growing microtubule plus ends could easily be distinguished from the more slowly growing minus ends (Fig. 2 B; no KIF2A). Adding mScarlet-KIF2A selectively suppressed minus-end growth already at low concentrations (Fig. 2 B; 2 nM, 5 nM KIF2A) and caused selective minus-end depolymerization of the GMPCPP seed at higher concentrations. At the same time, plus-end growth could be observed up to 100 nM of KIF2A (Fig. 2 B; 20 nM, 100 nM KIF2A) without affecting the plus-end growth speed (Fig. 2 C, top left). The plus-end growth speed was also unaffected by MCAK (Fig. 2 C, top right), in agreement with previous MCAK studies (Gardner et al., 2011; Montenegro Gouveia et al., 2010). At higher concentrations, KIF2A also accumulates along the microtubule lattice, triggering more frequently the transition from plus-end growth to shrinkage, thus increasing the plus-end catastrophe frequency (Fig. 2, B and C, bottom left).

The fraction of time spent growing is a convenient single parameter to capture differences in the dynamic state of microtubule plus and minus ends. For KIF2A, it highlights a strong asymmetry between plus and minus end behavior across all KIF2A concentrations studied (Fig. 2 D). In contrast, in the presence of MCAK, we observed only a very mild asymmetry for the fraction of time plus and minus ends spent growing (Fig. 2 D). These results establish that KIF2A also has a strong minus-end preference for dynamic microtubules and allows plus ends to grow up to quite high KIF2A concentration, which is characteristically distinct from its paralog MCAK.

Having observed that KIF2A preferentially accumulates at the minus ends of stabilized microtubules in the enzymatically inactive ADP-bound state (Fig. 1, F and G), we wondered what effect ADP-bound KIF2A has on dynamic microtubule minus ends. We found that ADP-KIF2A remarkably accumulated at growing minus ends, but not at plus ends, in contrast to MCAK, which did not visibly accumulate at any dynamic end (Fig. 2, E and F). Minus-end tracking by ADP-KIF2A slowed down, but did not completely block minus-end growth (Fig. 2, F and G), an effect not observed with MCAK (Fig. 2 G). This shows that the mechanism by which KIF2A recognizes minus ends in the ADP state partially interferes with tubulin addition to the growing minus end, but that complete suppression of growth requires ATP-dependent enzymatic kinesin activity.

Next, we wondered whether KIF2A can depolymerize the minus ends of microtubules that are nucleated by γTuRC, which typically nucleates microtubules in cells. We purified and immobilized human biotinylated and mBFP-tagged γTuRC on a functionalized glass surface via biotin–NeutrAvidin links, and observed by TIRF microscopy γTuRC-mediated microtubule nucleation in the presence of tubulin, as described previously (Consolati et al., 2020; Fig. 4 A). In the absence of mScarlet-KIF2A, microtubules nucleated with only the plus end growing and the minus end being tethered to an immobilized γTuRC (Fig. 4, B and C; 0 nM KIF2A; Video 4). Notably, after nucleation microtubules remained tethered to γTuRC throughout the entire duration of the experiment (Fig. S3 A), indicating that γTuRC stably caps minus ends after microtubule nucleation, in agreement with previous reports (Berman et al., 2023; Consolati et al., 2020; Rai et al., 2022,Preprint; Thawani et al., 2020; Wieczorek et al., 2021).

Next, we added mScarlet-KIF2A at different concentrations to the γTuRC nucleation assay. Microtubules still nucleated from γTuRC, but nucleation became less efficient at higher KIF2A concentrations (Fig. 4 B and Video 4). Quantifying the number of nucleated microtubules over time revealed that the nucleation rate (slopes in Fig. 4 D) decreased with increasing concentrations of mScarlet-KIF2A but to a considerably lesser extent than with mCherry-MCAK (Fig. 4 E), which was reported previously to inhibit γTuRC-mediated microtubule nucleation (Thawani et al., 2020). As observed for microtubules grown from GMPCPP-seeds (Fig. 2 C), KIF2A at higher concentrations increased the catastrophe frequency of the plus ends of γTuRC-nucleated microtubules (Fig. 4, C and F) without affecting their growth speed (Fig. 4 G). These results indicate that plus-end destabilization of the nascent microtubule nucleus forming on γTuRC negatively influences microtubule nucleation. Nevertheless, γTuRC-mediated nucleation tolerates much higher concentrations of KIF2A than MCAK, likely due to KIF2A’s end selectivity.

Remarkably, the large majority of the inherently more KIF2A-sensitive microtubule minus ends were now completely protected from the depolymerizing activity of KIF2A by the γTuRC cap (Fig. 4 C). Despite the presence of KIF2A, almost all microtubule minus ends remained protected by γTuRC for the entire duration of the experiment, as demonstrated by their minus ends remaining tethered to the functionalized glass surface (Fig. 4, C and H). This is in contrast to uncapped minus ends of dynamic microtubules (Figs. 2 and 3) or of occasionally spontaneously nucleated microtubules in the γTuRC-nucleation assay (Fig. S3 B), where the minus ends are depolymerized by KIF2A (Fig. 4 I). Nevertheless, we still sporadically observed microtubules detaching from γTuRC in the presence of KIF2A (mScarlet-KIF2A: Fig. 4 H; untagged KIF2A: Fig. S3 C). These microtubules first nucleated from γTuRC and remained minus-end capped for a period of time until KIF2A apparently destabilized the connection with the γTuRC and drove slow minus-end depolymerization (Fig. 4 J and Video 5). This was accompanied by minus end accumulation of KIF2A and resulted in treadmilling of the majority of γTuRC-released microtubules (Fig. 4 J and Video 5), as observed for seed-grown microtubules (Fig. 3 B). In the absence of KIF2A or in the presence of mCherry-MCAK, we did not observe a single released microtubule (Fig. 4 H and Fig. S3 C), demonstrating that although microtubule release from γTuRC is a rare event, it can be selectively triggered by KIF2A.

This raised the question of whether microtubule severases that have been reported to promote poleward microtubule flux in the spindle during mitosis (Guerreiro et al., 2021; Huang et al., 2021; Zhang et al., 2007) can promote the generation of uncapped microtubules and therefore KIF2A-accessible minus ends of γTuRC-nucleated microtubules. To be able to test this, we purified human spastin and fluorescently labeled mGFP-spastin (Fig. S1, F and G) and verified their microtubule severing activity by adding them to glass-surface immobilized GMPCPP-stabilized microtubules. Although mGFP-spastin was a little less efficient than the unlabeled severase, both caused microtubule disintegration within minutes (Fig. S4, A and B, top panel), as it was observed for related severases (McNally et al., 1996; Roll-Mecak and Vale, 2005; Vemu et al., 2018). However, the vast majority of γTuRC-nucleated microtubules did not show any severing events in the presence of spastin (∼99%, Fig. 5 A). This agrees with previous reports of microtubule severing being inefficient in the presence of free tubulin (Bailey et al., 2015; Jiang et al., 2017; Kuo et al., 2019; Vemu et al., 2018).

When we added mGFP-spastin and mScarlet-KIF2A in combination with stabilized microtubules, they were now severed considerably faster than in the presence of spastin alone (Fig. S4 B, bottom panel). Measuring the fluorescence intensity along the initial microtubule contour over time revealed that when only GFP-spastin was present, its binding initially increased as the microtubule started to become severed and then decreased as the microtubule disappeared (Fig. S4 C). In the additional presence of KIF2A, this process exhibited a similar pattern but at an accelerated rate (Fig. S4 C), suggesting that KIF2A supports severing by spastin and that both act in synergy. Indeed, KIF2A’s fluorescence intensity peaked right after the peak of the mGFP-spastin intensity (Fig. S4 C).

When we added both spastin and KIF2A to the γTuRC nucleation assay, we also observed an increased number of severing events along the lattice of γTuRC-nucleated microtubules compared with spastin being present alone (yellow asterisks, Fig. 5 B i and ii; severing observed in ∼7% of microtubules). After severing, the newly formed plus ends displayed dynamic instability behavior, switching between growth and fast shrinkage episodes (Fig. 5 B and Video 6), whereas the newly formed minus ends were slowly depolymerized by KIF2A (Fig. 5 B), resulting in treadmilling of the newly generated free microtubule (Fig. 5 B ii). Remarkably, in the combined presence of mScarlet-KIF2A and spastin, we observed also substantially more frequent detachments of microtubules from surface-tethered γTuRC, also causing them to treadmill (Fig. 5 C and Video 7). Microtubule release from immobilized γTuRC increased from ∼1% and ∼2% in the presence of spastin or KIF2A alone, respectively, to almost 40% when spastin and KIF2A were both present (Fig. 5 D). This release, induced by severing at or near the γTuRC/microtubule interface, is approximately five times higher than the percentage of microtubules severed at any other point along the lattice (Fig. 5 E), despite the much larger number of potential severing sites along the entire microtubule length compared with the single γTuRC cap at the minus end. Because we did not observe mGFP-spastin directly binding to surface-immobilized γTuRC (Fig. S4 D), we conclude that the γTuRC/microtubule interface is orders of magnitudes more sensitive to the combined action of KIF2A and spastin than an average microtubule lattice site, favoring γTuRC uncapping over lattice severing. Taken together, our results demonstrate that spastin and KIF2A act in synergy to trigger microtubule minus-end release from γTuRC and subsequent treadmilling, consistent with the proposed function of KIF2A and severases at spindle poles.

We discovered here that the kinesin-13 KIF2A is a microtubule minus-end depolymerase that can induce flux-like microtubule treadmilling. Although KIF2A acts mildly on plus ends, it has a strong preference for minus-end destabilization, establishing it as an asymmetric microtubule depolymerase. This property distinguishes KIF2A from its paralog MCAK which destabilizes both microtubule ends similarly. Moreover, we found that the presence of a γTuRC cap at minus ends inhibits minus-end depolymerization by both depolymerases. To induce treadmilling of γTuRC-nucleated microtubules, active uncapping of minus ends is required, which can be achieved by the combined activities of KIF2A and the severase spastin. Taken together, we show here that a γTuRC/KIF2A/severase module can provide the activities required for both microtubule nucleation and controlled microtubule minus-end depolymerization at spindle poles, in agreement with the requirements for the generation of microtubule flux in mitotic and meiotic spindles.

Minus-end recognition by KIF2A appears to be different from the generic microtubule end recognition mechanism of kinesin-13s, in which the motor domain in its ATP state binds to and stabilizes a curved microtubule protofilament (Asenjo et al., 2013; Benoit et al., 2018; Trofimova et al., 2018; Wang et al., 2017). We found that KIF2A binds preferentially to minus ends also in the presence of ADP, showing that enzymatic activity is not required for minus end recognition. Because ADP-KIF2A also slows down microtubule growth and because minus-end capping by iE5 and γTuRC prevents KIF2A minus-end accumulation, it is tempting to speculate that KIF2A’s minus-end recognition may involve binding to at least part of the exposed α-tubulin at microtubule minus ends. In the future, biochemical and structural studies will be required to identify the mechanism underlying KIF2A’s remarkable minus-end binding preference.

In interphase cells, individual microtubules have been observed to adopt a treadmilling state in which net plus-end growth is balanced by net minus-end depolymerization, which has been associated with loss of minus-end stabilizers (Goodwin and Vale, 2010; Rodionov and Borisy, 1997; Shaw et al., 2003; Waterman-Storer and Salmon, 1997). Remarkably, the kinesin-13 KLP10A, which in mitosis seems to share some functional homology with KIF2A (Laycock et al., 2006; Rath et al., 2009; Rogers et al., 2004), was observed to localize to shrinking minus ends and to induce treadmilling of microtubules in Drosophila cells when the minus-end stabilizing protein patronin was knocked down (Goodwin and Vale, 2010).

Early in vitro work with purified proteins investigated the possibility of treadmilling being an intrinsic property of microtubules, but later it became clear that the intrinsic dynamic behavior of microtubules is dominated by dynamic instability at both ends (Grego et al., 2001; Margolis and Wilson, 1978; Mitchison and Kirschner, 1984). We find here that purified KIF2A can trigger microtubule treadmilling in vitro, resulting as the consequence of the combination of two of its properties. First, it destabilizes minus ends more strongly than plus ends, so that in a certain concentration regime, plus ends grow dynamically whereas minus ends shrink; and second, in contrast to the intrinsically fast depolymerization after typical catastrophe events, KIF2A induces mainly slow minus-end depolymerization, while remaining end accumulated at the shrinking minus ends, which can then be balanced by (net) plus-end growth. The mechanism of this slow-down of minus-end depolymerization is unknown but might be related to the formation of higher-order structures formed by curved, depolymerizing protofilaments and KIF2A, as observed in electron microscopy with different members of the kinesin-13 family (Moores et al., 2006; Tan et al., 2006, 2008; Zhang et al., 2013).

A recent study reported the in vitro reconstitution of microtubule treadmilling with a combination of several vertebrate microtubule-binding proteins (Arpağ et al., 2020), evidence of which can also be seen in earlier in vitro work with the corresponding Drosophila orthologs (Moriwaki and Goshima, 2016). Because the symmetric depolymerase MCAK was used in this recent work, additional activities were required to generate an asymmetry in the control of microtubule dynamics. This was achieved by including the plus-end specific microtubule polymerase XMAP215 and the anti-catastrophe factor CLASP2, resulting in microtubule treadmilling with growing plus ends and shrinking minus ends. These additional factors were not required in our reconstitution due to the inherently asymmetric activity of KIF2A with respect to the two microtubule ends, simplifying the minimal requirements for microtubule treadmilling. In cells, KIF2A’s intrinsic minus-end preference can be enhanced by proteins that promote its minus-end accumulation (Guan et al., 2023).

Our study revealed that γTuRC and kinesin-13s compete in two different ways. First, we showed that microtubule nucleation by γTuRC is inhibited in a dose-dependent manner by both KIF2A and MCAK, however, to a much lesser extent by KIF2A. This suggests that kinesin-13 depolymerase activity destabilizes nascent microtubule plus-ends beginning to grow on a γTuRC template and that this destabilization is weaker for KIF2A due to its end selectivity. Second, once the microtubule has nucleated, γTuRC caps the minus end and now inhibits minus-end depolymerization by kinesin-13s, an activity reminiscent of the minus-end binding protein patronin/CAMSAP that however binds minus ends very differently from γTuRC (Atherton et al., 2017; Goodwin and Vale, 2010).

Our finding that spastin and KIF2A in combination, but not alone, can efficiently induce microtubule severing in the presence of free tubulin, thereby leading to KIF2A-mediated minus-end depolymerization, has several interesting implications. First, severase-induced lattice damage appears to provide an entry point for KIF2A to promote lattice destabilization and/or interfere with lattice repair by soluble tubulin (Schaedel et al., 2015; Vemu et al., 2018), thereby promoting complete severing. KIF2A’s severase-supporting effect may explain why in cytoplasmic extract severases appear much more active than purified severases in the presence of free tubulin in vitro (Vale, 1991; Vemu et al., 2018). α-Tubulins likely become exposed when spastin removes tubulins from the lattice where KIF2A might be able to act and increase the size of the damage site. Our results also highlight that severase activity in the presence of free tubulin can be promoted by mechanisms beyond the physical recruitment of severases to the microtubule, as shown for ASPM and katanin (Jiang et al., 2017).

Moreover, we observed that also γTuRC uncapping was promoted by the combined action of KIF2A and spastin, and remarkably was more frequent than microtubule severing, although (1) γTuRC prevents microtubule minus-end binding of KIF2A, (2) γTuRC and spastin do not directly interact, and (3) many more potential severing sites exist along the microtubule lattice than at the γTuRC/microtubule minus end interface. This implies that the γTuRC/minus end connection is much more sensitive to the combined action of spastin and KIF2A than the rest of the microtubule lattice. Although the γTuRC cap at minus ends is very stable in the absence of other proteins, it appears to become a structurally weak point in the presence of KIF2A and a severase, allowing preferential active minus-end uncapping, possibly not only to enable microtubule treadmilling but also to recycle γTuRC for new rounds of microtubule nucleation. γTuRC-uncapping by KIF2A and spastin is functionally distinct from recently reported uncapping by CAMSAPs that remove γTuRC by inducing slow minus-end polymerization, which is possibly more adapted to the function of microtubules in neurons (Rai et al., 2022 Preprint).

In conclusion, we characterized in vitro a minimal protein module that can be found at the poles of mitotic and meiotic spindles during cell division and that can control apparently antagonistic activities at this location. On one hand, microtubules need to be nucleated, which results in the stable capping of their minus ends by γTuRC. On the other hand, minus ends also need to be uncapped and slowly shrink for microtubule flux in the spindle to occur. γTuRC, KIF2A, and a severase in combination can provide this control of microtubule minus-end dynamics as required at spindle poles. A major element of this control is the asymmetric action of the kinesin-13 KIF2A, adding to the list of other plus or minus-end selective microtubule-associated proteins that control the functionally important asymmetric dynamic properties of the two microtubule ends in cells.

Human KIF2A (NCBI reference sequence: NP_004511.2; codon-optimized for insect cell expression by GeneArt) and MCAK (NCBI reference sequence: NP_006836.2) constructs were cloned similarly to previously published constructs (Roostalu et al., 2018). Briefly, coding sequences were amplified by PCR and cloned into a pFastBac expression vector for expression in Sf9 or Sf21 cells (Thermo Fisher Scientific), including a StrepTagII-coding sequence at the N-terminal of the construct, to generate the following constructs: StrepTagII-KIF2A, StrepTagII-mScarlet-G6S-KIF2A, and StrepTagII-mCherry-G5A-MCAK. Constructs included a region coding for a TEV protease recognition site between the StrepTagII sequence and the desired construct for cleavage of the tag. The monomeric red fluorescent protein variants mScarlet (Bindels et al., 2017) and mCherry (Shaner et al., 2004) were separated from the N-terminus of KIF2A and MCAK by flexible linker regions. The sequence coding for MCAK included an extra isoleucine residue after the initial methionine as a consequence of the cloning process.

The human spastin and mGFP-spastin constructs were cloned using the sequence corresponding to isoform 3 (UniProtKB reference sequence Q9UBP0-3; OriGene), a shortened variant derived from the use of an alternative start residue (Met-87), as identified and used in previous works (Claudiani et al., 2005; Roll-Mecak and Vale, 2005). The spastin sequence was amplified by PCR and cloned into a pETMZ and a pETMZ-mGFP vector for expression in Escherichia coli, generating the constructs His6-Ztag-TEV-spastin and His6-Ztag-TEV-mGFP-spastin. The sequence coding for spastin included an extra isoleucine residue after the initial methionine as a consequence of the cloning process.

The plasmids for bacterial expression of the αRep iE5 and the tandem repeat Design-Ankirin-Repeat-Protein (DARPin)— (D1)2—are described elsewhere (Campanacci et al., 2019; Pecqueur et al., 2012).

Tubulin was prepared from pig brains as previously described (Castoldi and Popov, 2003). Tubulin was further purified by recycling and some tubulin was labeled with AlexaFluor647-N-hydroxysuccinimide ester (NHS; Sigma-Aldrich), Atto647-NHS (Sigma-Aldrich), Cy5-NHS (Lumiprobe), or biotin-NHS (Thermo Fisher Scientific), as described (Consolati et al., 2022). Final concentrations and fluorescent labeling ratios were determined by UV absorption using NanoDrop One/OneC (Thermo Fisher Scientific; extinction coefficient for tubulin, 115,000 M−1 cm−1; molecular weight, 110 kD) after final ultracentrifugation, before aliquoting and snap-freezing. Aliquoted tubulin was then placed in liquid nitrogen.

KIF2A, mScarlet-KIF2A, and mCherry-MCAK constructs were expressed using Sf21 cells (Thermo Fisher Scientific) according to the manufacturer’s protocols. Cell pellets from 500 ml of culture were snap-frozen and stored at −80°C. Thawed pellets of cells expressing KIF2A and mCherry-MCAK constructs were resuspended in ice-cold KIF2A lysis buffer (50 mM Na-phosphate, 300 mM KCl, 1 mM MgCl2, 1 mM ethylene glycol bis(2-aminoethyl ether)tetraacetic acid [EGTA], 5 mM β-mercaptoethanol [β-ME], and 0.5 mM adenosine triphosphate [ATP], pH 7.5), supplemented with DNaseI (Sigma-Aldrich) and protease inhibitors (cOmplete EDTA-free; Roche). The lysate was homogenized in a glass douncer on ice (40 strokes) and subsequently clarified by centrifugation at 50,000 rpm in a Ti70 rotor (Beckman Coulter) at 4°C for 45 min. The supernatant was then run over a 5 ml StrepTrap HP or XT column (Cytiva) at 0.5 ml min−1 using a peristaltic pump at 4°C. Bound protein was washed with 10 ml of KIF2A lysis buffer supplemented with ATP to a total of 5 mM at 0.5 ml min−1 and then eluted in 0.5 ml fractions using KIF2A lysis buffer