Introduction

Microtubule is a major type of cytoskeleton in eukaryotic cells and supports many cellular processes, such as cell division, polarization and migration (Alberts, 2015). However, the properties of microtubules vary a lot in different cellular processes and structures (Janke and Magiera, 2020; Kapitein and Hoogenraad, 2015; Wittmann et al., 2001). The plasticity implies that cellular microtubules are under complex and accurate regulations. Because microtubule-associated proteins (MAPs) are important components in the regulatory networks of microtubules, their functions and working mechanisms are key to understand cellular regulations on microtubules.

Not only do kinesin superfamily members move cargoes along microtubules, they also act as regulators of microtubule dynamics, for example, the depolymerizing kinesins (e.g. the kinesin-8 and kinesin-13 family members). Kinesin-13 was initially identified in studying spindle functions and neurons (Aizawa et al., 1992; Noda et al., 1995; Walczak et al., 1996; Wordeman and Mitchison, 1995). Later, the functions of kinesin-13 were revealed in more cellular processes, including spindle assembly (Walczak et al., 2002), chromosome segregation (Kline-Smith et al., 2004), directional migration (Zong et al., 2021), ciliary length control (Piao et al., 2009; Vasudevan et al., 2015), and neuronal polarization (Ghosh-Roy et al., 2012; Puri et al., 2021). The loss of kinesin-13 often alters structural and dynamic properties of cellular microtubules, implying that the kinesin-13 class members play regulatory roles. In addition, the subcellular localizations of kinesin-13, such as growing microtubule ends, centrosome and centromere, are active regulatory sites of microtubule dynamics in cells, further suggesting the functions of the kinesin-13 class members in controlling the length, mass and polarity of microtubules in cells.

As a representative model for the kinesin-13 family, MCAK was shown to be a microtubule depolymerase and a catastrophe factor (Desai et al., 1999; Gardner et al., 2011; Hunter et al., 2003). Biochemical and structural analysis suggests that MCAK could deconstruct the protective end structure of dynamic microtubules, and it does so by bending and then breaking off the underlying substrate (i.e. terminal tubulin dimers or protofilaments) (Moores and Milligan, 2006). It was proposed that MCAK targets microtubule ends by rapid lattice diffusion or binding to the end-binding proteins (EBs) (Gouveia et al., 2010; Helenius et al., 2006; Honnappa et al., 2009). Furthermore, the direct binding of MCAK to growing microtubule ends was also observed (Gouveia et al., 2010), implying the presence of specific end-binding sites. Previous studies on the structures of MCAK, propose that MCAK targets to microtubule ends by recognizing the curved protofilaments (Asenjo et al., 2013; Benoit et al., 2018; Tan et al., 2008; Trofimova et al., 2018; Wang et al., 2017). This was mostly based on the findings on stabilized microtubules, short protofilaments or tubulin dimers. It still remains elusive how MCAK binds to dynamic microtubule ends.

In the present study, we characterize the direct binding of MCAK to growing microtubule ends using single-molecule imaging. The key finding is that MCAK binds to the entire GTP cap, including both the EB cap, where GDP‧Pi-tubulin dimers are thought to gather, and the distalmost cap, where XMAP215 binds to curved protofilament (mostly in GTP state). The high-affinity binding of MCAK to the EB cap was intriguing. Further analysis shows that MCAK strongly binds to GTPγS microtubules, thought to be the analogue of GDP‧Pi-tubulins. Therefore, we think that MCAK could recognize the nucleotide-dependent feature at dynamic microtubule ends. Moreover, the mutant studies show that the binding-preference of MCAK for GDP‧Pi-tubulins facilitates the end-binding and in turn the function of MCAK. Finally, we find that despite having partially overlapped binding regions, MCAK and XMAP215 could precisely specify their regulatory function at dynamic microtubule end, thereby providing novel insights in understanding how microtubule-associated proteins functionally orchestrate at growing microtubule ends.

Results

MCAK shows a binding preference for growing microtubule ends

We set off by measuring the single-molecule binding kinetics of MCAK at growing microtubule ends using the in vitro microtubule dynamics assay. This experiment was performed in BRB80 supplemented with 50 mM KCl and 1 mM ATP, providing a nearly physiological ion strength. In this condition, the individual binding events of MCAK can be observed at both the end and the lattice (Fig. 1A). By measuring the fluorescence intensity of these binding events, we confirmed that they reflected individual MCAK dimers (Fig. s1). Using an in-house software (Song et al., 2020), we determined the binding position of individual MCAK molecules at subpixel resolution by fitting a Gaussian function to their intensity profile along the long axis of microtubules (Fig. 1B). Using the position of the plus-end as the spatial reference, we overlaid the end-binding of 152 MCAK molecules (Fig. 1C). The distribution showed that the binding regions of MCAK had a length of 162 nm (FWHM, the Full Width at Half Maximum), about 20 layers of tubulin dimers. We defined the binding events of MCAK within this region as “end binding” and those on the lattice as “lattice binding” (Fig. 1A).

Figure 1.

To understand the binding kinetics, we first compared the apparent association rates of MCAK on the plus-end (k_on-P) and lattice (k_on-L) in the presence of ATP. The ratio (R_E/L) of k_on-P ((22. 6 ± 1.4) ×10^-6 s^-1 nM^-1, per tubulin dimer, hereinafter) and k_on-L ((2.8 ± 0.3) ×10^-6 s^-1 nM^-1) was 13 ± 1 (n=66 microtubules from 3 assays) (Fig. 1 D-E). We then measured the dwell time, which reflects the dissociation rate (k_off) (Fig. s1). The mean dwell time of MCAK on the lattice was shorter than that at the plus-end (plus end: 0.75 ± 0.02 s, n=966 events from 3 assays; 0.62 ± 0.02 s, n=702 events from 3 assays, p<0.001) (Fig. 1F). Using these kinetic parameters, we calculated the dissociation constant (K_d) of MCAK for the plus-end (K_d-P =69 μM) and lattice (K_d-L=1057 μM), respectively. Therefore, MCAK shows a clear end-binding preference. We also observed single-molecule binding events of MCAK to the minus-end (Fig. 1A). Given the higher physiological relevance, we focused on the plus-end binding events (referred to as “end-binding” hereinafter) in the following analysis.

To understand the influence of the nucleotide state, we measured the end-binding kinetics of MCAK‧AMPPNP, MCAK‧ADP or MCAK‧APO (the nucleotide-free state). First, the k_on-P and k_on-L of MCAK‧AMPPNP were both reduced in comparison to that of MCAK‧ATP (Fig. 1 D-E), resulting in an end-binding preference (R_E/L=11 ± 2, n=29 microtubules from 2 assays, p>1) that is similar to that of MCAK‧ATP. Mean-while, the end-binding of MCAK‧AMPPNP showed a longer dwell time (3.32 ± 0.24 s, n=231 events from 2 assays, p<0.001), indicating a lower k_off (Fig. 1F and Fig. s1). As a result of the proportionally reduced k_on-P and k_off, the end-binding affinity (K_d-P=61 μM) of MCAK‧AMPPNP was similar to that of MCAK‧ATP. Second, MCAK‧ADP showed a nearly unchanged k_on-P and an increased k_on-L, leading to a smaller R_E/L (2 ± 0.3, n=21 microtubules from 2 assays, p<0.001) and suggest a reduced, but not the absence of, end-binding preference (Fig. 1 D-E). Meanwhile, it showed a significantly shorter end-binding dwell time (0.46 ± 0.02 s, n=327 events from 2 assays, p<0.001) than MCAK‧ATP (Fig. 1 F). Therefore, MCAK‧ADP has a lower end-binding affinity (K_d-P=163 μM). Third, APO‧MCAK showed an increase in k_on-P and an even greater increase in k_on-L, thereby having a smaller R_E/L (2 ± 0.2, n=36 microtubules from 3 assays, p<0.001) (Fig. 1 D-E). It also showed a significantly longer end-binding dwell time (1.89 ± 0.23 s, n=71 events from 3 assays, p<0.001) (Fig. 1 F), suggesting a lower k_off (Fig. s1). Therefore, in comparison to MCAK‧ATP, MCAK‧APO possesses a higher end-binding affinity (K_d-P=1.8 μM) but partially loses the end-binding preference. Based on these measurements, we conclude that the end-binding preference and affinity of MCAK depend on its nucleotide state (see Discussion).

Having measured the end-binding kinetics, we were able to calculate whether the binding of MCAK on dynamic microtubule ends is due to direct end-binding (3D diffusion) or 1D lattice-diffusion (Gouveia et al., 2010; Helenius et al., 2006). Note that these two mechanisms are not mutually exclusive. Using our dataset, we measured that the lattice-diffusion coefficient of MCAK (D) on the lattice of dynamic microtubules was 0.023 ± 0.001 μm²‧s^-1 (Fig. 1G), close to the previously reported value (Cooper et al., 2010). Because the dwell time of the lattice-binding events was ∼0.6 s, the average length scanned by a MCAK molecule via 1D lattice-diffusion was ∼160 nm. Note that this is close to the length the GTP cap. Using a previously reported Fick’s first equation of diffusion (Helenius et al., 2006), we calculated the flux of MCAK reaching growing ends via 1D diffusion was 0.004 s^-1. This was significantly lower than the arrival rate of MCAK onto the plus end observed in our experiments (N_p= k_on‧C =0.067 s^-1. C: MCAK concentration). Therefore, in our experimental condition, 1D lattice-diffusion makes a minor contribution to the end-binding of MCAK (∼ 6%), and the end-binding of MCAK can be thought of as the direct interaction between diffusive MCAK and the end structure.

MCAK binds to the entire GTP cap

Having recorded and analyzed the end-binding events of MCAK, we wondered how MCAK could recognize the end of dynamic microtubules. Previous studies, mostly based on stabilized microtubules, proposed the idea that MCAK binds to microtubule ends by recognizing curved protofilaments (Asenjo et al., 2013; Benoit et al., 2018; Trofimova et al., 2018). This model predicts that the binding site of MCAK is located to the distalmost tip of growing microtubule ends, where the XMAP215 family members are thought to bind (i.e. the distalmost cap) (Ayaz et al., 2012; Brouhard et al., 2008). We tested this prediction by measuring the end-binding regions of MCAK, EB1 and XMAP215 (Fig. 2A). To our surprise, the binding region of MCAK was longer than that of XMAP215. It extended proximally towards the lattice, and covered the binding region of EB1 (i.e. the EB cap) (Maurer et al., 2012). To further understand if MCAK binds to the distalmost cap and the EB cap with different kinetics, we compared the dwell time of the distally located MCAK to that of the proximally distributed ones (Fig. 2B). Note that we defined the MCAK molecules localizing to the left of the FWHM of XMAP215 binding region as the proximally distributed, while those localizing to the right of the FWHM of EB1 binding region was distally distributed. No significant difference was found (τ_distal=0.8 ± 0.5, n=28 binding events; τ_proximal=0.7±0.6, n=45 binding events; p>0.3, the two-tailed Mann-Whitney U test with Bonferroni correction). Note that in all experiments, the morphologies of growing microtubule ends were comparable (Fig. 2C). Based on these results, we conclude that MCAK recognizes the entire GTP-cap. This result is intriguing as it suggests that in addition to the curved protofilaments, there are additional end-binding sites for MCAK. Moreover, this also provides an independent measurement for the size of the GTP cap at growing microtubule ends, which was previously determined using the tubulin dilution experiments or by measuring the binding region of EB1 (Duellberg et al., 2016).

Figure 2.

MCAK strongly binds to GTPγS microtubules in a nucleotide-independent manner

How could MCAK bind to the EB cap? The tubulin dimers in the EB cap are thought to be different from those in the distalmost cap (Maurer et al., 2011). For example, the distalmost tubulin dimers (mostly GTP-tubulin) often lack lateral constrains (Vitre et al., 2008) and adopt a bent shape (Hoeoeg et al., 2011; McIntosh et al., 2018), while those in the EB cap (mostly GDP‧Pi-tubulin) have formed at least some lateral contacts and had a more straight conformation (Guesdon et al., 2016; Maurer et al., 2012). To address this issue, we first examined that to what extent, MCAK may adopt a similar end-binding mechanism as EB1.

We tested the contribution of GDP‧Pi-tubulins, which underlies the end-binding preference of EBs (Maurer et al., 2011; Maurer et al., 2012). Here, we compared the binding affinity of MCAK‧AMPPNP on GTPγS, GMPCPP and GDP microtubules (Fig. 3A), which mimics tubulin dimers in different GTP hydrolysis states. Because MCAK in the concentration required for a full-lattice decoration would rapidly disassemble microtubules, we cannot directly compare the total binding of MCAK on microtubules as what was previously done for the EBs (Maurer et al., 2011; Maurer et al., 2012). Therefore, we took an alternative approach by recording the single-molecule binding of MCAK‧AMPPNP (1 nM) on microtubules and measuring fluorescence intensity summation over a period of time (1000 frames, 0.3 s per frame) (Fig. s2). Strikingly, we found that MCAK‧AMPPNP bound strongly to GTPγS microtubules (Fig. 3 A-C). Quantitative analysis revealed a significantly stronger binding on GTPγS microtubules than on GMPCPP or GDP microtubules (Fig. 3 B-C). Because the tubulin dimers in GTPγS microtubules were found to be the analogue of GDP‧Pi-tubulins (Maurer et al., 2011), this result suggests that MCAK‧ATP has a binding preference for GDP‧Pi-tubulins.

Figure 3.

Because the end-binding affinity and preference of MCAK depends on the nucleotide state, we wondered if the preference for GTPγS microtubules is similar. Therefore, we performed the same experiments using MCAK‧AMPPNP, MCAK‧ATP, MCAK‧ADP and MCAK‧APO. Although the MCAK variants bound to microtubules with different affinities (MCAK‧APO > MCAK‧AMPPNP > MCAK‧ADP > MCAK‧ATP) (Fig. 3B), they all showed a clear binding preference for GTPγS microtubules over GDP or GMPCPP microtubules (Fig. 3 B-C). These observations suggest that the binding preference to GTPγS microtubules (i.e. GDP‧Pi-tubulin) is, to a large extent, independent on the nucleotide state of MCAK.

MCAK binds to the end region of stabilized GMPCPP microtubules in a nucleotide state-dependent manner

We noted that MCAK showed a nucleotide state-dependence in the binding preference for dynamic microtubule ends, but the binding to GTPγS microtubules is nucleotide independent. Therefore, we wondered if the binding to curved protofilaments at the distalmost cap of dynamic microtubules could account for the nucleotide-state dependence. To address this issue, we measured the binding preference of MCAK at the end of stabilized GMPCPP microtubules (Fig. 4A), where the length and curvature of protofilaments are thought to be similar to those at the end of dynamic microtubules but has no GDP‧Pi-tubulins (Manka and Moores, 2018; McIntosh et al., 2018). We found that the MCAK in ATP or AMPCPP state showed a binding preference for the end of GMPCPP microtubules, while MCAK‧ADP and MCAK‧APO did not (Fig. 4B-C). This is similar to the observation that MCAK‧ADP and MCAK‧APO had a lower binding preference for dynamic microtubule ends, thereby providing an explanation for the nucleotide state-dependence of the end-binding of MCAK.

Figure 4.

The binding preference for GDP‧Pi-tubulins facilitates the end-binding of MCAK

Intuitively, the binding preference for the EB cap, in addition to the previously established preference for the curved protofilaments, would provide more high-affinity sites for MCAK, thereby facilitating the end-binding of MCAK. To tested this idea, we studied two mutants, MCAK^K524A (K524A in the α4 helix) and MCAK^V298S (V298S in the L2 loop). The α4 helix of MCAK interacts with the intra-dimeric interface of tubulin heterodimer (Fig. 5A), and the K524A mutation is expected to disrupt the electrostatic interaction between MCAK and tubulin heterodimer (Patel et al., 2016; Wang et al., 2017). The L2 loop of MCAK has a direct contact with the inter-dimeric interface of tubulin heterodimer (Fig. 5A), and the V298S mutation is expected to disrupt the hydrophobic interaction between the L2 loop and α-tubulin (Wang et al., 2015). We confirmed that both mutants were depolymerization-deficient (Fig. s3), as previously reported (Wang et al., 2017; Wang et al., 2015).

Figure 5.

Compared to MCAK, MCAK^K524A showed a significantly reduced binding intensity on microtubules. Meanwhile, it still showed a binding preference for GTPγS microtubules (Fig. 5B-C), but the ratios between the binding intensity on GTPγS microtubules and that on GMPCPP or GDP microtubules were both reduced (Fig. 5D). This suggests that MCAK^K524A had a lower preference for GTPγS microtubules. On the contrary, we found that MCAK^K524A showed a higher binding to the end than to the lattice of GMPCPP microtubules (Fig. 5 E-F), suggesting that the binding preference for curved protofilament is maintained. Therefore, MCAK^K524A could be a candidate variant to test how the preference for GDP‧Pi-tubulin contributes to the end-binding preference of MCAK. We found that MCAK^K524A could bind to dynamic microtubule ends, but its k_on-P and R_E/L were both reduced (Fig. 5 G-H). Meanwhile, the end-binding dwell time of MCAK^K524A was also reduced (Fig. 5I), suggesting a higher off-rate. These results are consistent with the idea that the higher binding affinity to GDP‧Pi-tubulin directly contributes to the end-binding preference of MCAK at growing microtubule ends.

By comparison, the binding preferences of MCAK^V298S for GTPγS microtubules and the end of GMPCPP microtubules were both reduced (Fig. 5 B-F), implying the reduced preferences for both the EB cap and the distalmost cap. In the dynamics assay, the k_on-P, R_E/L and dwell time of MCAK^V298S were decreased in comparison to those of MCAK (Fig. 5 H-I). We noted that the k_on-P, R_E/L and dwell time of MCAK^V298S were similar to those of MCAK^K524A (two-tailed Mann-Whitney U test with Bonferroni correction, p>0.05), suggesting that the binding to the EB cap makes a major importance for the end-binding preference of MCAK at dynamic microtubule ends.

Functional specification of MCAK and XMAP215

Having revealed the novel binding sites of MCAK in dynamic microtubule ends, we wondered how MCAK may work in coordination with other microtubule plus tip-binding proteins, for example EB1 and XMAP215. It has been shown that EB1 interacts with and recruits MCAK to growing microtubule ends (Gouveia et al., 2010; Honnappa et al., 2009; Lee et al., 2008), so it is straightforward that EB1 would facilitate the function of MCAK as a catastrophe factor. On the contrary, MCAK and XMAP215 were shown to antagonistically regulate the assembly of microtubules, and it was thought that these two molecules may compete directly for the growing ends (Tournebize et al., 2000). Here, we showed that the binding regions of MCAK and XMAP215 in growing microtubule ends are partially overlapping. In addition, previously resolved structural models suggest that MCAK and XMAP215 likely bind to similar regions on tubulin dimers (Fig. s4). Therefore, these two molecules could indeed compete for the binding sites in the distalmost cap. However, it is not yet clear to what extent, the competition from XMAP215 would antagonize the function of MCAK, especially given the recent finding that XMAP215 itself could promote the catastrophe of microtubules (Farmer et al., 2021). Therefore, it is intriguing what would be the collective effect of MCAK and XMAP215.

To address this issue, we studied how they together regulate microtubule grow rate and catastrophe frequency. Based on our pilot experiments, we chose 20 nM for MCAK and 50 nM for XMAP215 as the representative concentrations. In the presence of 20 nM MCAK, the dynamic microtubules could grow but showed a significantly increased catastrophe frequency and shorter lifetime (Fig. 6 A-B). Meanwhile, 50 nM XMAP215 increased growth rate by more than four times (control: 0.8 ± 0.1 μm min^-1, n=97 microtubules from 6 assays; 50 nM XMAP215: 3.7 ± 0.4 μm min^-1, n=50 microtubules from 3 assays). We noted that XMAP215 alone caused a mild shortening in the lifetime of dynamic microtubules (Fig. 6 A-B). This is consistent with the previous report and suggesting the enhanced fluctuations in the end structures (Farmer et al., 2021). We then added both 20 nM MCAK and 50 nM XMAP215. In this case, microtubule grew in a similar rate as if there was no MCAK (3.9 ± 0.4 μm min^-1, n=52 microtubules from 3 assays, p=0.5, the two-tailed Mann-Whitney U test with Bonferroni correction), suggesting that MCAK has no major effect on the function of XMAP215 as a polymerase. Meanwhile, the lifetime of dynamic microtubules became shorter than that with only MCAK or XMAP215 (Fig. 6B). To understand if this reflects the simple summation of the separate contributions of MCAK and XMAP215 or their synergistic effect, we calculated the probability-based catastrophe frequency in all conditions (Fig. 6C) (Gardner et al., 2011).We found that the effects of MCAK and XMAP215 on the probability of microtubule catastrophe were only additive, showing that there was no synergic effect in regulating catastrophe frequency (Fig. 6C).

Figure 6.

Discussion

This study provides mechanistic insights into understanding the end-binding mechanism of MCAK. The previously established view on the end-binding mechanism of MCAK is based on specific recognition of curved protofilaments at the distalmost tip of growing microtubules (Asenjo et al., 2013; Benoit et al., 2018; Tan et al., 2008). The findings in the present study add to the model by showing that MCAK strongly binds to the EB cap, and this binding preference could facilitate the end-binding and in turn the depolymerizing activity of MCAK. We now discuss our main findings and their implications.

MCAK preferentially binds to the GDP‧Pi-tubulin in a nucleotide state-independent manner

Our main finding is that MCAK preferentially binds to the EB cap where the GDP‧Pi-tubulins gather, in addition to the previously known binding preference for the curved protofilaments (Asenjo et al., 2013; Benoit et al., 2018; Tan et al., 2008). This conclusion is based on several lines of evidence. First, MCAK‧ADP or MCAK‧APO shows a significant reduction but not a complete loss in the binding preference for dynamic microtubule ends. However, their binding preference for stabilized microtubule ends is completely absent. The different binding behavior to dynamic and stabilized microtubule ends suggests that other end features, in addition to the curved protofilament, also contribute to the end-binding of MCAK. Second, the localization analysis of MCAK shows that MCAK binds to the entire GTP cap, including the EB cap and the distalmost cap. Third, MCAK shows a binding preference for GTPγS microtubules over GDP or GMPCPP microtubules. Because the tubulin in GTPγS microtubules is considered to be the analogues to the GDP‧Pi-tubulins (Maurer et al., 2011; Maurer et al., 2012), MCAK is expected to have a high affinity to the EB cap. Fourth, the loss of the binding preference of MCAK^K524A for GTPγS microtubules, but not for the ends of GMPCPP microtubules, suggests that the binding preference for GDP‧Pi-tubulins is required for the optimal end-binding affinity and preference.

Implications for the working mechanism of MCAK

We think that our findings add to the current working model of MCAK in four aspects. First, the nucleotide state-independent binding preference for GDP‧Pi-tubulins provides a mechanism for MCAK staying bound to the GTP cap after ATP hydrolysis. For example, in a speculative scenario (Fig. 7A), when one of the motor domains of MCAK is converted to the ADP or APO state after the breakage of the terminal tubulin dimers, it loses the binding affinity to the curved protofilaments (Fig. 7A). However, it still preferentially binds to GDP‧Pi-tubulins, so it has a larger chance binding back to the EB cap than to the lattice and thereby remains within the GTP cap (Fig. 7A). Then, while the ADP is being replaced by a new ATP, this motor domain could diffuse around to explore the working site for the next round of catalysis. In this way, the nucleotide state-independent binding feature of MCAK would facilitate the binding to the dynamic microtubule ends. This idea is consistent with the findings that MCAK^K524A, a mutant showing reduced preference for GDP‧Pi-tubulins, had disrupted end-binding affinity, preference and dwell time.

Figure 7.

Second, the previous studies showed that MCAK has a direct interaction with EB1, and EB1 increases the end-binding rate of MCAK (Gouveia et al., 2010; Honnappa et al., 2009; Lee et al., 2008). Intuitively, there is a gap here that after the arrival of MCAK at the binding region of EB1, how it could reach the high-affinity binding site or working site, previously known as the curved protofilament at the distalmost tip (Asenjo et al., 2013; Benoit et al., 2018; Tan et al., 2008). Our finding fills this gap by showing that MCAK also binds strongly to the GDP‧Pi-tubulins where the EBs bind (Fig. 7B). Moreover, MCAK could explore the entire length of GTP cap (∼160 nm) by 1D lattice diffusion within the binding dwell time. In this way, we argue that MCAK could find the precisely catalyzing site at dynamic microtubule ends by a combination of 3D diffusion (to first reach the GTP cap) and 1D lattice diffusion (short range, within the GTP cap).

Third, a previous study showed that the association on-rate is an important regulated parameter for the functions of MCAK (Cooper et al., 2010). The binding affinity and preference for the EB cap, in addition to the distalmost cap, would intuitively provide more binding sites, thereby further increasing the binding on-rate of MCAK. This idea is also consistent with the finding that MCAK^K524A, a mutant with a reduced preference for GDP‧Pi-tubulins, had a reduced end-binding on-rate.

Fourth, despite having partially overlapped binding regions, MCAK and XMAP215 provide independent controls for catastrophe frequency and growth rate of microtubules, respectively (Fig. 7B). The free combination of these two regulators suggests a straightforward strategy in the regulations on the length and mass of microtubules, i.e., simply by controlling their amount (i.e. expression level) or activity (e.g. via phosphorylation). In this sense, the working scenario of MCAK and XMAP215 provides a molecular paradigm to understand how the regulators of microtubules orchestrate their behaviors at the nanoscopic end structure of dynamic microtubules and reflects a clear logic in the regulatory network of microtubule dynamics.

Materials and Methods

Protein expression and purification

MCAK purification

GFP-MCAK was expressed using a modified BAC-to-BAC protocol. The purification was performed as previously described (Helenius et al., 2006). Briefly, the cells were lysed by dounce homogenization in the lysis buffer. The lysate was centrifuged at 40,000 rpm for 60 min at 4°C (XPN-100 ultracentrifuge, Type 45 Ti rotor, Beckman, USA). The supernatant was filtered through a 0.45 μm filter (Pall, 66229) and then purified using a cation-exchange column (HiTrap SP-HP, Cat.17115101, GE, USA). The protein was eluted using the elution buffer supplemented with a continuous salt gradient (0.15∼1.0 M NaCl). The peak fractions were pooled together and then loaded onto a Ni-sepharose column (QIAGEN, 30210). The column was then washed with the washing buffer I. After the washing step, 3C Protease (final concentration 10 µg/ml) (Thermo Fisher Scientific, 88946) was used to perform on-column his-tag cleavage overnight at 4°C. The eluate was further purified using a gel-filtration column (Superdex 200 increase 10/300 GL column, Cat. 28990944, GE, USA) and eluted using the washing buffer II. The purified MCAK was frozen in liquid N₂ and stored at −80℃. Lysis buffer: 50 mM HEPES pH 7.5, 150 mM NaCl, 5% glycerol, 0.1% Tween 20, 1.5 mM MgCl₂, 3 mM EGTA, 1 mM DTT, 0.5 mM Mg-ATP, 10 units/ml Benzonase. Elution buffer: 6.7 mM HEPES pH 7.5, 6.7 mM MES, 6.7 mM sodium acetate, 1.5 mM MgCl₂, 10 µM Mg-ATP. Washing buffer I: 50 mM NaPO₄ buffer pH 7.5, 300 mM NaCl, 10 mM imidazole, 10% glycerol, 1 mM MgCl₂, 10 µM Mg-ATP. Washing buffer II: BRB80, 300 mM KCl, 1 mM DTT and 10 µM Mg-ATP. BRB80: 80 mM PIPES/KOH pH 6.9, 1 mM MgCl₂, 1 mM EGTA.

EB1 purification

EB1-GFP was expressed and purified as previously described (Song et al., 2020). Briefly, EB1-GFP was expressed in the BL21 E. coli strain and lysed using sonication in the lysis buffer. The lysate was centrifuged at 40,000 rpm for 60 min at 4°C. The supernatant was filtered through a 0.45 μm filter and loaded onto a Ni-sepharose column. The column was then washed using the washing buffer. After the washing step, 3C Protease (final concentration 10 µg/ml) was used to perform on-column his-tag cleavage overnight at 4°C. The eluate was desalted into the storage buffer using the PD-10 desalting column (GE, 17085101). The final purified protein was frozen in liquid N₂ and stored at −80 °C. Lysis buffer: 50 mM NaPO₄ buffer pH 7.5, 300 mM NaCl, 10% glycerol, 10 mM imidazole, 1 mM DTT, 0.1% Tween 20. Washing buffer: 50 mM NaPO₄ buffer pH 7.5, 300 mM NaCl, 10% glycerol, 30 mM imidazole, 1 mM DTT. Storage buffer: BRB80, 100 mM KCl, 10% glycerol and 1 mM DTT.

XMAP215 purification

XMAP215-GFP was expressed using a modified BAC-to-BAC protocol. The purification was performed as previously described (Brouhard et al., 2008). Briefly, the cells were lysed by dounce homogenization in the lysis buffer. The lysate was centrifuged at 40,000 rpm for 60 min at 4°C. The supernatant was filtered through a 0.45 μm filter and then purified using a cation-exchange column. The protein was eluted using the elution buffer supplemented with a continuous salt gradient (0.15∼1.0 M NaCl). The peak fractions were pooled together and then loaded onto a Ni-sepharose column. The column was then washed with the washing buffer I. The protein was eluted using the washing buffer I supplemented with a continuous imidazole gradient (20-300 mM). The fractions with best purity were pooled together and further purified using a gel-filtration column (Superdex 200 increase 10/300 GL column, Cat. 28990944, GE, USA) and eluted using the washing buffer II. The final purified XMAP215 was frozen in liquid N₂ and stored at −80℃. Lysis buffer: 50 mM HEPES pH 7.5, 50 mM NaCl, 5% glycerol, 0.1% Tween 20, 1 mM DTT, 10 units/ml Benzonase. Elution buffer: 6.7 mM HEPES pH 7.5, 6.7 mM MES, 6.7 mM sodium acetate. Washing buffer I: 50 mM NaPO4 buffer pH 7.5, 300 mM NaCl, 10 mM imidazole, 10% glycerol, 1 mM MgCl₂, 10 µM Mg-ATP. Washing buffer II: BRB80, 150 mM KCl, 1 mM DTT.

Tubulin preparation

Tubulin was purified from porcine brain tissues by two cycles of polymerization and depolymerization followed by affinity purification using the TOG-based column, as previously described (Gell et al., 2010; Widlund et al., 2012). Tubulin was labeled with biotin (Thermo Fisher Scientific, 20217), TAMRA (Thermo Fisher Scientific, C1171) and Alexa Fluor 647 (Thermo Fisher Scientific, A20106) using the NHS esters according to the standard protocols (Gell et al., 2010).

Microtubule depolymerization assay

Microtubules (5% Alexa Fluor 647 labeled and 20% biotin labeled) were polymerized in the presence of GMPCPP (JenaBioscience, NU-405L) as previously described (Song et al., 2020). The GMPCPP-stabilized microtubules were immobilized on the surface of a cover glass using the biotin-NeutrAvidin protein links (Thermo Fisher Scientific, 31000). GFP-MCAK was then added into the flow cell in the imaging buffer. The sample was kept at 35°C using a temperature controller (Tokai Hit, Japan). Images were recorded using a total internal reflection microscope (TIRFM) (Olympus, Japan) equipped with a 100× 1.49 N.A. oil TIRF objective (Olympus, Japan) and an Andor 897 Ultra EMCCD camera (Andor, Belfast, UK). Images were recorded every 5 s with a 100 ms exposure. Imaging buffer: BRB80 supplemented with 1 mM ATP, 50 mM KCl, 80 mM D-glucose, 0.4 mg/ml glucose oxidase, 0.2 mg/ml catalase, 0.8 mg/ml casein, 1% β-mercaptoethanol, 0.001% Tween 20.

Microtubule dynamics assay

Microtubule dynamics assay was performed as previously described (Song et al., 2020). Briefly, the immobilized GMPCPP-stabilized microtubules were used as the template for microtubule growth. Tubulin dimers (13% TAMRA labeled) and the protein of interest were then added into the flow cell in the imaging buffer. The sample was kept at 35°C using a temperature controller (Tokai Hit, Japan). Images were recorded using a TIRFM (Olympus, Japan). To record microtubule dynamics, images were recorded every 5 s with a 100 ms exposure. To record single molecule binding events, images were recorded every 100 ms with a 50 ms exposure. Imaging buffer: BRB80 supplemented with 1 mM ATP, 2 mM GTP, 50 mM KCl, 0.15% sodium carboxymethylcellulose, 80 mM D-glucose, 0.4 mg/ml glucose oxidase, 0.2 mg/ml catalase, 0.8 mg/ml casein, 1% β-mercaptoethanol, 0.001% Tween 20.

The diffusion coefficient of MCAK on dynamic microtubule lattices

The diffusion coefficient of MCAK on dynamic microtubule lattices was calculated as previously described (Helenius et al., 2006). Firstly, we gained the position coordinates of MCAK at its molecular trajectory on dynamic microtubule lattices using the plugin of imageJ (Fiji), TrackMate (Tinevez et al., 2017). Note that the minimum track length was kept at 400 ms (100 ms/frame). Then, the mean-squared displacement (MSD) of MCAK at every molecular trajectory (203 trajectories from 3 assays) was calculated against the interval (100 ms/frame) by an in-house script of MATLAB. Lastly, the MSDs of MCAK was plotted against the time interval, and the diffusion coefficient, D was obtained by a linear regression analysis (<x²>=2Dt) in Origin 8.0 (OriginLab Corporation, USA).

The flux of MCAK reaching growing microtubule ends

The calculation of the flux of MCAK reaching growing microtubule ends by one-dimensional lattice-diffusion was performed according to Fick’s first equation of diffusion (Helenius et al., 2006):

where J is the flux of MCAK reaching growing microtubule ends, c(x, t) is the concentration of MCAK on microtubule lattice at position x from the growing microtubule end and time t.

Where

Where C is the concentration of MCAK. Therefore, we can calculate the flux by taking the measured parameters (D, k_on, C and k_off) into the equation.

Localization analysis at growing microtubule ends

The analysis on the single-molecule localization at growing microtubule ends was performed as previously described (Song et al., 2020). Briefly, we determined the position of growing microtubule end by considering the microtubule lattice as a Gaussian wall and its end as a half-Gaussian. This model was used to fit the microtubule end and determined the position of microtubule end in every frame. The peak position of the half-Gaussian was used as the origin of the positional axis, which was used as a reference to measure the localization of individual single-molecule binding events. We estimated that the accuracy of microtubule end tracking was ∼6 nm by measuring standard error of the distribution of the estimated error in microtubule end position (Demchouk et al., 2011). To measure the precise localization of individual binding events at growing microtubule ends, a Gaussian function was used to fit the intensity profile of individual molecules. We estimated that the accuracy of the measured position was ∼2 nm by measuring standard error of the fitting peak location (Maurer et al., 2014). The peak location was recorded as the position of each molecule in every frame. The average peak location over the entire binding period was recorded as the position of the molecule. The positional distributions of GFP-MCAK, EB1-GFP, XMAP215-GFP at growing microtubule ends were plotted along the longitudinal axis of microtubule. Finally, the plots of GFP-MCAK and XMAP215-GFP were fitted to a Gaussian function and the plots of EB1-GFP was fitted to an exponentially modified Gaussian function, from which we calculated the FWHM as an estimation for the length of the binding region of each molecule.

Microtubule polymerization

GMPCPP, GDP and GTPγS microtubules were polymerized as previously described (Manka and Moores, 2018). Briefly, GMCPP-microtubules were polymerized using a mixture of 10 μM tubulin, 1 mM GMPCPP (JenaBioscience, NU-405L) and 4 mM MgCl₂ in BRB80, which was incubated for 2 hr at 37°C. The polymerized microtubules were then collected using an Air-Driven Ultracentrifuge (Beckman, 340401), which were then resuspended in BRB80 and stored at 37°C. GDP-microtubules were polymerized using a mixture of 40 μM tubulin, 1 mM GTP (Roche, 10106399001), 4 mM MgCl₂ and 4% DMSO (Sigma, 276855). The mixture was incubated for 30 min at 37°C. The polymerized microtubules were collected using an Air-Driven Ultracentrifuge, resuspended in BRB80 with 20 μM taxol (Cell Signaling Technology, 9807) and stored at 37°C. GTPgS microtubules were obtained using two rounds of polymerizations. In the first round, a mixture of 40 μM tubulin, 2 mM GTPgS (Roche, 10220647001) and 4 mM MgCl₂ in BRB80 was incubated for 5 hours at 37°C. In the second round, 5 μl of the product from the first polymerization was mixed with additional 40 μM tubulin, 2 mM GTPgS and 4 mM MgCl₂ to make a final 25 μl reaction mixture. This mixture was incubated overnight at 37°C to obtain long GTPgS microtubules. The polymerized microtubules were collected using an Air-Driven Ultracentrifuge, resuspended in BRB80 with 20 mM taxol and stored at 37°C.

MCAK binding to different microtubules

Different nucleotide microtubules were first immobilized in flow cells. GFP-MCAK or its mutants were then added into the flow cell in the imaging buffer. Images were recorded every 300 ms with a 100 ms exposure. Imaging buffer: BRB80 supplemented with 1 mM ATP, ADP or AMPCPP, 20 μM taxol, 50 mM KCl, 80 mM D-glucose, 0.4 mg/ml glucose oxidase, 0.2 mg/ml catalase, 0.8 mg/ml casein, 1% β-mercaptoethanol, 0.001% Tween 20.

The binding intensity of GFP-MCAK on GMPCPP microtubule ends

To measure the binding intensity of GFP-MCAK on GMPCPP microtubule ends, an intensity profile of GFP-MCAK or the GMPCPP microtubule was plotted along the microtubule. The average binding intensity of GFP-MCAK of two pixels represents the end-binding intensity of GFP-MCAK at the corresponding position of the end, to the lattice from the midpoint of the microtubule intensity attenuation at the microtubule end.

Gamma fitting of microtubule lifetimes and catastrophe frequency analysis

The lifetime data of dynamic microtubules were fitted to the gamma density function using the dfittool toolbox in the MATLAB (Mathworks, USA). The probability-based catastrophe frequency is calculated as the ratio of the number of catastrophe events observed at a certain lifetime to the total number of microtubules that reached this lifetime (Gardner et al., 2011).

Acknowledgements

The authors thank Pengpeng Yu (school of life sciences and technology, TongJi University), Chunguang Wang (school of life sciences and technology, TongJi University), Peng Shi (school of basic medical sciences, Peking University) and Congying Wu (school of basic medical sciences, Peking University) for technical assistance. Special thanks to the light microscopy and protein facility in Tsinghua University. We acknowledge our funding from National Natural Sciences Foundation of China (32070704, 32370730) and IDG/McGovern Institute for Brain Research (Tsinghua University).

Supplementary materials

Figure s1.

Figure s2.

Figure s3.

Figure s4.