Members of the kinesin-13 sub-family, including KIF2C, depolymerize microtubules. The positive charge-rich ‘neck’ region extending from the N-terminus of the catalytic head is considered to be important in the depolymerization activity. Chemically synthesized peptides, covering the basic region (A182–E200), induced a sigmoidal increase in the turbidity of a microtubule suspension. The increase was suppressed by salt addition or by reduction of basicity by amino acid substitutions. Electron microscopic observations revealed ring structures surrounding the microtubules at high peptide concentrations. Using the peptide A182–D218, we also detected free thin straight filaments, probably protofilaments disintegrated from microtubules. Therefore, the neck region, even without the catalytic head domain, may induce lateral disintegration of microtubules. With microtubules lacking anion-rich C-termini as a result of subtilisin treatment, addition of the peptide induced only a moderate increase in turbidity, and rings and protofilaments were rarely detected, while aggregations, also thought to be caused by lateral disintegration, were often observed in electron micrographs. Thus, the C-termini are not crucial for the action of the peptides in lateral disintegration but contribute to structural stabilization of the protofilaments. Previous structural studies indicated that the neck region of KIF2C is flexible, but our IR analysis suggests that the cation-rich region (K190–A204) forms β-structure in the presence of microtubules, which may be of significance with regard to the action of the neck region. Therefore, the neck region of KIF2C is sufficient to cause disintegration of microtubules into protofilaments, and this may contribute to the ability of KIF2C to cause depolymerization of microtubules.
Kinesin is a molecular motor with ATPase activity that moves along microtubules using energy released upon ATP hydrolysis [1-3]. Kinesin and its related proteins have a well-conserved globular head domain and form the kinesin superfamily [3, 4]. Many members of the kinesin superfamily have their head at the N-terminus of the polypeptide and move toward the plus end of microtubules [1, 3, 4]. However, kinesin-14 sub-family members have their head at the C-terminus and move toward the minus end of microtubules [3-5]. It has been reported that the regions adjacent to the head play important roles in the characteristic motilities of these molecules .
Members of the kinesin-13 sub-family, including KIF2C (also referred to as mitotic centromere-associated kinesin, MCAK), differ significantly from other kinesins. The heads are located at the center of the polypeptide, so they are also called central kinesins or Kin I [3, 4, 7]. However, their most characteristic feature is that they do not show uni-directional movement along microtubules but depolymerize microtubules from both ends [8, 9]. It is natural to assume that, as the catalytic head is homologous to those of other kinesin motor proteins, other regions of the protein are important in the function of kinesin-13 sub-family members. In fact, there is a class-specific ‘neck’ region consisting of ~ 50 amino acid residues that extends from the N-terminus of the head domain [10, 11]. The neck region is rich in positive charges, and does not form a specific structure in the absence of microtubules, as this region is invisible in the X-ray crystal structure (Fig. 1A) . Indeed, the neck region is accessible to both the lateral and axial interfaces of the microtubule lattice when KIF2C protein is bound to microtubules (Fig. 1A–C).
It is thought that the KIF2C head domain is a prerequisite for depolymerization activity, and that the neck region enhances this activity [11, 15, 16]. Maney et al.  expressed serial recombinant proteins containing various lengths of the neck region together with the head domain. They reported that the depolymerizing activity of the recombinant protein in vivo was low with a short neck from the N-terminus of the head (D218–K231), but was high and comparable to that of full-length protein with the neck elongation (A182–K231). Recently, Wang et al.  analyzed in vitro the ATPase properties of a KIF2C head domain with a short N-terminal region, corresponding to R210–D246, as a minimal functional domain as determined by the decrease in turbidity of microtubule suspension, an indicator of microtubule depolymerization. However, Aoki et al.  reported that drugs targeting the neck region corresponding to N206–S219 inhibited depolymerization of microtubules by KIF2C both in vitro and in vivo.
We are interested in functions of the neck region of KIF2C, especially with respect to the issue of whether the neck region itself interacts with microtubules. We synthesized various peptide segments from the neck region and observed the interaction of these peptides with microtubules. We investigated two peptides, KF101 and KF501, with the sequences A182–E200 and A182–D218, respectively (Fig. 1F), in more detail. We discuss the possible mechanism for microtubule disintegration by these peptides.
Results and Discussion
Sigmoidal increase in the turbidity of a microtubule suspension by addition of the neck region peptide
To investigate interactions between the synthetic peptides and microtubules, we first examined the turbidity of a 1 μm (0.1 mg·mL−1) microtubule suspension in the presence of various concentrations of each of the peptides (for sequences, see Fig. 1D–G) by measuring the absorbance at 350 nm. In the case of KF501, A182–D218 of KIF2C, no obvious effects were detected up to 2 μm. However, at higher KF501 concentrations, the turbidity increased steeply, but the rate of increase in turbidity was much reduced at concentrations exceeding 5 μm. The overall profile was nearly sigmoidal. KF501 and the shorter peptides exhibited similar effects (Fig. 2A,B). The fitting parameters are summarized in Table 1.
Table 1. Summary of parameters for the turbidity change of a microtubule suspension in the presence of various peptides. Data in Fig. 2 were fitted to Hill's equation as a function of peptide concentration. Dissociation constants (K ) and Hill coefficients (n) were obtained from a fitting calculation, f(C) = a0 + [(a1 + b1C) − a0]/[(K/Cn) + 1], where C, f(C), a0 and (a1 + b1C) are the peptide concentration, the absorbance of the system at C, the absorbance for the unbound state, and the absorbance for the bound state defined hypothetically. After a steep increase, the increase in turbidity continued but at a much reduced rate. The peptide concentrations C10%, where , as the transition starting point, and Chalf, where , as the transition mid-point, were calculated. Dissociation constants were derived from the equation
Concentration of 10% transition (C10%/μm)
Concentration of half transition (Chalf/μm)
Hill coefficient (n)
No turbidity change
No turbidity change
No turbidity change
No turbidity change
No turbidity change
No turbidity change
No turbidity change
KF501 with subtilisin-treated microtubules
Transition mid-points for the increase in turbidity were 2.9 ± 0.2, 2.8 ± 0.1, 10.1 ± 0.6, 3.1 ± 0.1 and 3.4 ± 0.3 μm for KF101, KF201, KF301, KF401 and KF501, respectively. The Hill coefficients were 4.9 ≤ n ≤14.5 (Table 1), suggesting a high degree of apparent cooperativity. These findings suggest that the change in physical state reflected by the increase in turbidity required a certain number of peptide molecules per tubulin dimer. We also synthesized KF100, KF102 and KF103, comprising various portions of the KIF2C neck region, none of which induced an increase in the turbidity of a microtubule suspension (Fig. 2A and Table 1).
Various peptides based on KF101 but with amino acid replacements were also synthesized (Fig. 1G). KF101-rnd1, KF101-rnd2, KF101-Daa1 and KF101-Daa2 were prepared with the intention of interfering with possible formation of the native secondary structure of KF101. All of these peptides induced an increase in the turbidity of a microtubule suspension. The transition mid-points were 6.2 ± 0.3, 9.1 ± 0.7, 5.1 ± 0.5 and 6.1 ± 0.3 μm, respectively (Fig. 2C and Table 1). Higher peptide concentrations are required to cause a turbidity increase for each of these peptides, compared with the original KF101 (2.9 ± 0.2 μm).
KF101, which contains eight lysine or arginine residues of its 19 amino acid residues, is highly basic. In order to evaluate the importance of this basicity, we synthesized four peptides, KF101-his1, KF101-his2, KF101-his3 and KF101-his4, in which two lysine and/or arginine residues were replaced with histidine residues (Fig. 1G). These four peptides, with reduced basicity, failed to induce a turbidity increase (Table 1), suggesting that a certain level of basicity is indeed important.
We performed similar experiments in the presence of 90 mm KCl to observe the effects of reduced electrostatic interactions. In the presence of added salts, the increase in turbidity was totally or strongly suppressed for all peptides (Fig. 2B).
The above data suggest that electrostatic interactions between basic residues of the peptides and acidic residues of the microtubules are important for the turbidity increase. Additionally, the sequence fidelity of the peptides appears to have some significance. Ovechkina et al.  reported that a mutant protein of the KIF2C head domain, with either a truncated neck or an artificial neck in which some of the basic residues had been replaced with alanine residues, showed poor ability with respect to microtubule depolymerization. However, the activity was significantly recovered when a lysine-rich loop was added to the N-terminus of the truncated neck. Our results are in good agreement with theirs, but our results further indicate that neck region peptides probably interact with microtubules independently of the motor domain as well as constitute a part of the function for microtubule depolymerization (see below).
Electron microscopic observations: microtubule bundling and ring formation
To learn more about the effects of the synthetic peptides that induced a turbidity increase, KF501 (0–10 μm) was mixed with microtubules (1 μm) for negative staining electron microscopy (Fig. 3). While only typical microtubules were seen at 0.5 μm KF501, some bundles of microtubules were found at 1 μm, and extensive microtubule bundle formation was observed at 2 μm KF501 or more (Fig. 3B–F). We observed a similar bundling of microtubules for other peptides employed in this study (KF101, KF101-rnd2 and KF101-Daa2) (Fig. 4B–D). It should be noted that the peptides are cation-rich (~ 40% of their amino acid residues are lysines and arginines, see Fig. 1F,G). Electrostatic interaction between the peptides and tubulin molecules results in bundling, as discussed below with regard to interaction with subtilisin-treated microtubules.
It is also noteworthy that, upon addition of KF501, even at a low concentration, the microtubules showed a tendency to open up (Fig. 3B–E). It is considered that the interactions with peptides make it difficult for a microtubule to retain its tubular structure.
At 5 μm KF501 or above, ring-shaped structures surrounding microtubules were observed, similar to those in previous reports for peptides with catalytic heads  or cationic lipids . In the presence of 10 μm KF501, the rings covered the majority of the microtubule surface. Neither free rings nor double concentric rings around a microtubule were observed.
With regard to the ring structure, many of the rings were observed to serially surround the microtubules, almost in a packed manner (Figs 3E, F and 4A). Mulder et al.  observed stacked rings in the presence of a recombinant head domain with a neck region, and unstacked rings in the presence of a recombinant head domain without a neck region.
Ring structures were not uniformly formed. Even with a single microtubule, certain zones were covered with packed rings while others were bare. Similar ring structures were observed with KF101, and even with KF101-rnd2 and KF101-Daa2, which were designed to interfere with possible native secondary structure formation (Fig. 4B–D). In all cases examined, the ring structures were found at peptide concentrations higher than those showing steep increases in turbidity. More detailed analytical data concerning the rings in the presence of KF501 are available in Doc. S1.
Most interestingly, thin straight filaments were detected in the presence of 5 or 10 μm KF501 peptide (Figs 3E, F and 4A), as discussed below. In some cases, aggregations were found with 10 μm KF501 (Fig. 3F), as described below for subtilisin-treated microtubules.
Appearance of protofilaments: lateral disintegration of microtubules
As the thin filaments were detected only at high concentrations of KF501 accompanying ring formation, the appearance of filaments and rings appears to be correlated. Like the rings twining around microtubules, it is likely that these filaments are protofilaments produced as the result of the lateral disintegration of microtubules at high concentrations of KF501, because the suspension contained microtubules, peptides and components of the buffer only. On the other hand, no such filaments were detected, even by careful observation, when KF101, KF101-rnd2 or KF101-Daa2 was added. One possibility may be that KF101 and its derivatives are considerably shorter than KF501, and the former may not favor the existence of individual protofilaments.
In microtubules, it may be that electrostatic interaction plays important roles in lateral binding between tubulin dimers, as interfaces of two laterally associating tubulin dimers have both positive and negative regions in a complementary manner . It is possible that synthetic peptides derived from the neck region, such as KF501, are inserted into the lateral interface between two tubulin dimers to disintegrate microtubules. At low peptide concentrations, microtubules are opened up (Fig. 3B–E), whereas at high peptide concentrations, microtubules are laterally disintegrated, resulting in the appearance of free protofilaments and rings (Figs 3E, F and 4A).
However, we postulate that only part of the bound neck peptide is involved in the lateral disintegration. We found that four KF501 peptides bound to one tubulin dimer at a high peptide concentration (see details on secondary structure formation below). Some of the peptides bind to the lateral interface while the others bind to other sites. With the native KIF2C protein, the catalytic head domain binds to a tubulin molecule at the proper position, so that interaction of the neck region with the lateral interface between the protofilaments of a microtubule is facilitated (see also Fig. 1C).
Interaction of KF501 peptide with subtilisin-treated microtubules
As described above, for interactions between neck region peptides such as KF501 and microtubules, electrostatic factors are important. The peptide contains a number of basic residues and is thought to interact with acidic regions of tubulin. Both the α- and β-subunits of the tubulin dimer have flexible C-terminal regions, called ‘E-hooks’, which are anion-rich (Fig. 1B) . The anion-rich C-terminal regions may be selectively cleaved off by use of subtilisin . We prepared subtilisin-treated microtubules and investigated their interactions with KF501 peptide in order to evaluate the importance of tubulin C-termini.
Turbidity measurements were performed for mixtures of 1 μm subtilisin-treated microtubules and various concentration of KF501 peptide. Similar to the case with the intact microtubules, a steep increase in turbidity was induced with 2 μm or more of the peptide (Fig. 2D). The transition mid-point was 4.0 ± 0.4 μm, similar to that for intact microtubules (3.4 ± 0.3 μm, Table 1). The difference between the turbidity before and after the steep increase with the subtilisin-treated microtubules was approximately half of that with intact microtubules.
Next, electron microscopic observations were performed for the mixture of 1 μm subtilisin-treated microtubules and 10 μm KF501 (Fig. 5). Most of the subtilisin-treated microtubules were in bundles (Fig. 5A,C), as seen for the intact microtubules. In contrast to the intact microtubules, ring-shaped structures were found in only one view (Fig. 5D) of more than 50 views, and no thin straight filaments were detected. However, aggregations were often observed (Fig. 5D). They were also seen with the intact microtubules but rarely (see above and Fig. 3F). As for the rings described above, the aggregations are formed from protofilaments, as the suspension contained microtubules, peptides and components of the buffer only. This conclusion is also supported by the finding that ring structures and aggregations were directly connected in some micrographs (Figs 3F and 5D). The presence of tubulin C-termini somehow contributes to stabilization of the morphology of free protofilaments and the rings surrounding microtubules.
Thus, the KIF2C neck peptide induces lateral disintegration of microtubules into protofilaments, regardless of the presence or absence of the C-termini of tubulin dimers. Hertzer and Walczak  reported that, compared with the case for intact microtubules, KIF2C protein depolymerized subtilisin-treated microtubules into tubulin dimers with much reduced activity, but they also observed protofilaments in rings, sheets and aggregations. Our results are in line with their report; lateral disintegration of microtubules, an important step in depolymerization by KIF2C protein, is brought about by the neck region, even in the absence of the KIF2C head domain, and is not affected by the presence or absence of the anion-rich tubulin C-termini.
In the mixture of microtubules and a high concentration of the KIF2C neck region peptide, the absence of tubulin C-termini resulted in a reduced turbidity increase and a rare appearance of rings and free protofilaments. However, bundled microtubules were observed, regardless of the subtilisin treatment. Extensive bundling is responsible for part of the increase in turbidity . As the increase in turbidity was suppressed by the addition of salt, the bundling was considered to be induced by electrostatic interaction between the peptide and tubulin molecules. Although the KIF2C neck region peptide is cation-rich, our results suggest that the anion-rich C-terminal regions of tubulin subunits are not necessarily involved in the extensive microtubule bundle formation. In addition, ring formation is responsible for the remaining part of the increase in turbidity, partly due to augmented light scattering .
It should be noted that intact KIF2C protein functions on microtubules even at 40 nm, a concentration 50-fold lower than that at which our peptides showed their effects . However, intact KIF2C protein exhibits high affinity for tubulin molecules, so KIF2C may act on microtubules even at a concentration as low as 40 nm. Once the catalytic head of KIF2C binds to a tubulin molecule in microtubules, the local concentration of the neck region of KIF2C is much higher for the tubulin molecule, sufficient for interaction with microtubules. Binding of the free neck peptide to tubulin requires a much higher concentration, such as 2 μm.
Secondary structure formation of the neck region induced by the interaction with microtubules
It was assumed that the neck region of a kinesin-13 sub-family member does not form a specific structure in the absence of microtubules . CD spectra of the peptides derived from this region showed characteristics of random coils (Fig. S1). However, in the presence of microtubules, little information was obtained from CD spectra regarding the secondary structures of the peptides, due to the large light-scattering effects of the turbid suspension .
We therefore used isotope-labeling IR spectroscopy to obtain further structural information on KF501, which caused the lateral disintegration of microtubules (Figs 3E, F and 4A). Prior to performing IR spectroscopy, we evaluated the binding ratio of KF501 and a tubulin dimer. The samples for IR measurement were obtained by centrifugation of a mixture of 10 μm KF501 and 1 μm microtubules. The amount of KF501 sedimented with the microtubules was 2.3 ± 0.2 μm, as determined by reverse-phase HPLC analysis, while the amount of microtubules precipitated was 0.58 ± 0.05 μm. Thus, the binding molar ratio of KF501 for a tubulin dimer was 4.0 ± 0.4 under the conditions of the IR measurements.
IR experiments were performed for mixtures of microtubules and KF501 peptides with two or four 13C-labeled residues at various positions (Fig. 1H). An IR spectrum with non-labeled KF501 as a control showed relatively α-helix-rich features (Fig. 6A), consistent with X-ray crystallographic data . To detect a 13C-induced amide I band shift , difference spectra of 13C-labeled and non-labeled KF501 were obtained and analyzed by resolution with Gaussian curves. The obtained spectra and detailed descriptions of the analysis, as well as verification of the protocols for calculations, are given in Figs S2 and S3, Docs S2–S4 and Tables S1–S3. Representative difference spectra are shown in Fig. 6B, indicating the reconstituted spectra with Gaussian components. The peaks around 1610 cm−1 may be assigned to 13C-labeled amide bonds bearing β-structure, because band shifts by a 13C-isotope effect are ~ −20 cm−1, according to the values from our preliminary observations (Doc. S2). The peak height per 13C label was 0.5, 0.7, 1.0, 1.0 or 0.4 in 10−3 × absorbance units for KF501-13C1, -13C2, -13C3, -13C4 and -13C5, respectively. Therefore, it is highly likely that the segment K190–A204 (15 residues, including seven cationic residues) forms a β-structure upon interaction with microtubules.
On the basis of comparison of the peak areas between Gaussian curves 3 + 4 in Fig. 6A and Gaussian curve 3 for KF501-13C3 in Fig. 6B , the complex of a microtubule and KF501 is estimated to contain 222 residues of β-structure per tubulin dimer plus four KF501 molecules. As a tubulin dimer has 126 residues forming β-structure , 96 residues are newly formed β-structure. Among these 96 residues, 60 (four molecules x 15 residues) are assumed to be derived from KF501, such that the remaining 36 residues are new β-structure in the tubulin dimer.
The present study showed that the KIF2C neck region peptide itself, even in the absence of the head domain, interacts with microtubules, for which electrostatic interactions are important. At the same time, formation of β-structure upon interaction of the peptide and tubulin molecules, both in the peptide at K190–A204 and in tubulin, was suggested by IR measurements. It is considered that the interaction leads to opening up of the microtubule structure. At higher peptide concentrations, microtubules are laterally disintegrated, resulting in the appearance of free protofilaments and rings as well as aggregations, as often observed with subtilisin-treated microtubules. A schematic image of the lateral disintegration is presented in Fig. 7. The neck region itself contributes to an important step, lateral disintegration, in KIF2C function to depolymerize microtubules.
Synthesis of KIF2C peptides
The amino acid sequence of Cricetulus griseus (Chinese hamster) KIF2C was obtained from http://www.ncbi.nlm.nih.gov/protein/8488991. Peptides with sequences related to the neck region of KIF2C are shown with the prefix KF in Fig. 1E–H. Chemical synthesis of peptides was performed as described previously  (see Doc. S5). We also prepared artificially designed peptides and 13C-labeled peptides (Fig. 1G,H). The purified peptides were lyophilized and stored at −20 °C. Peptides were dissolved in buffer containing 100 mm MOPS/NaOH (pH 7.0) and 2 mm MgCl2 (MM buffer) prior to use.
Preparation of microtubules
Tubulin was isolated and purified from porcine brains as described previously  and stored at −80 °C. For polymerization of intact microtubules , appropriate amounts were thawed, and equimolar paclitaxel (Sigma-Aldrich, St Louis, MO, USA) was added. The mixture was incubated at 25 °C for 10 min. The suspension was layered on top of a cushion of 5% sucrose in MM buffer, and centrifuged at 23 000 g for 20 min. The pellet was resuspended in MM buffer at 200 μm (20 mg·mL−1 tubulin dimer subunits) with small amounts of paclitaxel. Protein concentrations were estimated by the Bradford method [].
For preparation of subtilisin-treated microtubules , 100 μg·mL−1 subtilisin (Sigma-Aldrich) was added to the polymerized microtubule suspension and incubated at 30 °C for 90 min. The reaction was terminated by addition of 5 mm phenylmethanesulfonyl fluoride. The subsequent steps from centrifugation to estimation of the protein concentration were identical to those for the intact microtubules. Cleavage was confirmed by polyacrylamide gel electrophoresis with Coomassie Brilliant Blue staining.
The experiments described below were performed at 25 °C. Microtubules (1 μm), either intact or subtilisin-treated, and various concentrations of peptides were mixed in 100 μl MM buffer and incubated for 15 min, as the turbidity reached a plateau in less than the first 5 min. The turbidity of the mixtures was estimated by measuring their absorbance at 350 nm with a 1 cm light path length.
Electron microscopic observations
Microtubules (1 μm) and KF501 (0–10 μm) or other peptides (10 or 15 μm) were mixed in MM buffer. A mixture (4 μL) was applied to a carbon-coated grid and negatively stained with 2% uranyl acetate. Samples were observed using an H-7000 electron microscope (Hitachi, Tokyo, Japan) operated at 75 kV. Micrographs were taken at a magnification of × 50 000.
Analyses for binding stoichiometry
Intact microtubules (1 μm) and KF501 (10 μm) were mixed in 500 μl MM buffer and incubated at room temperature for 15 min. The mixture was centrifuged at 13 000 g for 30 min on a sucrose cushion. The amount of KF501 in the precipitate was determined by reverse-phase HPLC using a YMC-Pack Pro C18 S-3 column (YMC, Kyoto, Japan) , and that of microtubules was estimated by the Bradford method.
The derivatives of KF501, with 13C labels at specific main chain amide groups or with their natural abundance of 13C (referred to as ‘non-label’), were used for the measurements. Intact microtubules (1 μm) and peptide (10 μm) were mixed in 1 ml MM buffer and incubated at room temperature for 15 min. The mixture was centrifuged at 13 000 g for 30 min on a sucrose cushion in MM buffer with deuterated water. The precipitate was resuspended in 10 μL MM buffer with deuterated water and used in the measurements. IR absorbance spectra were obtained using the attenuated total reflection method (see Doc. S3) .
We wish to thank Kazuhiro Oiwa (National Institute of Information and Communications Technology), Masaru Tanokura and Tsukasa Makino (Department of Applied Biological Chemistry, University of Tokyo) for valuable discussions. We are also grateful to Emiko Kobayashi (National Institute of Advanced Industrial Science and Technology) for her skillful assistance in electron microscopy.