Distinct Interaction Modes of the Kinesin-13 Motor Domain with the Microtubule

Abstract

Kinesins-13s are members of the kinesin superfamily of motor proteins that depolymerize microtubules (MTs) and have no motile activity. Instead of generating unidirectional movement over the MT lattice, like most other kinesins, kinesins-13s undergo one-dimensional diffusion (ODD) and induce depolymerization at the MT ends. To understand the mechanism of ODD and the origin of the distinct kinesin-13 functionality, we used ensemble and single-molecule fluorescence polarization microscopy to analyze the behavior and conformation of Drosophila melanogaster kinesin-13 KLP10A protein constructs bound to the MT lattice. We found that KLP10A interacts with the MT in two coexisting modes: one in which the motor domain binds with a specific orientation to the MT lattice and another where the motor domain is very mobile and able to undergo ODD. By comparing the orientation and dynamic behavior of mutated and deletion constructs we conclude that 1) the Kinesin-13 class specific neck domain and loop-2 help orienting the motor domain relative to the MT. 2) During ODD the KLP10A motor-domain changes orientation rapidly (rocks or tumbles). 3) The motor domain alone is capable of undergoing ODD. 4) A second tubulin binding site in the KLP10A motor domain is not critical for ODD. 5) The neck domain is not the element preventing KLP10A from binding to the MT lattice like motile kinesins.

Introduction

Kinesin-13s are microtubule (MT) depolymerases distinct from most other members of the kinesin superfamily that generate force and movement along MTs (1, 2). By inducing MT depolymerization, kinesin-13s modulate MT dynamics in a variety of important cell processes such as mitosis (3, 4, 5) cytokinesis (6), axonal branching (7), and ciliogenesis (8).

Like all kinesins, kinesin-13s have a highly conserved globular catalytic motor domain with ATPase and MT binding activity. The motor domain is located in the middle of the polypeptide chain, which gives origin to the alternate family name of “internal kinesins” or Kin-Is, distinct from N- or C-terminal kinesins (9). Residues outside the motor domain are required for dimerization, interaction with other proteins, and intracellular localization (10, 11, 12, 13, 14). Interactions between the C- and N-terminal domains may also play a role in modulating depolymerization activity and MT interaction (15, 16). Dimerization is not required for MT depolymerization activity but it does increase its efficiency (10, 17). For the kinesin-13 MCAK the minimal construct capable of near-full depolymerization activity is monomeric and consists of the motor domain and a family conserved positively charged stretch of ∼60 residues N-terminal to the motor domain called the neck domain (10, 17). However, the kinesin-13 motor domain alone is capable of inducing depolymerization and stabilizing curved tubulin protofilaments thought to be the key structural intermediate in the MT depolymerization process (18, 19, 20).

Despite their similar motor domains, kinesin-13s interact with tubulin very differently than motile kinesins. Motile kinesins alternate between strong and weakly bound states in coordination with ATP hydrolysis and stepping-related conformational changes. Kinesin-13s on the other hand interact weakly with the MT lattice and undergo direction unbiased one-dimensional diffusion (ODD) until reaching the end of the MT where they induce MT depolymerization (21, 22). ODD was first proposed for DNA binding proteins as an efficient mechanism to accelerate the rate by which an enzyme finds its substrate by reducing a three-dimensional diffusion search process to one-dimension (23, 24). Many MT-binding proteins and other kinesins are also reported to diffuse along MTs (25). All motile kinesins reported to diffuse along MTs combine diffusive behavior with the capacity for ATP hydrolysis coupled unidirectional stepping (26, 27, 28, 29, 30). Thus, kinesin-13s appear unique in being able to diffuse but not engaging in directional stepping.

It is not clear yet why kinesin-13s and motile kinesins interact with the MT lattice in these distinct manners, although kinesin-13-specific structural motifs are likely to be involved (19, 31). To address this issue and to investigate the mechanism of ODD, we used ensemble and single-molecule fluorescence polarization microscopy (FPM) to compare the behavior of distinct constructs of the kinesin-13 KLP10A bound to MTs.

Materials and Methods

Protein constructs and fluorescent labeling

Drosophila melanogaster

KLP10A protein constructs were expressed and purified as previously described (20, 32). The base neck and motor domain construct (NM) includes KLP10A residues T198 to I615 plus a 6His tag at the N-terminus. The following substitutions were made to remove surface exposed cysteines (cys-light) C282A; C339A; C389S; C548S; C594V and two residues were mutated to cysteines (T455C, D460C) for labeling with the bifunctional thiol-reactive fluorescence probe bis-((N-iodoacetyl) piperazinyl) sulfonerhodamine (BSR). The mutant constructs have additional residue replacements as follows: L2M: K317A, V318A, D319A; L8M: D444A, K446A; KT2M: K306A, K350A, and K399A. The motor domain alone cys-light construct (MDA) includes KLP10A residues I279 to I615. All the mutations were introduced in the plasmid vector using the Quick-change Lightning Site-directed Mutagenesis kit (Agilent Technologies, Stratagene, Santa Clara, CA). BSR labeling was done as described (33, 34), which results in 60–75% of labeled protein. Note that for the FPM experiments full labeling is not critical as only the fluorescently labeled proteins contribute to the data collected. A modified protocol with increased labeling efficiency (incubation with a 3.5:1 BSR/KLP10A molar ratio for 40 min at 25°C) was used to increase the labeled fraction (98%) for measuring the bulk depolymerization activity of the NM-BSR-labeled construct (Fig. S1 in the Supporting Material). Labeling efficiency was estimated from the ratio between concentrations of KLP10A and BSR after free BSR was removed from the solution. KLP10A concentration was estimated from Coomassie stained gels calibrated against a protein standard. BSR concentration was estimated spectrophotometrically at 549 nm using an extinction coefficient of 88,000 cm^-1M⁻¹. The attachment location of the probe (cross-linking residues 455–460) was verified by trypsin digestion followed by mass spectrometry. Pig brain tubulin was obtained from Cytoskeleton (Denver, CO).

FPM

Solutions and sample preparation

Flow chambers with MTs were prepared as described (35). Experiments were done in solution A12 (7.5 mg/ml bovine serum albumin (BSA), 2 mM MgCl₂, 1 mM EGTA, 20 μM placitaxel, 12 mM K−PIPES, pH 6.8) plus an antibleaching system (2.5 mM protocatechuic acid (PCA), 10 nM protocatechuate-3,4-dioxygenase (PCD), 1 mM Trolox (36)) at room temperature. Nucleotides were added according to the experimental condition to be tested. AMPPNP: 2 mM AMPPNP; ADP: 2 mM ADP; ADP-AlF₄⁻: 4 mM ADP, 2 mM AlCl₃, 10 mM KF; No Nucleotide (NN): 5 units ml⁻¹ apyrase with no added nucleotide. Typical concentrations of kinesin protein used in the experiments were ∼10 nM for ensemble polarization measurements and ∼0.1 nM for single-molecule recording. The flow chambers were placed in a custom-modified microscope for fluorescence polarization observation and data recording (33, 37).

Data analysis

Ensemble and single-molecule FPM data were recorded and analyzed as previously described (35, 37). Fluorescence emission digital movies (100 msec/frame, 48 nm/pixel) were collected with an Andor Ixon EM+897 EMCCD camera using Andor Solis software (Andor Technology, South Windsor, CT). Fluorescence excitation was induced with linearly polarized laser light illumination (λ = 532 nm) synchronized to the camera acquisition to collect successive frames corresponding to four excitation polarization directions (0°, 45°, 90°, and 135°). Fluorescence emission intensities from MTs or single molecules and each of the four polarization excitation direction were quantified from the digital movies using custom software (FRWIN31). Single-molecule position trajectories were selected and analyzed from kymograph images (37). The position of each molecule in each movie frame was determined by a semiautomatic method where the molecule trajectories were first selected manually in the kymographs and then refined using a custom Python script. The manually selected trajectory positions were refined first by calculating the fluorescence intensity center of mass within a 7 × 7 pixels region of interest centered in the manually determined position and then by fitting a two-dimensional Gaussian function to the intensity profile within a 25 × 25 pixel region of interest average of four consecutive frames (running average of the four images corresponding to the four polarization excitation direction used) using the center of mass coordinates as the initial guess for the Gaussian maximum location. Diffusing or stationary molecules where separated at the manual selection stage based on their corresponding kymograph traces (see Fig. 3 A). Traces showing displacement along microtubules were included in the diffusing group and the ones appearing as straight lines in the kymographs (D < 10⁻³ μm²·s⁻¹) in the stationary group. Diffusion constant (D) were calculated by linear fitting to the mean-square displacement (MSD) versus time interval (dt) relationship, MSD = 2·D·dt (21) with dt bins up to 2 s every 0.1 s, and excluding dt = 0 (see Fig. 3 C). A custom Python script was used to calculate the MSD from the position (x,y), and time data points. Data bins from individual molecule trajectories were pooled together or kept separate to calculate an overall D (D₁) or per molecule D (D₂), respectively (see Fig. 3, C and D, and Table S3). Lifetime distributions (Fig. S2) were calculated from data sets obtained under identical illumination conditions. The photobleaching rate was estimated from the observed fluorescence emission decay of microtubules decorated with BSR-labeled kinesin-1 construct in a strongly bound state (rigor) and otherwise similar buffer and illumination settings (Fig. S3). We also attempted to estimate the photobleaching rate from the KLP10A-BSR-labeled molecules stuck to the glass. However, their estimated mean lifetimes were shorter than the ones diffusing on MTs, suggesting relative rapid unbinding. Observed mean lifetimes (τ_o) were corrected for photobleaching to estimate the mean lifetime (τ) of the MT-bound state of the KLP10A constructs using the relationship 1/τ = (1/τ_o) + (1/τ_B), where τ_B is the photobleaching mean lifetime (21). Ensemble linear dichroism LD⁰, single molecule order factor r and projected angle relative to the MT axis α were calculated from the emission fluorescence intensity ratios: LD_0–90 = (I₉₀ − I₀)/(I₀ + I₉₀) and LD_45–135 = (I₁₃₅ − I₄₅)/(I₁₃₅ + I₄₅) (where the suffix indicates the excitation light polarization direction) as explained in (37). Ensemble FPM data were modeled as a combination of two populations of probes with distinct mobility cone angle parameters Γ of 32° (38) for the ordered population and 90° for the disordered population, both with an axial angle β of 93° (expected probe orientation if the KLP10A motor domain binds to the MT lattice like conventional kinesin (39)). Predicted LD⁰ were calculated using a custom numerical calculation program (convector) that modeled each dipole as a combination of 5000 vectors with random orientations around a cone axis with maximum angle set to the chosen cone angle Γ and distributing 10⁵ of these dipoles uniformly around a cylindrical symmetry axis with the cone axes set to the chosen axial angle β. Discrete LD⁰ data points corresponding to a different relative amount of disordered probes (0–100%, every 10%) were calculated and a continuous function (see dashed line in Fig. 4 C) estimated by fitting a second order polynomial to the estimated LD⁰ data points (R² = 0.996). Curve fitting and data analysis were done using GraphPad Prism version 6.00 for Windows (GraphPad Software, La Jolla, CA).

Figure 3.

Single-molecule fluorescence. (A) Kymographs of KLP10A molecules interacting with the MT. The left panel contains examples of molecules undergoing ODD while all the molecules in the right panel are stationary. Vertical scale bar: 7 μm, Horizontal scale bar: 15 s. (B) Mean lifetimes (τ) for diffusing and stationary molecules. Error bars: 99% confidence interval of the mean. (C) MSD versus time. To produce these plots the time intervals and corresponding displacements for all the molecules diffusing in a given condition were pooled together. The graph shows, for each experimental condition the mean MSD, standard error of the mean (SE), and linear regression fit to the data. (D) Distributions of one-dimensional diffusion coefficients D per molecule. The distributions are represented as box plots where the box spans the 25% to 75% percentiles, the middle line the median and the whiskers are the minimum and maximum values. Corresponding number of molecules in each distribution (N) and statistics listed in Tables S2 and S3. To see this figure in color, go online.

Figure 4.

Single-molecule FPM traces. (A) Diffusing molecule. (B) Stationary molecule. (C) Molecule with stationary and diffusing intervals. Top color traces: fluorescence intensities for four polarization directions of the excitation light. Bottom black trace: position along the MT. Calculated order r-factor and projected angle |α| for the indicated time intervals (horizontal black lines) are shown above the fluorescence intensity traces. (D) Order r-factor versus D of all NM molecule trajectories in the presence of ADP. The cluster of molecules with D values below and above 0.001 μm² × s⁻¹ correspond to the molecules in the stationary and diffusing groups, respectively. To see this figure in color, go online.

MT depolymerization assays

Depolymerization activity was assayed by pelleting and fluorescence microscopy assays. Pelleting assay: guanylyl 5′-α,β-methylenediphosphonate (GMPCPP) stabilized MTs were prepared by incubating tubulin for 2 h at 35°C in BRB80 solution (1 mM EGTA, 1 mM MgCl₂, 80 mM K-PIPES, pH 6.8) supplemented with 1.2 mM GMPCPP. MTs diluted to a final tubulin concentration of 4 μM were mixed with varying amounts of KLP10A (NM wild-type (WT) or BSR-labeled KLP10A NM cys-light) in reaction buffer (BRB80 solution supplemented with 5 mM ATP and 1 mM (2-carboxyethyl) phosphine) to a final volume of 20 μl. The range of concentrations of KLP10A NM tested were 0 to 0.8 μM. The reaction mixtures were incubated for 6 min at 25°C, followed by the addition of 20 μl of a sucrose cushion (20% sucrose w/v in BRB80) at the bottom of each reaction tube. The samples were subsequently centrifuged for 10 min at 31,3000 × g, 25°C in a Beckman (Brea, CA) TLX ultracentrifuge. After centrifugation, supernatant and pellet fraction were separated, resuspended to equal volumes, and mixed (1:1 v/v) with 2× sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) sample buffer. For each condition, 20 μl of the pellet and supernatant were loaded onto 15 wells 10% SDS-PAGE gels containing 10% acrylamide-bisacrylamide (37.5:1). After electrophoresis the gels were stained with Coomassie blue and scanned to quantify the relative amounts of protein in each band. The relative amount of depolymerized tubulin was estimated as the amount of protein in the supernatant fraction (minus unpolymerized tubulin) over the total (depolymerized + microtubule in pellet). Unpolymerized tubulin (typically ∼30%) was estimated from the supernatant fraction remaining in the microtubule solution with no kinesin added.

Fluorescence microscopy assay

GMPCPP stabilized MTs were prepared with a 10:1 mixture of unlabeled tubulin and HiLyte Fluor 488 fluorescently labeled tubulin (Cytoskeleton). The MTs were flowed and immobilized in a flow chamber made with silanized glass (27) covered with beta-1 Tubulin antibody (SAP.4G5, ThermoFisher, Waltham, MA). The flow chamber with attached MTs was first rinsed with blocking buffer (BRB80 supplemented with 1% Pluronic F-127, 7.5 mg/ml BSA), and then either KLP10A NM or Cys-light BSR-labeled KLP10A NM construct (100 nM) in reaction buffer (BRB80 supplemented with 0.5% Pluronic F-127, 7.5 mg/ml BSA, 2 mM ATP, 10 mM DTT, 2.5 mM PCA, 10 nM PCD, 1 mM Trolox) was flowed into the chamber and digital image/movies were recorded. Control experiments were performed in the same conditions but flowing reaction buffer with no KLP10A. MTs movie/images were acquired with an Hamamatsu Orca Digital camera (Hamamatsu, Bridgewater, NJ) on an inverted Ultraview spinning-disc confocal system confocal microscope (Perkin Elmer, Waltham, MA) mounted onto a Nikon Eclipse Ti microscope and a high numerical aperture objective (60×, NA = 1.49) (Nikon, Melville, NY), using 488 nm laser excitation at 1 s exposure time, 1.105 s per frame for 10 min. Digital movies were recorded as 16-bit images using Volocity software (Nikon). Image pixel size was 0.11 μm. Kymographs were generated from the movies by drawing a segment line along MTs using the Multiple Kymograph plugin from ImageJ (J. Rietdorf and A. Seitz, European Molecular Biology Laboratory, Heidelberg, Germany). Depolymerization rates were calculated from the kymograph images as the change in position of each microtubule end in the first 3 min after KLP10A was flowed into the chamber. Kymograph scale units used were 0.11 μm/pixel along the position axis and 0.0184 min/pixel along the time axis.

Electron microscopy

Tubulin-ring-KLP10A complex

The Cys-light BSR-labeled KLP10A NM (3 μM) construct was incubated with unpolymerized tubulin (1.5 μM) in BRB80 solution supplemented with 1 mM AMPPNP, layered onto ultraviolet-treated carbon-coated electron microscope grids and imaged by negative staining electron microscopy as described (19).

Microtubule-KLP10A complex

4 μL of taxol stabilized microtubules (1.5 μM tubulin) in BRB80 solution were layered onto ultraviolet-treated carbon-coated grids. The grids were then partially blotted before applying 4 μL of a solution with KLP10A constructs (5 μM, Cys-light BSR-labeled KLP10A NM or WT KLP10A NM) in BRB80 supplemented with 2 mM ADP or AMPPNP. Grids were negatively stained and imaged in an electron microscope as described (19). Image power spectrums and layer line filtered images were calculated with ImageJ. Layer line filtered images were produced by applying rectangular masks at the equator, 1/4 nm⁻¹and 1/8 nm⁻¹ layer lines to the image Fourier transforms and calculating the corresponding inverse transform.

Results

Average configuration of MT-bound KLP10A constructs

To investigate how different kinesin-13 family-specific residues and structural motifs affect the interaction of kinesin-13 with the MT lattice, we created several mutant constructs of the Drosophila melanogaster kinesin-13 KLP10A. Previous work has shown that the minimal constructs retaining characteristic kinesin-13 functionality include the motor and neck domains (10, 17). Therefore, we concentrated here on such minimal constructs to dissect key structural elements that may account for the distinct interaction mode of kinesin-13s with the MT lattice. For site-specific fluorescent labeling, we created cys-light versions with two unique solvent exposed cysteines (T455C, D460C) to be cross-linked with the thiol reactive bifunctional fluorescent probe BSR (see Materials and Methods section). The resulting labeled constructs have the probe located on the motor domain with the absorption dipole moment oriented as shown in Fig. 1 A. This location and orientation is equivalent to those previously used in several FPM studies on kinesin-1 (33, 34, 38, 40). It aligns the fluorescent absorption dipole close to perpendicular to the MT axis when the motor domain is tightly bound to the MT lattice through the putative kinesin-tubulin binding site.

Figure 1.

Probe location and KLP10A constructs. (A) Model of the KLP10A motor domain bound to a tubulin protofilament (PDB: 3J2U) showing the location and predicted orientation of the BSR probe fluorescence transition dipole moment (magenta double arrow) and several key residues (dark blue atom spheres). α-tubulin (light-yellow); β-tubulin (dark-yellow); KLP10A motor domain (cyan). The C-β carbons of the two unique cysteines of the cys-light KLP10A constructs are shown as magenta spheres. (B) KLP10A constructs used in the study. NM, base cys-light construct including the neck and motor domains. L2M, construct with the residues KVD at the tip of loop-2 mutated to alanines. L8M, construct with the residues D444 and K446 in loop-8 mutated to alanines. KT2M construct with the residues K306, K350, and K399 that form an additional tubulin binding site (20, 42) mutated to alanines. MDA, construct including the motor but no neck domain residues. The KLP10A motor domain model is shown as a ribbon representation with the residues mutated to alanines in the different constructs shown as black spheres. One letter residue name code for the WT residues are indicated in the NM construct and the alanines substitutions in each of the mutant constructs are indicated with the letter A. The N-terminal neck domain is represented as a wiggly line to the left of the motor domain. L2 in (A) and (B) indicates the location of the tip of the KLP10A motor domain L2. To see this figure in color, go online.

We verified that the cys-light BSR-labeled construct retained the ability to induce curved tubulin protofilaments and depolymerize MTs (Fig. S1). The introduction of the cys-light mutations in KLP10A however was not completely inconsequential; in a MT depolymerization assay the cys-light BSR-labeled construct showed lower apparent MT affinity than an equivalent KLP10A WT construct (Fig. S1 A). Despite this caveat the fact that the cys-light KLP10A construct retains typical kinesin-13 functionality makes it a useful tool for FPM structure and function studies, particularly when comparing the same construct in different conditions, or different mutations in the same KLP10A cys-light background.

We created five different constructs with specific amino acid deletions, or with key charged residues changed to alanines (Fig. 1). The construct NM corresponds to the base cys-light construct and includes the neck and motor domains with no additional mutations other than the cys substitutions. The constructs L2M and L8M are similar to the NM construct but with added mutations in the motor domain loop-2 (L2) and loop-8 (L8), respectively. These two loops are part of the binding interface of the KLP10A motor domain with curved tubulin (19). The elongated L2, with the KVD residues at its tip, is a kinesin-13 family-specific feature, whereas L8 is part of the generic putative kinesin tubulin binding site. In addition, we made a construct with the mutations K546A, E547A, C548A in helix-4 (H4), which is also part of the putative kinesin-tubulin interface. Consistent with previous studies (20, 41) these mutations in H4 resulted in very poor MT binding making this construct unsuitable for further FPM analysis. The L2M construct has the three family conserved contiguous residues, KVD, mutated to alanines. These mutations have been shown to have a strong deleterious effect on MT depolymerization activity (20, 31, 41). The L8M construct has the mutations D444A and K446A. The construct KT2M (kinesin-tubulin binding site 2 mutant) has three lysines mutated to alanines (K306A, K350A, K399A). These lysines are part of an additional tubulin binding site on the kinesin-13 motor domain that is on the opposite side of the putative kinesin-tubulin interface (42). The MDA construct includes KLP10A motor domain residues without the neck domain.

To determine the overall KLP10A motor domain MT binding configuration (orientation and angular disorder) of the different constructs, we measured the fluorescence linear dichroism (LD⁰) values of each construct bound to MTs. The LD⁰ reports on the axial orientation and angular disorder of the MT attached probes ensemble. Positive or negative values correspond respectively to probes perpendicular or parallel to the MT axis with possible values ranging from −1 to 1. Increased angular disorder or axial angles closer to 54.7° result in LD⁰ values closer to zero (37). FPM experiments were carried out in the absence of nucleotides (NN) or in the presence of different nucleotide analogs (AMPPNP, ADP+AlF₄⁻, or ADP) to mimic different points of the motor domain ATPase cycle.

Most of the LD⁰ values measured were positive (Fig. 2 B; Table S1), as previously reported for a more limited set of conditions (19). This indicates that the motor domain of these constructs binds to the MT lattice in such a way that the probes are oriented close to perpendicular to the MT axis. This is consistent with the KLP10A motor domain binding to the MT lattice in a similar configuration as other kinesins. However, the observed LD⁰ values for the KLP10A constructs are slightly lower (closer to zero) than reported values for equivalent monomeric kinesin-1 constructs in nucleotide conditions that induce tight binding (38), suggesting a greater degree of angular disorder for KLP10A relative to kinesin-1. Another important difference from kinesin-1 is that the LD⁰ value of the KLP10A NM constructs is only slightly reduced in the presence of ADP relative to the other nucleotide conditions. On the other hand, ADP on kinesin-1 induces a state with high rotational mobility (defined as the amount of probe angular wobble within the measurement timescale) with a near zero LD⁰ value (34, 38).

Figure 2.

Ensemble FPM. (A) Images of MTs decorated with many BSR-labeled KLP10A construct (NM construct in the presence of AMPNP). The green double arrows indicate the polarization direction of the fluorescence excitation light. (B) Ensemble FPM LD⁰ of KLP10A constructs in four nucleotide conditions. Error bars: SD. Corresponding number of MT measurements (N) and statistics shown in Table S1. (C) Predicted LD⁰ of a model with variable relative amounts of ordered and disordered probe configurations. The mobility cone angle Γ of the ordered and disordered configurations were set to 32° and 90°, respectively, and the axial angle β to 93° in both cases. The averaged LD⁰ in the four nucleotide conditions for each construct and the corresponding relative amount of disordered molecules in the model are indicated. To see this figure in color, go online.

Comparing the LD⁰ values of the NM with the other mutant constructs informs on whether the mutated structural elements are important to maintain the orientation of the motor domain when bound to the MT lattice. To interpret the observed changes we compared the observed LD⁰ values with the ones predicted from a model combining two populations of probes with low and high rotational disorder (Fig. 2 C). Mutations in L2 and deletion of the neck domain produce the largest reductions in LD⁰ (97% and 85% of the molecules being rotationally disordered, respectively, according to the model) indicating that interactions between these kinesin areas and the MT are critical for maintaining the KLP10A motor domain in a fixed orientation relative to the MT.

More modest but statistically significant reductions in LD⁰ were observed for the L8M and KT2M constructs (p < 0.05 in t-tests between mutants and NM constructs in the same nucleotide conditions) indicating that residues in these areas have a small effect in setting the orientation of the motor domain when KLP10A is bound to the MT lattice. The reduction in LD⁰ for the L8M and KT2M relative to the NM construct were equivalent to an ∼10% increase in the disordered population according to the model (Fig. 2 C). L8 is part of the putative kinesin-tubulin binding site, so it is not unexpected that mutations in this area affected kinesin binding. However, the mutations in the KT2M constructs are located on the opposite side of the motor domain (Fig. 1), so they cannot simultaneously form part of the binding interface with the MT lattice. This could indicate an indirect effect of the KT2M mutations on the putative tubulin binding site or alternatively, this result could be explained if at any one time ∼10% of the NM constructs were MT bound through the second tubulin binding site and this population becoming disordered in the KT2M mutant (the orientation of the probe is expected to be similar whether the motor domain binds through the putative or second tubulin binding site, with axial angles β of 93° and 87°, respectively).

Single-molecule behavior of MT-bound KLP10A constructs

We used time-resolved single-molecule FPM to further characterize the interaction of the KLP10A constructs with MTs. Fig. 3 A shows kymographs of typical single-molecule records. We observed molecules either undergoing ODD or staying stationary, as well as several alternating between diffusive and stationary modes (Figs. 4 and S8). The ability to engage in ODD on MTs is a characteristic activity of kinesin-13s (21) but the stationary events could in principle represent molecules nonspecifically stuck to the surface. However, the stationary events are very likely to also represent a genuine KLP10A MT binding state because 1) they are selected from the same MT regions where diffusing events occur; 2) molecules diffusing and stationary are found within the same group of molecules; 3) the same molecule can alternate between diffusive and stationary behavior (Figs. 4 C and S7); 4) the binding pattern of many of the stationary molecules (orientation and spacing along the MT, see next section) is similar to other kinesins.

All the constructs were capable of undergoing ODD, indicating that none of the potential KLP10A-MT interaction sites, mutated or deleted, are essential for this behavior. A possible exception to this rule is helix-4 in the motor domain, given that mutations in this area resulted in very poor MT binding (see above). ODD events were observed in all nucleotide conditions (AMPPNP, ADP, ADP+AlF₄⁻) but in the absence of NN relatively few events were observed.

The FPM experiments were performed in relatively low ionic strength to improve kinesin-MT binding. Increasing ionic strength augmented the amount of background fluorescence precluding the acquisition of reliable single-molecule polarization intensity measurements. However, at higher ionic strength we also detected stationary and diffusing KLP10A molecules, and we did not observe major changes in the overall orientation of MT-bound KLP10A molecules as measured by ensemble FPM. Similar LD⁰ values were obtained for the NM construct at several ionic strengths (Fig. S6). Therefore, the overall configuration of MT-bound KLP10A molecules at the low ionic strength used is likely to be similar at higher (physiological) ionic strength.

To investigate possible differences in MT binding between constructs, we looked at the lifetime distribution of diffusive and stationary events of each KLP10A construct in the presence of ADP (Fig. S2). The mean lifetimes (τ) calculated from these distributions are shown in Fig. 3 B and Table S2. The lifetimes were significantly longer for the diffusive than for the stationary events for all the constructs. In addition, the lifetimes corresponding to the four mutant constructs (L8M, L2M, KT2M, and MDA) were shorter than for the NM constructs. These results indicate that although none of the mutated/deleted residues are essential for ODD, they all play a role in keeping KLP10A attached to the MT during ODD. Our interpretation of this finding is that during ODD the KLP10A motor domain rocks back and forth or tumbles over the MT lattice, alternating between several possible tubulin-interaction sites, none of which are essential for MT binding. A rocking-tumbling motor domain would result in high angular disorder consistent with the relative low LD⁰ observed in the ensemble FPM experiments (previous section). Direct support for this model comes from single-molecule FPM records (next section).

We also found significant differences in the speed of diffusive displacements between constructs. The MSD during the diffusive events changed linearly over time (Fig. 3 C), as expected for purely diffusive movement (21). We estimated the diffusion constant (D) corresponding to each construct (Table S3) from the slope of the MSD versus time interval relationships. The data presented in Fig. 3 C was obtained by pooling together all the time intervals from all the ODD events for each construct. This provides a measure of the average diffusion constant for each construct. We also obtained distributions of the D per ODD molecule event to account for observed molecule-to-molecule variability (Fig. 3 D). Despite the high variability observed (up to two orders of magnitude), we detected significant (p < 0.01, t-test) differences in the average D values between the NM and the mutant constructs. The average D of the KT2M, L2M, and NM constructs were relatively similar and ∼3× lower than the L8M and MDA constructs. This result indicates that interactions between the MT lattice and L8 and the neck domain slow down ODD.

High and low rotational mobility interaction modes of KLP10A NM with the MT lattice

To further understand the configuration of KLP10A constructs when interacting with the MT lattice, we analyzed the fluorescence polarization signals to determine the orientation and rotational mobility of individual KLP10A molecules. Fig. 4 shows simultaneous fluorescence intensity and position along the MT traces of three selected representative molecules. A diffusing molecule (Fig. 4 A) shows very little fluorescence anisotropy (similar average intensity values for each of the four excitation polarization directions: 0°, 45°, 90°, and 135°) indicating high rotational mobility. On the other hand a stationary molecule (Fig. 4 B) shows very high anisotropy, indicative of low rotational mobility. The traces on Fig. 4 C correspond to a molecule with alternating diffusing and stationary periods. High and low rotational mobilities are associated with the diffusive and stationary periods, respectively. From the ratio between the fluorescence intensities, we calculated the order r-factor, which provides a measure of the rotational mobility of each molecule (0 = no anisotropy and high rotational mobility, 1 = maximum anisotropy and no rotational mobility). A scatter plot of the order r-factor versus the diffusion constant of all the NM molecules in the presence of ADP (Fig. 4 D) reveals two well-separated clusters based on their diffusion constant. All the molecules in the high diffusion group have low order r-factor values (high rotational mobility), whereas the ones in the stationary low diffusion group (D < 0.001 μm²·sec⁻¹) have a wider range many reaching r-values over 0.5 (low rotational mobility). We found all the diffusing molecules, regardless of construct or nucleotide condition, to have very high rotational mobility, whereas the ones with low rotational mobility were found only in the low diffusion stationary group.

Fig. 5 shows single-molecule order r-factor and orientation frequency distributions for the NM and the L2M constructs. Additional single-molecule FPM data are presented in Table S4 and Fig. S4. For all the constructs analyzed the diffusing molecules had r-factor distributions heavily skewed toward low values (median r = 0.1. Fig. 5, A and D, and Table S4) indicating high rotational mobility. The high rotational mobility of the diffusing molecules provides direct evidence that the KLP10A motor domain rocks back and forth or tumbles, changing orientation rapidly during ODD.

Figure 5.

Single-molecule FPM distributions. (A) Order r-factor of NM diffusing molecules. (B) Order r-factor of NM stationary molecules. (C) Projected angle |α| of NM stationary molecules (with r > 0.3). (D) Order r-factor of L2M stationary molecules. (E) Order r-factor of L2M diffusing molecules. (F) Projected angle |α| of L2M stationary molecules (with r > 0.3). Number of molecules included in each distribution (N) and additional data in Table S4 and Figs. S4 and S5. To see this figure in color, go online.

Stationary molecules ranged between high and low rotational mobility cases shifting the corresponding r-factor distributions toward higher values (Fig. 5 B; Table S4). For the NM construct the angular distributions (projected angle relative to the MT axis, |α|, of the less rotationally mobile stationary molecules (r > 0.3) is skewed toward high angular values with peak frequencies between 70° and 90° (Fig. 5 C) indicating a probe orientation near perpendicular to the MT axis. On the other hand, all the mutant constructs (L2M, L8M, KT2, and MDA) showed higher angular variability without a clear peak and distribution medians of lower angular values (Table S4). In other words, all the diffusing molecules were rotationally disordered and only a fraction of the stationary molecules were rotationally ordered with the probe oriented perpendicular to the MT axis. This ordered fraction was reduced or absent in the mutant constructs.

The near perpendicular orientation of the probe of the stationary low-rotational-mobility NM molecules is consistent with the motor domain being bound to the MT like motile kinesins. Further supporting this view is the fact that many NM molecules (Cys-light-BSR labeled or WT) can bind to MTs with the typical 8 nm periodicity observed for many other kinesins (Fig. S9).

The single-molecule FPM results support the model used to interpret the observed ensemble FPM LD⁰ values (Fig. 2 C). The observed positive LD⁰ derive from the population of rotationally ordered molecules with the probe oriented near perpendicular to the MT axis and the differences in LD⁰ derive from differences in the relative amounts of rotationally ordered and disordered molecules. However, for some constructs the amount of angular disorder derived from the single-molecule records is higher than expected from the ensemble measurements. For example, the r-factor and angular distributions for the KT2M construct were similar to those of the L2M construct, even though the LD⁰ of the former was higher. A higher than expected number of disordered molecules may be the result of a fraction of molecules, particularly stationary ones, being nonspecifically bound to the coverslip surface. In our current setup it is not possible to distinguish molecules bound to a MT in random orientations from the ones bound to the surface but close to a MT (<300 nm from MT axis) without some additional revealing property such as displaying movement along the MT. Another possible contributing factor to higher LD⁰ values than expected from the single-molecule FPM distributions could be a cooperative effect causing a more uniform angular orientation when many molecules are bound simultaneously to the MT in the ensemble FPM experiments. To investigate this possibility we obtained single-molecule FPM data for the KT2M constructs in the presence of an unlabeled KLP10A construct (Fig. S5). We found that the addition of unlabeled KLP10A reduced the amount of angular disorder supporting the existence of cooperative ordering in this case. The mechanism of this cooperative effect needs to be further investigated but one possibility is that in the presence of many neighboring molecules ODD is restricted and more molecules adopt a stationary and low angular mobility binding mode.

There is an interesting and unanticipated negative correlation between angular order and the mean lifetimes of the single-molecule distributions (Fig. S7), which is particularly noticeable when comparing the distribution of diffusing and stationary molecules. One possibility that may account for this observation is that the diffusing and highly mobile molecules make transient but more numerous contacts with the MT resulting in relatively longer lifetimes.

We did not observe a simple relationship between the average orientation of the constructs and their corresponding D values. Although the L2M construct showed the largest decrease in LD⁰ (Fig. 2), it had only a slight reduction in the average D relative to the NM construct (Fig. 3 C). On the other hand, the MDA construct despite also showing a decrease in LD⁰, had an increased D relative to the NM construct. This may indicate that some of the binding interactions between KLP10A and the MT lattice in the stationary and diffusing states are different. For example, in the stationary mode L2 may interact with the tubulin interdimer interface, as it does when bound to curved tubulin protofilaments (19), but interact transiently with other parts of tubulin during ODD. Accordingly, mutations in L2 could produce a drastic reorientation of the motor domain in the stationary state but have a more limited effect during ODD as observed.

Discussion

We have used ensemble and single-molecule FPM to characterize the interaction of the kinesin-13 KLP10A with the MT. The data support a model in which KLP10A molecules interact with the MT in at least two modalities (Fig. 6): 1) a low rotational mobility mode where the motor domain binds to the MT in a well-defined orientation. 2) A highly mobile mode where the motor domain rocks back and forth or tumbles over the MT lattice. ODD occurs in this second binding mode. Binding to the MT lattice in both modes is likely mediated by the putative kinesin motor domain tubulin binding site as mutations at or near this site have a strong effect on MT binding or the configuration of MT bound KLP10A constructs.

Figure 6.

KLP10A-MT interaction modes model. (A) Stationary mode. The KLP10A motor domain (red ribbons) with key charged residues marked as (dark blue spheres) binds to the MT lattice (green and yellow ribbons) in a configuration similar to other MT-bound kinesins, which orients the BSR probe fluorescent dipole (arrow in magenta) perpendicular to the MT axis. Interactions with the MT involving L2 in the motor domain and the neck domain (light blue) are critical to keep the motor domain in place and with a fixed orientation. (B) Mobile and diffusive mode. The molecule binds to the flexible tubulin CTTs (green wavy lines) through any of its multiple charged residues. This binding mode allows rapid changes of orientation of the motor domain (represented as two overlapping motor domain orientations) and binding-unbinding events to the same or other tubulin subunits leading to ODD. To see this figure in color, go online.

High rotational mobility is not a necessary consequence of movement along the MT lattice. Kinesin-1 keeps its two motor domains attached to the MT with very low rotational mobility for most of the stepping cycle, while moving processively at saturating [ATP] (35, 38). On the other hand, a highly mobile interaction of the motor domain with the MT lattice is not unique to the MT depolymerase kinesin-13s. In the presence of ADP, the motor domain of the kinesin-1, KIF5B, can also bind to the MT with high rotational mobility (34). We interpret these highly rotationally mobile MT-bound states, as states in which the kinesin motor domain is prevented from forming a tight interaction with tubulin. A possibility is that binding in this mode is mediated mainly by the disordered C-terminal tails of tubulin (CTTs) through electrostatic interactions. An involvement of the CTTs in kinesin MT-bound states with high rotational mobility is consistent with the CTTs disordered structure (43) and the deleterious effect on kinesin-MT interactions (processive run length, ODD, depolymerization activity) caused by their removal (18, 21, 44, 45, 46, 47). It is also reported that removal of the CTTs decreases kinesin-1 motor domain MT binding flexibility (48). An interesting possibility to be investigated is whether the transition between the low and high rotational mobility MT interaction modes of kinesin-13 is modulated by the presence of distinct CTT tubulin isotypes or posttranslation modifications (49, 50). This could potentially be a mechanism to regulate how kinesin-13s bind to the MT and reach their ends.

The configuration of the KLP10A motor domain in the low rotational mobility resembles the one of motile kinesins strongly bound to MTs as modeled from electron microscopy three-dimensional reconstructions (51). In both cases the putative kinesin tubulin binding site mediates MT binding, the motor domain is similarly orientated relative to the MT axis and many KLP10A molecules decorate the MT following the 8 nm axial repeat of the tubulin heterodimer. However, the binding interactions are not identical because our data indicates that KLP10A requires kinesin-13 family class-specific residues inside and outside the motor domain (L2 and neck domain, respectively) to bind to the MT in a fixed orientation. The data also imply that the KLP10A motor domain alone, without the neck domain or other parts of the molecule, is adapted to bind to the MT lattice differently than motile kinesins. It has been proposed that the kinesin-13 family-specific neck domain sterically prevents the motor domain from interacting with the MT lattice as motile kinesins do (31), and kinesin-13 constructs that include only the motor domain can bind to the microtubule with a similar configuration than other kinesins (52). However, if the neck domain were to prevent the motor domain from forming a stereo-specific interaction with the MT, we would then have expected the KLP10A motor domain to have improved orientational order when bound to the MT and impaired ODD relative to the KLP10A NM construct; however, the opposite was observed. This last result is also consistent with the fact that eliminating charged residues in the neck domain does not prevent MCAK ODD (53). It would also not have been expected that disrupting other kinesins-13 family-specific residues, such as the KVD motif in L2, would result in a more disordered interaction of the motor domain with the MT lattice. The results then support a model in which a specialized kinesin-13 putative motor-domain-tubulin binding site, without the neck domain or other parts of the molecule, is adapted to undergo ODD and not to form a motile kinesin-like interaction with the MT lattice. This is consistent with electron microscopy results indicating that the putative tubulin binding site of KLP10A is adapted to bind and fit curved and sheared tubulin protofilaments, presumably at MT ends, instead of straight tubulin as found in the MT lattice (19). This model does not exclude other roles for the class conserved kinesin-13 neck domain, such as increasing the binding rate to the MT (53), bridging adjacent tubulin protofilaments (54), or being part of a regulatory switch (15). The fact that deleting the neck domain resulted in higher angular disorder suggests the additional role for the neck domain of steering the kinesin-13 molecule into the binding configuration required for tubulin bending at the MT ends.

The fact that ODD was observed in all the KLP10A mutant construct investigated (except the H2 mutant that had very poor MT binding) rules out ODD mechanisms mediated by interactions through any of these groups exclusively. It rules out the neck domain or the second tubulin binding site on the motor domain as the key MT binding sites mediating ODD. On the other hand, the fact that deleting or neutralizing charges had a measurable effect on the diffusing molecules, together with the high degree of rotational mobility, suggest that the mutated areas and perhaps others may alternatively interact with the MT during ODD. The high angular mobility of the motor domain is also contrary to models in which ODD is mediated only by specialized domains at or near the motor domain putative kinesin tubulin binding site, such as the K-loop (loop-12) in KIF1A (55), or that a unique orientation must be adopted to diffuse along the MT (25). Our results instead suggest a mechanism for ODD where a KLP10A molecule is able to rotate and exchange binding interactions between tubulin subunits without dissociating from the MT. Kinesin-13s are particularly suited for such a mechanism as they have many positively charged residues all over the motor and neck domains that could form transient electrostatic interactions with the negatively charged MT surface. Such a mechanism could be shared with other MT diffusing proteins, although the binding groups would necessarily be different given the wide variety of proteins reported to diffuse along MTs (25). It is also interesting to note the negative correlation observed between angular order and lifetimes. Whether this correlation holds for other interacting macromolecules remains to be investigated. In the case of kinesin-13s it may represent an adaption to favor diffusive interactions with the MT lattice.

From this discussion a slightly different view emerges regarding some of the structural adaptations giving kinesin-13s distinct functionality from motile kinesins. Rather than possessing a specialized domain or particular group of residues responsible for ODD, kinesin-13s have a motor domain uniquely adapted to elude forming a tight interaction with the MT lattice at any point of the ATPase cycle. In this situation weak electrostatic interactions conductive to ODD would prevail allowing kinesins-13s to diffuse to the MT ends where a distinct tubulin conformation (e.g., curvature and/or lack of interprotofilaments lateral contacts) would promote ATP-hydrolysis coupled MT depolymerization activity.

Author Contributions

C.C. designed and expressed protein constructs, designed and performed FPM experiments, and analyzed data. M.P.B. expressed protein constructs, designed and performed FPM experiments and depolymerization assays, and analyzed data. J.D.D.V. performed depolymerization assays. A.B.A. designed and expressed protein constructs and performed EM experiments. G.J.S., D.J.S., and H.S. planned and discussed experiments. H.S. supervised the project, analyzed data, and wrote the article. All authors edited and contributed to the final manuscript version.

Editor: Jennifer Ross.