A Chimeric Kinesin-1 Head/Kinesin-5 Tail Motor Switches between Diffusive and Processive Motility

Abstract

Homotetrameric kinesin-5 motors are essential for chromosome separation and assembly of the mitotic spindle. These kinesins bind between two microtubules (MTs) and slide them apart, toward the spindle poles. This process must be tightly regulated in mitosis. In in vitro assays, Eg5 moves diffusively on single MTs and switches to a directed mode between MTs. How allosteric communication between opposing motor domains works remains unclear, but kinesin-5 tail domains may be involved. Here we present a single-molecule fluorescence study of a tetrameric kinesin-1 head/kinesin-5 tail chimera, DK4mer. This motor exhibited fast processive motility on single MTs interrupted by pauses. Like Eg5, DK4mer diffused along MTs with ADP, and slid antiparallel MTs apart with ATP. In contrast to Eg5, diffusive and processive periods were clearly distinguishable. This allowed us to measure transition rates among states and for unbinding as a function of buffer ionic strength. These data, together with results from controls using tail-less dimers, indicate that there are two modes of interaction with MTs, separated by an energy barrier. This result suggests a scheme of motor regulation that involves switching between two bound states, possibly allosterically controlled by the opposing tetramer end. Such a scheme is likely to be relevant for the regulation of native kinesin-5 motors.

Introduction

Microtubules (MTs) are the most versatile functional elements of the cytoskeleton. Their complex dynamics are an integral part of cellular machineries such as the mitotic spindle (1,2). MTs in functional structures such as cilia, flagella, and the mitotic spindle are organized by a large variety of MT-binding proteins (3,4), an important class of which includes kinesin and dynein motor proteins (5). Some of these motors drive cargo transport along the MTs, whereas others serve to organize MT superstructures. Motor proteins are typically allosterically regulated on several levels to consume only ATP and produce force when needed. First of all, ATPase activity is suppressed when kinesins are not bound to MTs (6). In addition, several kinesins have been found to be cargo activated (7–9). In dimeric kinesin-1 motors, backfolding of the tail was shown to suppress ATPase activity (7,10–12). In the homotetrameric kinesin-5 Eg5 from Xenopus laevis, a modified version of cargo activation was found. Kinesin-5 motors are plus-end directed, bind between two MTs, and can slide MTs relative to each other (13). In mitosis, they act in the midzone of the spindle to exert poleward force (14,15). In vitro experiments demonstrated that Eg5 motors are only activated between two MTs (16), and that activation in physiological buffer conditions requires the interplay of eight MT-binding sites, four head-MT interactions, and nonATP dependent interactions between the four C-terminal tails and the MTs (17). The Saccharomyces cerevisiae kinesin-5 Cin8 adds a further regulation feature: it can reverse directionality from minus-end directed on single MTs to plus-end directed when bound between MTs (18,19). The plus-end motility of Eg5 on single MTs under low-salt conditions differs from kinesin-1 motility. It consists of short runs interrupted by one-dimensional (1D) diffusion (16). It was shown that diffusive motility does not require ATP hydrolysis (20). The diffusive mode of Eg5 appears to be due to the second binding site in the tail domain (17,21,22). The complex regulation of kinesin-5 motor proteins remains far from understood.

One useful approach is to create truncated constructs that lack domains or form dimers (17,18,23). A further approach is to generate chimeras by fusing parts of well-studied kinesins to parts of kinesin-5 motors. Here, we characterized a chimera, dubbed DK4mer, constructed by replacing the motor domain and neck linker of Eg5 with homologous parts of Drosophila melanogaster kinesin-1. The rationale was to create a fast tetrameric MT-sliding motor without the kinesin-5 tail-head regulation, but with the additional tail binding sites and bipolar structure. This should make it possible to study how such a motor binds between two MTs and how this changes its behavior. An important open question is how motors are turned on in the right place at the right time in the cell. For example, during cell division, kinesin-5 motors localize largely to the spindle poles and less to the spindle midzone (24). However, the spindle midzone is where kinesin-5 motors are believed to exert force, whereas they may simply be parked at the spindle poles. This requires an on-off switch that is sensitive to the local environment, such as the presence of antiparallel MTs. From earlier studies, it is clear that kinesin-5 tails are involved in sensing binding between antiparallel MTs (21). Even if the kinesin-1 heads are not regulated by the Eg5 tails, the behavior of the tails in conjunction with the motor heads should be informative. We studied the chimera’s motility and capability to slide MTs in in vitro assays at different ionic strengths. DK4mer showed a fast and highly processive motility as expected from the kinesin-1 heads, and MT cross-linking and sliding as expected from the dumbbell geometry. The intriguing result was that DK4mer performed extraordinarily long runs and switched in a mutually inhibitory manner between two modes of interaction with the MT: a processive one and a diffusive one.

Materials and Methods

Cloning and Protein Purification

The construction of DK4mer and a dimeric derivative was based on an analysis of the transition between the motor domain and neck coiled-coil in Eg5 and Drosophila kinesin-1 (DmKHC). The first 345 amino acids (aa) of Drosophila kinesin-1 were fused to Xenopus kinesin-5 (Eg5) at aa 370. Additional constructs contained either a 6-his tag (DK4mer-his) or a green fluorescent protein (GFP)-6his cassette as described previously (25). The exact sequence is shown in Fig. 1 A. The dimeric derivative DK511 was truncated at aa 511 (Fig. 1 D). Donor plasmids were kindly provided by W.O. Hancock (DmKHC) and T.M. Kapoor (FL-Eg5-GFP, BK006). We used a nested PCR approach (25) to extend the motor domain and neck linker of DmKHC (DK) with sequences providing an uninterrupted transition with selected restriction sites to a neck/stalk/tail-fragment of Eg5 (EK). The integrity of motor protein constructs was confirmed by commercial sequencing (Seqlab; Goettingen, Germany). DK was amplified with a forward primer, DK1 fwd, providing an NdeI site (flanked by additional SalI and XmaI-sequences), and two reverse primers, DKrev1 and DKrev2, providing transition sequences of the Eg5 neck up to an AflI site that was generated using a silent mutation in the natural sequence (further flanking regions provide more restrictions sites, such as NotI, for subcloning of fragments). The EK fragment was generated using a forward primer, EKfwd1, and two reverse primers, EKrev1 and EKrev2, providing the same restriction sites at the N-terminus and a cassette containing a sequence containing an AscI and XmaI site, followed by a 6his box and a stop codon, followed by the cutting sites NotI, Sal, and XhoI. The AscI/XmaI site allowed us to insert a previously used GFP-6his casette from pT7-7-GFP-his. The resulting PCR fragments were initially parked in a pTOPO-XL vector (Invitrogen; Darmstadt, Germany) before they were subcloned into a pFastBac vector (Invitrogen) for expression in Sf9 cells. Truncation constructs (DK511-GFPhis as a truncated version of DK4mer-GFP, and D421-GFPhis as a truncated version of DmKHC) were generated using a simple PCR approach and subsequent subcloning in a pET-21b(+) vector (Novagen-EMD Biosciences; Darmstadt, Germany). Expression and purification were performed as described previously (25). Chemicals were purchased from Sigma (Munich, Germany) if not otherwise stated.

Figure 1.

Construction and purification of the tetrameric chimera DK4mer-GFP. (A) Details of the junction between the Xenopus laevis Eg5 neck coiled-coil and the Drosophila melanogaster DmKHC motor domain (the numbering refers to the amino acid numbering in the respective wild-type motor sequences as indicated by gray boxes). (B) SDS-PAGE gel showing DK4mer at ∼130 kDa and DK4mer-GFP at ∼157 kDa in comparison with molecular weight markers (lane 3). (C) Cartoon of the overall geometry of the bipolar homotetrameric chimera DK4mer. (D) Cartoon of the overall geometry of a truncated version of DK4mer, the dimeric chimera DK511, lacking the Eg5 tail domains.

Multimotor Surface-Gliding Assays

Multimotor surface-gliding assays were performed as described previously (25). The motor proteins were adsorbed nonspecifically to coverslips. Assay chambers were made from coverslips, microscope slides, and double-stick tape. Chambers were flushed with approximately three chamber volumes of motility assay mix (BRB80⁺) based on BRB80 buffer (80 mM PIPES/KOH, pH 6.8, 1 mM MgCl₂, 1 mM EGTA) containing 10 μM taxol (paclitaxel), 2 mM ATP, 4 mM MgCl₂, 10 mM DTT, 0.08 mg/ml catalase C40, 0.1 mg/ml glucose oxidase, and 10 mM glucose. For multimotor surface-gliding assays, 0.022 mg/ml tetramethylrhodamine (TMR)-labeled MTs, polymerized as described previously (25), were added to BRB80⁺. Motility was observed in a fluorescence microscope (Axiovert 200, EC Plan-Neofluar 100x 1.3 NA oil objective (Zeiss; Goettingen, Germany)). The temperature was T = 22°C. Images were recorded with a digital CCD camera (CoolSnap ES; Roper Scientific; Planegg, Germany) at a frame rate of two frames/s and analyzed with ImageJ (National Institutes of Health; Bethesda, MD).

Single-Molecule Fluorescence Assays

Coverslips were plasma cleaned (PDC-002; Harrick Plasma, Ithaca, NY) and silanized with 3-[2-(2-Aminoethylamino) ethylamino]propyl-trimethoxysilane (DETA; Sigma; Germany) for MT immobilization (25). TMR-labeled MTs were attached with 5 min incubation, followed by 5 min incubation with 0.1 mg/ml casein in BRB80. Finally, DK4mer-GFPhis (DK4mer-GFP) diluted in BRB80⁺ to ∼150 nM was introduced. Fluorescence was observed in a custom-built total internal reflection fluorescence (TIRF) microscope (T = 22°C). The TIRF setup was similar to one described previously (26) with modifications: we used two lasers (473 nm and 532 nm; Viasho; Beijing, China) to excite GFP and TMR. The lasers were expanded and coupled via a multiwavelength beam splitter (z474/488/532/635rpc; Chroma; Bellows Falls, VT) off-axis into an objective (SFluor 100x, 1.49 oil; Nikon; Duesseldorf, Germany). Emitted fluorescence was split into GFP and TMR signals by a dichroic mirror (525/50; Chroma), passed through bandpass filters (530/50 for GFP and 605/70 for TMR; Chroma), and finally directed to separate areas of the detector area of a frame-transfer EMCCD camera (Cascade 512B; Roper Scientific) controlled with WinSpec32 (Princeton Instruments; Trenton, NJ). In a further modification (not used here), a commercial image splitter (Optosplit III; Cairn Research; Kent, UK) was integrated (Fig. S1 in the Supporting Material). For analysis, the separated signals were superimposed using the OptoSplit ImageJ plugin (Cairn Research). Experiments were recorded at two frames/s and subsequently analyzed for velocities and run lengths using kymographs generated with a custom-written LabView (National Instruments; Austin, TX) routine. Average velocities were estimated from linear fits to straight segments of runs (judged by eye). Runs < 2 s were not scored. Statistical analysis was performed with OriginPro (OriginLab; Northampton, MA). A mean-squared-displacement (MSD) analysis was done with a custom-written MATLAB (The MathWorks; Natick, MA) routine based on a previously published algorithm (16).

For measurements at different salt concentrations, KCl was added to BRB80⁺ buffer. Measurements with <80 mM salt concentration were done in motility assay mix as described above, but based on P30 buffer (30 mM PIPES/KOH, pH 6.8, 1 mM MgCl₂, 1 mM EGTA) termed P30⁺. To calculate the total ionic strength of the different buffers, one has to take into account the fact that PIPES buffer is a diprotonic acid (pKa₁ < 3, pKa₂ = 6.8). For example, to adjust 80 mM of PIPES free acid to pH 6.8, the addition of 120 mM KOH is required. At pH 6.8, in addition to the K⁺ and Cl⁻ ions, an equimolar amount of each acid group is in solution, and the OH⁻ ions are buffered by water. Ionic strength can be calculated as I = 1/2 Σ c_i z_i², where c_i is the concentration and z_i the valence of ion type i. For example, BRB80 plus 10 mM KCl has an ionic strength of I = 1/2 [40 mM PIPES_-1 × (−1)² + 40 mM PIPES_-2 × (−2) ² + (120 + 10) mM K₊₁ × (1) ² + 10 mM Cl_-1 × (−1) ²] = 170 mM.

Relative Sliding of Polarity-Marked MTs

Polarity-marked MTs were assembled in two steps: First, TMR-labeled tubulin (3.33 mg/ml) was polymerized to short seeds with 0.4 mM guanosine-5′-[(α,β)-methyleno]triphosphate (GMPCPP; Jena Bioscience; Jena, Germany); and second, the seeds were diluted and extended by further polymerization in a solution of 0.54 mg/ml TMR-labeled tubulin and 0.72 mM GTP in P30 buffer. Short polarity-marked MTs were polymerized by increasing the seed concentration and shortening the incubation time. MTs were stabilized in P30 containing 10 μM taxol (paclitaxel; Sigma). DETA-silanized assay chambers were prepared as described above. Long MTs were adsorbed for ∼4 min, followed by ∼6 min incubation with 0.5 mg/ml casein in P30 buffer. Finally, DK4mer-GFP diluted in 10 μl of P30⁺ buffer to ∼600 nM and short MTs were added. To obtain different salt concentrations, KCl was added to P30⁺ or BRB80⁺. Fluorescence was observed with a Zeiss Axiovert/Coolsnap ES camera (T = 22°C). Experiments were recorded at two frames/s, and images were analyzed using kymographs generated with custom-written LabView (National Instruments) routines for MT velocity and binding geometry.

Single-Molecule Imaging of Motors during MT Sliding

Single-motor molecules were imaged during relative sliding of polarity-marked MTs in the custom-built TIRF setup (T = 22°C). The assay was performed and recorded as described above, but with 320 nM DK4mer-GFP. The TMR and GFP signals were aligned using the OptoSplit ImageJ plug-in provided by Cairn Research, and analyzed for motor protein and MT velocity as described above.

Results and Discussion

The tetrameric kinesin-1/kinesin-5 chimera we cloned (Fig. 1 C) was intended to combine fast, robust motility with the capability to cross-link and slide MTs. The motor domain and neck linker of the fast and processive DmKHC should guarantee the former, and the stalk and tail of Eg5, with additional MT binding sites, the latter.

DK4mer/DK4mer-GFP/DK511 in Multimotor Surface-Gliding and Single-Molecule Fluorescence Assays

To characterize DK4mer, we performed in vitro motility assays. In surface-gliding assays, we first confirmed that DK4mer and DK4mer-GFP are competent MT motors. In BRB80⁺ buffer, the gliding velocity was ∼550 nm/s with and without the C-terminal GFP tag (Fig. S2, A and B). This indicates that the GFP tags did not interfere with motor function, consistent with the literature (17). At saturating ATP, the velocity was ∼40% lower than that of DmKHC (Fig. S2, A and B) but 10- to 20-fold higher than that of Eg5 in similar conditions (13). To control for differences in ATP affinities, although both motors share the same motor domain, we performed surface-gliding assays with DK4mer-GFP and kinesin-1 at different ATP concentrations (Fig. S2 C). For both motors, the ATP dependence of the gliding velocity was well fit by a Michaelis-Menten equation, resulting in reaction constants of K_M = 0.056 mM for DK4mer-GFP and K_M = 0.061 mM for DmKHC. The results suggest that the affinity of the kinesin-1 motor domain for ATP is unchanged in the context of the DK4mer-GFP chimera. Therefore, the discrepancy in velocity may be caused by different structures outside of the motor domain, such as the neck linker or the Eg5 tail domain. A difference in velocities may also be caused by the fact that in surface-gliding assays, the MT-gliding velocity is produced by an ensemble of motors that might include motors in the diffusive state, adsorbed in different geometries or damaged.

In surface-gliding assays, nonprocessive motor proteins also produce motility. To determine whether individual DK4mer-GFP chimeras are processive, we next performed single-molecule fluorescence assays with DK4mer-GFP (25). In BRB80⁺ buffer, we observed single fluorescent spots interacting with and moving along surface-adsorbed MTs for tens of seconds toward the plus end of the MTs (Fig. 2 A and Fig. S3), indicating processive motility of DK4mer-GFP. Micrometer-long unidirectional movements were occasionally interrupted by unbiased diffusive episodes (Figs. 2 A and 3 A). Here we use the term “processive motility” strictly for periods of motion with a constant nonzero velocity that presumably are driven by continuous hand-over-hand motion of one pair of kinesin heads. We thus exclude the diffusive pauses. During processive phases, single DK4mer-GFP molecules moved at an average velocity of ∼500 nm/s (Fig. 2 B, solid columns), in close agreement with the results from the surface-gliding assays. To make sure that we were not observing aggregates of motors, we recorded and analyzed the intensity time course of individual fluorescent spots. We observed bleaching of moving spots in at most four consecutive steps, as expected for tetramers (Fig. S4).

Figure 2.

Single-molecule fluorescence motility assay on surface-immobilized MTs with 2 mM ATP in BRB80⁺ buffer. (A) Kymograph of DK4mer-GFP motility along a selected MT, showing diffusive and processive periods. (Inset) Zoom on run showing diffusive pauses. (B) Velocity distribution of the tetrameric DK4mer-GFP and the dimeric DK511-GFP fitted with Gaussians. Average velocity: 500 ± 10 nm/s (N = 212, number of processive segments) for DK4mer-GFP, and 530 ± 10 nm/s (N = 226) for DK511. Velocities were scored from straight segments of motion longer than 2 s.

Figure 3.

Selected kymographs of single DK4mer-GFP motility on surface-immobilized MTs. Switching of DK4mer-GFP between diffusive and processive motility modes in dependence of buffer ionic strength. (A) DK4mer-GFP motility at 2 mM ATP in P30, BRB80, and BRB80 buffer with an additional 40 mM KCl. Arrows mark the transitions between motility modes. (B) DK4mer-GFP motility at 2 mM ADP in P30, BRB80, and BRB80 buffer with an additional 40 mM KCl. Horizontal scale bars: 5 s; vertical scale bars: 1600 nm.

The overall run lengths were remarkably long, on average ∼9 μm in BRB80⁺ buffer (Fig. 4 D and Table S1). This behavior is in stark contrast to what was observed with full-length Eg5 (20), in which directed and diffusive motility was difficult to discern, and the average run length was ∼600 nm in similar conditions. For dimeric kinesin-1 motors, run lengths typically range between 1 and 3 μm (27–30) (also see D421 data in Fig. 4 D and Table S1). An earlier study of a tail-less dimeric kinesin-1 tail/kinesin-5 head chimera (Eg5Kin) showed a run length of 1.9 μm (25). Consistent with that result, we observed a run length of ∼1 μm for an Eg5 dimer, truncated at aa 511 (data not shown). Valentine et al. (23) reported a run length of only ∼68 nm for a truncated version of human Eg5 (Eg5-513-5His) in high-salt buffer. Thus, the evidence indicates that bare motor domains of Eg5, at least at intermediate ionic strengths, produce run lengths similar to those observed for kinesin-1. The short run lengths of full-length Eg5 motors are known to be caused by tail-mediated cargo regulation, which at physiological ionic strength completely inhibits the motors unless they are bound between two MTs (16). Lower ionic strength can override the inhibition, and BRB80⁺ buffer is close to the margin of this effect. The long run length of DK4mer-GFP observed here indicates that the kinesin-1 motor domains are not susceptible to this type of regulation.

Figure 4.

Statistics of processive (runs) and diffusive (pauses) periods in the motility and overall run lengths of DK4mer-GFP and control constructs, from single-molecule fluorescence assays on surface-immobilized MTs at 2 mM ATP and different buffer ionic strengths (I, given in legend). (A) Distributions of pause times in the motion of single DK4mer-GFP molecules in buffers as listed in the legend; N = number of pauses (fitted with single-exponentials, using the number of events as weights). Characteristic times are given in Table 1. (B) Distributions of run times in buffers as listed in the legend of (A), fitted with single-exponentials, using the number of events as weights; runs < 2 s were not scored, and first points were not included in the fit. Characteristic times are given in Table 1. Inset: Percentages of events terminating with unbinding from pauses and runs at different buffer ionic strengths. (C) Distributions of overall run lengths before unbinding of DK4mer-GFP in buffers as listed in the legend (fitted with single-exponentials, using the number of events as weights; runs < 1 μm were not scored, and first points were not included in the fit). (D) Average overall run lengths plotted against buffer ionic strength: DK4mer-GFP (open squares), DKmer-GFP/only processive segments (solid squares), truncated DmKHC Kinesin-1 (D421) (circles), and truncated DK4mer-GFP (DK511) (triangles). Dashed lines serve to guide the eye.

The distinct separation between runs and pauses allowed us to determine four transition rates: 1), from run to pause (k_rp); 2), from pause to run (k_pr); 3), unbinding from run (k_ru); and 4), unbinding from pause (k_pu; Fig. 4, A and B, and Table 1). The pause exit rates to both run and unbinding remained largely constant with increasing ionic strength of the buffer (Fig. 4 A and Table 1). We found an average pause time before a continued run of 3.0 ± 0.3 s in low ionic strength (P30⁺ buffer) and 3.9 ± 0.6 s in high ionic strength (BRB80⁺ buffer + 40 mM KCl), corresponding to k_pr = 0.33 s⁻¹ and k_pr = 0.26 s⁻¹, respectively. This low transition rate cannot be due to a diffusional search, because the motor is kept near the track by tail binding. Rather, the diffusional state promoted by tail binding and the processive state driven by the heads appear to be separated by an energy barrier, possibly associated with a conformational change in the tetramer. Weinger et al. (17) showed that the two binding modes are cooperative for full-length Eg5 in MT sliding assays, presumably due to simultaneous binding. From our data, we cannot exclude the possibility that the tail remains bound during processive periods, or that the heads remain interacting with the MT during the diffusive periods. Tail binding during processive runs must be weak compared with the force the motors exert, because we observed no significant difference in motor velocities between tetrameric DK4mer and dimeric DK511 lacking the tails (see below). Likewise, we did not observe a significant difference in unbinding rates from processive periods between DK4mer-GFP and DK511. In earlier work (21), we showed that headless tetrameric kinesin-5 mutants are sufficient to bundle MTs. Therefore, it is reasonable to assume that the tails dominate the adhesive interaction during diffusive periods.

Table 1.

Transition rates of DK4mer-GFP and DK511 under different buffer ionic conditions

Motor and buffer	tpbr [s]	kpr [1/s]	N	trbp [s]	krp [1/s]	N	trbu [s]	kru [1/s]	N	tpbu [s]	kpu [1/s]	N
DK4mer-GFP P30+	3.0 ± 0.3	0.33	39	6.2 ± 0.2	0.16	74	6.4 ± 0.4	0.16	56	6.0 ± 0.8	0.17	20
DK4mer-GFP BRB80+	4.5 ± 0.6	0.22	48	6.5 ± 0.2	0.15	81	8.6 ± 0.7	0.12	51	6.6 ± 1.0	0.15	12
DK4mer-GFP BRB80+ + 40 mM KCl	3.9 ± 0.6	0.26	65	3.1 ± 0.1	0.32	66	2.8 ± 0.5	0.36	75	6.0 ± 0.5	0.17	20
DK511 P30+	-	-	-	-	-	-	6.0 ± 4.5	0.17	61	-	-	-
DK511 BRB80+	-	-	-	-	-	-	5.1 ± 3.5	0.20	72	-	-	-

t_pbr: pause time before run; k_pr: rate pause → run; t_rbp: run time before pause; k_rp: rate run → pause; t_rbu: run time before unbinding; k_ru: rate run → unbinding; t_pbu: rause time before unbind; k_pu: rate pause → unbind.

Transitions from pause to run were about twice as likely to occur as transitions from pause to unbinding (Table 1). This means that the barrier to unbinding is higher than the barrier to the processive state. Transition rates from runs to pause and to unbinding were roughly equal at all ionic strengths, but both decreased ∼2-fold between the lowest- and highest-ionic-strength buffers (Fig. 4 B, Table 1). This is consistent with the general finding that kinesin-1 run lengths decrease with ionic strength (Fig. 4 D). During processive motion, kinesin heads cycle through a series of nucleotide binding and conformational states, so there are many possible ways in which ionic strength can influence the cycle. Consistent with this, at all ionic strengths, unbinding from a run was 3–4 times more likely to occur than unbinding from a pause state. Taken together, these findings make it seem likely that the tail-head and tail-MT interactions are at the core of the intriguing cargo regulation of Eg5.

In BRB80⁺ with ADP, purely diffusive interactions were observed (Fig. 3 B), much as previously reported for Eg5 (16,20). We determined effective 1D diffusion constants by MSD analysis. The 1D-diffusion constant of DK4mer-GFP in BRB80⁺ with 2 mM ADP was 1.2 × 10⁴ nm²s⁻¹ (Fig. 5 B), which is ∼10-fold higher than that reported for Eg5 (0.7–1.1 × 10³ nm²s⁻¹) (16,20). Diffusion constants measured with other diffusive kinesins lie in the same range as that determined for DK4mer (kinesin-13 MCAK: 38 × 10⁴ nm²s⁻¹ (31), and kinesin-8 kip3p: 0.43 × 10⁴ nm²s⁻¹ (32)). With ionic strength, the diffusion coefficient of DK4mer-GFP increased ∼4-fold, reflecting a weaker interaction between the motor and MT (Fig. 5 B).

Figure 5.

Diffusive motion of DK4mer-GFP on surface-immobilized MTs in 2 mM ADP at buffer ionic strengths up to 20 mM added KCl. (A) MSD of single DK4mer-GFP motors in buffers as listed in the legend, fitted by power laws (straight lines); N = number of analyzed motor traces. Gray line: power-law slope = 1. (B) 1D-diffusion constants of DK4mer-GFP motors determined from the data shown in (A) on MTs at 2 mM ADP (crosses) as a function of buffer ionic strength. The dotted line serves to guide the eye. Diffusion constant of DK4mer-GFP (circle, N = 11 pauses) during pauses between processive runs (marked with arrows in Fig. 3) with 2 mM ATP in BRB80⁺ buffer.

Truncated Dimer, DK511

To ascertain whether the tail domains are indeed responsible for the diffusive attachment, we generated a shortened dimeric construct on the basis of DK4mer-GFP, truncated at Eg5 residue 511 and thus lacking the Eg5 tail domains (DK511; Fig. 1 D). We chose the truncation site such that monomers should form dimers but not tetramers. It is reasonable to assume that the most C-terminal parts of the stalk before the tail, which normally are embedded in a tetrameric coiled-coil, do otherwise not influence the motility characteristics of the dimeric end. Single-molecule fluorescence assays confirmed that DK511 produced uninterrupted and unidirectional processive motility at an average velocity of ∼530 nm/s (Fig. 2 B and Table S1), similarly to DK4mer-GFP, and with an average run length of ∼2.8 μm (Table S1). These results support the following conclusions: 1) The Eg5 tail has no major influence on the velocity produced by the kinesin-1 motor domains, and thus the 40% reduction in velocity of DK4mer-GFP compared with kinesin-1 is not caused by friction introduced by the Eg5 tail domains, but instead may reflect a mismatch in the neck linker region, which can strongly influence the communication between the motor domains (33–35). 2) The pauses observed in DK4mer-GFP, but not in DK511 motility, are likely to be introduced by the Eg5 tail, as are the extraordinarily long run lengths of DK4mer-GFP. These results also show that the C-terminal GFP tags are not responsible for the diffusive attachment and also do not inhibit tail binding, consistent with the findings of Weinger et al. (17).

Motility of Single DK4mer-GFP Molecules at Different Ionic Strengths

Ionic strength can have a strong influence on the motility of motor proteins by modifying the motor-MT interaction (13,16,18,20). To further explore the influence of the tail-MT interaction, we analyzed the effect of increasing ionic strength on the motility of the DK4mer-GFP motor construct. DK4mer-GFP is better suited for such experiments than native Eg5 because transitions between diffusive pauses and runs are very clearly detectable. An increase of ionic strength and Debye screening is predicted to reduce the interaction strength because the tail domain likely interacts via electrostatic interactions with the C-terminus of tubulins (29,36–38).

Consistent with this expectation, increasing the ionic strength led to reduced overall run lengths in our experiments, from 9.6 ± 2.4 μm in P30⁺ buffer to 1.7 ± 0.8 μm in BRB80⁺ buffer plus 60 mM KCl (Figs. 3, 4C and D, and Table S1). The decrease in total run length, however, was caused mainly by shorter pauses with an increased likelihood for detachment, whereas the processive segments of runs appeared less affected (Figs. 3 A and 4, A and B; and Table S1). From 20 mM added KCl on, motility was purely processive with a run length of 2.6 ± 0.7 μm, comparable to that of D421 (2.9 ± 0.8 μm) and DK511 (2.3 ± 0.9 μm; Fig. 4 D and Table S1). We thus found that an increase in ionic strength of the buffer mostly reduced tail-MT interactions. This result is consistent with that reported by Weinger et al. (17) for Eg5. It is instructive to compare the lengths of processive runs of DK4mer-GFP at low salt with those of DK511 and D421, which do not have the second motor dimer (Fig. 4 D). The finding that those run lengths are about equal argues that the second pair of heads is not bound during processive runs, even at low ionic strengths. If binding of the second pair could switch the motor into a diffusive state (a possible scenario), processive run lengths of DK4mer-GFP should be shorter, and if binding of the second pair could support processive movement, run lengths should be longer than those of the dimeric motors. This result compares in an interesting way to the behavior of DNA-coupled artificial kinesin tetramers reported in previous studies (39,40), where it was found that run lengths were extended, but less than expected, due to negative interference between the two dimers. Possibly due to the rigid nature of the tetrameric coiled-coil stalk, this negative interference appears to be maximal for kinesin-5. Note that the run length (Fig. 4 D) shows a sharp switch-like rather than gradual transition from mixed motility at low ionic strength to purely processive motility at high ionic strength. Furthermore, increasing the ionic strength from P30 buffer to BRB80 buffer plus 40 mM KCl led to an ∼4-fold reduction in the duration of interactions in the presence of ADP (Fig. 3 B).

In ADP buffer, MSD analysis showed that DK4mer-GFP interactions with single MTs were purely diffusive (Fig. 5 A). Because the duration of attachment decreased with increasing salt, the statistics were not sufficient to calculate MSDs for buffers with >20 mM added KCl. Increasing the ionic strength led to a 4- to 5-fold increase in the diffusion constant, from 0.5 ± 0.04 × 10⁴ nm²/s to 2.2 ± 0.7 × 10⁴ nm²/s, consistent with what has been reported for Eg5 (16) and arguing for an electrostatic component (Fig. 5 B). Interestingly, however, the diffusion constant of DK4mer-GFP during pauses between processive runs at 2 mM ATP was significantly lower (0.4 ± 0.2 × 10⁴ nm²/s) than during ADP-diffusive interactions (1.2 ± 0.04 × 10⁴ nm²/s) under otherwise identical buffer conditions (Fig. 5 B, circle). These results suggest that the two diffusive modes of DK4mer-GFP, i.e., the one in the presence of ATP and the one in the presence of ADP, are determined by somewhat different interactions with the MT.

Increasing ionic strength had little effect on the average velocity of directed motility (Fig. S5). This supports the notion that, once the motor is moving, the kinesin-1 motor domains work independently of and undisturbed by the Eg5 tail. With increasing ionic strength, slow velocities appeared in the velocity distribution of DK4mer-GFP. This slow fraction is most likely explained by the limited resolution of the experiments. With increasing ionic strength, the relative effect of the inaccuracy of determining the beginning and end of runs will increase. For 40 mM added KCl, most of the runs are ≤5 s. This results in a spread of the distribution toward lower values.

Truncated Dimer DK511 at Different Ionic Strengths

To confirm the result that only the tail domain-MT interaction is influenced by ionic strength, we also measured the motility of DK511 at low ionic strength (P30⁺ buffer). We found purely processive motility with an average run length of 3.4 ± 0.4 μm, which is only ∼30% of the run length of DK4mer-GFP (9.6 ± 2.4 μm) under the same buffer conditions. Interestingly, the DK511 run length was very similar to the run length of the truncated dimeric kinesin-1 construct D421 of 2.7 ± 0.4 μm (Fig. 4 D and Table S1). This result confirms that the long run lengths found for DK4mer-GFP are caused by the Eg5 tail domains. In BRB80⁺ buffer, the run lengths for both D421 and DK511 were only slightly shorter than those obtained in low-ionic-strength buffer, confirming that the kinesin-1 motor domains are only slightly influenced by ionic strength.

Relative Sliding of MTs Driven by DK4mer-GFP at Different Ionic Strengths

We reported previously (16) that individual Xenopus laevis Eg5 motors are able to cross-link MTs and drive relative sliding at twice the velocity (2v) of an individual motor on a single MT. This velocity (2v model; Fig. 6 B) can only be reached when relative sliding is driven by the simultaneous action of both pairs of motor domains, with each one driving motility at the single-motor velocity v. We performed relative sliding assays with DK4mer-GFP as described previously (18) and found that DK4mer-GFP was able to cross-link and slide MTs (Fig. 6 A) at a higher than single-motor velocity (Fig. 6 C). This result confirms that DK4mer-GFP is a fully functional tetrameric motor.

Figure 6.

DK4mer cross-linking and sliding a mobile MT over a surface-attached MT. (A) Series of snapshots of a fluorescence recording showing TMR-speckle-labeled MTs in P30⁺ buffer with 70 mM KCl added. DK4mer motors were attached to the substrate and moved between the MTs. First, a long MT moved on the surface with velocity v. After 99 s, a short MT bound to a surface-attached MT and started moving at 2v (see also Movie S4). The distance moved by the MT in given time intervals is marked by arrows. (B) Cartoon of possible motor-binding situations when the sliding and resulting velocities are driven against the substrate: (a) motor on the substrate, trying to move mobile MT with v (v = motor velocity on the MT); (b) motors binding in the overlap region between the MTs, which themselves move with v while trying to move the short MT with 2v; and (c) motors attached to substrate moving the short MT with v. (C) Histogram of mobile MT velocities before (green) and during (red) overlap with the other MT from 10 independent sliding recordings.

To analyze the effect of ionic strength on the motility of DK4mer-GFP between cross-linked MTs, we performed relative sliding assays at sufficiently high motor concentrations to ensure that motility was driven by multiple motors. We found that relative sliding occurred over a wide range of salt concentrations (Fig. S6), even at high ionic strengths (BRB80⁺ buffer + 40 mM KCl, I = 200 mM) at which switching to tail-mediated binding on single MTs, i.e., diffusive intervals, were no longer seen (Fig. 4 D). This result suggests that at high ionic strength, motor-domain-driven interactions alone were able to maintain cross-linking and simultaneously drive relative sliding even with diminished tail binding.

The velocity of relative sliding at low ionic strength remained below the expected value of twice the single motor velocity (2v), but increasing the ionic strength shifted the distribution toward 2v (Fig. 6, B and C, and Fig. S6). The relative sliding speed was ∼700 nm/s in P30⁺ buffer, but ∼1100 nm/s in P30⁺ buffer plus 70 mM added KCl. A possible explanation for the reduced sliding velocity is drag induced by diffusive motors. This seems unlikely, though, because 1), motors appear to switch to largely processive motility when bound between two MTs; and 2), we observed no ionic-strength-dependent velocity in the surface gliding assays, which should be subject to the same drag. Thus, a more likely explanation for the reduced sliding velocity is that sliding MTs also interact with surface-adsorbed motors, which would slow them down. In P30⁺ buffer with 80 mM KCl added (I = 140 mM) and higher ionic strengths, we could not observe any relative sliding. In single-molecule assays, we could still see processive runs on single MTs in BRB80⁺ buffer + 60mM KCl (I = 220 mM). One possible explanation for this discrepancy is that the probability of a motor binding to a MT decreases at high ionic strength. One motor binding to two MTs simultaneously at high ionic strength then has an even lower probability, which might make it difficult to observe relative sliding.

Conclusions

Recently, much attention has been focused on the regulation of motor proteins in cellular contexts. Kinesin-5, as an essential actuator of the mitotic spindle, is a prime example of a highly regulated mechanoenzyme. Here, we used a kinesin-1 head/kinesin-5 tail chimera to obtain a well-behaved model system in which to investigate details of the multiple binding modes of tetrameric kinesins to MTs, which are believed to be involved in turning this motor on and off. Several earlier approaches used mutated and chimeric constructs, mainly in the experimental context of the mitotic spindle, observing modifications of its elongation dynamics (41,42). A chimera similar to ours was independently described by Cahu and Surrey (41) and used in in vivo experiments in Xenopus egg extracts. In those experiments, it was shown that the chimeric motors mislocalized in mitotic spindles and consequently caused spindle collapse into tightly bundled MT arrays.

Our tetrameric chimera exhibited fast, processive motility and was able to cross-link and slide MTs, as expected from the properties of the parent kinesins. Additionally, our results clearly show that alternating diffusive and processive episodes were separated by an energy barrier, with processive interaction being driven by the motor heads and diffusive attachment being dependent on the tails. This finding makes it tempting to speculate that tail binding may cause a conformation of the heads with respect to the MT that makes motility impossible, and, vice versa, that head binding brings the tail-binding sites either out of contact with the MT or into a position where they do not inhibit processive motility. It is an important point with regard to biological function that an additional interaction of the motor with the MT does not necessarily simply enhance processive motility. Our findings demonstrate a more subtle and versatile option for motor regulation to on the one hand localize the motor on the track by a nonspecific interaction, and on the other hand make the start of processive motility dependent on some further conformational change. The obvious way this might play out for kinesin-5 motors would be for the tail interaction to place the motor on one MT, where it could possibly (perhaps with the help of other factors, such as the bundling protein ase1) diffuse in search for the overlap with another MT, and then click into a processive motility state when it binds between two MTs. It has been shown in vitro and in Xenopus egg extract that phosphorylation of threonine 937 in the C-terminus of Eg5 strongly enhances MT binding (43). One can speculate that the cell could use phosphorylation as a further regulatory element in the control of these motors and, for example, turn on spindle assembly by switching on the kinesin-5 motors.

The robustness of the kinesin-1 motility of our chimeras contrasts with what was found for native Eg5, the heads of which appear to be particularly sensitive to regulation by the tails. In Eg5, the tail domain is required for proper and efficient orientation and function of Eg5 when it cross-links and orients MTs in vivo and in vitro (17). It is not clear yet whether the diffusive interaction mediated by the tail domains has a physiological function by itself, such as guiding motors to the spindle midzone without the expense of ATP. Further studies of specifically altered or chimeric constructs are necessary and will yield further insights into the structural changes involved in regulating the transition from the off to the on state of kinesin-5 motors.