Motile microtubule crosslinkers require distinct dynamic properties for correct functioning during spindle organization in Xenopus egg extract

The dynamic organization of the microtubule cytoskeleton depends on the activities of a variety of molecular motors, kinesins (Lawrence et al., 2004) and dynein (Numata et al., 2008). During cell division the cytoskeleton reorganizes, which results in assembly of the mitotic spindle that ultimately segregates the chromosomes into the two new daughter cells. Kinesins from different subfamilies play different roles in this self-organization process (Walczak and Heald, 2008). In a simple in vitro system it has been shown that the kinetic properties of microtubule crosslinking motors can determine the final structure of a self-organized motor-microtubule system (Surrey et al., 2001). However, to what extent differences in mechanochemical properties of mitotic kinesins, belonging to different subfamilies, determine their specific organizing function during spindle assembly in a physiological context is not understood.

Kinesin-5 is one of the main organizers of spindle microtubules during mitosis (Hoyt et al., 1992; Mayer et al., 1999; Sawin et al., 1992; Sharp et al., 1999b). It is a plus-end-directed, bipolar motor (Cole et al., 1994; Kashina et al., 1996; Sawin et al., 1992) with an architecture that allows it to crosslink and slide microtubules (Kapitein et al., 2005; Kashina et al., 1996; Miyamoto et al., 2004; Sharp et al., 1999a). Kinesin-5 members from a number of different species have been shown to be slow compared with other kinesins (Cole et al., 1994; Sawin et al., 1992) and to be only weakly processive (Valentine et al., 2006). Whether the characteristic motor properties of Kinesin-5 are required for its function, or whether the simple presence of plus-end-directed motors that can crosslink microtubules is sufficient to ensure its essential role in spindle formation is unknown. We have tested whether plus-end-directed microtubule crosslinkers, consisting of the major part of the Kinesin-5 sequence, but with motor domains from kinesins of other subfamilies (and hence with motile properties different from those of natural Kinesin-5) can substitute for natural Kinesin-5 function. This addresses the question of the robustness of spindle design with respect to variations in the mechanochemical properties of this crucial mitotic motor.

We generated chimeric motor constructs of Kinesin-5 by domain swapping. Our goal was to change the conserved properties of its motor domain without, however, changing its directionality, and to preserve the overall architecture of Kinesin-5, which is required for its crosslinking activity. Therefore, we replaced the motor domain of Xenopus laevis Kinesin-5 (also called Eg5) by the motor domains of kinesins from other subfamilies, leaving intact the Eg5 backbone, which is required for its tetrameric oligomerization state (Tao et al., 2006). We chose the Xenopus chromokinesin Kid (a Kinesin-10) (Antonio et al., 2000; Funabiki and Murray, 2000) as a source for an unprocessive motor that is a little faster than Eg5 (Brouhard and Hunt, 2005; Yajima et al., 2003), and Drosophila Kinesin-1 (Kin1) (Saxton et al., 1988) as a source for a very processive motor that is much faster than Eg5 (Block et al., 1990; Hancock and Howard, 1998). The replacement of the mechanically active part of Eg5 included, in addition to the motor domain, the neck-linker, a flexible stretch of amino acids that connect the motor domain with the backbone of the molecule and that is known to be important for the stepping mechanism of kinesins (Kalchishkova and Bohm, 2008; Tomishige et al., 2006; Vale and Milligan, 2000). The exact position for the exchange was determined from an alignment of multiple kinesin sequences (Fig. 1A). Both chimeric constructs were additionally tagged with a C-terminal green fluorescent protein (GFP), generating the two chimeric constructs Kid-Eg5-GFP and Kin1-Eg5-GFP. These constructs were expressed in insect cells and purified (Fig. 1B).

Despite their artificial nature, purified Kid-Eg5-GFP and Kin1-Eg5-GFP were active in microtubule gliding assays and moved microtubules at a velocity of 4.74 μm/minute and 19.5 μm/minute, which is about two times and eight times faster than wild-type Eg5, respectively (Fig. 1C). These velocities are in the range of the velocities of the motor domains in their native molecule (Hancock and Howard, 1998; Yajima et al., 2003), demonstrating that the original kinetic properties of the grafted motor domains were mostly maintained in the chimeric constructs.

To ensure that the bipolar architecture of wild-type Eg5 was indeed preserved in the engineered constructs, size-exclusion chromatography was performed at low and high salt concentrations. The elution volumes and Stokes radii of the wild-type and chimeric constructs were found to be similar (Fig. 1D) and in the range of tetrameric Eg5-GFP, as reported previously (Kwok et al., 2006). This shows that both Kid-Eg5-GFP and Kin1-Eg5-GFP are most probably tetramers like wild-type Eg5-GFP and that, as expected, the bipolar architecture of the constructs is conserved independently of the salt concentration in the buffer.

Consequently, like native Eg5, both chimeric constructs could bundle microtubules in vitro indicating that the microtubule crosslinking property of Eg5 (Kapitein et al., 2005; Kashina et al., 1996; Sharp et al., 1999a; Tao et al., 2006) was also preserved in these chimeric constructs (Fig. 1D). The ability to form bundles depended on the ionic strength in a manner characteristic for each construct. Whereas the bundles formed by Eg5 and Kin1-Eg5-GFP disassembled when the KCl concentration was increased from 50 mM to 150 mM, Kid-Eg5-GFP retained its bundling activity at 150 mM KCl. This indicates an increased affinity of Kid-Eg5 for microtubules, which is probably a consequence of the higher isoelectric point of the Kid motor domain as compared to the Eg5 or Kinesin-1 motor domains. These results demonstrate that the characteristic properties of the grafted motor domains were successfully preserved, generating novel bipolar motors with properties different from wild-type Eg5 regarding both motility and affinity for microtubules.

To test the importance of the specific properties of the Eg5 motor domain for spindle assembly, we replaced endogenous Eg5 with the chimeric motors in Xenopus egg extract and studied spindle assembly around sperm nuclei (Hannak and Heald, 2006). This was achieved by immunodepletion of Eg5 from the extract and subsequent addition of recombinant protein. We then studied the consequences of, effectively, a motor domain exchange on spindle formation (Fig. 2A). In control experiments, we first demonstrated that wild-type Eg5-GFP added at 400 nM (Kapoor and Mitchison, 2001) to Eg5-depleted extract rescues spindle formation (for an example of a spindle, see Fig. 2C bottom right), as reported previously (Cahu et al., 2008; Kwok et al., 2006; Uteng et al., 2008). Strikingly, although both chimeras are bipolar crosslinkers and plus-end-directed motors, bipolar spindles did not form when the chimeric motors were added to Eg5-depleted extract at this concentration. Monopolar asters that form in the absence of Eg5 activity (Mayer et al., 1999; Sawin et al., 1992) were also not observed. Instead, both chimeric constructs induced the formation of a new phenotype, namely collapsed spindles ultimately transforming into bundled microtubule structures that were mostly composed of a single microtubule bundle with one of the focussed ends of the bundle being connected to chromatin (Fig. 2C). In these bundles, Kid-Eg5-GFP and Kin1-Eg5-GFP localized along microtubules being enriched towards the end to which the chromatin was attached and where most probably microtubule plus-ends predominate (Fig. 2C). The length of these bundles was on average 50 μm, almost twice as long as normal spindles with an average length of 30 μm (Fig. 2C). In conclusion, although both plus-end-directed chimeric motors crosslink microtubules in mitotic egg extract, they are unable to promote normal spindle formation, instead causing a peculiar defective spindle phenotype, probably as a consequence of motor properties unsuitable for spindle assembly.

We then tested whether a change in the concentration of the chimeric motors could compensate for their different motor properties and restore spindle bipolarity. We varied the concentrations of Kid-Eg5-GFP and Kin1-Eg5-GFP that were added to Eg5-depleted extract within the range of 66 nM to 800 nM (Fig. 2D). Interestingly, whereas the percentage of monopolar spindles that predominate in the absence of Eg5 decreased with an increasing concentration of either of the added chimeric constructs, bipolar spindle formation could not be rescued at any of the concentrations tested (Fig. 2D). Instead, the percentage of bundled structures increased. This demonstrates that although Kid-Eg5-GFP and Kin1-Eg5-GFP are bipolar plus-end-directed motors their motor domains cannot substitute for the function of the motor domain of wild-type Eg5. Consequently, bipolarity and plus-end-directed motor activity of a microtubule crosslinking motor alone are not sufficient for promoting spindle assembly in Xenopus egg extract. The genuine properties of the motor domain of Eg5 appear therefore to be fine-tuned for its conserved organizational function in spindle assembly.

We next added 400 nM chimeric constructs to preformed spindles in Xenopus egg extract in order to test whether they can override the action of endogenous Eg5 (Fig. 3A,B). Spindles were fixed at different times after addition of chimeric constructs and observed by fluorescence microscopy. Interestingly, both chimeric constructs caused the destruction of bipolar spindles transforming them into structures that were again very different from the typical monoasters that usually form in the absence of Eg5. At early time points after addition of chimeric constructs, distorted bipolar spindles predominated. With time, the percentage of these distorted spindles decreased and correspondingly, in the case of Kid-Eg5-GFP addition, the percentage of both collapsed spindles and bundles increased (Fig. 3C, left), whereas in the case of Kin1-Eg5-GFP addition, only an initial small fraction of collapsed spindles could be detected and bundles dominated already at early time points (Fig. 3C, right). This indicates that collapsed spindles are an intermediate structure transforming later into tightly bundled structures, possibly by fusion of the two half-bundles in a collapsed spindle. These final bundled structures resembled the ones formed after the addition of the chimeric construct to Eg5-depleted extract. Interestingly, the time of transformation of distorted spindles into final bundled structures was different for the two chimeric constructs, being faster for Kin1-Eg5-GFP than for Kid-Eg5-GFP (Fig. 3C). This probably reflects the difference in velocity between these two chimeric constructs (Fig. 1C), indicating that excessive antiparallel microtubule sliding contributes to spindle destruction because of the unnatural motor properties of the chimeric molecules.

Because addition of Kid-Eg5-GFP transformed spindles more slowly than addition of Kin1-Eg5-GFP, distorted spindles could be observed for about 10 minutes after the addition of the slower of the two chimeric constructs. Closer characterization of these distorted spindles promised to provide insight into the mechanism of spindle collapse. We therefore measured their length after addition of Kid-Eg5-GFP during the first minutes before they transformed into collapsed spindles. We found that distorted bipolar spindles elongated continuously, leading to a 50% increase in spindle length after 9 minutes (Fig. 4A). Control spindles, to which only buffer was added, stayed constant in length. This observation suggested an enhanced activity of chimeric Eg5 in the spindle midzone where Eg5 most probably slides overlapping antiparallel microtubules apart.

Therefore, we investigated the distribution of Kid-Eg5-GFP and of wild-type Eg5-GFP along the spindle axis 5-7 minutes after their addition to preformed spindles. Strikingly, Kid-Eg5-GFP preferentially localized to the spindle midzone, whereas Eg5-GFP was enriched at spindle poles (Fig. 4B), as reported previously for endogenous Eg5 (Sawin et al., 1992). These observations support the suggestion that the abnormal accumulation of Kid-Eg5-GFP in the midzone contributes to spindle elongation by excessive antiparallel microtubule sliding (Fig. 4A). Taken together, our results indicate that the pathway of spindle collapse in the presence of the chimeric Eg5 leads from normal bipolar spindles via elongating distorted spindles to collapsed spindles with chromatin in the center, which then completely collapse into bundled microtubule structures with the chromatin mass at one end of the bundle (Fig. 4C).

Our observations have demonstrated that targeting Eg5 variants with drastically different motor domain properties to spindle microtubules is incompatible with spindle assembly and maintenance in Xenopus egg extract. Hence, the microtubule crosslinking activity and plus-end directionality of Kinesin-5 alone are not sufficient for the function of Eg5. Which exactly are the properties of the Eg5 motor domains that need to be conserved to ensure spindle formation still needs to be elucidated. Both excessive binding to spindle microtubules as a result of an increased affinity for microtubules, as in the case of the Kid motor domain, and an increased velocity, as in the case of the Kinesin-1 motor domain, might contribute to the internal imbalance of the spindle. There might therefore be an upper limit for the velocity of Eg5 and also for the amount of Eg5 in the midzone that is compatible with spindle maintenance. Interestingly, single point mutations in the Eg5 motor domain that reduced its speed were previously shown to affect only the assembly time, but not the efficiency of bipolar spindle formation or the spindle length (Kwok et al., 2004). Moreover, it is also possible that force-induced unbinding from the microtubule that appears to be typical for Eg5 (Valentine et al., 2006), in contrast to the force-induced stall that is characteristic for example for Kinesin-1 (Svoboda and Block, 1994), is an important requirement for spindle maintenance. Finally, the motor domain of Eg5 might also specifically interact with other binding partners, as was recently proposed (Eckerdt et al., 2008).

In conclusion, we have shown that the properties of the motor domain of one of the most important motors for spindle assembly appear to be fine-tuned for the specific task of this motor during spindle assembly in Xenopus egg extract. Changing these properties by motor domain swapping leads to a characteristic spindle defect that is caused by unnatural interactions of the re-engineered motile crosslinkers with microtubules. Our results illustrate how the conservation of motor domain sequence, and hence of the mechanochemical properties that are typical for individual kinesin subfamilies, might be related to conserved function of these motors.

Wild-type and GFP-tagged Eg5 were cloned as described (Uteng et al., 2008). Chimeric Eg5 molecules consisted of the first 343 amino acids of Drosophila melanogaster Kinesin 5 and the first 369 amino acids of Xenopus laevis Kid fused to the last 698 amino acids of Xenopus laevis Eg5 followed by the GFP sequence. In order to generate Kid-Eg5, the Kid motor domain was amplified by PCR using 5′-GCTGCCATGGTTCTTACTGGGCCTCTC-3′ as 5′ primer and 5′-GGTGGTTTCCTGGCTGAAAG-3′ as 3′ primer, introducing a NcoI site at the 5′ end. The stalk and tail region of Eg5 was amplified by PCR using 5′-ACAAAGAAGGCACTCATCAAGGAG-3′ as 5′ primer and 5′-TTCGAAAGCGGCCGCTC-3′ as 3′ primer, respectively, introducing a NcoI site at the 3′ end. The PCR products were digested, phosphorylated and ligated into pFastBacHTa vector (Invitrogen) thereby connecting also the blunt 3′ end of the Kid motor domain and the blunt 5′ end of the Eg5 fragment, generating Kid-Eg5. To introduce a C-terminal GFP, a fragment encoding for GFP was inserted using the SacI and XhoI sites, generating Kid-Eg5-GFP. The same cloning strategy was used to generate Kin1-Eg5-GFP. The primers used for the Kin1 motor domain were 5′-GCTGCCATGGCAATGTCCGCGGAACGAGAG-3′ and 5′-AAGCTCCTCGTTAACGCAG-3′.

All Eg5 constructs were expressed in Sf9 insect cells and purified essentially as described (Cahu et al., 2008). Chimeric constructs were purified using a Protino Ni-TED resin (Macherey-Nagel, Düren, Germany) and buffers containing 350 mM KCl. The final storage buffer was 350 mM KCl, 50 mM imidazole, 0.5 mM EGTA, 10% sucrose, 5 mM β-mercaptoethanol, pH 7.0). Size-exclusion chromatography of the purified Eg5 proteins was performed using a Superose 6 (GE Healthcare) in storage buffer containing in total either 50 mM or 350 mM KCl. Two runs were performed per construct and condition. The measured elution volumes were averaged. Polyclonal anti-Eg5 antibody was purified as described (Cahu et al., 2008) and tubulin was purified as described (Castoldi and Popov, 2003).

A flow chamber with a volume of about 5 μl was filled first with BRB80 (80 mM PIPES, 1 mM MgCl₂, 1 mM EGTA, pH 6.8) and then with motor mix (3 μM motor protein and 10 mM MgATP). After incubation on ice for 5 minutes, motor mix was replaced by wash buffer (10 mg/ml BSA, 10 mM MgATP in BRB80) and after 5 additional minutes by microtubule mix [Alexa-Fluor-568 (Invitrogen)-labeled microtubules (Cahu et al., 2008) in BRB80 supplemented with 10 mg/ml BSA, 10 mM MgATP, 0.25 mg/ml glucose oxidase, 0.12 mg/ml catalase, 25 mM glucose, 10 mM β-mercaptoethanol] at room temperature.

Time-lapse epifluorescence microscopy was performed on a wide field Zeiss AxioVision Cell Observer (63′ oil-immersion objective, AxioCam MRm camera, 5- or 2-second time intervals, 100 or 300 mseconds exposure time). Microtubule gliding velocities were measured from kymographs using ImageJ. Average velocities were obtained from Gaussian fits to velocity distributions.

Bundling experiments were performed by mixing Alexa-Fluor-568-labeled microtubules with 100 nM of Eg5 construct and 5 mM MgATP in storage buffer containing in total either 50 mM KCl, 150 mM KCl or 250 mM KCl. After 2 minutes samples were placed between two coverslips and observed by epifluorescence microscopy.

The preparation of cytostatic factor arrested extract (CSF extract) from Xenopus laevis eggs (Murray, 1991) and spindle reconstitution experiments were performed as described (Cahu et al., 2008). For the analysis of spindle structures in Figs 2 and 4, spindles were fixed and centrifuged onto coverglasses (Cahu et al., 2008). For the time course experiments shown in Fig. 3, spindles were first allowed to form for 30 minutes, then chimeric motors were added and the extract was incubated for various times as indicated before being mixed with fixing solution (48% glycerol, 11.1% formaldehyde, 5 mg/ml Hoechst). Images were taken directly using an epifluorescence microscope. For the quantification of the percentages of bipolar spindles, monoasters, distorted spindles, collapsed spindles and bundled structures, randomly selected structures were counted from at least three different experiments and the results per condition were averaged. `Collapsed spindles' were defined as bipolar structures with two tight microtubule bundles of similar length symmetrically connecting a mass of chromatin (in contrast to normal spindles that are wide in the center and focused at the poles); the angle between the two bundles of a collapsed spindle measured at the chromatin was at least 45°C. By contrast, `bundled structures' were composed either of a single bundle or several bundles of varying length. Structures with two bundles of differing length or with an angle of less than 45°C between them were considered as `bundled structures', as illustrated in Fig. 3C. Average spindle lengths in Fig. 4A were obtained from at least six structures. For the measurement of the distribution of GFP tagged Eg5 constructs in spindles, fluorescence intensity profiles were generated from line scans along the spindle axis.