S. pombe Kinesins-8 Promote Both Nucleation and Catastrophe of Microtubules

The kinesins-8 were originally thought to be microtubule depolymerases, but are now emerging as more versatile catalysts of microtubule dynamics. We show here that S. pombe Klp5-436 and Klp6-440 are non-processive plus-end-directed motors whose in vitro velocities on S. pombe microtubules at 7 and 23 nm s−1 are too slow to keep pace with the growing tips of dynamic interphase microtubules in living S. pombe. In vitro, Klp5 and 6 dimers exhibit a hitherto-undescribed combination of strong enhancement of microtubule nucleation with no effect on growth rate or catastrophe frequency. By contrast in vivo, both Klp5 and Klp6 promote microtubule catastrophe at cell ends whilst Klp6 also increases the number of interphase microtubule arrays (IMAs). Our data support a model in which Klp5/6 bind tightly to free tubulin heterodimers, strongly promoting the nucleation of new microtubules, and then continue to land as a tubulin-motor complex on the tips of growing microtubules, with the motors then dissociating after a few seconds residence on the lattice. In vivo, we predict that only at cell ends, when growing microtubule tips become lodged and their growth slows down, will Klp5/6 motor activity succeed in tracking growing microtubule tips. This mechanism would allow Klp5/6 to detect the arrival of microtubule tips at cells ends and to amplify the intrinsic tendency for microtubules to catastrophise in compression at cell ends. Our evidence identifies Klp5 and 6 as spatial regulators of microtubule dynamics that enhance both microtubule nucleation at the cell centre and microtubule catastrophe at the cell ends.


Introduction
Microtubule dynamics allow cells to rapidly assemble, remodel, or disassemble polarized arrays of microtubules. Pure tubulin in vitro shows intrinsic dynamic instability, whereby microtubules spontaneously nucleate, grow steadily, and then spontaneously and rapidly depolymerise [1]. Dynamic instability is well described by four parameters: growth rate, shrinkage rate, catastrophe frequency (how often microtubules switch from growth to shrinkage) and rescue frequency (how often they switch from shrinkage to growth) [2]. In cells, these parameters are all heavily regulated [3].
Amongst the regulators are members of the kinesin-13, -14 and -8 families, which modulate the catastrophe frequency [4]. MCAK and related kinesins-13 can diffuse along the microtubule lattice to the growing microtubule tip and drive tubulin subunits to dissociate, either alone or in complex with the MCAK [5], driven by an ATPase cycle that is distinct from that of translocating kinesins [6]. Kar3, a kinesin-14, is a minus end directed translocase that heterodimerises with Cik1, targets the plus ends of taxol-stabilised microtubules, and depolymerizes them at a rate dependent on the taxol concentration [7]. The kinesins-8 are plus end-directed translocases [8,9,10,11,12], at least one of which, S. cerevisiae Kip3, can step processively along the lattice of GMPCPP microtubules to their plus ends, where it enhances the off-rate of GMPCPP tubulin heterodimers [10]. The tail of Kip3 contains a microtubule and tubulin-heterodimer binding site that enhances its processivity and microtubule end binding [13]. Although, like MCAK, Kip3 ATPase is stimulated by tubulin heterodimers [9] the mechanism of depolymerisation is different. Accumulation of Kip3 at MT ends depends on Kip3 translocase activity and requires cooperation between Kip3 molecules, in that Kip3tubulin complexes are displaced from the microtubule tip by the arrival of another Kip3 [14]. The cell biology of Kip3 is consistent with this type of length-dependent catastrophe mechanism operating in vivo. The velocity of Kip3 in vivo is 47-73 nm s 21 [9,10], similar to its single molecule velocity of 50 nm s 21 on brain microtubules in vitro [10], and is sufficient to account for its accumulation at the ends of microtubules growing at 23 nm s 21 in vivo [9,15]. Deletion of kip3 leads to unusually long spindles [16] [17], longer cytoplasmic microtubules, effects on chromosome congression [18] and a decrease in microtubule catastrophe frequency [9] consistent with microtubule depolymerase activity. However in vivo Kip3 also increases microtubule growth rate, rescue frequency and pause duration whilst decreasing shrinkage rate, suggesting Kip3 has a wide range of effects on dynamic microtubules [9]. The tail of Kip3 is required in vivo for the increase in microtubule rescue frequency and reduction in microtubule shrinkage rate. These effects may result directly from binding of the Kip3 tail to microtubules since in vitro the tail reduces the shrinkage rate of GDP microtubules [13].
The cell biology of other kinesins-8 is also only partially consistent with their having solely a microtubule depolymerase activity. Some observations are consistent with depolymerase activity. RNAi knockdown of Klp67a in Drosophila S2 cells produces unusually long spindles [19] as do Klp67a mutants in Drosophila embryos [20] [21] and mislocalisation of Klp67a to the cytoplasm of interphase cells causes a shortening of microtubules [22], all consistent with depolymerase activity. However in Drosophila primary spermatocytes, although Klp67a destabilizes microtubules during pre-anaphase, it is subsequently required to stabilize the central spindle [23] suggesting Klp67a can have both stabilizing and destabilizing effects. Kif18a, a mammalian kinesin-8, accumulates in an ATP-dependent manner at the plus ends of kinetochore microtubules [11,24], and affects the dynamics of kinetochore oscillations [11,24,25]. Like Kip3, processivity and microtubule end binding by Kif18a depends on a microtubule binding site in the Kif18a tail that is essential for its effects in vivo [26,27]. Kif18a depletion was originally reported to slow chromosome movement [11], but subsequent studies suggest that chromosome movement actually speeds up [24,25]. Stumpff and colleagues [20] found that Kif18a increased the frequency of kinetochore directional switching, whereas Jaqaman and colleagues [21] found no such effect but suggested instead that Kif18a may promote depolymerisation specifically on the trailing face of the kinetochore to slow chromosome movement. Over-expression of Kif18a in the interphase cytoplasm, in addition to increasing the catastrophe frequency as expected for a depolymerase, also increases microtubule growth rate, rescue frequency and the duration of pauses whilst slowing shrinkage, overall reducing microtubule dynamicity [28]. Intriguingly, although Kif18a depletion produces long microtubules, inhibition of its motor activity by a drug does not [29], suggesting that long microtubules may not arise from loss of Kif18a function but from loss of its interactions with another protein such as CenpE [30]. Depletion of Kif19, a kinesin-8 present in higher eukaryotes, produces monopolar spindles [31] but not other Kif18a depletion phenotypes [19,32]. The human kinesin-8 Kif18b destabilises microtubules during mitosis, targeting microtubule ends through a combination of motor activity and an EB1 binding site in its tail [33]. However, Kif18b has no significant direct effect on microtubule dynamics, but rather acts by transporting the kinesin-13 MCAK to microtubule ends and forming a complex with EB1 to enhance its end binding [34].
The fission yeast S. pombe has two kinesins-8: Klp5 and Klp6, and no kinesins-13 [35], suggesting it may rely heavily on Klp5 and Klp6 to drive microtubule dynamics. Klp5 and Klp6 are both essential for meiosis [36] but non-essential during interphase and mitosis. Deletion of either Klp5 or Klp6 alone or in combination produces unusually long interphase microtubules [36,37], suggesting that both Klp5 and Klp6 are depolymerases. In mitosis, Klp5 and Klp6 enter the nucleus separately [38] and then localise to the kinetochores and the spindle midzone [36,37]. Deletion of either Klp5 or Klp6 produces longer mitotic spindles [39], a delay in establishing the metaphase plate, lagging chromosomes and chromosome mis-segregation [37,39,40,41]. Some of these effects may arise from the role of Klp5 and 6 in chromosome attachment [40]. The non-motor domains of Klp5/6 also form part of a complex important for spindle checkpoint silencing that does not depend on Klp5/6 motor activity [42]. Tischer et al [43] carefully mapped the position of interphase microtubule catastrophes in S. pombe and found that they largely occur at cell ends, with a frequency that depends on Klp5/6 activity. They also found that Klp5/6 accumulate at microtubule ends, and that catastrophe frequency increased with increasing microtubule length, consistent with a Kip3-like depolymerase mechanism [43,44]. However, as well as increasing catastrophe frequency, Klp5/6 in vivo is also reported to increase the rescue frequency and growth rate [38]. Overall, the effects of kinesin-8 on interphase microtubules in budding yeast [9], fission yeast [38] and human cells [28] suggest that different members of the kinesins-8 family can increase not only the catastrophe frequency but also the rescue frequency, growth rate and duration of pauses of dynamic microtubules in vivo [28]. Furthermore, these phenomena can arise through either direct or indirect effects of the kinesins-8.
Resolving the mechanisms responsible for the diverse in vivo properties reported for the kinesins-8 requires in vitro reconstitution experiments to distinguish direct from indirect effects. Perhaps influenced by Kip3 work, most reconstitution experiments so far have focused on depolymerisation of brain microtubules stabilized with taxol or GMPCPP. Kip3 depolymerizes GMPCPP brain microtubules [9,10,13,14] but any effect on dynamic microtubules remains to be demonstrated in vitro. For Kif18a the ability to depolymerise GMPCPP stabilised microtubules in vitro is itself controversial [11] [28]. Kif18a also fails to depolymerise Taxol stabilised microtubules, although in the presence of a nonhydrolysable ATP analogue it does sequester tubulin into ring structures similar to those formed by MCAK, suggesting it shares some of MCAKs functional as well as structural features [45]. In assays on dynamic microtubules Kif18a has no effect on the stability of existing microtubules [28], but does block polymerisation of microtubules by acting as a capping protein [26,28]. In vitro experiments on full-length Klp5 and 6, expressed as a complex in baculovirus, revealed plus-end-directed sliding of brain microtubules at 39 nm s 21 , whilst Klp6 motor domain alone drives sliding at 56 nm s 21 [12]. Klp5/6 is reported to have no in vitro depolymerase activity, either on GMPCPP or Taxol stabilised brain microtubules or on shrinking GDP brain microtubules [12].
Since the biochemical behaviour of kinesin can be different for microtubules from different species [46] we have examined the effect of Klp5/6 on dynamic S. pombe microtubules, both in vitro and in vivo. Our data suggest a new working model for Klp5/6 in vivo. We propose Klp5/6 tubulin complexes initially promote the birth of new microtubules, and thereafter continuously land on the growing microtubule tip, attempting to keep pace with the tip as it grows, but only succeeding at cell ends, where the microtubule tips lodge and their growth slows down in compression. Only then, within the context of the cell end, does Klp5/6 promote microtubule catastrophe. This mechanism, which allows Klp5/6 both to promote nucleation and to amplify an intrinsic tendency for microtubule tips to catastrophise under compression, may have relevance to other kinesins-8.

Results
Klp5 436 and Klp6 440 are both plus-end-directed microtubule translocases Since it was unclear if Klp5 is, like Klp6, a microtubule translocase, we expressed Klp5 and 6 alone and in combination. We purified co-expressed full-length Klp5 and Klp6 or full-length Klp6 alone from E. coli, but were unable to obtain full length Klp5 alone. Co-expressed proteins were purified by sequential affinity purification using the different tags on Klp5 and 6. After the final affinity purification the ratio of Klp5 to Klp6 present was not 1:1, suggesting that Klp5 and 6 can form a complex of variable composition ( Figure S1). We found evidence that the tail of full length Klp6 mediates an auto inhibition with respect to tubulin heterodimer binding (see below and Table S1), similar to the fulllength Klp5/6 complex [12]. In order to address the individual biophysical mechanisms of Klp5 and Klp6 with respect to microtubule dynamics, we focused on two tail-less constructs, Klp5 436 GST and Klp6 440 His. Both constructs contain the motor head domain plus the putative coiled coil dimerisation domain ( Figure 1A). We expressed these two constructs separately in bacteria, and purified them by tag-affinity chromatography ( Figure 1B). Superose 12 gel filtration of Klp6 440 His confirmed that it forms homodimers ( Figure 1C). A construct equivalent to our Klp6 440 His could rescue an in vivo klp6 deletion [47], showing that this short Klp6 construct is functionally competent in vivo.
Microtubule sliding assays using polarity-marked taxol-stabilized microtubules assembled from mammalian brain tubulin showed that both Klp5 436 GST and Klp6 440 His are plus end directed motors (Movies S1 and S2). Sliding velocities for GMPCPP stabilized S. pombe microtubules were 6.563.6 nm s 21 with Klp5 436 GST and 23612 nm s 21 with Klp6 440 His (Table 1, Figure S2). These velocities predict that in vivo, Klp5 and Klp6 will be too slow to keep pace with the plus ends of interphase microtubules growing at ,50 nm s 21 [48]. Klp6 could slide taxolstabilized brain microtubules at 87 nm s 21 (Table 1), showing that our preparation can in principle move faster. Our value for Klp6 on brain microtubules is on the order of the fastest reported velocity for a kinesin-8 of 174 nm s 21 for Kif18a [13] (Table S2). Despite Klp5 having about 4-fold slower microtubule sliding velocity than Klp6, the V max and K m for the S. pombe microtubuleactivated ATPases of Klp5 436 GST and Klp6 440 His are similar ( Table 2).

Multivalent Klp6 beads are processive
To determine if Klp6 homodimers can move processively, we used beads coated with Klp6. An optical trap was used as a micromanipulator to place Klp6 beads on to immobilized microtubules. Beads carrying multiple copies of Klp6 440 His translocated smoothly towards the plus ends of taxol-stabilized brain microtubules at 42610 s (6) nm s 21 ( Figure 2, Table S3). On reaching the microtubule plus ends, the Klp6 440 His beads dwelt for 42624 s (mean 6 SD, n = 6) before dissociating ( Figure 2, Table S3). Beads prepared at lower concentrations of Klp6 440 His failed to attach to microtubules. We obtained similar results in microtubule sliding assays on surfaces coated with Klp6 440 His. Reducing the motor density on the surface caused an abrupt transition from smooth microtubule gliding without angular fluctuations at a motor density of 6, 900 mm 22 to no microtubules attaching to the surface at 5, 500 mm 22 . With a kinesin-1 motor rkin430 [49] at a surface density of 300 mm 2 microtubule movement with angular fluctuations and detachment at microtubule ends was observed, consistent with microtubules sliding over a low density of processive motors [50,51]. These data suggest that unlike Kip3, Klp6 440 His homodimers are nonprocessive and need to operate in a team of linked multiple motors in order to move processively along microtubules, as has been observed with nonprocessive kinesin constructs attached to beads [52,53].
Klp5 and Klp6 do not affect the in vitro dis-assembly of GMPCPP-brain microtubules Klp5 or Klp6 deletion strains of S. pombe contain unusually long microtubules [36,37], suggesting that Klp5 and 6 destabilise microtubules. To test this hypothesis, we looked for effects of Klp5 and Klp6 on microtubules in vitro. Kip3 is known to accelerate the off-rate of tubulin from GMPCPP-stabilised brain microtubules [10,14] but it is unclear if this occurs with other kinesins-8 [11,28]. We found no significant effect of either Klp5 436 GST or Klp6 440 His or combinations of them on the shrinkage rate of pig brain GMPCPP-microtubules by direct observation in the  microscope, by light scattering in solution, or in quantitative sedimentation assays ( Figures S3, S4, S5, Table S4). Under the same conditions, our positive control, the bona fide depolymerase MCAK, did show a robust depolymerase activity ( Figure S3, S5b). We conclude, in agreement with observations on complexes of full length Klp5/6 [12], that Klp5 436 GST and Klp6 440 His do not accelerate depolymerisation of GMPCPP stabilised brain microtubules in vitro.

Klp5 and Klp6 do not affect the in vitro dynamic instability of S. pombe microtubules
There are no reports on the influence of kinesin-8 motors on dynamically unstable microtubules in vitro. We used direct observation of dynamically unstable, unlabelled S. pombe microtubules to assess the effect of Klp5 and Klp6 on microtubule dynamics. In extensive experiments, we found no significant effect on microtubule growth, shrinkage, catastrophe or rescue at either the fast or slow growing ends of dynamic S. pombe microtubules ( Figures S6, S7, S8, Tables S5, S6, S7, S8, S9, S10). The only exception was Klp5 436 GST, which at very high concentrations corresponding to almost a 1:1 molar ratio of motor heads to tubulin heterodimers caused a reduction in the shrinkage rate at the faster growing (plus) end (Table S9). Over-expression of a construct with two Klp5 heads in vivo produced a similar effect [38] suggesting that Klp5 activity in our in vitro assays is similar to its activity in vivo at high expression levels. Intriguingly, Kip3 at physiological levels also slows shrinkage in vivo [9].

Klp6 forms stable complexes with tubulin heterodimers
We examined the ability of Klp6 440 His to form complexes with tubulin heterodimers using gel filtration chromatography. In the presence of the slowly hydrolysable ATP analogue AMPPNP the single protein peak eluted earlier from the column than with tubulin alone (figure 3a). SDS-PAGE analysis of the eluted fractions showed that in addition to the a1 and b S. pombe tubulins, Klp6 440 His was also present in the peak (figure 3b). Quantification of the fluorescently stained gels and correction for differences in the protein staining showed that Klp6 440 His formed a stable complex with tubulin heterodimers with ,1.0 tubulin heterodimer bound to each kinesin head in the Klp6 440 His homodimer.

Tubulin heterodimers competitively inhibit the microtubule stimulated ATPase of Klp5 and 6
To learn more about the mechanisms by which Klp5 and Klp6 interact with tubulin and microtubules, we examined the tubulinactivated and microtubule-activated ATPases. Grissom [12] found that full length Klp5/6 had no tubulin-activated ATPase. We confirmed this is also so for full length Klp6 (Table S1). However, our truncated constructs Klp5 436 GST and Klp6 440 His do have tubulin-activated ATPases ( Table 2), suggesting that full length Klp5/6 self-inhibits. Kip3 and Kif18a have substantial tubulinactivated ATPase activity and bind more tightly to tubulin than to microtubules [9] [11]. In contrast Klp5 436 GST and Klp6 440 His bind more tightly to S. pombe microtubules than to S. pombe tubulin ( Table 2). The V max for Klp5 ATPase is 8 fold lower with tubulin than with microtubules whilst the V max for tubulin activation of Klp6 is 250 fold lower than with microtubules. These values predict that in a solution of dynamic microtubules, Klp5 and Klp6 will proportionate between the tubulin-bound and microtubule lattice-bound pools. In assays of microtubule stimulated ATPase activity, addition of tubulin heterodimers enhances the total ATPase activity of Klp5 436 GST, whilst addition of tubulin heterodimers inhibits the total ATPase activity of Klp6 440 His (fig. 4a), with K i of 175 nM for pig brain and 121 nM for S. pombe tubulin ( Figure 4B, C). This inhibition can be explained by sequestration of Klp6 440 His from the microtubule lattice by free tubulin heterodimers. The re-proportionation of Klp6 440 His between lattice-bound and tubulin-bound pools requires only a few tens of seconds to complete following a perturbation ( Figure  S10).

Klp5 and Klp6 drive nucleation of new S. pombe microtubules in vitro
Whilst the in vitro dynamic instability parameters of S. pombe microtubules were unaffected by Klp5 or Klp6, it was clear from our experiments that Klp5 and 6 motors very strongly stimulated the formation of microtubules in free solution, independent of the axoneme nucleation centres used in these experiments ( Figure S6c, S7c, S8d). We investigated this effect further using GMPCPP to stabilise microtubules under conditions where nucleation is limiting. Addition of Klp6 440 His under these conditions caused the formation of numerous short microtubules (Figure 5a), at a rate dependent on the Klp6 440 His concentration (Figure 5b, c). We could mimic the effect of Klp6 440 His by adding stabilized seeds (very short pre-formed microtubules) to a duplicate sample of tubulin (data not shown), indicating that Klp6 440 His stimulates microtubule assembly by inducing the formation of nuclei at tubulin concentrations where nucleation is normally limited. Klp5 had a similar, though less potent, effect ( Figure S8). Both motors promoted microtubule nucleation in both the presence and absence of ATP (not shown). To ensure the nucleation effect was not caused by protein aggregates that might form during freezing and storage of the proteins, Klp6 440 His was gel filtered and an aliquot taken from the peak of monodisperse protein ( fig. 6a). This repurified protein produced both a dose-dependent increase in the number of microtubules formed from S. pombe tubulin and a dosedependent reduction in mean microtubule length ( fig. 6b, c), consistent with it promoting microtubule nucleation. We considered two possible classes of nucleation mechanism. First, Klp5 and Klp6 binding to tubulin heterodimers might alter their conformation to favour assembly. Second, dimeric Klp5 and Klp6 might link tubulin heterodimers together to stabilise nascent microtubule nuclei. To distinguish these possibilities, we tested Klp6 415 His, a single-headed Klp6 construct ( fig. 1c), which has microtubule stimulated ATPase activity similar to Klp6 440 His. Klp6 415 His did not detectably enhance microtubule nucleation (data not shown), showing that the binding of Klp6 heads to free tubulin does not in itself drive microtubule nucleation. Instead, enhancement of nucleation requires a dimeric construct that can then link two tubulin heterodimers.
We also found that double headed Klp5 and Klp6 constructs, either individually or when co-expressed, could crosslink microtubules and cause bundling of both dynamic microtubules ( Figure  S6c, S7c, S8d) and of stable preformed GMPCPP-microtubules   Figure S9), similar to the bundling reported for full length Klp5/6 [12]. Bundling was absent when microtubules were assembled by adding Klp5/6 to free tubulin ( Figure 5, 6), possibly because both tubulin binding sites are then occupied, leaving no sites free to support microtubule cross linking. Single-headed Klp6 415 His does not drive bundling ( Figure S5) suggesting that microtubule crosslinking occurs by the same mechanism that permits binding of two heterodimers.

Klp5 & Klp6 stimulate microtubule catastrophes in vivo
To learn more about the mechanisms of full length Klp5/6 in vivo and in context, we revisited live cell imaging of Klp5/6 deletion mutants, taking data at higher time resolution than has previously been achieved (Movies S3, S4, S5). Catastrophe events in wild type S. pombe cells occur almost entirely in the end zone of the cell [43,48,54]. Figure 7A is a cumulative frequency plot showing the fraction of cell-end-resident microtubule tips that undergo catastrophe within a specified period. The data reveal a clear, statistically significant difference between microtubule lifetimes in the end zone of wild type cells and Dklp5 cells, and between wild type and Dklp6. In both deletion strains, the half-life (the time for 50% of microtubules in contact with the cell end to undergo catastrophe) increased from 29 to 42 seconds (p,0.05). The median dwell time of microtubule tips at cell ends increased from 36 seconds in wild type to 52 seconds in both Dklp5 and Dklp6 ( Figure 7B). A similar trend was reported by Unsworth et al. [38]. We thus confirm that in vivo and within the context of the end zone of wild type S. pombe cells, Klp5 and Klp6 are accelerators of microtubule catastrophe.

Deletion of Klp6 reduces microtubule number in vivo
We found that deletion of klp6 also causes a reduction in the number of interphase microtubule assemblies (IMAs [48]) from 2.860.1 (mean 6 SEM, n = 77) in wild type to 2.460.1 (n = 89) in klp6 deletions (P = 0.01), whilst deletion of klp5 had no significant effect (2.760.1, n = 48) (Figure 7c, d, e). This reduction in IMA number in vivo is consistent with our observation that Klp6 had a strong nucleation effect in vitro whilst the effect of Klp5 was weaker. IMAs normally have a bipolar microtubule arrangement so that microtubules depolymerise to the centre of the cell [48,55,56]. However in Dklp6 mutants we found that some depolymerisations continued along the entire cell length (Movie S6), suggesting a disrupted IMA structure with unipolar microtubules as observed in Dklp2 [57].

Klp5/6 promote microtubule depolymerisation in vivo but not in vitro
Our observations show that Klp5/6 destabilise interphase microtubules in S. pombe cells by reducing the time between the growing microtubule contacting the cell end and undergoing catastrophe, similar to effects reported in previous studies [38,43]. What remains unclear is the underlying molecular mechanism. Acceleration of GMPCPP-stabilised microtubule depolymerisation in vitro by Kip3 [9,10] has led to the working assumption that all kinesins-8 would act on microtubules by a similar mechanism. However demonstration of a Kip3-like destabilisation of microtubules in vitro by other kinesins-8 is limited to a single report for Kif18a [11]. Subsequent studies have found no effect of either Kif18a [28] or Klp5/6 [12] on in vitro microtubule stability.
A common feature of microtubule depolymerases is tubulin heterodimer stimulated ATPase activity [9,58], which was absent in the full-length constructs of Klp5/6 used in previous studies  [12]. We confirmed that full length Klp6 lacked tubulin stimulated ATPase but found that the truncated tail-less Klp5 and Klp6 did show tubulin-activated ATPase, suggesting that full length Klp5/6 may auto-inhibit. A Klp6 truncation equivalent to ours can rescue the loss of microtubule destabilising activity in klp6 deletions [47], demonstrating that the truncated construct is functionally competent in restoring normal microtubule lengths in vivo. We find that despite having microtubule translocase activity, neither full length nor truncated Klp6 constructs promote depolymerisation of GMPCPP brain microtubules, or of GMPCPP S. pombe microtubules. Even on dynamic S. pombe microtubules, no significant effect on microtubule stability was observed using truncated Klp5 and 6 constructs alone or in combination, despite all of the constructs having active ATPase and microtubule translocase activity. Translocase activity was observed directly in the assays confirming the presence of active kinesin. We conclude that Klp5 and Klp6 lack an in vitro microtubule depolymerase activity.
Klp5/6 in vivo will only keep pace with growing MT tips at cell ends Klp6 and Klp5 translocate S. pombe microtubules in vitro at 23 nm s 21 and 7 nm s 21 respectively. Kip3 moves with similar velocities in vitro and in vivo [9,10] at about twice the rate of microtubule elongation in vivo [9,15], which is sufficient to account for the accumulation of Kip3 at growing microtubule ends by translocation along the microtubule. If the velocity of Klp5 and Klp6 in vivo is similar to their microtubule sliding velocities in vitro they would be too slow to keep pace with the microtubule tip  [38,43,48,55,59,60,61,62]. This implies either that a factor exists in vivo which accelerates the Klp5/6 microtubule translocase activity above the velocity observed in vitro, or that Klp5/6 track microtubule ends by a different mechanism. We also found in both microtubule sliding assays and with Klp6 coated beads moving along microtubules that a double headed Klp6 motor domain construct at low densities is non-processive. Only when moving in multiple motor teams is processive movement along microtubules possible, as observed with other non-processive kinesin constructs [52,63]. Redistribution in 10-20 seconds between microtubules and tubulin heterodimers during ATPase assays also suggests that both Klp5 and Klp6 motor domains are either non-processive or have very limited processivity. At present it is unclear if Klp5 and 6 have a microtubule binding region in their tail domain as found in Kip3 and Kif18a [13,26], which might confer processivity on full length proteins.
Non-processivity of Klp5/6 would rule out a Kip3-like mechanism, which requires high processivity coupled to a translocation velocity that is higher than the microtubule growth rate [10,14]. However Klp5/6 has been shown to accumulate at the ends of growing microtubules in vivo [43]. Since we observed that Klp5/6 motor domains could bind tubulin heterodimers and form stable complexes we propose that Klp5/6 could accumulate at microtubule ends in complex with tubulin heterodimers as they incorporate at the microtubule tip rather than by procession along the lattice. After incorporation the Klp5/6 complex would be left behind by the growing microtubule tip and the motor would rapidly dissociate from the microtubule lattice. This cycle can form the basis of a tip-tracking mechanism. In vivo, tip tracking is likely to involve other proteins: for example Alp14, a member of the TOG/XMAP215 family of tip-tracking proteins has a role in S. pombe in localising Klp5 to kinetochores [40].

Klp5/6 drives MT nucleation in vivo and in vitro
In cells Klp5/6 promote microtubule growth, catastrophe and rescue, thereby increasing overall microtubule dynamicity [38]. We observed a significant reduction in the IMA number in cells lacking Klp6 suggesting a role for Klp6 in IMA formation or stabilisation. No effect on IMA number was observed in klp5 deletions. This also shows that the Klp6 effect does not simply arise as a consequence of the altered free tubulin level in the deletion strain since klp5 and klp6 deletions both affect microtubule catastrophe frequencies yet only klp6 deletions had altered IMA numbers. Some of the IMAs formed in klp6 deletions had abnormal monopolar arrangements of microtubules previously only observed in deletions of klp2 [57], consistent with Klp6 influencing IMA formation.
In vitro we found that Klp5/6 promotes microtubule nucleation without increasing growth rate. This is a hitherto-undescribed activity for kinesin motors. Xenopus XKlp1, a kinesin-4, has microtubule-stabilizing activity [64], but also inhibits growth and shrinkage, whereas Klp5 and Klp6 do not affect growth and Klp5 only affects shrinkage at very high concentrations. Our data suggest that Klp5/6 homodimers may promote microtubule nucleation by cross-linking tubulin heterodimers. Potentially, Klp5/6 might grip the lattice of nascent microtubule nuclei and link the subunits together, inhibiting subunit dissociation until a stable microtubule structure forms. Our data show that the latticebound Klp5 and Klp6 molecules then dissociate and exchange back into the free tubulin pool within 10-20 seconds.
A new working model for the in vivo mechanism of Klp5/ 6 A model for Klp5/6 activity in vivo has to account not only for microtubule destabilisation activity but also for other apparently contradictory roles including both enhanced microtubule rescue [38] and enhanced IMA formation. Our observations on Klp5/6 are incompatible with MCAK-like or Kip3-like mechanisms, in which the motor depolymerises the GTP cap of dynamic microtubules in solution. Instead, we favour a model in which Klp5/6 are dynamases that first promote the birth of new microtubules, and then shorten their lifetime by accelerating catastrophe in the end zone of the cell ( figure 8). In our model, Klp5/6 activity promotes the formation of new microtubules. Thereafter, Klp5/6 will continuously land on the growing microtubule tip as a complex with GTP-tubulin heterodimers, consistent with the known tendency of Klp5/6 to enrich at microtubule tips [43]. The GTP-tubulin will incorporate into the microtubule lattice and Klp5/6 will dissociate after a few seconds of residence time. According to our data, Klp5/6 is not fast enough to keep pace with the growing tip. Furthermore dimers of Klp5 or Klp6 are non-processive and would only be able to translocate along the microtubule if they assemble into a multiple motor complex. Only at the cell ends, where the microtubule tip engages with the inner cell membrane and its growth rate approximately halves [48,55] do we predict that Klp6 motility would be similar to that of the microtubule tip and Klp6 would have the potential to enrich more substantially than during microtubule growth through the cytoplasm. The extent of enrichment by translocation along the microtubule would depend on formation of multimotor complexes or attachment of the motors to other proteins that might enhance processivity. Our data show that Klp5/6 does not cause catastrophe of dynamic microtubules in vitro whereas Klp5/6 in its natural in vivo environment does accelerate microtubule catastrophe. Clearly the cellular context is required for the catastrophase activity of Klp5/ 6. It is possible that our Klp5/6 constructs lack an activating posttranslational modification or binding partner found in cells. However, we doubt this is the case since our constructs are active, in that they have microtubule and tubulin stimulated ATPase activity and can translocate microtubules. Instead, we suggest that the missing factor in our in vitro assays is mechanical compression of the microtubule tip. In S. pombe, microtubule catastrophes occur almost exclusively in the end zone of the cell, with the microtubule tip in contact with the cell wall and in compression [43,48,54]. Mechanical force can trigger microtubule catastrophe in vitro [65] and a role for force has been suggested in S. pombe [44,48]. The growing plus end of a dynamic microtubule is thought to carry a small sheet of protofilaments that has yet to close into a tube, that may resemble the nucleus formed at the outset of microtubule assembly [66]. We speculate that Klp5 and Klp6 would stabilise the sheet in the absence of external forces by linking heterodimers, thus promoting nucleation. However at cell ends Klp5 and 6 would no longer stabilise but rather destabilise the microtubule. This might occur through the same property of gripping the tipsheet if Klp5/6 were, for example, to link to the cell end and apply enhanced force to the already compressed microtubule. We propose that the effect of this additional destabilising force exerted by Klp5/6 on the microtubule is sufficient to overcome the stabilising effect of Klp5/6 binding, further increasing the catastrophe frequency.
Our working model for Klp5/6 is part substance and part speculation. As in the Kip3 model [43,44], catastrophe is microtubule length-dependent, but in our model control of microtubule length is achieved by amplifying the intrinsic tendency for microtubules to catastrophise in compression in the end zone of the S. pombe cell. Testing these proposals will require the development of improved in vitro reconstitution systems that include other microtubule or Klp5/6 interacting proteins, and can determine the influence of mechanical force on microtubule dynamics.

Protein expression
Proteins were expressed in Rosetta TM 2 (DE3) E. Coli (Novagen) grown overnight 37uC, diluted 1:50 in 26YT medium containing 34 mg/ml chloramphenicol and either 100 mg/ml ampicillin for Klp6 or 30 mg/ml kanamycin for Klp5 constructs, or all three antibiotics for co-expression. Expression was induced by 1 mM IPTG and incubating for a further 4 h at 25uC.

Protein purification
Bacterial pellets were re-suspended in column buffer containing 0.5 mM DTT, Complete Protease Inhibitor EDTA Free Cocktail tablets (Roche), made 0.1 mg/ml lysozyme, and incubated on ice 20 min. The lysate was made 0.05% (v/v) Triton X-100, 10 mM MgCl 2 , 40 mg/ml Deoxyribonuclease I, incubated 10 minutes on ice and clarified by centrifugation (47800 g, 4uC, 35 min). Klp6 His-tagged constructs, were loaded on a HisTrapH column (GE

Protein storage and quantification
Kinesin-containing fractions were pooled and small aliquots flash-frozen in liquid nitrogen. All preparations were subsequently either desalted into the appropriate buffer using Zeba spin columns (Pierce) or microtubule affinity purified and flash frozen again. Proteins were analysed by SDS-Page (NuPAGE 10% Bis-Tris gels run in MOPS buffer, Invitrogen) stained with either InVision TM His-tag fluorescent dye (Invitrogen) for Klp6 440 His or Sypro Red (Invitrogen) for Klp5 436 GST. Gels were imaged (Molecular Imager FX, BioRad) and concentrations determined by comparison with standards (Quantity one software (BioRad)).
For use in microtubule dynamics assays HisKlp6 FL , Klp5 436 GST/Klp6 440 His and Klp5 436 GST purified proteins were mixed with a 5 fold molar excess of NV10 (Novexin Ltd) and desalted (0.5 ml Zeba column, Pierce) into K-PEM (100 mM PIPES, 1 mM MgSO 4 , 2 mM EGTA, adjusted to pH 6.9 with KOH [2]) containing 50 mM ATP and 1 mM DTT. Proteins were frozen in small aliquots in liquid nitrogen. Freezing reduced microtubule stimulated Klp6 ATPase activity by 7%. Protein concentration was measured using a calculated e 280 [69]

Microtubule affinity purification of Klp5 and Klp6
Klp6 440 His or Klp5 436 GST were incubated 10 minutes 25uC with an excess of Taxol (paclitaxel) stabilized pig brain microtubules (typically 7. Purification of Tubulin S. pombe tubulin was prepared from the wild type strain of S. pombe (972 h 2 ) containing the a1b and a2b isoform or a single isoform strain containing a1b isoform tubulin heterodimers as described in [71,72]. Pig Brain tubulin was prepared as described in [73]. Purified tubulins were desalted into K-PEM buffer (100 mM PIPES, 1 mM MgSO 4 and 2 mM EGTA adjusted to pH 6.9 with KOH [2]) containing 20 mM GDP before storage in liquid nitrogen. Protein concentrations were determined by OD 280 nm in 6 M GuHCl, assuming full nucleotide occupancy and using a calculated extinction coefficient [69] e = 105838 M 21 cm 21 for pig brain tubulin and 108,390 M 21 cm 21 for S. pombe tubulin.
Gel filtration chromatography of Klp6-tubulin complexes 15 mM S. pombe single isoform tubulin (a1b tubulin) or mixtures of 15 mM tubulin and 10.7 mM dimeric Klp6 440 His (equivalent to 21.4 mM kinesin heads) in 80 mM PIPES, 2.5 mM MgCl 2 , 1 mM EDTA, 1.5 mM GMPCPP, 1 mM DTT pH 6.9 were incubated at 4uC for 30 min, centrifuged at 20 000 g for 10 min, 4uC then 150 ml of the supernatant separated by gel filtration on a Superose-12 HR 10/300 GL column (GE Healthcare) held at 4uC and running at 0.5 ml min 21 with 80 mM PIPES, 1.2 mM MgCl 2 , 1 mM EDTA, 1.2 mM GMPCPP pH 6.9, in an Ä KTA purifier 10 system (GE Healthcare). Samples from column fractions were separated by SDS-PAGE using 4-12% gradient Nu-PAGE gels in MOPS buffer (Invitrogen) run at 200 V for 80 min and visualized by staining with Krypton fluorescent protein stain (Thermoscientific) and imaged using an Odyssey laser gel scanner (Li-COR Biosciences). Preferential staining of tubulin was corrected by comparison with BSA protein standards.

Steady state ATPase
The microtubule or tubulin heterodimer stimulated ATPase activity of microtubule affinity purified kinesin was measured at 25uC using a linked assay as described [73] except the PIPES concentration was increased to 80 mM and ATP to 1 mM. Microtubules assembled from pig brain tubulin and GTP or S. pombe tubulin and GMPCPP were pelleted and resuspended in BRB80, 20 mM Taxol or BRB80 respectively before dilution into the linked assay buffer (80 mM PIPES, pH 6.9, 5 mM MgCl 2 , 1 mM DTT, 0.1 mg/ml BSA). For measurements with pig brain microtubules 20 mM taxol was added to the buffer. NADH absorption was monitored at 340 nm in a Cary50 spectrophometer (Varian) using 70 ml disposable cuvettes (Eppendorf). Values for V max and K m were obtained by least squares fitting of the Michaelis-Menten equation to plots of ATPase versus tubulin heterodimer concentration using Prism (GraphPad Software Inc.). ATPase values are per kinesin head. IC50 was obtained by fitting the one site competition equation Y = heterodimer V max +(MT V max 2heterodimer V max )/1+10 (X-logIC50) ) to the plot of ATPase activity versus tubulin heterodimer concentration using Prism software. IC 50 is the concentration of Tubulin heterodimer inhibiting 50% of activity. K i is calculated from the IC 50 value according to [74].
In vitro motility assay Pig brain microtubules were stabilised by 20 mM Taxol and S. pombe microtubules by 1 mM GMPCPP. Coverslips were coated with anti-His tag antibody (Qiagen) or polylysine and anti-GST antibody (Santa Cruz biotechnology Inc.) then blocked with 5 mg/ml BSA. Microtubule affinity purified kinesins were flushed into the chamber and incubated 10 min, followed by microtubules diluted in 80 mM Pipes, pH 6.9, 1 mM EDTA, 1 mM MgCl 2 , 5 mM DTT, 2 mM ATP. The buffer also contained 20 mM taxol for pig brain microtubules and 50 mM KCl for Klp5 436 GST. S. pombe microtubule dilution buffer contained 50 mM KCl for both Klp6 440 His and Klp5 436 GST. Microtubules were visualised by VE-DIC [75] and their velocity analysed using RETRAC software (http://www.mechanochemistry.org).
Microtubule motility assays on surfaces with low kinesin density used affinity purified Klp6 440 His diluted in buffer containing K-PEM, 100 mM KCl, 1 mM ATP, 1 mM MgCl 2 , 0.5% (v/v) NP40, 1 mg/ml BSA then flushed into a flow cell coated with anti-His antibody and incubated for 5 min to allow binding of the motor to the surface. The chamber was about 100 mm deep and complete absorption of kinesin to the surface is expected within a few minutes [50]. The chamber was washed with K-PEM, 1 mM ATP, 1 mM MgCl 2 , 5 mM DTT, 20 mM Taxol. Then the same buffer with 0.05 mM pig brain microtubules stabilised with Taxol. In assays where no binding of microtubules to the kinesin surface was observed non-binding was confirmed by flushing in 0.25 mM pig brain microtubules to increase the probability of microtubules colliding with the few kinesins present. Control experiments used the processive rat kinesin-1 rkin430GST [49] bound directly to the flow cell. Surface motor densities were calculated from the dimer kinesin concentrations in the dilutions and assuming all added motor bound to both the upper and lower flow cell surfaces [50].
5 ml of Klp6 440 His from a frozen microtubule affinity purified stock was mixed with 2 ml of beads and 8 ml of BRB80, 1% (w/v) BSA, 0.05% (v/v) Tween20 and incubated for 15 min at room temperature then kept on ice. For laser trap experiments, 3 ml of the bead suspension was mixed with 20 ml of BRB80, 1 mM DTT, 1 mM Mg 2+ ATP, 20 mM Taxol, 3 mg/ml glucose, 100 mg/ml glucose oxidase, 20 mg/ml catalase and the solution flushed into a flow chamber. A laser trap [76] was used to place the beads onto selected microtubules, and the trap then turned off to allow the bead to move freely. Because of the tendency of Klp6 to bind free tubulin, it was necessary to flush the flow cell prior to the experiment, in order to minimize the free tubulin concentration.
Microtubule polymerisation assay S. pombe or pig brain tubulin was diluted to 1.5 mM final concentration in BRB80, 1 mM DTT, 1 mM GMPCPP in a quartz cuvette. Microtubule polymerisation was assayed by 90u light scattering at 350 nm using a Cary Eclipse fluorometer (Varian) with a Peltier temperature controller at 25uC. After establishment of a baseline, desalted Klp6 440 His was added at different final concentrations. Samples were taken at t = 0 and t = 37 min for examination by dark field microscopy.

Darkfield assays of microtubule stability
To create a sensitive assay for acceleration of depolymerisation, GMPCPP (Jena Bioscience GmbH) stabilised pig brain microtubules were diluted to 350 nM (Klp6 440 His bundling) or 100 nM (other experiments) in K-PEM pH 6.9, 100 mM KCl, 0.1% (v/v) NP40, 1 mM DTT, 2 mM Mg.ATP, 25uC, causing spontaneous depolymerisation. Kinesin was added, and time-point samples removed and mounted under an untreated or poly-L-Lysine coated coverslip and sealed with nail polish.
Coverslips (22622 mm, No 1.5 Menzel-Glä ser) and slides (1-1.2 mm Menzel-Glä ser) were prepared by ultrasonic cleaning in 3% (v/v) Neutricon detergent (Decon Laboratories Ltd) using a 600 W ultrasonic bath (Ultrawave Ltd), followed by extensive rinsing and ultrasonic treatment in ultra pure water to create a hydrophilic surface. Then rinsed in 80% (v/v) ethanol and dried using a coverslip spinner (Technical Video Ltd). 40 ml of 0.1 mg/ ml poly-L-lysine in H 2 O (Sigma .300, 000 mw) was added to the coverslip, and then excess poly-L-lysine removed using 36150 ml H 2 O washes and a coverslip spinner before air-drying.
Microtubules were imaged using an Ixon DU-897E EM-CCD camera (Andor Technology PLC) and Metamorph software (Molecular Devices), E800 microscope, 1.4NA darkfield condenser, Plan Fluor 10061.3NA iris objective, Green interference filter (Nikon) with 100 W mercury vapour lamp illumination via fibre optic light scrambler and cold mirror (Technical Video Ltd). Microtubule lengths were measured using Metamorph and analysed using Kaleidagraph (Synergy Software) and Prism software (Graphpad Software Inc).

Microtubule dynamics assays
Microtubule dynamics assays [2] used flow cells of hydrophilic, ultrasonically cleaned 22622 mm coverslips (No 1.5 Menzel-Glä ser) and glass slides (1-1.2 mm Menzel-Glä ser) separated by Parafilm spacers (Pechiney Plastic Packaging). Flow cells were flushed with Echinus esculentus sea-urchin sperm tail axoneme fragments [77], 5 chamber volumes of dynamics buffer (K-PEM pH 6.9, 50 mM K-Acetate, 1 mM MgSO 4 , 1 mM ATP, 1 mM GTP, 100 mg/ml BSA, 10 mM phosphocreatine, 50 mg/ml Creatine kinase), 5 volumes of dynamics buffer containing S. pombe tubulin and kinesin and sealed using VALAP (1:1:1 by weight Vaseline, lanolin and soft paraffin wax). Microtubules were imaged at 2 sec intervals by video enhanced DIC microscopy at 25uC using a video rate CCD camera (C3077, Argus 20 video processor, 2 frame averaging, Hamamatsu) and LG3 frame grabber card (Scion Corp) on a Mac G4 computer running NIH image software (US, National Institutes of Health, http://rsb.info. gov/nih-image/). Microtubule end positions were digitised manually (MT length measure custom Macro for NIH image). Periods of growth and shrinkage in plots of microtubule length against time were fitted by linear regression using Kaleidagraph (Synergy Software). Length changes ,0.24 mm were classified as pauses. Catastrophe and rescue frequencies are total number of events divided by total growth or shrinkage time. Re-growth from an axoneme was not counted as a rescue.

Live cell imaging
For live cell analysis, 1 ml of a log-phase culture of S. pombe in EMMG medium, 4 mM thiamine and appropriate amino acids was concentrated 100 times by centrifugation and a small drop mounted for microscopy as described [48]. Cells were observed using a Nikon Eclipse E800 fluorescence microscope equipped with an Andor iXon EMCDD camera (Andor technology), Nikon PlanApo 6100/1.40 NA oil objective lens and bandpass EGFP filter set (Chroma). Images were collected using MetaMorph software (Molecular Devices) and subsequently deconvolved by Autoquant software (MediaCybernetics). In the presence of Klp6 440 His, there was a decrease in light scattering to a level higher than with microtubules alone. This level remains constant until the addition of CaCl 2 which causes a slower rate of decrease in light scattering than the one observed for microtubules alone (green trace). Samples of the Klp6 440 His assay examined by VE-DIC microscopy contain microtubules before adding CaCl 2 (A) and fewer, mostly bundled, microtubules after adding CaCl 2 (B). Scale bar: 5 mm. (TIF) Figure S4 Depolymerase pelleting assay with GMPCPP stabilised pig brain microtubules. A pelleting assay was used to determine the effect of Klp5 436 GST on GMPCPP stabilised pig brain tubulin microtubules. 350 nM of GMPCPP pig brain microtubules were incubated with increasing concentrations of Klp5 436 GST at 25uC for up to 60 min before pelleting the microtubules. The plot of the percentage of total tubulin in the supernatant shows that under the assay conditions the GMPCPP stabilised pig brain tubulin microtubules are spontaneously depolymerising. Addition of Klp5 436 GST causes this rate of depolymerisation to decrease suggesting that rather than accelerating depolymerisation under these conditions, Klp5 430 GST appears to stabilise the microtubules. Klp6 440 His and tubulin have similar migration on SDS-PAGE. Therefore, in pelleting assays although we could exclude large effects of Klp6 440 His or Klp6 440 His/Klp5 436 GST on MT depolymerisation we could not exclude effects that are more modest. shown to same scale with scale bar (in D) equivalent to 5 mm. Scale bar in E is also equivalent to 5 mm. Pig brain tubulin GMPCPP microtubules were diluted to 350 or 100 nM concentration. In these conditions the microtubules spontaneously depolymerise so any effects of the kinesins upon this rate of depolymerisation should be easily detected. Klp5 and Klp6 were added to the microtubules in solution then aliquots removed and examined by darkfield microscopy. We found that the constructs Klp5 436 GST (B), Klp6 440 His (E) or Klp5 436 GST/Klp6 440 His (C) all caused bundling of the preformed microtubules under conditions where microtubules alone did not bundle. Removal of the GST tag from Klp5 436 (D) or reduced ionic strength by omitting 100 mM KCl from the buffer or omitting ATP from buffer (data not shown) did not prevent bundling by this construct. Only Klp6 415 His, which omits the predicted dimerisation domain, did not cause any significant bundling even upon prolonged incubation (supplementary fig S5). These results suggest that the bundling activity, at least for Klp6 depends upon crosslinking via both of the kinesin heads in multiheaded constructs. It also supports Klp6 forming functional dimers or multimers. (TIF) Figure S10 Klp6 440 His ATPase assay for competition between microtubule and tubulin heterodimer binding. 100 nM Klp6 440 His and 73 nM of Taxol stabilised pig brain microtubules were added to an ATPase assay where the decrease in absorbance at OD 340 is directly linked to the ATPase activity in the assay. After incubation for 75 seconds pig brain tubulin heterodimers were added to 1 mM final concentration, with a break in monitoring during mixing. A new steady state of ATPase activity is then established rapidly. The decrease in slope of the new line (by 66% in the example shown) compared to the initial conditions (indicated by the dashed line) shows the inhibitory effect of the tubulin heterodimers in competition with the microtubules for activation of Klp 440 His ATPase. Typically the microtubule activated ATPase is inhibited within 20 sec following addition of an excess of pig brain tubulin and a new steady state of lower ATPase activity (due to the very low ATPase activation by the tubulin heterodimer) is established within ,100 s.

(TIF)
Movie S1 Microtubule sliding assay of Klp5. Time-lapse movie (1 frame /40 s) of Klp5 driven motility of Rhodamine tagged pig brain microtubules imaged by fluorescence microscopy at 25uC. The microtubules are polarity marked by more intensely stained microtubule seeds at their minus end. Pixel size is 0.130 mm.  Table S3 Velocity and dwell times of Klp6 440 His coated beads on pig brain tubulin microtubules. An optical trap was used to position Klp6 440 His coated beads on Taxol stabilised microtubules assembled from pig brain tubulin (figure 2). The velocity of bead translocation along the microtubule and subsequent dwell time at the microtubule tip was measured for 6 beads. The mean dwell time was 42610 s (6) and velocity 4767 nm/s (5) (mean 6 SEM (n)). The average velocity is slower than is observed in motility assays (87618 nm/s (1085)), which may be caused by different buffers or surface densities of Klp6 440 His in the two assays. Bead movement was processive at high motor concentrations. At low motor densities beads failed to attach and move along microtubules. (DOC)  figure S9) prevented detailed analysis of their effect upon microtubule stability in solution by microscopic methods, apart from noting that even in incubations lasting up to 70 min numerous bundled microtubules were still present at the end of the incubation period. To test if this bundling activity might be masking a depolymerase activity the effect of klp5 436 GST upon individual microtubules was tested by first binding single microtubules to a poly-lysine coated surface before addition of the kinesin and imaging by dark field microscopy. Time-lapse movies were recorded then analysed using the kymograph function of Metamorph software. Spontaneous depolymerisation of the GMPCPP microtubules at 0.40 nm s 21 was still observed. Although addition of Klp5 436 GST caused no bundling, depolymerisation was not significantly enhanced suggesting that bundling was not masking a depolymerase activity. (DOC)