Kinesin-5 Kip1 is a bi-directional motor that stabilizes microtubules and tracks their plus-ends in vivo

Mitotic chromosome segregation is mediated by the bipolar microtubule (MT)-based mitotic spindle, which goes through a well-defined sequence of morphological changes in each mitotic cycle. Mitotic spindle dynamics is governed by two interdependent mechanisms: (1) by the activity of accessory proteins, including cross-linkers, tracking proteins and motors; (2) by the dynamics of the three different groups of spindle MTs: the kinetochore MTs (kMTs) that capture the chromosomes (Maddox et al., 2003; Sprague et al., 2003), the cytoplasmic MTs (cMTs) that facilitate spindle positioning (O'Connell and Wang, 2000; Sheeman et al., 2003) and the interpolar microtubules (iMTs) which overlap in the central midzone region, creating a platform for accessory proteins that control spindle stability, iMT plus-end dynamics and spindle elongation (Bouck and Bloom, 2005; Buvelot et al., 2003; Fridman et al., 2009; Khmelinskii et al., 2007; Schuyler et al., 2003).

Spindle MT function is partially controlled by proteins that preferentially localize to the MT plus-ends, the plus-end tracking proteins (+TIPs). +TIPs comprise several unrelated families that are conserved in eukaryotes and include motor and non-motor proteins (reviewed by Galjart, 2010). One of the suggested mechanisms for +TIPs accumulation at growing MT ends is through a direct recognition of the GTP cap or the unique curvature of growing MT plus-end. The majority of +TIPs are indirectly recruited to the MT plus-ends through interactions via the conserved ‘core’ plus-end protein (Honnappa et al., 2009; Weisbrich et al., 2007). Most +TIPs exclusively track growing MTs (Akhmanova and Steinmetz, 2008). Exceptions are the budding yeast +TIPs Bim1, Stu2, Bik1, Kip2, which track both growing and shrinking MTs (Carvalho et al., 2004; Caudron et al., 2008; Wolyniak et al., 2006).

Spindle MT functions are also controlled by the activity of the conserved mitotic bipolar kinesin-5 family members. Kinesin-5 motors crosslink and slide apart antiparallel MTs (Kapitein et al., 2005; Kashina et al., 1996). Kinesin-5 homologs in S. cerevisiae and higher eukaryote cells aid in spindle assembly, maintenance of the bipolar spindle structure and anaphase B spindle elongation (Asada et al., 1997; Blangy et al., 1995; Roof et al., 1992; Saunders et al., 1997b; Sharp et al., 1999; Straight et al., 1998; Touïtou et al., 2001). S. cerevisiae cells express two kinesin-5 homologs, Cin8 and Kip1. In spite of their overlapping function in spindle assembly, anaphase B spindle elongation and kinetochores clustering (Chee and Haase, 2010; Roof et al., 1992; Saunders and Hoyt, 1992; Saunders et al., 1995; Tytell and Sorger, 2006), a number of studies have indicated that these two kinesin-5 homologs also perform unique mitotic roles. First, Cin8, but not Kip1, is essential for spindle assembly at 37°C (Hoyt et al., 1992). Second, the functions of Cin8 and Kip1 are important during different phases of anaphase-B spindle elongation, Cin8 being important for the initial fast phase, while Kip1 is important for the second slow phase of anaphase B (Straight et al., 1998). Third, in cells impaired for sister-chromatid separation, the deletion of Cin8 but not of Kip1 induces nuclear migration to the daughter cell, most likely by affecting the dynamics of cytoplasmic MTs (de Gramont et al., 2007). Finally, Kip1, but not Cin8, is essential for mitotic segregation of the 2-micron plasmid and its clustering near the spindle poles, a process that precedes anaphase significantly (Cui et al., 2009). Interestingly, Kip1-mediated 2-micron plasmid segregation requires transport of the 2-micron plasmid toward the spindle pole bodies (SPBs), i.e. in the minus-end direction of the spindle MTs.

Research over recent years has focused mainly on Cin8. It was found to destabilize the plus-end of MTs prior to and following the onset of anaphase (Fridman et al., 2009; Gardner et al., 2008). Most remarkably, Cin8 is the first N-terminal kinesin shown to move processively to the minus-end of the microtubule (Gerson-Gurwitz et al., 2011; Roostalu et al., 2011; Thiede et al., 2012). It was further shown to switch directionality as a function of ionic strength conditions and relative MT orientation (Gerson-Gurwitz et al., 2011; Thiede et al., 2012). The switching mechanism is not known yet in detail, and the change of direction has been suggested to be induced either by the unusually large loop 8 of Cin8 (Gerson-Gurwitz et al., 2011; Thiede et al., 2012), and by binding of both motor ends to antiparallel MTs (Gerson-Gurwitz et al., 2011; Thiede et al., 2012) or by motor coupling (Roostalu et al., 2011). Unlike Cin8, the mitotic functions and motile properties of Kip1 are largely uncharacterized. Thus, the mechanisms by which Kip1 performs its multiple mitotic roles remain unclear. Here we set out to investigate these properties with a combination of in vivo and in vitro experiments.

To examine the function of Kip1 during anaphase spindle elongation, we first followed the localization of Kip1–3GFP to anaphase spindles. Similarly to its homolog Cin8 (Avunie-Masala et al., 2011), Kip1 occupied short anaphase spindles in their whole length. As spindles elongated, Kip1 became concentrated at the midzone, as well as near the SPBs (Fig. 1A), probably at the kinetochores (Tytell and Sorger, 2006). Kip1 localization to the midzone was hardly apparent in cells deleted for the function of Ase1 (supplementary material Fig. S1), the conserved MT-binding midzone-organizing protein (Fridman et al., 2009; Schuyler et al., 2003), indicating that Ase1-mediated organization of the midzone is necessary for Kip1 recruitment to this region.

In contrast to Cin8, which detaches from the spindle midzone at late anaphase (Avunie-Masala et al., 2011), Kip1 remained attached to this region, during late anaphase stages. This difference in localization was confirmed by co-expressing Kip1–3GFP and Cin8–tdTomato: in the majority of late-anaphase spindles (>6 µm), Kip1 was enriched at the midzone compared to Cin8 (Fig. 1B). In addition, measurement of normalized fluorescence intensity along spindles in cells co-expressing Kip1–3GFP and Cin8–tdTomato revealed that while in short–intermediate anaphase spindles the intensity profiles of Kip1 and Cin8 were similar, in late-anaphase spindles, the intensity of Kip1 in the middle region of the spindle was significantly higher than that of Cin8 (Fig. 1C,D). This suggests that at the end of anaphase, Kip1 performs a unique function at the spindle midzone, which is not shared with Cin8.

High temporal resolution imaging revealed that at the end of anaphase, the Kip1 population split into two clusters, which translocated in the minus-end direction of the iMTs towards the two SPBs (Fig. 1E–G; supplementary material Movie 1). In wild-type (WT) cells, translocation of Kip1 towards the SPBs was continuous and spanned 3–5 µm, with an average rate of ∼3.0 µm/minute (Table 1). In 90% of the spindles (18/20), the movement was symmetrical, towards both SPBs. In 80% of the spindles (16/20), the poleward movement of Kip1 coincided with a decrease in the distance between the two SPBs (Fig. 1E,G) suggesting that Kip1 moves towards the SPBs during spindle breakdown (sb) (Fig. 1E–G). We occasionally observed Kip1 movement from the poles towards the midzone, in the plus-end direction of iMTs (Fig. 1G, arrowhead). Thus, similarly to Cin8 (Gerson-Gurwitz et al., 2011), we observed movement of Kip1 in both plus-end and minus-end directions of the spindle iMTs.

To test whether the observed poleward movement of Kip1 is coupled to iMT depolymerization, we examined the iMT plus-end dynamics during late anaphase in cells expressing tubulin–GFP. We observed two types of iMT plus-end dynamics during anaphase. First, consistent with a previous report (Fridman et al., 2009), we saw growth and shortening of individual iMTs, extending from and retracting to the midzone (Fig. 2A,B, asterisks and arrowheads). These events were limited in length, up to ∼3 µm, and are likely to represent sporadic growth and shortening of the single iMTs. Second, in accordance with a previous report that long anaphase S. cerevisiae spindles contain four iMTs (Winey et al., 1995), we observed complete depolymerization of four iMTs, starting from the midzone and terminating at the SPBs (Fig. 2A,B, arrows).

In spindles of WT cells, iMT depolymerization occurred in two stages. The first stage, during which one pair of anti-parallel iMTs depolymerized, preceded spindle breakdown by 90±6 seconds (mean ± s.e.m.; n = 27, Table 1); the second stage, during which the second pair of anti-parallel iMTs depolymerized, coincided with spindle breakdown (Fig. 2; supplementary material Movies 2,3). We termed the first stage laMTD (late anaphase MT depolymerization) and the second sbMTD (spindle breakdown MT depolymerization). Interestingly, the average iMT depolymerization rates during laMTD and sbMTD were significantly different, 7.7±0.3 µm/minute (n = 52) and 5.4±0.2 µm/minute (n = 53), respectively (P<0.0001, Table 1; supplementary material Fig. S2A). Moreover, in 90% of the examined cells, the ratio (measured per cell) between laMTD and sbMTD rates was larger than one (n = 40, supplementary material Fig. S2B). These results indicate that the two stages of iMT depolymerization, laMTD and sbMTD are differentially regulated.

Simultaneous observation of iMT dynamics and Kip1 movements in cells co-expressing tubulin–GFP and Kip1–tdTomato from 2-micron plasmid revealed that, although iMTs depolymerization occurred in two distinct stages, Kip1 movement to the SPBs coincided only with the final iMT depolymerization during spindle breakdown, sbMTD (Fig. 2C). Interestingly, the rates of sbMTD in cells expressing Kip1–3GFP were slower than those obtained from tubulin–GFP imaging (Table 1). This difference could be caused by GFP tagging of Kip1 and suggests that the rate of final iMT depolymerization is affected, at least in part, by Kip1–3GFP attachment to the plus-ends of the iMTs.

To address the possibility that, during poleward movement, Kip1 is carried to the SPBs by other motors, we followed Kip1 movements in cells deleted for each of the five S. cerevisiae mitotic motors: four kinesin-related proteins, Cin8, Kip3, Kar3, Kip2 and cytoplasmic dynein heavy chain, Dyn1. We found that, in contrast to the confined and focused localization of Kip1–3GFP to the midzone in WT cells (Fig. 1E,G), in cin8Δ, kip3Δ, kip2Δ and dyn1Δ cells, localization of Kip1 was less well defined and occupied a larger portion of the spindle (supplementary material Fig. S3A,B,D,E). We also observed differences in Kip1 movements caused by CIN8 and KIP3 deletions (supplementary material Fig. S3A,B; Table 1). In cin8Δ cells, Kip1 poleward movement spanned shorter distances compared to WT cells (supplementary material Fig. S3A), which may result from MT stabilization in cin8Δ cells (Fridman et al., 2009; Gardner et al., 2008). Deletion of KIP3 significantly increased the velocity of Kip1 poleward movements (Table 1). In addition, KIP3 deletion also affected the iMT plus-end dynamics during anaphase (supplementary material Fig. S2C). These results are consistent with the established role of Kip3 in MT dynamics (Gardner et al., 2011; Saunders et al., 1997a). Nonetheless, despite the effect of the mitotic motor deletions on Kip1 localization or movement, none of the deletions abolished the Kip1 poleward movement (supplementary material Fig. S3A–E, arrows), indicating that Kip1 moves to the SPBs independently of other motors.

We next asked if poleward movement of Kip1 is dependent on major plus-end binding/tracking proteins. Deletion of the core plus-end tracking protein Bim1(EB1) (Jiang and Akhmanova, 2011) or expression of Kip1 mutated in the EB1-binding motif [SxIP, I1022N,P1023N, Kip1–NN–3GFP (Honnappa et al., 2009)] did not abolish Kip1 movement to the SPBs (supplementary material Fig. S3F,G). We also examined whether Cik1 and Vik1 proteins, which target the minus-end-directed Kar3 to MT plus-ends (Sproul et al., 2005) or Bik1(CLIP-170), which targets dynein heavy chains to the plus-end of cMTs and participates in MT–kinetochore attachment (Lin et al., 2001; Markus et al., 2009; Miller et al., 2006), affect Kip1 spindle movements. Deletion of these three proteins affected the midzone localization of Kip1 (supplementary material Fig. S3H–J) and CIK1 and BIK1 deletions affected the rate of SPB-directed Kip1 movements (Table 1). However, while some plus-end interacting proteins affected Kip1 localization to the spindle, they did not abolish the ability of Kip1 to track plus-ends of depolymerizing iMTs (supplementary material Fig. S3H–J).

We often observed that following spindle breakdown, Kip1 remained attached to the iMT plus-ends and followed their growth and shrinkage. To examine this phenomenon in detail, we monitored MT dynamics and Kip1 localization simultaneously in cells, which expressed tubulin–GFP and mildly overexpressed Kip1–tdTomato from a 2-micron plasmid. Of these cells, ∼10% exhibited collapsed spindles with highly dynamic MTs. These cells were chosen to examine Kip1 plus-end tracking. High temporal resolution imaging revealed that Kip1 was persistently localized to the plus-ends of MTs throughout the duration of our experiments (480–600 seconds), and its movements closely followed MT plus-ends during growing and shrinking phases (Fig. 3A,B; supplementary material Movies 4,5).

To address the question whether nuclear factors are necessary for Kip1 plus-end tracking, we followed localization of 3GFP-tagged Kip1 variant in which the nuclear localization sequence (NLS) was mutated (K1084A, R1085A, R1086A) (Kip1^nls––3GFP), and which remained outside the nucleus (Gordon and Roof, 2001). We found that Kip1^nls––3GFP accumulated in two distinct locations (Fig. 3C). First, in all examined cells, Kip1^nls––3GFP accumulated near the SPBs (Fig. 3C). Second, in about 20% of cells, we also observed the accumulation of Kip1^nls––3GFP in bright spots in the cytoplasm, likely at the plus-ends of the cMTs (Fig. 3C, arrows). To examine whether these Kip1^nls––3GFP spots track the plus-ends of cMTs, we visualized their movements in cells expressing tubulin–GFP (Fig. 3D,E). Although in these cells, Kip1 and tubulin were visualized by the same fluorophore (GFP), the accumulation of Kip1 could be easily identified at the plus-ends of the cMTs in G1 cells and mitotic anaphase spindles (for comparison, see supplementary material Fig. S4A,B, tubulin–GFP only).

Our measurements revealed that Kip1^nls––3GFP tracks the plus-ends of growing and shrinking cMTs (Fig. 3D,E). The rate at which Kip1^nls––3GFP tracked cMTs during shrinkage was 2.8±0.2 µm/minute (n = 14) and during growth 1.2±0.1 µm/minute (n = 15). These rates are comparable to those previously reported for shrinkage and growth rates of cMTs (Gupta et al., 2002; Gupta et al., 2006; Kosco et al., 2001; Tirnauer et al., 1999), indicating that nuclear factors are not required for the plus-end tracking activity of Kip1. We occasionally observed detachment of Kip1^nls––3GFP from the cMTs (data not shown) and found that, in general, Kip1^nls––3GFP tracking of the cMTs was less processive than that of endogenously expressed Kip1–3GFP (Fig. 1E,G) or of moderately overexpressed Kip1–tdTomato (Fig. 3A,B). Finally, we observed plus-end localization and tracking of Kip1^nls––3GFP in cells expressing a temperature-sensitive mutant of the plus-end tracking protein Stu2, stu2-10, at the non-permissive temperature (Kosco et al., 2001) (supplementary material Fig. S5). This result indicates that Kip1^nls– tracking of the plus-ends of cMTs is Stu2 independent. Thus, our data strongly suggest that Kip1 is a plus-end tracking protein, which is able to persistently track the plus-ends of MTs during growing and shrinking phases.

Several lines of evidence presented here raise the question whether Kip1 is a bi-directional motor: first, Kip1 moves towards the midzone as well as towards the SPBs (Fig. 1E,G and supplementary material Fig. S3) and second, it exhibits plus-end tracking activity (Fig. 3). Furthermore, the closely related homolog Cin8 is able to switch directionality depending on salt conditions and MT binding geometry (Gerson-Gurwitz et al., 2011; Roostalu et al., 2011). To address this point, we examined the in vitro activity of Kip1–3GFP in a single-molecule fluorescence assay on taxol-stabilized MTs (Gerson-Gurwitz et al., 2011; Kapitein et al., 2008). Experiments were performed with Kip–3GFP–6HIS purified from S. cerevsiae cells, and in whole extracts of cells expressing Kip1–3GFP from its own promoter on a 2-micron plasmid (supplementary material Fig. S6). We found that, similarly to the homologous Cin8 (Gerson-Gurwitz et al., 2011; Roostalu et al., 2011), single molecules of purified Kip1 moved towards the minus-ends of the MTs at high ionic strength (Fig. 4A,B; supplementary material Movie 6), with an average rate of −20.6±2.2 nm/second (n = 234, supplementary material Table S1). The minus-end-directed motility was observed in both purified and whole-extract samples (Fig. 4D). The average travel distance (here termed run length) of Kip1–3GFP was shorter in whole extracts (∼560 nm) than in purified samples (∼850 nm, supplementary material Table S1). This may be explained by the presence of MT-binding proteins in the extracts that may create obstacles on the MT and reduce processivity, as was previously shown for Kinesin-1 (Telley et al., 2009). However, the average velocities of Kip1–3GFP from whole extracts and purified samples were similar, ranging between −14 and −21 nm/second (Fig. 4D; supplementary material Table S1). MSD analysis of the minus-end motility of Kip1 in purified and whole-extract samples quantitatively confirmed a diffusive component (supplementary material Fig. S7A) with a diffusion coefficient of ∼2×10³ nm²/second (supplementary material Table S1), similar to values reported for other kinesin-5 homologs (Gerson-Gurwitz et al., 2011; Kapitein et al., 2008). To validate that Kip1 motility is not diffusion-driven only (Kikkawa et al., 2001), we tested the ATP-dependence of the motility. We found that the average velocity was close to zero in the presence of ADP and that the average velocity in the presence of 0.01 mM ATP was lower by a factor of two than that at saturating ATP conditions (Fig. 4B; supplementary material Table S1). Based on these findings, we conclude that the minus-end-directed motility of Kip1 is ATP dependent. Thus, Kip1 is the second N-terminal kinesin motor that can move processively towards the minus-ends of the MTs under high-ionic-strength conditions.

We next examined whether, similarly to the homologous Cin8 (Gerson-Gurwitz et al., 2011), single Kip1 molecules can switch directionality by lowering ionic strength. We found that, as ionic strength decreased, Kip1 gradually became bi-directional (Fig. 4C). While at a high ionic strength of 0.240 M, Kip1 moved ∼90% of the time in the minus-end direction (Fig. 4B), at an ionic strength of 0.180 M, Kip1 was minus-end directed for only 65% of the time (Fig. 4C). However, in contrast to Cin8, we observed neither a clear nor a complete switch to plus-end-directed motility. In ionic strengths lower than 0.114 M, Kip1 moved ∼50% of the time in plus- and minus-end directions (Fig. 4C), resulting in close-to-zero average velocity (supplementary material Fig. S7B). Nonetheless, the bi-directional motility of Kip1 in low-ionic-strength buffer was ATP dependent (supplementary material Fig. S7C,D), indicating that single molecules of Kip1 can move actively to both minus- and plus-ends of MTs. Furthermore, we also found that in multi-motor gliding assays at different ionic strength conditions, Kip1 exhibited exclusively plus-end-directed motility (supplementary material Fig. S8; Movies 7,8). This is consistent with previous reports on Cin8 that was shown to be plus-end directed in MT gliding assays (Gerson-Gurwitz et al., 2011; Gheber et al., 1999; Roostalu et al., 2011). Our results show that Kip1 is a bi-directional kinesin-5 motor, producing plus- and minus-end motility in vitro, depending on assay conditions. It is possible that the bi-directional movements of Kip1 in cells can be attributed, at least in part, to the ATP-dependent bi-directional motor activity of Kip1.

To examine whether translocation of Kip1 to the SPBs is dependent on its motor function, we examined the in vivo and in vitro functions of the mutant Kip1-T148N, in which a conserved threonine (T) in the ATP-binding P-loop is mutated to an asparagine (N). Kinesin motors that carry this mutation were shown to be immobile due to a persistent attachment to the MTs (Blangy et al., 1998; Meluh and Rose, 1990; Nakata and Hirokawa, 1995). As expected, this variant did not exhibit any in vitro motility (Fig. 4E). To examine the in vivo phenotype of this mutant, we expressed Kip1-T148N–3GFP from the endogenous Kip1 promoter in kip1Δ cells and followed its localization on anaphase spindles. We found that localization of Kip1-T148N to the midzone was reduced compared to WT Kip1 (Fig. 5A; Fig. 1G; supplementary material Fig. S9). However, although Kip1-T148N was present at the spindle midzone region (Fig. 5A,E), no poleward movement of this mutant occurred (Fig. 5A). Instead, Kip1-T148N disappeared from the midzone region before spindle breakdown (Fig. 5A). In cells co-expressing Kip1-T148N–3YFP and tubulin–GFP, sbMTD was still observed (data not shown), indicating that the movement of Kip1 towards the SPBs is not required for sbMTD. Thus, our results indicate that while the catalytic motor activity of Kip1 is not required for spindle breakdown or sbMTD, it is required for the translocation of Kip1 towards the SPBs.

We noticed that, in contrast to the WT Kip1, the localization of Kip1-T148N in the anaphase spindles was not symmetrical but rather shifted to the spindle pole located in the bud (Fig. 5B; supplementary material Fig. S9). To quantify this effect, we measured the ratio between GFP fluorescence intensities of bud-located pole to mother-located pole (Fig. 5C). Indeed, while in cells expressing WT Kip1–3GFP this ratio was 0.71±0.04 (n = 90), it was much larger in Kip1-T148N–3GFP-expressing cells: 3.6±0.4 (n = 74) (Fig. 5C). This result indicates that motility of Kip1 is required for equal distribution of Kip1 on the spindle apparatus. Unlike Kip1^nls––3GFP, when the rigor mutant, Kip1-T148N–3GFP, was targeted outside the nucleus by NLS mutagenesis, it was not observed on the cMTs and did not accumulate at the plus-ends of the cMTs (Fig. 5B). This result indicates that motor activity of Kip1 is required for its plus-end tracking of the cMTs. Interestingly, targeting the rigor Kip1 mutant outside the nucleus exacerbated the asymmetry of bud-to-mother pole localization of Kip1, increasing the ratio between Kip1–3GFP fluorescence at the two poles to ∼11 (Fig. 5C). This phenomenon requires further investigation.

Next, we examined the influence of the rigor Kip1-T148N mutant on cell cycle progression and anaphase spindle elongation. We found that kip1Δ cells expressing Kip1-T148N–3HA were not viable. The reason for this lethality is unclear. Therefore, we examined a 3GFP-tagged version of Kip1-T148N. Following the distribution of cell and spindle morphologies in cycling cells, we found that cells expressing Kip1-T148N were delayed in mitosis and accumulated in pre-anaphase and anaphase stages, compared to cells expressing WT Kip1 (Fig. 5D). Moreover, we found that cells expressing Kip1-T148N had considerably shorter anaphase spindles (Fig. 5E), of 3.24±0.25 µm (n = 31) compared to 8.12±0.18 µm (n = 25) in cells expressing WT Kip1. Therefore we conclude that motor function of Kip1 is required not only for its plus-end tracking activity but also for proper cell cycle progression and spindle elongation.

It remained to find out what the function of Kip1 in late anaphase is. To address this question, we first examined the spindle morphologies in WT versus kip1Δ cells. We performed fluorescence-intensity profile analysis of spindle tubulin–GFP (Fig. 6A). This analysis revealed that while in spindles of 5–6 µm, tubulin–GFP fluorescence profile was similar in the WT and kip1Δ cells (Fig. 6A, top), significant differences in tubulin–GFP fluorescence were apparent in the middle region of long anaphase spindles (6–7 µm) (Fig. 6A, bottom). This result suggests that at the late anaphase, iMT dynamics may be different in WT and kip1Δ cells.

To address the possibility that Kip1 controls iMT dynamics at late anaphase, we first examined iMT dynamics by high temporal resolution imaging in kip1Δ cells expressing tubulin–GFP (Fig. 6B). These experiments revealed that compared to WT cells (Fig. 2A; Fig. 6C), in kip1Δ cells, the iMT overlapping region of long anaphase spindles was less well defined (Fig. 6B). In addition, we observed more sporadic iMT growth and shortening events, not related to laMTD or sbMTD (Fig. 6B, right panel). These phenotypes point to more dynamic iMTs in anaphase spindles of kip1Δ cells. We further assayed the kinetics of fluorescence recovery after photobleaching (FRAP) in the midzone region of anaphase spindles in cells expressing tubulin–GFP (Fridman et al., 2009). An illustration of the FRAP procedure in tubulin–GFP-expressing cells is presented in supplementary material Fig. S10. Consistently with previous results (Fridman et al., 2009), we found that in WT cells, the first order rate of FRAP kinetics was 0.043 second⁻¹ and 0.013 second⁻¹ for intermediate and long anaphase spindles, respectively (Table 2). Similarly to WT cells (Fridman et al., 2009), FRAP traces of anaphase spindles in kip1Δ cells exhibited first-order kinetics (Fig. 6D). However, in long anaphase spindles in kip1Δ cells, the average FRAP first order rate constant was significantly higher (0.028 second⁻¹) than those in WT cells (0.013 second⁻¹) (Table 2). These results indicate that in kip1Δ cells iMT plus-ends are less stable than in WT cells, namely that during late anaphase Kip1 stabilizes iMTs.

In agreement with the notion that Kip1 affects iMT dynamics at late anaphase, we found that KIP1 deletion also affected some characteristics of the iMT depolymerization during laMTD and sbMTD (Fig. 6B; supplementary material Table S2). First, a single stage of iMT depolymerization during anaphase, instead of two, was observed in ∼38% of kip1Δ cells (Fig. 6B, right), compared to only ∼7% of WT cells (supplementary material Table S2). The occurrence of only one stage of iMT depolymerization is likely due to less well organized spindles and/or simultaneous depolymerization of all iMTs in kip1Δ cells. Second, spindle breakdown that occurred before sbMTD was observed in ∼20% of kip1Δ cells (Fig. 6B, middle), compared to only ∼3% of WT cells (supplementary material Table S2).

Interestingly, while affecting some aspects of iMT depolymerization during laMTD and sbMTD, deletion of KIP1 did not affect the rate of iMT depolymerization in these stages nor the time interval between them (Table 1). Similarly, deletion of the S. cerevisiae kinesin-8 Kip3, which has an established role in controlling MT dynamics and timing of spindle disassembly (Gupta et al., 2006; Straight et al., 1998; Varga et al., 2006; Woodruff et al., 2010) also did not affect the rates of laMTD and sbMTD (Table 1). We thus suggest that partially independent mechanisms control iMT plus-end dynamics in late anaphase versus the complete iMT depolymerization that leads to spindle breakdown. Nonetheless, our results indicate that KIP1 deletion compromises spindle stability in late anaphase as well as the orderly iMT depolymerization. Hence, the persistent binding of Kip1 to the midzone region of late anaphase spindles (Fig. 1) is required to stabilize the iMT plus-ends and to control the sequential events of iMT depolymerization, prior to- and during spindle breakdown.

Minus-end-directed, albeit non-processive motility was until recently observed only for kinesin-14 family members, which are structurally distinct from all other kinesin subfamilies in that they carry the conserved motor domain at the C-terminus instead of the N-terminus (Block, 2007; deCastro et al., 2000; McDonald et al., 1990; Walker et al., 1990). The N-terminal homotetrameric S. cerevisiae kinesin-5 Cin8 (Gerson-Gurwitz et al., 2011; Roostalu et al., 2011; Thiede et al., 2012) has, however, recently been shown to be a bi-directional motor: single Cin8-motors were directly observed in vitro to move processively to the minus-end of MTs at high ionic strength and to switch to plus-end directionality under low-ionic-strength conditions and when binding between two antiparallel MTs (Gerson-Gurwitz et al., 2011) or as a function of motor density (Roostalu et al., 2011). The data presented here show direct evidence that, similarly to Cin8, single Kip1 molecules also move towards the minus-end of the MTs at high ionic strength (Fig. 4). Lowering the ionic strength leads to a gradual increase of the fraction of motors moving towards the plus-end of the MT. These findings are consistent with our own previous experiments with Cin8 mutants: when the unique and unusually large loop 8 of Cin8 was replaced by the short loop 8 of Kip1, the minus-end directionality was preserved (Gerson-Gurwitz et al., 2011). We also report here that, when numerous Kip1 molecules interact with the same MT in a multi-motor surface gliding assay, Kip1 exclusively displays plus-end-directed motility at all ionic-strength conditions tested (supplementary material Fig. S8). Thus, Kip1 is the second N-terminal kinesin motor protein that is shown to produce both minus-end and plus-end-directed activity in vitro. Although the mechanism by which Cin8 and Kip1 produce and switch between both minus- and plus-end-directed activity in vitro is not clear, these findings indicate that bi-directionality may be a common feature of several N-terminal kinesin motors.

The function of kinesin-5 motors in spindle assembly (Blangy et al., 1995; Mayer et al., 1999; Saunders and Hoyt, 1992) and in anaphase spindle elongation (Movshovich et al., 2008; Saunders et al., 1995; Straight et al., 1998) is attributed to their ability to provide plus-end-directed motility and outwardly directed pushing forces between overlapping antiparallel MTs of the spindle (Saunders and Hoyt, 1992). Thus, the role of the minus-end-directed activity of the S. cerevisiae kinesin-5 motors is less obvious. In addition to the role in kinetochore clustering and fast clearance from the spindle that was suggested for Cin8 (Gerson-Gurwitz et al., 2011), the minus-end-directed motility of Kip1 may be linked to its intracellular plus-end tracking activity on depolymerizing MTs (Figs 1–f02,3). Furthermore it might be linked to its additional, completely independent role in the segregation of the 2-micron plasmid during mitosis (Cui et al., 2009), which also requires minus-end-directed movement on the spindle MTs.

The data presented here demonstrate that in S. cerevisiae cells, Kip1 processively tracks the plus-ends of growing and shrinking MTs (Figs 1–f02,3). The plus-end tracking activity of Kip1 is independent of other mitotic motors, major plus-end tracking proteins (supplementary material Figs S3, S5) or nuclear localization factors (Fig. 3C–E). Although we cannot completely rule out direct or indirect effects of other proteins on the movement of Kip1, these findings suggest that Kip1 may track MT plus-ends by an entirely autonomous activity.

Thus far, Kip1 is the only kinesin-5 motor for which intracellular plus-end tracking activity has been demonstrated. This raises the question of the physiological importance of this activity. It has been recently demonstrated that Kip1 physically interacts with the 2-micron plasmid partitioning system and that the 2-micron plasmid fails to segregate in kip1Δ cells (Cui et al., 2009). Since segregation of the 2-micron plasmid requires its attachment to the MT plus-ends and movement in the minus-end direction of the MTs, it may be linked to the unique plus-end tracking activity of Kip1 in vivo and/or its minus-end-directed movement in vitro.

The mechanism by which Kip1 tracks the plus-ends of the MTs requires further investigation. Tracking of growing MTs is more common than tracking of shrinking MTs (reviewed by Akhmanova and Steinmetz, 2008). Several mechanisms have been suggested for tracking the plus-ends of depolymerizing MTs. The S. cerevisiae kinetochore Dam1 complex was shown to track MT plus-ends in an ATP-independent manner, by forming oligomers of fully closed rings or non-ring patches around the shrinking MTs (Gao et al., 2010; Mennella et al., 2009; Westermann et al., 2006). The open conformation of the MT plus-end during depolymerization was suggested to push the ring-like complexes toward the MT minus-end (Efremov et al., 2007; Miranda et al., 2005; Westermann et al., 2006). In contrast to the Dam1 complex, the Klp5/6 hetero-dimers do not form a ring around the MTs. Klp5/6 was shown to use ATP to translocate towards the plus-end of the MTs, but tracking of shrinking MTs was demonstrated to be ATP independent (Grissom et al., 2009).

We have shown here that Kip1 is a bi-directional motor in vitro (Fig. 4; supplementary material Fig. S8) and that a rigor mutant of Kip1 cannot produce plus-end tracking in vivo (Fig. 5A). Therefore, it is possible that during plus-end tracking, Kip1 actively moves toward the minus and/or plus end of the MT. In addition, it has been shown that a P-loop motor-domain mutant of Kip1 failed to support the segregation of the 2-micron plasmid (Cui et al., 2009), supporting the notion that the plus-end tracking of Kip1 in cells is ATP dependent. It may, however, be possible that only some aspects of the ATPase cycle are required for Kip1 plus-end tracking. For example, it had been demonstrated for the plus-end tracking Drosophila kinesin-13 KLP10A and KLP59C, that a ring-structure formation required for the plus-end tracking, was favored specifically by the ATP-bound state (Moores et al., 2003; Varga et al., 2006). Finally, it has been recently shown that a non-motor MT binding site in kinesin-5 is required for MT crosslinking (Weinger et al., 2011) and that the tail of Kip3 promotes its accumulation at the MT plus-ends (Su et al., 2011). Therefore, the tail of Kip1 may be involved in plus-end MT tracking, in addition to the role of its motor domain. Direct in vitro studies of MT plus-end tracking by Kip1 should be able to distinguish between the different possibilities.

It has been proposed recently that anaphase spindle disassembly in S. cerevisiae cells is composed of three sub-processes: disengagement of the spindle halves, arrest of spindle elongation, and initiation of iMT depolymerization (Woodruff et al., 2010). Data presented here indicate that iMT depolymerization at the end of anaphase does not occur in a single step, but is divided into two stages: the first taking place ∼90 seconds before spindle breakdown (laMTD) and the second coinciding with spindle breakdown (sbMTD). Previous electron microscopy and reconstitution analysis showed that long anaphase S. cerevisiae spindles contain only four iMTs, two emanating from each pole (Winey et al., 1995). Visualization of the depolymerization of individual iMTs presented here (Fig. 2A) indicates that, before final spindle disassembly, the S. cerevisiae anaphase spindles consist of only two iMTs, one emanating from each pole. The rate of the final iMT depolymerization, sbMTD, is slower than that of laMTD (Table 1) indicating that the two depolymerization events are differentially regulated. In addition, the deletion of the kinesin-8 Kip3, which was previously shown to prolong the duration of anaphase (Straight et al., 1998) and to destabilize MTs in vitro and in vivo (Gardner et al., 2011; Gupta et al., 2006), increases the time period between the two stages (Table 1), supporting the notion that the two iMT depolymerization events are differentially regulated. Regulation mechanisms are likely to include the mitotic exit network (MEN), reviewed by Rock and Amon (Rock and Amon, 2009), the chromosome passenger aurora-like complex that localizes to the spindle at the end of anaphase (Buvelot et al., 2003; Woodruff et al., 2010), and the anaphase-promoting complex, APC (Sullivan and Morgan, 2007; Woodruff et al., 2010) which were shown to regulated spindle disassembly and exit from mitosis.

Data presented in this study indicate that the kinesin-5 Kip1 performs unique functions in anaphase B, not shared with the homologous Cin8. At early–mid anaphase, during the first fast stage of anaphase spindle elongation, both Cin8 and Kip1 localize to the spindle with Cin8 contributing the majority of the SPB separation force (Movshovich et al., 2008; Saunders et al., 1995; Straight et al., 1998). At mid–late anaphase, Cin8 detaches from the spindles due to phosphorylation of Cdk1 targets located in its motor domain (Avunie-Masala et al., 2011) while Kip1 remains attached to the spindle midzone (Fig. 1). The differences in regulation of Cin8 and Kip1 at this stage may be due to the presence of the long loop 8 of Cin8, possibly exerting its timely function through phosphorylation of two Cdk1 phosphorylation sites, which are absent in Kip1. Detachment of Cin8 from the spindle slows down spindle elongation rate (Avunie-Masala et al., 2011) giving rise to the second, slower stage of anaphase spindle elongation, mediated by the motor activity of Kip1 (Straight et al., 1998). Kip1 remains attached to the midzone region until the spindle breaks down. It does not detach from the midzone when two of the four iMTs depolymerize during laMTD, but rather remains attached to the two last iMTs. The high affinity of Kip1 to the iMT plus-ends may facilitate its attachment to the middle region of the spindle at the end of anaphase. The presence of Kip1 at this location and time is crucial to stabilize the iMT plus-ends since in its absence the MTs are highly dynamic (Fig. 6). We have previously shown that iMT plus-end dynamics is altered during the progression of anaphase and that, as spindles elongate, iMT plus-ends become more stable (Fridman et al., 2009). The data presented here strongly suggest that one of the major factors that stabilize iMTs at late anaphase is the presence of Kip1 at the midzone. This iMT plus-end stabilization can be one of the factors that arrests spindle elongation at the end of anaphase (Rozelle et al., 2011; Woodruff et al., 2010) and provides stability to the anaphase spindles at the final stages of mitotic spindle dynamics.

To summarize, Kip1 is one of the key spindle-stabilizing factors at the end of anaphase which crosslinks iMTs and stabilizes their plus-ends. It is likely to be the key regulator of the separation of the two phases of iMT depolymerization. The bi-directional motility of Kip1 and its plus-end tracking activity are likely to enable Kip1 to perform its multiple intracellular functions in segregation of the 2-micron plasmid, anaphase spindle elongation and final spindle disassembly.

S. cerevisiae strains, plasmids and primers used in this study are described in supplementary material Tables S3–S5. S. cerevisiae cell growth, protein expression and DNA manipulation were performed using standard procedures (Sambrook et al., 1989). Mutants were generated by PCR site-directed mutagenesis as described (Ko and Ma, 2005). All PCR-cloned products and mutagenesis were sequenced.

For Kip1–3GFP, Kip1-T148N–3GFP–6HIS and Kip1–3GFP–6HIS overexpression, Kip1 ORF was amplified from yeast genomic DNA was cloned to the p425GAL1 plasmid (Mumberg et al., 1994). A unique NotI site was introduced before the stop codon and 3GFP, 3GFP–6HIS fragments were sub cloned on it resulting in plasmids that termed pVF26 and pVF78 respectively. For mild Kip1–3GFP overexpression from its own promoter in 2-micron plasmid, a GAL1 promoter of the pVF26 plasmid was substituted by the Kip1 promoter (from −543 bp to −1 bp upstream to ORF) resulting in pVF39 plasmid. For endogenous Kip1 tagging with 3GFP, a Kip1–3GFP promoter-less fragment was sub-cloned from pVF26 plasmid to pRS305 plasmid (Sikorski and Hieter, 1989), resulting in a pVF34 plasmid. For construction of a plasmid for Kip1–3GFP expression from its own promoter in the LEU2 locus, Kip1 promoter was PCR-amplified and sub-cloned in to the pVF34 plasmid before KIP1 ORF.

Live-cell imaging was done using a spinning-disc confocal microscope at the Ilse Katz Institute for Nanoscale Science and Technology (Zeiss Axiovert 200M, UltraView ESR, Perkin Elmer, UK), equipped with a FRAP module, as previously described (Avunie-Masala et al., 2011; Fridman et al., 2009; Gerson-Gurwitz et al., 2011). Z-stacks of 0.2–0.3 µm separations were acquired with frame time of 2–4 seconds/frame for in vivo imaging of MT dynamics, and 60 seconds/frame for spindle elongation measurements. FRAP experiments to follow iMT plus-end dynamics in tubulin–GFP-expressing cells were done as described previously (Fridman et al., 2009). Image processing and kymograph preparation was performed using ImageJ and MetaMorph software. Rates of iMT depolymerization and Kip1 movement on the spindles were determined from kymographs. During spindle collapse, these velocities were corrected for decreasing spindle length.

The ratio of bud to mother spindle pole intensity in each cell was calculated by measuring fluorescence in a circle of 5 pixels in diameter at the poles, after subtracting the background of each cell. Line-scan fluorescence intensity analysis was performed using the ‘line-scan’ plug-in in the Metamorph software. Spindle intensities were measured from mother to the bud and the background of the same cell was subtracted. Fluorescence intensity along each spindle was normalized to its maximal intensity. Spindle length was normalized and interpolated by the Origin software to 50–100 equally spaced distance points.

Motility experiments were performed on two forms/preparations of Kip1: (1) Kip1–3GFP was expressed in kip1Δ cells from the endogenous promoter on a 2-micron plasmid. Then crude extracts were prepared from these cells as previously described for Cin8 (Gerson-Gurwitz et al., 2011); (2) Kip1–3GFP–6HIS or Kip1-T148N–6HIS were overexpressed on a 2-micron plasmid from a GAL1 promoter for 4–5 hours in a protease-deficient yeast strain and affinity purified on a Ni-NTA column (supplementary material Fig. S6).

Motility assays of Kip1–3GFP were performed on taxol-stabilized MTs as previously described (Gerson-Gurwitz et al., 2011). Unless otherwise stated, ATP was added to a saturating concentration of 3 mM. To generate buffers of various ionic strengths, various concentrations of PIPES buffer and NaCl were combined as follows: PIPES 80 mM, NaCl 60 mM (P80-60), ionic strength, IS = 0.24 M; PIPES 80 mM, pH = 6.8, NaCl 30 mM (P80-30), ionic strength, IS = 0.21M; PIPES 80 mM, NaCl 0 mM (P80-0), ionic strength, IS = 0.18 M; PIPES 50 mM, NaCl 0 mM (P50-0), ionic strength, IS = 0.114M; PIPES 12 mM, NaCl 25 mM (P12-25), ionic strength, IS = 0.055 M.

Single-molecule fluorescence data were collected on a Zeiss Axiovert 200M, HBO 100 Mercury Illuminator, cooled CCD (SensiCam, PCO), frame time 1–1.5 seconds. Data were processed using ImageJ and MetaMorph (MDS Analytical Technologies) software. MSD (mean square displacement) of Kip1 movement was obtained by tracking Kip1 location using the ImageJ Spot Tracker plug-in. Values of velocity and diffusion coefficient were derived from a second-order polynomial fit to MSD data (Gerson-Gurwitz et al., 2011; Kapitein et al., 2005). Velocity histograms were assembled by drawing lines through consecutive 10.5 second segments of kymograph traces (ImageJ Grid plug-in) and measuring velocity in each segment. A stagnation period of two segments or less between two movement episodes was taken in account. To determine run length distributions, directional episodes for which both beginning and end were apparent in the kymograph and stall period, if it existed, was shorter than 15 seconds. Motility assays at high motor densities were performed following standard procedures (Gerson-Gurwitz et al., 2011; Howard et al., 1989).

The statistical significance of differences was determined using a two-tail Student's t-test.

We thank Prof. Tim Huffaker for providing stu2 mutant S. cerevisiae strains.