Upon MT polymerization, the detached kinetochores return to the poles and become bi-oriented

When arrested nda3 cells are returned to 32°C, their MTs polymerize within a few minutes; the SPBs then separate, and the cells proceed through mitosis (Hiraoka et al, 1984; Kanbe et al, 1990). To examine the process by which ‘lost' kinetochores are ‘retrieved' by spindle MTs, we analyzed the progression of these cells through mitosis by immunostaining (Figure 1). The percent of lost kinetochores decreased fast (τ_1/2~10 min) from the initial 81% (Figure 1A and B). Concomitant with kinetochore–SPB re-association, there was a transient increase in the percent of kinetochores positioned between the SPBs (Figure 1C and E). Approximately 6 min after the shift to 32°C, cells began to exit mitosis (τ_1/2~19 min), as judged by a decrease in the percent of cells with condensed, clustered chromosomes (Figure 1D). By 40 min more than 80% of the cells had exited mitosis having succeeded in retrieving the detached kinetochores. Thus, after initiation of MT polymerization, the lost kinetochores are first drawn to the separated poles (Figure 1E (b–d)); they then bi-orient and go on to anaphase (Figure 1E (e–g)), behaviors that are strikingly similar to those found in normal vertebrate mitosis.

This conclusion was tested by visualizing the poles and the kinetochores of chromosome 2 in living cells with GFP-marked centrosomes and centromeres (see Materials and methods). At restrictive temperature, the detached SPBs and kinetochore underwent seemingly random motions (Supplementary Video 1). The kinetochore frequently split into two, indicating a transient separation of sister chromatids, presumably due to thermal motions and chromatin visco-elasticity. Following the shift to permissive temperatures, the SPBs separated in 2.9±0.3 min (N=41). About 7.0±1.7 min later (N=30), the tagged kinetochore was drawn towards one of the poles (Figure 1F). Bi-orientation (defined as the first time when the kinetochore was roughly midway between SPBs) occurred 6.9±0.7 min (N=17) after the kinetochore's arrival at the SPB. It was frequently preceded by a rotation of the kinetochore from its orientation during the approach to the pole, allowing it now to face the opposite pole. By 30 min after the shift anaphase had occurred in ~2/3 of the cells. Thus, the results from live imaging and from immunofluorescence were fully consistent. Together they demonstrate that a lost kinetochore can move poleward and bi-orient successfully. Presumably, similar motions occur during normal mitosis, albeit over shorter distances.

The efficiency of kinetochore retrieval is enhanced by the stability of kinetochore–MT attachment and the processivity of the ensuing movements

We then asked about the effectiveness of this kinetochore retrieval process. With increasing distance between the detached kinetochore and poles, the time for retrieval increased slightly (Supplementary Figure 2A, red curve); at distances comparable to the diameter of a fission yeast nucleus, fewer kinetochores were recovered (Supplementary Figure 2B). Nonetheless, ~80% of the ‘lost' kinetochores were pulled to the pole within 15 min, so long as they had scattered <1.6 μm from the poles. Thus, the MT capture and kinetochore retrieval was effective at these distances.

The shortness of the MTs that connected SPBs with retrievable kinetochores made their visualization difficult, whether by immunofluorescence with antitubulin or in live cells with tubulin-GFP (Supplement). To circumvent these difficulties and gain more information about the kinetics of kinetochore retrieval, we have collected the three-dimensional (3D) coordinates of both poles and the kinetochore. The resulting data were analyzed with custom computer programs (see Materials and methods). For example, Figure 2A shows consecutive kinetochore positions (red X's connected with lines) relative to the poles (green circles) as a 2D image, displayed in the plane that contains these three objects. In the cold (red line) kinetochore movements appeared random (Supplementary Video 2); its position fluctuated around a mean with a standard deviation (s.d.) of 0.318 μm. About 6 min after the shift to 32°C, the kinetochore's movement became directed towards one of the poles (Figure 2A, blue line). The s.d. for a line fitted to this part of the trajectory was 0.077 μm, significantly less than the s.d. determined from the same data fit to the hypothesis of continued random movement (0.544 μm); this phase of movement was essentially linear.

Open in a separate window

Figure 2

Kinetics and structure of initial chromosome–MT interaction. (A, B) Trajectories of kinetochore movements prior to (red line) and after (blue line) the temperature shift to 32°C in two control cells. The images show movements in the plane that contained the kinetochore (red X's) and both spindle poles (see Materials and methods). Blue arrows point to the onset of the final movements that brought these kinetochores to the pole. Purple arrows point to pauses. Final positions for the kinetochores and poles are shown as red and green circles, respectively. The grid squares are 0.5 μm. (C) A slice from a tomogram, built from nda3 cell frozen during early stage of spindle formation. (D–F) Reconstructions of three spindles (frozen 3–7 min after the shift to 32°C), one of which is shown in (C). MTs nucleated from the different poles are colored green and red, spindle poles are yellow, and the nuclear envelope is gray. Minus ends are shown with orange dots, plus ends with purple dots; ends that lie between the sections are not marked. Occasionally, there were MTs that continued beyond the reconstructed sections; their most distal (plus) ends were marked with green dots. Bar: 0.1 μm.

We have used this data format to analyze both the persistence of kinetochore–MT attachments and the processivity of kinetochore P-movement. If a kinetochore frequently captured and lost MTs that grew from opposite poles, one would expect the kinetochore's trajectory to consist of segmented movements toward different poles. However, all persistent motions were directed towards only one of the poles (Figure 2B). They occurred at average rate of 0.6±0.1 μm/min (range 0.2–2.8 μm/min, N=8), and they were quite processive. The kinetochore moved without interruption for 1.0±0.1 μm (N=12), with a longest uninterrupted run of 1.5 μm. The pauses (Figure 2A and B; purple arrows) lasted for 124±30 s, which could have resulted either from a stalled movements or from kinetochore–MT detachment, followed by recapture.

P-movement in S. pombe occurrs via pole-originated MTs and is driven by kinetochore-dependent mechanisms

We measured the angles between the direction of the kinetochore's final trajectory and the spindle axis. Their range was broad (82±25°; M±s.d., N=28), suggesting that during prometaphase the SPBs nucleated MTs in a wide range of directions, broad enough to cover most pole-proximal regions in the nucleus. This view differs from the published description of more mature, metaphase spindles, which comprised a narrow bundle of MTs (Ding et al, 1993). We therefore used electron tomography (ET) to examine the MT distributions in the forming spindles. Early in prometaphase the duplicated SPBs enter a fenestra in the nuclear membrane (Ding et al, 1997; Uzawa et al, 2004). As the spindle grew in nda3 cells recovering from the cold block, the SPBs initially lay almost parallel to each other (Figure 2C and D). Their separation was probably driven by the few overlapping MTs that ran oblique to the SPB's nuclear surface (Supplementary Video 3). Other MTs projected in diverse directions into the nucleoplasm, which should maximize the chance of their encountering kinetochores (Sagolla et al, 2003). As the spindle matured and its length increased, the poles turned to face each other, and the number of overlapping MTs grew. There were still, however, many MTs of various lengths that grew in a variety of directions, reminiscent of a vertebrate spindle (Figure 2E; Supplementary Video 4). We found no examples in which MTs from both poles pointed towards the same, pole-distant region of the nucleus, as one might expect during merotelic chromosome attachment (Maiato et al, 2004). This is consistent with the suggestion that a lost kinetochore attaches initially to only one pole (in either a monotelic or syntelic configuration). The exact mode of attachment cannot, however, be determined, due to insufficient contrasting of yeast kinetochores in the EM. Although some regions near the MTs plus ends had structural features that made them candidates for kinetochores, these features were difficult to quantify (Supplementary Figure 3A). All such areas were at the ends of highly splayed MT plus ends, implying that the kinetochores interact with the polymer end, rather than with the MT's lateral surface.

In all spindles reconstructed by ET (N=24), every MT minus end lay in the vicinity of the SPBs (average distance 86±5 nm, N=70) (Supplementary Figure 3B). We found no evidence for MTs initiated at kinetochores, as happens in some systems (Khodjakov et al, 2003). At least 91% of the minus ends (N=483) were capped by a roughly conical structure (Supplementary Figure 3C), which resembled the γ-tubulin ring complex in both budding yeast and nematode spindles (O'Toole et al, 1999, 2003). These results suggest a stable attachment of the minus MT end to the SPB, consistent with the apparent lack of MT dynamics at this end (Mallavarapu et al, 1999). Thus, it is unlikely that pole-associated kinesins in S. pombe contribute to P-movements by destabilizing the MT minus ends (Rogers et al, 2004). We conclude that kinetochore P-movement in fission yeast are caused by MT-dependent, kinetochore-associated, minus-end directed motor(s) and/or by forces generated by depolymerization at the MT's plus end.

Neither dynein nor Pkl1p is required for normal kinetochore retrieval

We have used this experimental system to examine the contributions of different minus-end directed motors. Strains lacking the gene for Dhc1p, Pkl1p or Klp2p (all in the nda3 background) were synchronized as described above. By immunofluorescence of fixed cells, the overall parameters of the arrest were similar for all strains, indicating that none of these deletions interfered with the cell cycle block or kinetochore scattering (Figure 1A). In the dynein deletion strain, the kinetics of kinetochore retrieval and bi-orientation were not statistically different from those of the control strain (Figure 1B–D). Live cell imaging has also revealed normal kinetochore retrieval in this strain (N=23), though the mean retrieval time was slightly longer than in control cells (Supplementary Figure 4A).

Pkl1p is a S. pombe motor of the kinesin-14 family; it is normally found along the spindle and is enriched at the SPBs (Pidoux et al, 1996). During the initial stages of mitotic progression in pkl1Δ cultures (<5 min), the kinetics of kinetochore retrieval, bi-orientation and exit from mitosis were indistinguishable from those of the control strain (Figure 1B–D). However, in pkl1Δ the fraction of cells in prometaphase and metaphase decreased prematurely (Figure 1C), which led to a correspondingly higher percent of unretrieved kinetochores at the later times (5–15 min; Figure 1B). The slight advancement in mitotic exit suggested a precocious anaphase onset (Figure 1D); apparently kinetochores are retrieved normally in pkl1Δ cells, but there is a defect later in mitosis (Grishchuk and McIntosh, manuscript in preparation).

We have traced the 3D coordinates for kinetochores and poles in 5 pkl1Δ and 6 dhc1Δ cells. The trajectories of pole-directed kinetochore movements showed no significant deviations from those of control cells (Supplementary Figure 4B and C). The variations in the onset of kinetochore movement to a pole and its speed are almost as great within each of these genotypes as they are between the genotypes (Figure 3A). We conclude that normal P-movements can occur in the absence of either dynein or Pkl1p.

Open in a separate window

Figure 3

Movements during chromosome retrieval in live cells. (A, B) Example traces of the kinetochore-to-pole distances versus time after the shift to 32°C (see also Supplementary Figure 6). (C) Comparison of the slopes of final kinetochore approaches for the klp2Δ cell 1 (Supplementary Video 5), klp2Δdhc1Δ cell (Supplementary Video 12), one control cell and 4 pkl1Δ klp2Δ dhc1Δ cells.

Deletion of Klp2 leads to a delay in kinetochore retrieval, due to an increased heterogeneity in the kinetics of kinetochore movements

S. pombe's Klp2p is another member of the kinesin-14 family; it is the only minus-end directed motor that is known to localize to mitotic kinetochores (Troxell et al, 2001). Nda3 cells lacking Klp2p had a reproducible increase in the percent of lost kinetochores, suggesting that Klp2p contributes to normal kinetochore–pole attachments (Figure 1A). Following the return of these cells to 32°C, their kinetochores re-established association with SPBs significantly more slowly than cells of the three other genotypes (τ_1/2 20 min; Figure 1B). There was also a delay in the accumulation of kinetochores in prometaphase and metaphase (Figure 1C) and a slower mitotic exit (τ_1/2 35 min; Figure 1D), indicating normal operation of the mitotic checkpoint.

By following live mitosis in klp2Δ cells (N=30), we identified two factors that contribute to slower kinetochore retrieval. First, klp2Δ cells retrieve fewer distant kinetochores (1.2 μm from the poles) than control cells (Supplementary Figure 2B). Second, the average time required for kinetochore recovery from shorter distances was increased: 11.8±2.9 min in klp2Δ cells versus 6.8±1.1 min in control (Supplementary Figure 4A). Since the Klp2 motor is localized at kinetochores, it might contribute directly to the P-movement. One would then expect that the slower average rate of kinetochore recovery was due to a slower kinetochore movement. However, direct visualization of kinetochore motion showed that this was not so: some of the klp2Δ kinetochores were recovered efficiently and moved with the highest speed we observed (~3 μm/min; Supplementary Video 5). Movements of other kinetochores were greatly delayed, even though they were initially close to the poles (Supplementary Video 6). Thus, in the absence of Klp2p kinetochores can still move rapidly, but there is an increased probability of slower recovery. Since the klp2⁺ gene is deleted from these cells (Troxell et al, 2001), this behavior could not be due to variations in the levels of Klp2p. We have therefore explored other possibilities.

Klp2p promotes processivity of kinetochore movement by provoking kinetochore fiber shortening and/or preventing its elongation

To analyze mechanisms that led to movement heterogeneity in klp2Δ cells, we traced kinetochore–pole distances over time. Some records showed periods when the distance was unchanged or even increased (e.g. cells 2–5 in Figure 3B). This suggested that klp2Δ kinetochores were either defective in MT attachment and moved randomly until a new capture could occur, or that their P-movement stalled frequently and even reversed. To distinguish between these possibilities, we examined the 2D kinetochore trajectories. For example, the retrieval path for the kinetochore in cell 3 can be divided into two segments (Figure 4A). Initially, the trajectory resembles a circular arc with its center at the right pole. This model can be evaluated by the s.d. of the data fit to this hypothesis (0.141 μm), while the s.d. for a model based on random movement is 0.410. Such motions could arise in part if the kinetochore's movement were restricted by its proximity to the nuclear envelope. Alternatively, the kinetochore might have attached to one or more stable MTs originating from the right pole. The later segments of the path fit well to a line pointing directly toward that pole (s.d. 0.128 μm, s.d. for random fit 0.422), consistent with the latter model. These movements were not, however, processive; the kinetochore moved forward and back before finally arriving at that pole (Supplementary Video 7). The backward movement could have resulted from the chromatin recoiling after its kinetochore detached from an associated MT(s), but the conspicuous absence of random kinetochore motions, which one would expect for a detached chromosome, suggests that during retrieval this klp2Δ kinetochore maintained a stable attachment to a kinetochore MT. This prediction was tested by performing ET on 4 prometaphase spindles from nda3 klp2Δ cells delayed in mitotic progression. All of them contained 1–2 MT bundles that diverged from the spindle axis, each consisting of 2–3 MTs (a pombe kinetochore binds 2–4 MTs (Ding et al, 1993; Figure 4B and C; Supplementary Video 8). Although individual kinetochores cannot be located in these tomograms, the positions of the MT plus ends in these bundles (identified through ET) were similar to the positions of kinetochores during their retrieval in klp2Δ cells, as seen via live imaging. Together, these results suggest that klp2Δ kinetochores capture MTs early after the shift to 32°C, but they pause in their P-movement and frequently move away from that pole.

Open in a separate window

Figure 4

Klp2p facilitates kinetochore retrieval. (A) 2D trajectory of kinetochore retrieval in cell 3. The best fitting circular arc (magenta) for frames 1–50 (Supplementary Video 7); a best linear fit (blue) for frames 51–96. Tomographic slice (B) and 3D model (C) of nda3klp2Δ cell delayed in prometaphase (frozen 15 min after the shift to 32°C). Arrow in (B) points to one of the MTs in a two-MTs bundle growing from the right pole. Bar: 100 nm. (D) Average numbers of MTs per spindle pole for control nda3 cells (4 poles), nda3dhc1Δ (10), nda3pkl1Δ (6) and delayed nda3klp2Δ cells (6). (E) ET slices of the MTs plus ends. The first five images were collected from a control nda3 cell; all the rest are from six klp2Δ cells. There is a great variability in the appearance of the plus ends in both normal and mutant cells, which makes it difficult to quantify differences. Usually, the plus MT ends are splayed open (upper row), but in klp2Δ cells they frequently contain occlusions (red arrows). The images in middle row may represent intermediates that lead to ‘scarred' MTs with altered polymerization behavior; the bottom row suggests a possible pathway for the rescue of rapid shortening at an MT abnormality, followed by renewed MT polymerization. Bar: 100 nm. (F) Percent of MTs plus ends that show darker ‘bars' inside the MT lumen. Number of ends examined: control 68, klp2Δ 239. (G) Selected live images (30 s apart) from an nda3 klp2Δ dhc1Δ cell undergoing anaphase A. Bar: 2 μm.

The inability of these cells to sustain P-movements and the tendency of their chromosomes to switch to movements away from the pole appear to be the primary causes of their delayed chromosome retrieval. For example, the kinetochore in cell 4 (Figure 3B) was initially close to the unseparated poles. Immediately after the shift to 32°C, this kinetochore began moving away from the pole at 0.5 μm/min before moving back again and bi-orienting (Supplementary Video 9). In cell 5 (Figure 3B) the kinetochore began with movement towards the pole, then abruptly switched its direction and moved away (Supplementary Video 10). Our ultrastructural analysis of spindles in delayed klp2Δ cells revealed a small increase in the number of MTs growing from each pole relative to wild type (Figure 4D). This difference might result from a longer time spent in mitosis, or the MTs in these cells may show altered MT dynamics (Troxell et al, 2001). Close examination of the MT ends in this strain showed that their protofilaments tended to be less splayed than in other strains (Figure 4E). Many MTs showed occlusions, for example, dark ‘bars' across their lumens (Figure 4F). These structures probably represent proteinaceous material, perhaps protofilament pieces, trapped in a growing MT (Figure 4E, gray arrows). Such irregularities may arise during normal MT dynamics, forming ‘scarred' MTs whose depolymerization is partially impeded. Perhaps Klp2p is required for MT shortening when depolymerization encounters such abnormalities; in its absence, depolymerization rescue and subsequent MT elongation at the kinetochore may be more common.

Live imaging

The centromere of chromosome 2 was visualized using the cen2-GFP construct (Supplementary Table 1), kindly provided by A Yamamoto (Yamamoto and Hiraoka, 2003). The immunolocalizations of the cen2-GFP marker and of the kinetochore component Mis12-HA are always either co-localized or immediately adjacent (e.g. see Supplementary Figure 1g). Spindle poles were visualized with Pcp1p-GFP (Flory et al, 2002). The microscope stage was precooled to 14°C with a custom made water jacket of metal tubing wrapped around an oil-immersion × 100 NeoFluar objective and oil-immersion condenser. Temperature shift occurred within 1–2 min after switching the tubing of the water jacket from a 14°C to a 32°C bath. Images were acquired as stacks of 7–12 planes (0.25 μm step, 150 ms exposure) every 15 s with a Photometrics Cascade 650 CCD camera (Roper Scientific) controlled via the Metamorph (Universal Imaging). The videos are the averaged, depixelated XY-projections manually aligned to minimize the image drift brought about by the abrupt temperature changes. In all videos, except Video 1, the temperature was shifted to 32°C after the third frame. Since S. pombe nuclei/spindles frequently rotate around randomly oriented axes, the changes in apparent distance between objects in these 2D projections do not necessarily reflect actual changes in separation, and additional image processing was necessary to obtain accurate distances between objects in the images.

Image analysis

All computer programs were custom written using MatLab 6.5. Briefly, the XYZ coordinates of the labeled kinetochore and of the two poles at each time point (collected manually by selecting the apparent centers of the brightest signals in each stack) were transformed, so one of the poles was always at the center of a Cartesian coordinate system. The direction of the X-axis was chosen to run through the second pole; the Y-axis lay in the plane that contained all three objects. The resulting images display the true 3D distances. This program also eliminates movements caused by stage drift or spindle rotations. In the trajectories the points are moving averages of four consecutive points.

The ET was carried out in the Boulder Laboratory for 3D EM of cells, and was supported by RR000592 to JRM. M Morphew prepared the samples for EM and acquired TEM images of nda3 cells at 18°C; E O'Toole and other members of this lab helped with data acquisition and reconstruction. The temperature control for light microscope was built by V Sarbash and F Ataullakhanov. Strains and reagent were kindly provided by Drs I Hagan, P Silver, K Gull, Y Hiraoka, M Yanagida and T Davis. We are grateful to I Spiridonov for technical assistance, J Becerril for help with collecting 3D coordinates, to M Molodtsov and F Ataullakhanov for the software used to analyze kinetochore trajectories, and to M Winey for helpful discussions. This work was supported in part by GM33787 to JRM.