Kinetochore motility after severing between sister centromeres using laser microsurgery: Evidence that kinetochore directional instability and position is regulated by tension

When a chromosome first associates with microtubules during mitosis in vertebrate cells, the kinetochore attached to microtubules (MTs) predominately exhibits poleward (P) motion. At some position nearer to the pole, the attached kinetochore exhibits abrupt switches between similar durations of constant velocity P and away from the pole (AP) motility states termed directional instability (Skibbens et al., 1993; Cassimeris et al., 1994). It has been shown in mitotic newt cells that, when the unattached sister kinetochore captures MTs from the opposite pole, the centromere congresses to the equator, by coordinated sister kinetochore motion (one kinetochore in P movement, the other in AP movement)(Skibbens et al., 1993). When sister chromatids disjoin at anaphase onset, kinetochores persist in P movement, pulling their chromatids poleward. Kinetochore AP motion, and the majority of P motion, is tightly coupled to the growth and shortening of kinetochore microtubules (kMTs) at their plus-end attachment sites (Mitchison et al., 1986; Wise et al., 1991) and occurs along relatively stationary kMTs (reviewed by Rieder and Salmon, 1994), although kinetochore P motility is enhanced by a slow flux of the kMT lattice (Mitchison and Salmon, 1992).

To understand how chromosomes congress to the spindle equator in prometaphase and segregate to the poles in anaphase, it is important to know how switching between kinetochore P and AP motility states is controlled for each kinetochore and how the motility of sister kinetochores is coordinated (Mitchison, 1988; Skibbens et al., 1993; Rieder and Salmon, 1994; Murray and Mitchison, 1994). Obtaining the answers to these questions requires understanding what kinetochores ‘sense’ within the spindle and how this information is used at the molecular level to control switching between P and AP motility states.

We have hypothesized that tension at the kinetochore is a key factor which controls switching between P and AP motility states and that tension across the centromere coordinates sister kinetochore directional instability (Skibbens et al., 1993; Cassimeris et al., 1994; Rieder and Salmon, 1994). In our Kinetochore Motor/Polar Ejection model, kinetochore tension is generated by the stretch of the centromere chromatin which attaches sister kinetochores together and tethers kinetochores to their chromosomes arms. The amount of stretch (or compression) of the centromere depends on the motion of sister kinetochores relative to each other and to the chromosome arms. Tension is viewed as an ‘upstream’ regulator of molecular mechanisms, such as kinetochore phosphorylation, which have been postulated to control the direction of kinetochore motion (Hyman and Mitchison, 1991; Gorbsky and Ricketts, 1993; Davis et al., 1994; McIntosh, 1994). The key premises of this tension model are that: (i) the velocities of kinetochore P and AP motility are insensitive to tension and are independent of position relative to the spindle poles; (ii) the durations of kinetochore P and AP motility states are sensitive to tension where high tension promotes switching from P to AP motility and low tension (or compression) promotes switching from AP to P motility (directional instability is due to hysteresis); and (iii) polar microtubule arrays push chromosome arms AP (polar winds or ejection forces) and resist poleward chromosome movement with a magnitude which increases with microtubule density nearer the poles. Consistent with this model, reducing the amount of tension exerted on a kinetochore by ablating its sister kinetochore on a bi-oriented chromosome results in the undamaged sister pulling the chromosome poleward away from the spindle equator (McNeill and Berns, 1981; Rieder et al., 1986; Hays and Salmon, 1990). Similarly, severing the bulk of the arms from a mono-oriented chromosome produces poleward migration of the centromere region and ejection of the acentric arms (Rieder et al., 1986; Ault et al., 1991). On the other hand, increasing the amount of tension at a kinetochore by stretching one of its chromosome arms results in kinetochore movement away from the pole (Nicklas, 1977; Skibbens, 1994).

In this paper, we investigated how kinetochore directional instability is affected when the centromeric chromatin that tethers sister kinetochores together and to the bulk of the chromosome arms is severely weakened. The kinetochore-tensiometer model predicts that lowering the stiffness of the centromeric chromatin on bi-oriented chromosomes should: (1) produce ‘anaphase-like’ sister kinetochore separation in prometaphase and disrupt mitotic kinetochore congression since the centromere must stretch much further than normal to achieve the tension necesssary to induce P to AP switching; (2) reduce the coordination of sister kinetochore directional instability because the compliant linkage between sisters reduces the influence of the motion of one sister on the other; and (3) change the durations but not the velocities of kinetochore P and AP motility states because the tension needed to induce P to AP switching is much less than that needed to stall kinetochore P motility. To test these predictions, we used a 0.3 µm diameter green light laser microbeam (Brenner et al., 1980; McNeill and Berns, 1981; Rieder et al., 1986; Hays and Salmon, 1990; Ault et al., 1991) to severely ablate the centromeric chromatin that tethers sister kinetochores together and to the bulk of the arms on bi-oriented chromosomes in mitotic newt lung cells. This microsurgery left sister kinetochores intact and produced two mono-oriented, single kinetochore-chromatin fragments attached to the bulk of the arms only by thin, compliant chromatin strands. The motility of these single kinetochore-centromere fragments relative to a spindle pole and to each other were then analyzed using the semi-automatic digital image tracking system previously employed to track kinetochore directional instability on intact mono-oriented and bi-oriented chromosomes (Salmon et al., 1991; Skibbens et al., 1993).

Newt lung cells were cultured as previously described (Rieder et al., 1986; Rieder and Hard, 1990). Briefly, Oregon newts (Taricha granulosa) were sacrificed and the lungs excised, minced and placed into Rose chambers containing L-15 medium supplemented with FCS and antibiotics. After about 10 days, a monolayer of cells had spread from the explant that contained mitotic cells suitable for experimentation.

Experiments were performed at the NIH sponsored LAMP facility at Irvine, CA. Laser ablations were performed as previously described (Rieder et al., 1986) using the equipment detailed by McNeill and Berns (1981). Briefly, 532 nm light from a Neodynium YAG (yttrium, aluminum, garnet) laser was focused to a 0.3 µm diameter microbeam and used to sever the chromatin between sister kinetochores and between kinetochores and the bulk of the chromosome arms. Each energy pulse contained about 130 nJ and was emitted at 10 pulses/second. A maximum of 49 µJ (assuming continuous ablation for 38 seconds) was required to effectively sever between sister kinetochores with the exposure usually being much less.

Phase contrast images were obtained using a Zeiss Axiomat microscope equipped with a Zeiss 100×/1.33 NA Neofluar Phase 3 or 63×/1.4 NA Planapochromat Phase 3 objective, matching condenser, Omega 605 nm, 50 nm bandpass filter, heat cut and heat reflecting filters and a 60 W tungsten light source. Video images were generated using a Video Standard camera (Sierra Scientific, Mountain View, CA) and recorded in real time on either a K inch U-matic tape (Sony TVO 9000 time-lapse video cassette recorder) or G inch S-VHS tape (Panasonic TL AG-6750A recorder, VCR).

Images from the VCR were time lapsed (one frame every 2 seconds) on a Panasonic TQ2028F Optical Memory Disc Recorder (OMDR) for analysis. Data used to generate distance vs time plots were obtained using the semi-automated tracking device and motion analysis program previously described (Skibbens et al., 1993). Briefly, a computer generated 8×8 pixel cursors was placed over each object to be tracked (typically a kinetochore-chromatin fragment and the associated pole) on images displayed from an OMDR. The gray values in each 8×8 cursor were stored as templates. The computer then advanced the OMDR one frame and searched each new image for an 8×8 pixel array that best matched the original template. Positions and time intervals were stored and read out later for analysis. A program written in-house was used to convert position coordinates to distance between objects. Velocities and durations were obtained via regression analysis from distance vs time plots generated using another program developed in-house, Single Frame Movement (SFM). For analysis of sister kinetochore coordination, sister kinetochore movement was plotted and superimposed on one time axis. Switching was then analyzed in overlapping 10 second windows; indeterminate kinetochore movements (durations where neither P nor AP movement were identifiable) were excluded from the data set.

To ensure that laser microsurgery of the centromere did not damage either the kinetochores or their kinetochore fibers, we initiated cutting only when both sister kinetochores were in P movement, stretching the centromere as shown in the examples in Figs 1 and 3. As the centromere became stretched, a central phase-translucent cleft (Rieder, 1990) usually became apparent (Fig. 1a and b, 0 seconds). The laser beam path was centered over this cleft and several laser pulses were fired into this region. As kinetochore P movement further separated the kinetochores, we often continued cutting until much of the chromatin linking the sisters together and to their arms was severed. We analyzed only those cells (a total of 8) where we were sure that the laser microbeam never hit the sister kinetochore regions. As found in the previous studies using the Bern’s laser microbeam apparatus (McNeill and Berns, 1981; Rieder et al., 1986; Hays and Salmon, 1990), the effects of the laser ablation appeared limited only to the targeted chromatin without affecting spindle integrity, anaphase or cytokinesis. For comparison in the following experiments, the rest length between sister kinetochores of unattached chromosomes is about 1.1 µm and the centromere on bi-oriented chromosomes near the equator is rarely stretched beyond 3 µm before one sister switches to AP movement, relieving the stretch (R. V. Skibbens, J. C. Waters and E. D. Salmon, unpublished observation).

An example of cutting the central centromere region of a bioriented chromosome near the equator is shown in Figs 1 and 2. In this experiment, only a few pulses were required to sever the region between sister kinetochores as evidenced by the immediate migration of each sister kinetochore toward its associated pole (Fig. 1a and b, 89 seconds). This procedure generated two mono-oriented single kinetochore-chromatin fragments that were tethered to the bulk of the arms by thin compliant chromatin strands (Fig. 1a). No tether directly linking the two sister kinetochores, independent of the arms, could be detected. During and immediately following laser microsurgery, sister kinetochores exhibited predominantly P movement and eventually became separated by about 20 µm; about half of the 44 µm interpolar distance (Figs 1b and 2, 40400 seconds). At some distance closer to their poles, each sister again began to oscillate between similar durations of P and AP movement (Fig. 2, 400-1140 seconds). During this time, the arms of the chromosome remained near the spindle equator(Fig. 1b, 89-914 seconds). During kinetochore P movements, the chromatin strands that tethered each kinetochore to the arms (most noticeable in Fig. 1b, 89 seconds) became stretched 10 µm or more before the kinetochore switched back to AP motion. As the kinetochores moved AP, the thin chromatin strands tethering each kinetochore to the bulk of the arms shortened, but did not noticeably buckle before the kinetochores switched back to P movement.

In addition to increasing the average distance between sister kinetochores, severing the central centromere region also inhibited the congression of the sister kinetochores, but not their arms, to the spindle equator. An example of the movements of sister kinetochores on a congressing chromosome during and after central centromere ablation are shown in Figs 3-6. In this cell, centromere congression was followed until both sister kinetochores switched to P movement and stretched the centromere (Fig. 3a and b). We then used the laser microbeam to cut through the centromere and further sever the chromosome in half (Fig. 3a and b, 13-90 seconds). During this ablation, both kinetochores persisted in P movement.

The lower kinetochore initially moved closer to the pole, pulling the severed arms into the half spindle (Fig. 3b, 13-90 seconds). Ablation just behind the kinetochore (Fig. 4, 100-188 seconds) severely decreased the amount of chromatin that tethered the single kinetochore-chromatin fragment to the bulk of the severed arms (chromatin strands most noticeable in Fig. 3b, 421 seconds). During this second ablation, both the lower kinetochore-chromatin fragment and the severed arms exhibited AP movements (Figs 3 and 4). After microsurgery, the kinetochore progressively moved closer to the pole due to longer P movements, compared to AP movements (Fig. 4). Furthermore, the motilities of the kinetochore and severed arm appeared largely uncoupled; the arm slowly moved AP toward the metaphase plate and did not oscillate in parallel with the kinetochore. This uncoupled motility is especially clear where the arm moved slightly AP while the kinetochore moved P (Fig. 4, 550 seconds).

After the initial cuts, the kinetochore tethered to the upper half-spindle became obscured as it moved into the mass of chromosomes at the metaphase plate (Fig. 3b, 90-833 seconds). This kinetochore migrated about 30 µm from the initial laser site as it moved through the chromosome mass at the spindle equator and part of the way into the upper half spindle (Fig. 5, 1265-2021 seconds). Similar to the lower kinetochore, once this single kinetochore-chromatin fragment moved to about 5 µm from its pole, it exhibited directional instability, oscillating between P and AP phases of motion of similar durations (Figs 5 and 6, 1650-2121 seconds).

To quantitatively assay for the coordination between sister kinetochores freed from the normal centromeric linkage by laser microsurgery, we tracked the movements of several sister kinetochores, post-ablation, relative to one pole and superimposed these movements on one time axis (Figs 2 and 3). Of the 8 cells where each sister kinetochore and at least one spindle pole were visible after ablation, 4 cells contained sequences of extended duration that were amenable to high resolution tracking. These measurements exemplify the motility observed in the other cells.

As previously described for kinetochores on intact chromosomes (Skibbens et al., 1993), the degree to which sister kinetochore movements were coordinated was based on two criteria: the percent of time both kinetochores exhibited coordinated out-of-phase movement (one in P and the other in AP) relative to in-phase movement (both P or both AP); and the number of sister kinetochore switches that occurred within any given 10 second window (synchronous) versus switches that did not occur within 10 seconds of each other (independent). Analysis of 8 sister kinetochore-chromatin fragments by the first criteria show that, following laser microsurgery on the centromere, sister kinetochores moved out-of-phase 35.3% of the time because of the persistent inphase P movement of both sisters following centromere severing. In contrast, the sisters on intact bi-oriented chromosomes moved out-of-phase 75% of the time (Table 1). When the initial persistent P movement following severing is excluded from the data set, the analysis reveals that sister kinetochore movements were indeed mostly uncoordinated in that they moved out-of-phase 44.1% of the time; there was still a slight bias for net P movement by both sisters (Table 1). Analysis of 4 sister kinetochore-chromatin fragments in 2 cells by the second criteria showed that single kinetochore-chromatin fragments exhibited 19.5% out-of-phase switches and 29.3% in-phase switches within 10 seconds of each other in comparison to sister kinetochores on intact bi-oriented chromosomes which exhibited 38.5% out-of-phase switches and 7.7% in-phase switches (Table 1). The incidence of single switches (no corresponding switch by the sister kinetochore within a 10 second window) exhibited by kinetochores on bi-oriented chromosomes did not change after centromere ablation (48.8% vs 46.2%) (Table 1).

A fundamental premise of the tension model for controlling directional instability is that the durations but not the velocities of kinetochore P and AP motility are sensitive to tension (Nicklas, 1965; Skibbens et al., 1993, and see Rieder and Salmon, 1994). To test this hypothesis, we performed regression analyses on the plots of movements of 12 single kinetochore-chromatin fragments in 9 cells. The velocities and durations of several such single kinetochores were then compared to those values measured previously for kinetochores on bi-oriented chromosomes prior to anaphase and on segregating chromosomes through mid-anaphase A (Skibbens et al., 1993). The data show that the P and AP velocities of kinetochore-chromatin fragments were nearly identical to the velocities of kinetochores on intact chromosomes during prometaphase, metaphase and the first half of anaphase, about 2 µm/minute on average (Table 2). Immediately after severing between sister centromeres during prometaphase, sister kinetochore-chromatin fragments moved toward their associated poles in a similar fashion as the kinetochores on segregating chromosomes in early anaphase A: by prolonged durations of P movement. However, once new positions around which kinetochore-chromatin fragments oscillated were established, those fragments exhibited durations and velocities of P and AP movement indistinguishable from those of kinetochores on intact mono- and bi-oriented chromosomes (Table 2).

Following the initial cutting procedure, kinetochore-chromatin fragments often exhibited P and AP displacements of 4 µm or more before abruptly switching direction (Figs 2, 4 and 6). These P and AP displacements were noticeably greater than the 2 µm average displacements exhibited by sister kinetochores on intact chromosomes (Table 2). One explanation for these extended displacements is that, compared to an intact centromere, the compliant chromatin strands that tether the kinetochores to the bulk of the arms require a greater amount of stretch to generate sufficient tension to promote switching to AP movement and a corresponding greater amount of shortening before the tension is reduced enough to promote switching back to P movement. To test this possibility, we further ablated the chromatin strands just behind the lower kinetochore-chromatin fragment described in the second cell (Fig. 4, 690-770 seconds). Ablation just behind the kinetochore resulted in the net migration of the kinetochore further toward the pole and a corresponding increase in the stretch of the chromatin tether (Fig. 3, 883 seconds, and Fig. 4, 800-970 seconds). Both types of switch events were effected by this cutting: distances from the pole at which P-to-AP switches occurred decreased 1.2 µm (from 4.5 to 3.3 µm) and AP-to-P switches decreased 2.4 µm (from 7.2 to 4.8 µm) (Fig. 4).

The single kinetochore-chromatin fragment attached to the upper half-spindle became visible only after it had migrated through the tangle of chromosomes at the metaphase plate (Fig. 5). Compared to the kinetochore in the lower half-spindle, this kinetochore remained tethered to one of the chromosome arms by a more phase-dense chromatin linkage and pulled the arm part way through the metaphase plate and into the upper half spindle (Figs 5 and 6, 1390-2230 seconds). As observed for the lower kinetochore, even though the motilities of the kinetochore-chromatin fragment and chromosome arms appeared uncoupled (Fig. 6, 1390-2230 seconds), ablation behind the kinetochore resulted in movement of the arm toward the metaphase plate (at 0.9 µm/minute) and further movement of the single kinetochore toward its associated pole (Fig. 6, 22502460 seconds). The distance from the pole where P-to-AP switches occurred was not greatly altered, possibly because the kinetochore was already close to the pole. However, by weakening the chromatin tether, the distance from the pole where AP-to-P switches occurred decreased markedly - from 7 to 4 µm (Fig. 6). Thus, further destruction of the chromatin tethering the single kinetochore-chromatin fragments to the bulk of the chromosome arms altered the durations of both the P and AP excursions.

To further demonstrate that the chromatin linkage regulates kinetochore congression and position, we tried to generate kinetochore-chromatin fragments with variable linkages back to the chromosomes arms. Because of the high severing capability of the laser microbeam, quantitative severing was difficult to achieve. A single example follows. We observed the movements of the centromere of a bi-oriented chromosome during congression (note the initial proximity to the pole - Fig. 7, 0-170 seconds). Each sister kinetochore initially moved P as a result of our first attempt at severing between sister centromeres (Fig. 7, 172-228 seconds). However, chromatin strands, independent of the chromosome arms, still tethered the sisters together and allowed congression of the kinetochores to proceed (Fig. 7, 230-370 seconds); the kinetochore associated with the upper pole becoming untrackable as it entered the mass of chromosome arms near the spindle equator. We attempted to sever this chromatin (Fig. 7, 370-402 seconds), but succeeded in only decreasing the density of the chromatin strands that directly linked the sister kinetochores together. After this second ablation, the kinetochore associated with the lower pole stopped its net motion toward the metaphase plate and moved P for about 4 µm. The kinetochore-chromatin fragment then oscillated P and AP at a position about 12 µm from the pole and 8.5 µm from the center of the metaphase plate (Fig. 7, 4021000 seconds). This position is much further from the pole than the positions around which many intact mono-oriented chromosomes and other single kinetochore-chromatin fragments, freed from their sister kinetochores, oscillate (Figs 1-6).

An important prediction of our tension-based model is that a stiff centromere linkage is required to coordinate sister kinetochore activity. Coordinated motility (sister kinetochores on intact bi-oriented chromosomes exhibited coordinated motility 75% of the time) was previously postulated to be essential for kinetochore congression based on the observation that uncoordinated sister kinetochore motility resulted in either stretching or compressing the centromere without achieving net centromere displacement (Skibbens et al., 1993). The data obtained in this study shows that severely weakening the chromatin between sister kinetochores on bi-oriented chromosomes near the metaphase plate results in sister kinetochore motilities becoming uncoordinated (Table 1), even after the initial P-biased movement that typically occurs after centromere ablation is excluded from the data set.

One consequence of this loss of coordination is that kinetochore congression to the metaphase plate, as normally observed for vertebrate cell mitosis, was also inhibited by severing the centromere (Figs 3 and 4). This result is consistent with laser ablation studies where disruption of one sister kinetochore on bi-oriented chromosomes was followed by chromosome movement off the metaphase plate and toward the pole attached to the undamaged kinetochore (McNeill and Berns, 1981; Rieder et al., 1986; Hays and Salmon, 1990). Similarly, when grasshopper spermatocyte sister chromatids were prematurely separated via micromanipulation during meiosis II (but that still retained phase translucent chromatin tethers between the sisters - Bruce Nicklas, personal communication), those chromatids moved a short distance off the metaphase plate and remained there until anaphase onset - at which point the chromatids segregated the rest of the way toward the poles (Nicklas, 1967). Thus, the positioning and maintenance of kinetochores on bi-oriented chromosomes near the metaphase plate in mitosis requires tightly linked opposing sister kinetochores.

Based on these results, kinetochores tethered by chromatin whose elasticity is intermediate between stiff mitotic centromeres and laser-induced compliant chromatin strands should exhibit an intermediate level of coordination. The kinetochore regions on meiotic chromosomes provide a natural case study. In contrast to mitotic chromosomes, meiotic chromosomes are held together by one or more chiasmata that are located at variable positions along the chromosome arms but seldom occur near the centromere. Thus, homologous kinetochore regions of bivalents are not linked by short stiff centromeres, but through long chromosome arms. Seto et al. (1969) have previously recorded the movements of homologous kinetochore regions in meiotic newt spermatocytes. These kinetochore regions are tethered together through long chromosome arms - which provide a link intermediate in stiffness to intact mitotic centromeres and the thin compliant strands generated in this study. Our analysis of the movements of the homologous kinetochore regions, from plots published by Seto and coworkers (1969), shows that these kinetochores exhibit coordinated motility 60% of the time (based on 1,925 seconds of homologous kinetochore motility). This value falls between the per cent of time that tightly linked sister kinetochores on bi-oriented chromosomes (75%) and either loosely tethered sister kinetochores on bi-oriented chromosomes or unlinked adjacent kinetochores on intact mono-oriented chromosomes (about 50%) exhibited coordinated motility (Table 1). This analysis, and the results presented in this study (see Fig. 7), show that a continuum of coordination exists for kinetochores linked together and that the level of coordination directly depends on the compliance of the tether.

The second test of our tension-based model is that decreasing the stiffness of the centromere would cause sister kinetochore-chromatin fragments to move apart and establish a new position around which to oscillate in the spindle. Our results show that sister kinetochores on bi-oriented chromosomes pre-maturely separate and move poleward when their centromere regions are experimentally disjoined (Figs 1 and 2). This P displacement occurred in the absence of the cell cycle biochemical changes associated with the onset of anaphase (i.e. inactivation of MPF - see Murray, 1992, for review) and concurrent changes in spindle dynamics (Snyder et al., 1991), kMT turnover (Gorbsky and Borisy, 1989) and tenacity of kinetochore/MT attachment (Nicklas and Staehly, 1967). Thus, kinetochore P movement during anaphase requires only the physical disjunction of sister chromosomes (or ablation of one of the sister kinetochores - McNeill and Berns, 1981; Rieder et al., 1986; Ault et al., 1991) and does not require progression of the cell cycle such as exiting mitosis. A similar finding was reported by Holloway et al. (1993) and Surana et al. (1993) who showed that kinetochore segregation can occur during elevated levels of MPF activity in cycling Xenopus extracts and budding yeast, respectively.

An important premise to our model is that this shift in position will occur by regulating the durations without altering the velocities of kinetochore P and AP motility. Nicklas has shown for grasshopper meiosis I spermatocytes that the velocity of kinetochore poleward movement is insensitive to drag forces on the arms (1965) and to artificial stretching of the arms over a wide range of load force (1983). From this work, Nicklas concluded that kinetochore velocity in meiotic spermatocytes is ‘load independent’. Our results extend this view to include vertebrate mitotic cells and updates the concept of velocity-load independence to include both P and AP states for kinetochore-chromatin fragments relatively devoid of the bulk of the chromosome arms. We found no differences in either P or AP velocities over extended excursions (Figs 3,6,8 and 9) of single kinetochore-chromatin fragments in comparison to the velocities of kinetochores on intact chromosomes (Table 2), even though intact metacentric chromosomes were a minimum of 17 times longer in total length. These results clearly show that directional instability is a property of the kinetochore, separable from the motility of chromosome arms, such that kinetochores move at constant P and AP velocities similar to kinetochores on intact chromosomes - regardless of chromosome size or proximity to the spindle pole.

In our severing experiments, we found that single kinetochore-chromatin fragments, tethered to the bulk of the arms by thin compliant chromatin strands, immediately moved off the metaphase plate and eventually oscillated around a distance that was about twice as far from the pole as kinetochores on intact mono-oriented chromosomes in the same cells (data not shown). One explanation for why the single kinetochore-chromatin fragments remained further away from the pole than intact chromosomes is that the chromatin strands between the kinetochore-chromatin fragments and the severed arms acted to keep the kinetochores within the high density MT region of the central spindle where strong ejection forces exist (Cassimeris et al., 1994; Rieder and Salmon, 1994). This interpretation is consistent with the observation that mono-oriented chromosomes do not have such tethers and are free to swing out of the spindle into regions where MT density and ejection forces are low and there they are able to move closer to the pole (Rieder and Salmon, 1994). Conversely, mono-oriented chromosomes located within the spindle region and embedded in a high density of polar MTs can be as far from the pole as bi-oriented chromosomes that have congressed to near the metaphase plate (Cassimeris et al., 1994). These observations suggest that ejection forces associated with polar MT arrays, even in the absence of chromatin tethers, impose load at the kinetochore/MT attachment site and thereby regulate kinetochore directional instability.

An equally likely explanation for why the kinetochore-chromatin fragments remained further away from the pole than intact chromosomes is that kinetochore directional instability is predominantly regulated by the elasticity of the chromatin tether: i.e. when the tether is stretched taut, the kinetochore switches to AP motion regardless of its position relative to the spindle pole. This explanation is supported by the observations that: (i) severing the chromatin between sister kinetochores inhibited kinetochore congression (Figs 3 and 4); (ii) kinetochore-chromatin fragments immediately moved P after severing between kinetochores (Figs 2 and 4) regardless of the dense array of polar MTs that appears normally associated with chromosomes (Cassimeris et al., 1994); (iii) kinetochore-chromatin fragments were occasionally observed to pull their associated severed arms well into the half-spindle (data not shown); and (iv) kinetochore-chromatin fragments moved closer to their poles (within 2-3 µm) when their chromatin tethers back to the arms were further weakened by ablation (Figs 4 and 6).

In our severing experiments, kinetochore-chromatin fragments moved off the metaphase plate and towards the pole despite the maintenance of the spindle geometry: the bulk of the arms remained at the metaphase plate and the integrity of the spindle apparatus remained unchanged. These results add to the list of evidence that kinetochore motility cannot be explained solely by a chemical gradient of regulatory factors maintained between the chromosomes at the metaphase plate and the spindle poles (reviewed by Skibbens et al., 1993; Cassimeris et al., 1994; Rieder and Salmon, 1994). Because the 0.3 µm microbeam was used to ablate centromeric chromatin only when sister kinetochores were well separated, it is unlikely that kinetochore movement off the metaphase plate resulted from altering the number of kinetochore microtubules (Hyman and Mitchison, 1991) or by altering a gradient of regulatory factors located along kinetochore microtubules (McIntosh, 1994). Our results, in combination with the observations that stretching chromosome arms with microneedles induces centromere movement in the direction of stretch (Nicklas, 1977; Skibbens, 1994), show that kinetochore tension is a major factor controlling kinetochore directional instability during congression and segregation.

How do kinetochores sense tension in order to switch between motility states? Starting at rest, kinetochore attachment results in P movement and stretching of the centromere until the kinetochore switches to AP motility. AP motility reduces the stretch until the kinetochore again switches to P motility. Proposed mechanisms must account for this hysteresis which is typical of kinetochore directional instability. In addition, any model must also explain how kinetochores abruptly switch between constant velocity P and AP motility states. Finally, switching between motility states must be explained in terms of both P directed forces (kMT depolymerization and/or minus-end directed MT motors or tethering devices) and AP directed forces (kMT polymerization and/or plus-end directed MT motors) (see Rieder and Salmon, 1994; Murray and Mitchison, 1994, for reviews).

There are two categories of potential tension-sensing mechanisms: (i) one where tension directly regulates the P and AP force producers located at the kinetochore; and (ii) one where a molecular complex within the kinetochore detects tension and in turn regulates the P and AP force producers located at the kinetochore via post-translation modifications such as phosphorylation. The first model, previously described by Skibbens et al. (1993), Rieder and Salmon (1994), and Murray and Mitchison (1994), is consistent with evidence that directional instability for MT gliding in vitro is produced by the activity of one type of motor inhibiting the activity of the other (Vale et al., 1993) and that modulating kMT assembly dynamics in vivo and in vitro alters the kinetochore motility state (Bajer, 1982; Cassimeris et al., 1990; Koshland et al., 1988; Coue et al., 1991; Ault et al., 1991; Shelden and Wadsworth, 1992). In the second model, presented below (Fig. 8), we postulate that a kinetochore tensiometer protein complex senses tension and in turn regulates directional instability.

We propose that at low tension, a tension-sensitive protein complex located at the kinetochore attachment site spontaneously adopts an energetically favored closed conformation. This conformation switches kinetochores to P motion by putting kMTs into the shortening state (curved tubulin-GDP protofilaments) with concurrent or subsequent net minus-end directed motor activity (Fig. 8A). Short high-energy bonds within the tension sensor maintains the closed conformation during the P motion which stretches the centromere (Fig. 8B). At some threshold force, tension overcomes the short high-energy bonds of the tension sensor and opens up the protein conformation. The open conformation switches the kinetochore to AP motion by putting kMTs into the growth state (straight tubulin-GTP capped ends) with concurrent net plusend directed motor activity (Fig. 8C). The kinetochore persists in AP motion, reducing the centromere stretch (Fig. 8D) until tension is low enough to allow weak long-range forces within the tension sensor to pull it into a conformation such that the high-energy short bonds snap the tension sensor back into the closed conformation, switching the kinetochore into the P state (Fig. 8A). One way to mechanically model this tension sensor is to orient a pair of bar magnets so that they attract each other.

Once together (closed conformation), it takes a relatively high force to pull the magnets apart, compared to the force necessary to keep them apart (open conformation). When the magnets are brought sufficiently close together, the strong short-range magnetic forces snap the bars back together (closed conformation). Whether the tension sensor effects kinetochore motility by post-translation modification of factors that control the assembly reactions of kMT plus-ends or MT motors is unknown.

Although such a tension sensor that effects kinetochore motility by post-translation modification has yet to be identified, analogous protein signaling complexes are well documented in other systems (see Sachs, 1991; Hudspeth and Gillespie, 1994, for reviews). In addition, evidence exists for both tension-sensing proteins at the kinetochore and for changes in phosphorylation in kinetochore components. Hyman and Mitchison (1991) previously showed for isolated chromosomes that plus-end directed MT gliding (AP motility) across kinetochores occurs at elevated thiophosphorylation levels while minus-end directed motility occurs at low phosphorylation levels. Thus, regulation of kinetochore MT motor activity may involve a series of kinase and phosphatase reactions that are downstream of a tension-sensing device. Complicating the elucidation of such a pathway is evidence that the kinetochore also participates in a cell cycle checkpoint that regulates the exit from mitosis (Rieder et al., 1990; Hoyt et al., 1991; Li and Murray, 1991) and that this checkpoint may also involve phosphorylation (Rattner et al., 1990; Shero and Hieter, 1991; Nicklas et al., 1993; Gorbsky and Ricketts, 1993) and dephosphorylation of kinetochore components (Gorbsky and Ricketts, 1993) in a tension-sensitive fashion (Li and Nicklas, 1995; Nicklas et al., 1995). Thus, there may be several mechanisms by which kinetochores sense (Mitchison, 1989) and chemically respond to tension-related cues provided by the spindle apparatus.

The authors are indebted to Drs Bruce Nicklas and Gary Gorbsky for sharing with us their unpublished results, J. Waters for sharing her measurements of interkinetochore distances and Dr L Cassimeris for sharing her studies of these cells. We also appreciate Drs A. Murray, D. Koshland, B. Nicklas, R. Walker, M. Kenna, L. Cassimeris and comrades of the Salmon lab, especially S. Parsons and J. Waters, for their input and criticisms of this manuscript and M. Salmon for his artistic rendering of the model. In addition, we thank the staff in Mike Bern’s laboratory and the NIH LAMP facility at the Beckman Laser Institute, University of California, Irvine, where the initial portion of this study was performed, for providing us with their technical expertise and aid. This work was supported by National Institutes of Health grants GM-40198 to C.L.R. and grant GM-24364 to E.D.S. and NCRR/BTR P41-RR01192 which supports the University of California at Irvine Laser Microbeam Program as a National Biotechnological Resource.