Force on spindle microtubule minus ends moves chromosomes

During cell division, the mitotic spindle assembles itself from its constituent parts. Spindle microtubule minus ends are focused into two poles, and these poles dictate where duplicated chromatids are transported at anaphase. Forces that focus microtubules into poles are crucial to spindle organization and function. Cytoplasmic dynein, a minus end–directed microtubule motor, clusters parallel microtubules into spindle poles (Verde et al., 1991; Heald et al., 1996) and transports the microtubule-binding protein NuMA to build poles (Merdes et al., 2000). At poles, dynein and NuMA tether microtubules (Gaglio et al., 1995; Merdes et al., 1996; Heald et al., 1997; Dionne et al., 1999), and pole structure remains robust despite rapid microtubule turnover (Saxton et al., 1984) and opposing tension on kinetochore fibers (k-fibers) from kinetochore-based forces (Gordon et al., 2001; Manning and Compton, 2007; Silk et al., 2009). Thus, poles must both oppose force and be constantly rebuilt (Gaglio et al., 1997; Goshima et al., 2005). This engineering challenge highlights a long-standing paradox: how can the spindle maintain its structure and mechanical integrity and yet remain dynamic, flexible, and architecturally plastic, as its functions require?

For the spindle to preserve its structural integrity, it must be able to continuously rebuild poles by recognizing and sorting new microtubule structures. Indeed, during spindle assembly, poles can integrate both new peripheral microtubules (Rusan et al., 2002; Tulu et al., 2003) and kinetochore-nucleated k-fibers (Khodjakov et al., 2003; Maiato et al., 2004). Established spindles can move short microtubule seeds to poles (Heald et al., 1996, 1997) and reincorporate k-fibers severed by ablation as microtubules grow back (Snyder et al., 1991; Chen and Zhang, 2004; Maiato et al., 2004), and poles from different spindles can fuse together (Gatlin et al., 2009). Although dynein and NuMA are either demonstrated or suspected to mediate these observations of dynamic microtubule integration into poles, it is not clear which microtubule structures serve as dynein cargo, where on them force is exerted, or how strong that force is. We do not know how forces that maintain poles compare to other spindle forces or on what timescale they contribute to spindle architecture. In large part, this is because the response of the established spindle to detached microtubules is challenging to study: k-fiber minus ends are already embedded in the spindle and free microtubules within the spindle body are difficult to image.

Here, we use laser ablation to challenge the spindle’s architectural steady state by detaching microtubules from poles and we probe cellular forces exerted on, and molecules recruited to, these microtubules. We show that detached microtubules are rapidly identified by dynein/dynactin and NuMA and transported toward poles, overpowering opposing forces on microtubules and chromosomes to repair spindle architecture. Force is generated by localized pulling on new minus ends, which powers a newly identified mechanism of chromosome movement at mitosis, independent of kinetochore forces. We propose that rapid detection and dominant poleward transport of free minus ends by dynein maintains spindle integrity throughout mitosis, making k-fiber anchorage and spindle pole structure robust to component turnover and mechanical challenges.

We used pulsed laser ablation to sever microtubules and detach them from poles (Fig. 1 A) in mammalian GFP–α-tubulin PtK2 cells. In metaphase spindles, we ablated k-fibers a mean distance of 2.5 ± 0.1 µm from the kinetochore, leaving a short k-fiber “stub” that terminated with new microtubule minus ends. In the brief first response phase after ablation, the previously stretched centromere rapidly relaxed, causing the k-fiber stub to move toward the chromosome, and the uncapped (and unstable) microtubule plus ends created by k-fiber severance quickly depolymerized (Spurck et al., 1990; Maiato et al., 2004; Brugués et al., 2012; Sheykhani et al., 2013; Fig. 1 B and Video 1). These mechanical cues verified that ablation severed k-fibers (see Materials and methods).

Then, beginning after a mean delay of 15 ± 4 s following ablation, the k-fiber stub was rapidly transported back toward the spindle pole to which it was previously connected, dragging its attached chromosome and stretching the centromere back to preablation levels (Fig. 1 B, Fig. 2 A, Table 1, and Video 1). This poleward transport comprised a distinct, second response phase and was unexpected given the loss of a direct connection of the k-fiber to the pole. The severed k-fiber often began moving before its nonablated sister k-fiber and the centromere stretched rather than compressed during movement, both indicating that movement was powered by pulling forces on the severed k-fiber rather than by pushing forces from the sister. In some cases, the initiation of movement coincided with the k-fiber stub end visibly contacting a neighboring microtubule or k-fiber. Chromosomes moved poleward at a mean speed of 2.2 ± 0.2 µm/min after ablation (Table 1), significantly faster than normal poleward chromosome speeds at metaphase (1.3 ± 0.2 µm/min; Table 1). Transport persisted for an average of 60 ± 7 s and 2.1 ± 0.1 µm, indicating processive force generation. After each discrete episode of poleward movement in response to ablation, the k-fiber stub was reincorporated into the spindle and chromosomes resumed normal metaphase oscillations. Cells entered anaphase even after numerous ablations. Thus, k-fiber severance reveals that chromosomes can be pulled poleward (and centromeres can be stretched) without direct k-fiber stub connection to the pole, suggesting a force-generating attachment of the stub to the spindle body.

To probe the magnitude of the poleward transport force and its structural impact on the spindle, we compared its poleward pull on chromosomes to other spindle forces, such as opposing tension from a sister k-fiber (Fig. 1 A). We ablated k-fibers during anaphase, when chromatids are no longer strongly attached to their sisters. As at metaphase, k-fiber stubs and attached chromatids were pulled toward spindle poles (Fig. 1 C and Video 2). In fact, chromatids with severed k-fibers moved poleward significantly faster than those with uncut k-fibers (Fig. 2 B), suggesting that a different mechanism powered their segregation. In addition, transport speeds were very similar from trace to trace (Fig. 2 B), consistent with a single mechanism powering the poleward transport response. Finally, transport speeds were faster at anaphase than at metaphase (2.7 ± 0.3 µm/min; Table 1), indicating that opposing force from a sister k-fiber can slow the poleward transport response during metaphase.

To test whether other interactions with the sister half-spindle can meaningfully oppose the poleward transport force, we generated monopolar spindles (where such interactions are eliminated) by inhibiting Eg5 with 5 µM S-trityl-l-cysteine (STLC). After ablation, severed k-fibers and associated chromosomes were abruptly pulled poleward (Fig. 1 D). As in metaphase and anaphase bipoles, chromosomes in monopolar spindles ultimately resumed normal movements (Fig. 2 C). However, poleward transport speeds after ablation (3.7 ± 0.3 µm/min) were significantly higher than in metaphase or anaphase bipoles (Fig. 1 D, Fig. 2 C, Table 1, and Video 3). In an anaphase bipolar spindle, therefore, the sister half-spindle still contributes resistance to poleward transport. In established bipolar spindles treated with STLC, the poleward transport force pulled chromosomes at 2.7 ± 0.3 µm/min (15 ablations in six cells): this is faster than in untreated bipolar spindles (Table 1), indicating that cross-linking by Eg5 may oppose transport in bipolar spindles, but is significantly slower than in monopolar spindles (Table 1), indicating that a monopolar architecture lacking sister half-spindle forces permits faster poleward transport independent of Eg5 activity. Moreover, the response in monopolar spindles, without sister half-spindle forces, confirms that the response force pulls, rather than pushes, on chromosomes.

We next investigated the strength of the poleward transport force relative to forces exerted directly on chromosomes as they move through the spindle. We eliminated the latter by ablating bundles of nonkinetochore microtubules in monopolar spindles, where these bundles were easiest to target. After severance, non–k-fiber microtubule bundles were pulled poleward with their minus ends leading, just as ablated k-fiber stubs were (Fig. 1 E and Video 4). Therefore, the poleward transport response does not require molecular loading at the kinetochore or k-fiber–specific proteins (Royle et al., 2005; Bird and Hyman, 2008; Meunier and Vernos, 2011; Sturgill and Ohi, 2013): it not only acts on all spindle architectures tested (Fig. 1, B–D), but also on all microtubule populations tested (Fig. 1, B and E). Non–k-fiber bundles moved poleward more than threefold faster than k-fiber stubs (13 ± 1 µm/min; Table 1), suggesting that forces exerted directly on chromosomes (e.g., from drag or chromokinesins) significantly oppose poleward transport. Thus, the poleward transport force is similar in magnitude to opposing forces on chromosomes, but can overpower them to dictate transport direction and preserve the mechanical connection between chromosome and pole.

Collectively, the characteristics of the poleward transport response suggest it is powered by a minus end–directed microtubule motor: after a delay, force generation begins abruptly and is active, fast, and processive. We suspected that dynein, which can slide microtubules to poles, powered the poleward transport response. Indeed, functional inhibition of dynein by overexpression of a dominant-negative p150 fragment (Quintyne and Schroer, 2002) resulted in absent or long-delayed incorporation of k-fiber stubs after ablation (Fig. 3 and Video 5). Free minus ends were tracked for an average of 122 ± 24 s after ablation without displaying poleward movement or minus end reincorporation (Table 1). Both direct dynein inhibition and perturbed spindle microtubule architecture (e.g., a perturbed microtubule network for dynein to walk along) may contribute to the loss of poleward transport. Thus, dynein powers the poleward transport response.

To investigate the physical mechanism of dynein-powered poleward transport, we probed whether and how the response (displacement, timescale, and speed) varied spatially with the location of new minus ends. We focused on monopolar spindles because they displayed the most marked response to ablation and are in an experimentally tractable steady state. The response to new minus ends was robust: we observed fast poleward transport of 37 out of 40 ablated k-fibers in 19 cells. The magnitude of poleward movement correlated with the distance of the cut site from the pole (Table 2 and Fig. S1 A), suggesting that the poleward transport force pulls k-fiber stubs to the spindle pole.

Consistent with the idea that polar microtubule tracks guide post-ablation transport, the initiation of poleward movement sometimes coincided with visible contact between a k-fiber stub and neighboring microtubules (Fig. S2 and Video 2). In addition, very short stubs (<1 µm) displayed a longer delay between ablation and poleward transport initiation (Table 2 and Fig. S1 B). Because they sterically sample smaller volumes, short k-fiber stubs may have a lower probability of contacting microtubule tracks for dynein to walk along. For longer stubs, this correlation was absent, but the existence of an average 8-s delay (Table 1) suggested that the poleward force was not present on the k-fiber before ablation; if that were the case, movement would have begun immediately.

The delay is consistent with ablation-triggered recruitment of force-generating machinery to new, ablation-created structures such as the new minus ends. This suggestion that motor machinery is targeted to specific microtubule structures raised the question of where on transported microtubules force is exerted. A model of microtubule sliding powered by dynein at points of lateral microtubule contact would predict force generation all along the microtubule (Merdes et al., 2000; Radulescu and Cleveland, 2010). In contrast, we found that the speed of the poleward transport response does not correlate with the length of the k-fiber stub (Table 2 and Fig. S1 C), suggesting that the force driving transport is not generated all along the stub. (Slower speeds of ablated k-fibers compared with non–k-fiber bundles suggest that dynein can move faster under a lighter load, so that an increase in number of motors and corresponding decrease in load per motor would be capable of increasing the transport speed.) Because speed does not increase with stub length, force may be generated at its minus end, a structure singular to each ablated microtubule bundle. Consistent with force at minus ends dominating mechanical interactions along the k-fiber, the severed stub rapidly pivoted 180° during rare (<5% of cases) observations of post-ablation movement toward the opposite pole (similar to observations by Dick and Gerlich [2013]), likely along an atypically polarized microtubule track (Fig. S3 and Video 6). Given these observations, we hypothesized that the dynein-powered poleward transport force is in fact a minus end poleward transport force, which identifies and pulls free minus ends as cargo.

To test whether force generation at minus ends is necessary for k-fiber stub poleward transport, we performed a second ablation to destroy the stub minus end, after poleward transport began but before the minus end was indistinguishable from surrounding microtubules. Ablation of the stub minus end abolished movement toward the pole, both in bipolar and monopolar spindles; movement began again after a delay (Figs. 1 F and 2 E and Video 7). Although current data cannot resolve whether few or most minus ends within a k-fiber productively engage during poleward force generation, the loss of force generation after minus end ablation suggests that mechanical engagement at the k-fiber minus end is essential for the poleward transport response. Thus, although forces may also be generated all along the k-fiber (Nicklas et al., 1982; Maiato et al., 2004; Vladimirou et al., 2013), they were not sufficient to power the rapid stub and chromosome transport we observed.

If dynein and its binding partners identify and pull free microtubule minus ends, rather than nonpreferentially exerting sliding forces all along microtubule lengths, we would expect them to specifically localize to new minus ends. To test for such localization, we fixed metaphase bipolar cells immediately after ablation and performed immunofluorescence staining. We first stained for NuMA (Fig. 4, A and B), which showed striking localization at new minus ends as early as 15 s after ablation, the fastest we were able to fix after ablation. When we ablated non–k-fiber bundles, we found that NuMA also localized specifically to these newly created minus ends (Fig. 4, C and D). Thus, the free minus ends of both k-fiber and non–k-fiber microtubule bundles recruit NuMA.

To investigate whether dynein accompanied NuMA and was selectively enriched at new minus ends, we stained after ablation for p150 (also called p150^Glued), a component of the dynein adapter complex dynactin (Gill et al., 1991), which is required for NuMA to associate with dynein (Merdes et al., 2000). Indeed, p150 colocalized with NuMA selectively at minus ends of severed fibers (Fig. 4, E–G). The fact that NuMA and dynactin specifically localize to minus end structures is consistent with data suggesting that dynein generates force at the minus end itself, as opposed to all along the microtubule (Tables 1 and 2; Figs. 1 F, 2 E, S1 C, and S3; and Videos 6 and 7).

After immunofluorescence indicated that NuMA and dynactin specifically mark new minus ends as motor cargo, we wondered how quickly they identified minus ends and if and where they subsequently moved. We turned to live imaging to understand the timescale of dynein/dynactin and NuMA recruitment and their dynamics and processivity relative to the minus end poleward transport response. Another component of the dynactin complex, Arp1A (Lees-Miller et al., 1992), was quickly recruited to sites of ablation in metaphase bipoles, within an average of 14 ± 3 s (Fig. 5, A and B; Table 1; and Video 8). GFP-Arp1A puncta moved at speeds (2.6 ± 0.4 µm/min; Table 1) similar to the metaphase k-fiber poleward transport response, and they continued processively until they became indistinguishable from the pole. To determine if Arp1A puncta colocalized with minus ends throughout their trajectories, we ablated k-fibers in cells expressing mCherry-tubulin and GFP-Arp1A. Arp1A puncta appeared at new minus ends and moved together with them toward poles (Fig. 5, C and D; and Video 9).

GFP-NuMA was visible at ablation sites an average of 8 ± 1 s after ablation (Table 1; Fig. 5, E and F; and Video 10) in metaphase bipoles, more quickly than GFP-Arp1A recruitment was detectable. This difference may reflect temporal recruitment order or may result from higher contrast GFP-NuMA signal. The recruitment of GFP-NuMA occurred before the initiation of poleward movement of severed k-fibers in metaphase bipoles (15 ± 4 s; Table 1). In some cases, GFP-NuMA was detectable at the ablation site for up to 40 s before it moved poleward (Fig. 5 G and Fig. S4). Like GFP-Arp1A, recruited GFP-NuMA moved poleward at speeds in keeping with the minus end transport response until it merged with NuMA structures at the pole (1.9 ± 0.1 µm/min; Table 1 and Fig. 5, E and F). The speed of movement decreased as the recruited NuMA neared the spindle pole (Fig. 5, F and G). Notably, the speed of microtubules in Xenopus laevis extract spindles decreases near poles in a dynein-dependent manner (Burbank et al., 2007; Yang et al., 2008). Consistent with our characterization of k-fiber movement during minus end poleward transport, the poleward movement of GFP-NuMA in monopolar spindles was significantly faster (2.9 ± 0.3 µm/min; Table 1) than in bipolar spindles. Lastly, in cells expressing mCherry-NuMA and GFP-Arp1A, NuMA and Arp1A puncta colocalized and moved together to poles (Fig. S5).

In sum, the data indicate that throughout mitosis, dynein/dynactin and NuMA rapidly and specifically identify free microtubule minus ends and pull them as cargo to the spindle pole, generating enough force to maintain k-fiber anchorage and move chromosomes.

The maintenance of focused poles is central to spindle architecture and function. Here, we sever spindle microtubules using laser ablation to create detached k-fiber and non–k-fiber microtubules. We find that dynein/dynactin and NuMA rapidly, efficiently, and specifically identify the free minus ends of these detached microtubules and processively pull minus ends as cargo to poles. Although this force acts at minus ends, it can overpower opposing pulls all along microtubules and on chromosomes and power spindle self-repair, prioritizing the structural reintegration of minus ends. We demonstrate not only that a force exerted on microtubule minus ends can move and segregate chromosomes but also that this poleward minus end force overpowers canonical forces on chromosomes exerted at plus ends. Minus end reintegration maintains spindle architecture on a faster timescale (seconds) than chromosome movement (minutes), ensuring that chromosomes remain bioriented and are transported to poles at anaphase.

The data herein suggest a model for spindle architecture maintenance by efficient cellular identification and poleward transport of free microtubule minus ends (Fig. 6 A). Within seconds of their creation, new minus ends recruit NuMA and dynein/dynactin. Minus ends remain detached from the spindle until they come into contact with a neighboring spindle microtubule track seconds later. Upon contact, dynein walks processively poleward along this track, pulling its microtubule minus end cargo (Fig. 6 B). Together, complexes of NuMA, dynactin, and dynein may act as a physical tether between a microtubule track (bound to dynein) and microtubule cargo (bound to NuMA) and power the poleward transport of spindle microtubule minus ends. Estimation of the frictional force experienced by a microtubule bundle moving at the speeds we observe suggests that multiple dynein motors are likely responsible for this poleward force on minus ends (Nicklas, 1988; Shubeita et al., 2008; Shimamoto et al., 2011). Targeting dynein molecules specifically to minus ends, rather than all along the microtubule, provides a simple and intuitive mechanism for clustering minus ends into poles (Burbank et al., 2007).

NuMA and dynein are thought to incorporate kinetochore-nucleated k-fibers during spindle formation (Merdes et al., 2000; Khodjakov et al., 2003). Yet when NuMA is inhibited, two focused spindle poles still form; only later do poles defocus (Silk et al., 2009). The present study helps reconcile these findings. Based on the observation that NuMA rapidly and robustly localizes to new minus ends, we hypothesize that spindle maintenance involves not only NuMA’s canonical role as a tethering “glue” at poles (Merdes et al., 1996; Silk et al., 2009) but also NuMA-mediated capture and poleward transport of spindle microtubule minus ends (Khodjakov et al., 2003) that are created or lost from poles. NuMA may act as a tether in two places: between a minus end and its microtubule track, and between that track and the pole. In retrospect, earlier visualization of NuMA along the length of kinetochore-nucleated k-fibers, not only at their tips (Khodjakov et al., 2003), may reflect the tapered nature of these fibers. By creating synchronized minus ends, the targeted ablation approach used here allows us to dissect the location of dynein force generation and reveals specific recruitment and activity of dynein/dynactin and NuMA at microtubule minus ends (Fig. 6 B). Together, the creation of synchronized minus ends by ablation and their subsequent molecular marking by the cell offer a way to image spindle microtubule minus ends—a long-standing challenge.

This study places an upper bound of 8 s on the time required to establish distinct biochemistry at free minus ends and it raises the question of how microtubule minus ends are specifically identified within the spindle body. We consider two models and cannot currently distinguish between them. First, dynein could walk along microtubules until it runs out of track, delivering dynactin and NuMA to minus ends. In this scenario, dynein molecules first use the k-fiber as a track, pile up at the minus end without falling off—the basis for minus end selective identification—and later hold on to the minus end as cargo. Alternatively, soluble NuMA and dynein/dynactin may be recruited selectively to minus ends from the cytoplasm, binding either directly or indirectly to minus ends. NuMA binds microtubules directly (Du et al., 2002; Haren and Merdes, 2002), and previous work has suggested that NuMA can localize to minus ends independently of dynein (Heald et al., 1997). After ablation, then, NuMA may recruit dynactin and dynein specifically to minus ends, consistent with the faster recruitment we observe for NuMA than dynactin (Table 1). If indirect, binding may occur through one of the two known direct minus end–interacting partners, the γ-tubulin ring complex (Zheng et al., 1995) or the recently characterized CAMSAP family (Meng et al., 2008), which localizes to interphase microtubule minus ends created by ablation (Jiang et al., 2014).

This work identifies an unexpected mechanism of chromosome translocation mediated by a poleward force on k-fiber minus ends. The minus end poleward force is similar in magnitude to other forces on chromosomes but capable of directionally dominant transport (Fig. 6 C), allowing it to (a) work in tandem with other forces without being so strong as to disrupt spindle architecture but (b) prioritize correct integration of free minus ends. Minus ends are relatively stable and thus preserve structural memory, and a uniquely fast cellular response to free minus ends may preserve bipolar structure and chromosome biorientation. As such, minus end transport forces may be tuned to competing spindle forces—strong and fast enough to ensure the accurate integration of k-fiber minus ends into poles given the strength and dynamics of opposing forces on chromosomes.

Spindle microtubules need to be continuously organized as minus ends are created or exposed by microtubule nucleation, microtubule-severing enzymes, augmin-mediated branching events (Goshima et al., 2008), or detachment from previous structures (Lancaster et al., 2013; Petry et al., 2013). The findings herein open new questions: whether NuMA and dynein localize to the minus ends of all spindle microtubules and whether all populations of spindle minus ends are rapidly transported by this poleward force. If the poleward transport machinery recognizes only a subset of minus ends, it could discriminate on the basis of structure (e.g., minus ends of bundled microtubules) or biochemistry (e.g., “naive” ablation-created minus ends with exposed interfaces not yet masked by minus end binding proteins).

Paradoxically, the mitotic spindle is a highly dynamic and yet mechanically robust macromolecular machine. Spindle poles, for example, persist for lifetimes longer than their components and are both flexible and mechanically strong, as illustrated by manipulation experiments (Nicklas and Staehly, 1967; Begg and Ellis, 1979; Charlebois et al., 2011). Rapid-acting, strong forces that continuously maintain spindle integrity could resolve this paradox. The robust minus end transport response we observe represents one such mechanism for maintaining chromosome anchorage and spindle architecture by dynamic self-repair: rapid error detection (of free minus ends) and resolution (via dominant poleward transport).

PtK2 GFP–α-tubulin cells (stable line expressing human α-tubulin in pEGFP-C1; Takara Bio Inc.; a gift from A. Khodjakov, Wadsworth Center, Albany, NY; Khodjakov et al., 2003) and PtK2 cells were cultured in MEM (Invitrogen) supplemented with sodium pyruvate (Invitrogen), nonessential amino acids (Invitrogen), penicillin/streptomycin, and 10% qualified and heat-inactivated fetal bovine serum (Invitrogen). Cells were plated on #1.5 25-mm coverslips (HCl-cleaned and poly-l-lysine coated) and imaged in Leibovitz’s L-15 medium with l-glutamine without phenol red (Invitrogen) with antibiotics and serum as for cell culture. PtK2 cells were transfected with mCherry-tubulin (human α-tubulin in pmCherry-C1; Takara Bio Inc.; a gift from M. Davidson, Florida State University, Tallahassee, FL), GFP-NuMA(N1) (human NuMA in pEGFP-N1; Takara Bio Inc.; a gift from D. Compton, Dartmouth Medical School, Hanover, NH; Kisurina-Evgenieva et al., 2004), mCherry-NuMA (from GFP-NuMA(N1); we replaced GFP with mCherry between AgeI and NotI sites), or GFP-Arp1A (human Arp1A in a pBABE variant; Addgene; a gift from I. Cheeseman, Whitehead Institute, Cambridge, MA; Kiyomitsu and Cheeseman, 2012) using Fugene6 (Promega) and imaged 36–48 h after transfection.

To make monopolar spindles, STLC (Sigma-Aldrich) was added to cells in MEM (see previous paragraph) 20 min before imaging at 5 µM (10 mM DMSO stock). Cells were imaged in L-15 (see previous paragraph) containing 5 µM STLC. To probe bipolar spindles with inhibited Eg5, we added 5 µM STLC after spindle formation, which preserves bipolarity (Cameron et al., 2006).

To perturb dynein via dominant-negative overexpression, DsRed-p150^217–548 (amino acids 217–548 of chicken p150 in pDsRed-N1; Takara Bio Inc.; a gift from T. Schroer, Johns Hopkins University; Baltimore, MD; Quintyne and Schroer, 2002) was transfected with Fugene6 72–96 h before imaging. Dynein-inhibited cells were identified via DsRed fluorescence and spindle morphology: cells with fragmented, multipolar spindles (Quintyne and Schroer, 2002) indicative of dynein loss of function were specifically selected. We mildly compressed (Dumont and Mitchison, 2009) some of the dynein-inhibited spindles imaged (including the one in Fig. 3 and Video 5) to keep them in focus and to increase contact probabilities between new minus ends and neighboring microtubules. Both k-fiber and non–k-fiber bundles were laser ablated.

For immunofluorescence of individual cells after ablation, cells were live imaged on coverslips photoetched with a labeled grid (Thermo Fisher Scientific). After ablation, cells were fixed in 95% methanol with 5 mM EGTA for 3 min. The time between laser ablation and fixation was usually ∼30 s, but could be as fast as 15 s. The following antibodies and dyes were used: mouse anti–α-tubulin DM1α (1:1,000; Sigma-Aldrich), rabbit anti-NuMA (1:300; Novus Biologicals), mouse anti–p150-Glued (1:500; BD), human anti-centromere protein (CREST; 1:25; Antibodies, Inc.), mouse anti–α-tubulin DM1α conjugated to Alexa 488 (1:50; Cell Signaling Technology), fluorescent secondary antibodies (1:500; Invitrogen), and Hoescht 33342 (Sigma-Aldrich). After staining, we identified the ablated cell using the coverslip grid.

Live imaging was performed on an inverted microscope (Eclipse Ti-E; Nikon) with a spinning disk confocal (CSU-X1; Yokogawa Electric Corporation), head dichroic Semrock Di01-T405/488/561GFP (except Di01-T488 for GFP-NuMA-only, GFP-Arp1A-only, Fig. 1 F, and Video 5), 488-nm (120 mW) and 561-nm (150 mW) diode lasers, emission filters ET525/36M (Chroma Technology Corp.) for GFP (except ET500LP for GFP-NuMA-only, GFP-Arp1A-only, Fig. 1 F, and Video 5) or ET630/75M for mCherry, and an iXon3 camera (Andor Technology). Cells were imaged by phase contrast (400–500-ms exposures) and fluorescence (75–500-ms exposures) every 3.5–14 s with a 100× 1.45 Ph3 oil objective through a 1.5× lens yielding 105 nm/pixel at bin = 1 (Metamorph 7.7.8.0; Molecular Devices). Figs. 5 C and S5 and Video 9 were collected at bin = 2. Cells were imaged at 29–31°C in a homemade heated aluminum coverslip holder, using the Perfect Focus System (Nikon). Targeted laser ablation (several 3-ns pulses at 20 Hz) using 551- (if only GFP imaged) or 514-nm (if GFP and mCherry imaged) light was performed using a galvo-controlled MicroPoint Laser System (Photonic Instruments) operated through Metamorph. Based on fluorescence imaging, we estimate the diameter of the ablation site at ∼1 µm. K-fiber ablation was verified by observed loss of mechanical tension across the centromere, depolymerization of the uncapped microtubule plus end, and free reorientation of the severed k-fiber stub after ablation (Fig. 1 B). For cells fixed after ablation, loss of tubulin staining at the ablation site confirmed k-fiber and non–k-fiber severance (Fig. 4).

Chromosome position data were generated by manual tracking of k-fiber plus ends, ablation sites, and spindle poles in live-imaged GFP–α-tubulin PtK2 cells, using overlaid GFP–α-tubulin and phase-contrast time-lapse videos in a home-written MatLab (R2012a Version 7.4) program. As controls, we also tracked neighboring k-fiber plus ends (of unmanipulated chromosomes in metaphase and monopolar spindles, and of the paired sister chromatids in anaphase spindles). We manually selected the start and end times of the poleward transport response (for ablated k-fibers) or the first poleward movement after ablation (for control chromosomes) by examining plots of k-fiber plus end position over time and choosing the segment over which poleward motion was processive, and we used these times and positions to calculate mean speeds. We calculated the delay times as the difference between the first frame after ablation and the first frame of this sustained poleward response. For non–k-fibers, we manually tracked the position of their new minus ends as long as possible.

Line scan analysis of immunofluorescence colocalization was performed using the plot profile function of ImageJ with a line width of 1 pixel.

Kymographs of GFP-Arp1A and GFP-NuMA puncta and pole position over time were generated in ImageJ. A second home-written MatLab program generated fluorescence intensity line scans for each frame from the kymograph. Using each sequence of line scans, the peaks indicating the positions of GFP-Arp1A/GFP-NuMA puncta and the spindle pole were manually selected, with the intensity maxima of these peaks used to indicate puncta positions. We defined the initial recruitment time of GFP-Arp1A/GFP-NuMA puncta as the first frame in which a clear peak was visible in the line scan. We continued to track puncta until their intensity peaks could not be clearly separated from those of the spindle poles. To calculate the distance of these peak positions from the ablation sites, we used the ablation targeting coordinates from Metamorph. For comparing the intensity of GFP-NuMA puncta recruited to sites of ablation to other puncta, we calculated the fold difference in the mean integrated intensity of at least four puncta in each cell at sites/times without ablation and the integrated intensity of puncta recruited after ablation.

Data are expressed as mean ± SEM. Calculations of correlation coefficients (Pearson’s r) and p-values were performed in MatLab. For calculating mean traces in Fig. 5 G, data from all traces were collected into 10-s-wide bins in time and the mean position within this bin was calculated.

To detect the details of the cellular response to ablation while minimizing photodamage from frequent imaging, cells were imaged less frequently before ablation and more frequently immediately after ablation. For the videos to play back at a constant 30× speed relative to real time, frames are shown in the videos for a time inversely proportional to the rate at which they were collected.

Fig. S1 shows that ablation position affects the magnitude of movement and the delay before poleward transport, but not the transport speed. Fig. S2 uses an example to illustrate that the initiation of poleward transport immediately follows contact with neighboring microtubules. Fig. S3 shows the rare observation of a 180° pivot of the k-fiber stub and poleward transport to the opposite pole, suggesting that poleward force is not exerted all along the k-fiber. Fig. S4 shows that the recruitment of NuMA sometimes precedes its poleward movement. Fig. S5 shows that GFP-Arp1A and mCherry-NuMA colocalize at sites of ablation and move together to poles. Video 1 shows that laser ablation of a k-fiber triggers poleward chromosome movement in a metaphase bipolar spindle. Video 2 demonstrates that the poleward transport response persists into anaphase to segregate chromosomes. Video 3 shows that the poleward transport response is faster in monopolar spindles than in metaphase or anaphase. Video 4 illustrates that the poleward transport response occurs and is fastest for non–k-fiber microtubule bundles. Video 5 demonstrates that dynein function is required for the poleward transport response. Video 6 shows the rare observation of a 180° pivot of the k-fiber stub and poleward transport to the opposite pole, suggesting that force is not exerted all along the k-fiber. Video 7 suggests that the poleward transport response requires force generation at minus ends because ablation of the minus ends stops the response. Video 8 shows that GFP-Arp1A is recruited to sites of ablation and moves all the way to poles. Video 9 shows that GFP-Arp1A localizes specifically to new minus ends and moves with them. Video 10 shows that GFP-NuMA is recruited to sites of ablation and moves all the way to poles.

We thank Alexey Khodjakov for PtK2 GFP–α-tubulin cells; Duane Compton, Iain Cheeseman, Michael Davidson, and Trina Schroer for the GFP-NuMA, GFP-Arp1A, mCherry-tubulin, and DsRed-p150^217-548 constructs, respectively; Rudolf Oldenbourg, Wendy Salmon, and the Nikon Imaging Center at Harvard Medical School for technical advice; and Tim Mitchison, David Morgan, Ted Salmon, Anita Sil, Marvin Tanenbaum, Ron Vale, and the Dumont Laboratory for discussions and critical reading of the manuscript. Special thanks to Alexey Khodjakov for sharing unpublished data and for insightful discussions.

This work was supported in part by National Institutes of Health (R00GM09433 and U54CA143836) and a National Science Foundation Graduate Research Fellowship (C.L. Hueschen). M.W. Elting is a Damon Runyon Fellow supported by the Damon Runyon Cancer Research Foundation (DRG-2170-13). S. Dumont is an Alfred P. Sloan Research Fellow, Kimmel Scholar, and Rita Allen Foundation Milton E. Cassel Scholar, and receives support from the Searle Scholars Program.

The authors declare no competing financial interests.