|Home | About | Journals | Submit | Contact Us | Français|
Mitotic spindle assembly requires the combined activity of various molecular motor proteins, including Eg5  and dynein . Together, these motors generate antagonistic forces during mammalian bipolar spindle assembly ; what remains unknown, however, is how these motors are functionally coordinated such that antagonism is possible. Given that Eg5 generates an outward force by crosslinking and sliding apart antiparallel microtubules (MTs) [4–6], we explored the possibility that dynein generates an inward force by likewise sliding antiparallel MTs. We reasoned that antiparallel overlap, and therefore the magnitude of a dynein-mediated force, would be inversely proportional to the initial distance between centrosomes. To capitalize on this relationship, we utilized a nocodazole washout assay to mimic spindle assembly. We found that Eg5 inhibition led to either monopolar or bipolar spindle formation, depending on whether centrosomes were initially separated less than or greater than 5.5μm, respectively. Mathematical modeling predicted this same spindle bistability in the absence of functional Eg5, and required dynein acting on antiparallel MTs to do so. Our results suggest that dynein functionally coordinates with Eg5 by crosslinking and sliding antiparallel MTs, a novel role for dynein within the framework of spindle assembly.
Before exploring the functional coordination of Eg5 and dynein, we first confirmed the antagonistic nature of these motors in LLC-Pk1 cells, as well as the relevance of this antagonism to spindle bipolarity . To accomplish this, metaphase cells expressing GFP-tubulin (LLC-Pk1α ) were treated with monastrol to inhibit Eg5  or injected with p150-CC1 to inhibit dynein . Immediately following monastrol treatment, bipolar spindles shortened by ~30% (Figures S1A and S1C), but complete collapse into monopoles was not observed. Conversely, spindles lengthened by ~30% after dynein inhibition (Figures S1B and S1C). This spindle elongation was not a consequence of centrosome dissociation from spindle poles or mislocalization of Kif2a (unpublished data). Together, spindle shortening and lengthening following Eg5 and dynein inhibition, respectively, confirm the presence of an antagonistic relationship between these two motors in LLC-Pk1 cells.
We next monitored the response of Eg5-inhibited monopolar spindles to inhibition of dynein. LLC-Pk1α cells were treated with monastrol prior to nuclear envelope breakdown, and mitotic cells containing monopolar microtubule (MT) arrays were then injected with p150-CC1. In ~50% of such cells (13 of 30 cells), monopolar spindles reorganized into bipolar spindles (Figure S1D and Movie S1), defined here and subsequently as fusiform MT arrays with the majority of chromosomes aligned between two distinct poles separated by a minimum of 5μm. Bipolar spindles that formed following inhibition of Eg5 and dynein were morphologically and functionally equivalent to controls (Figure S2). Spontaneous bipolarization of monastrol-induced monopolar spindles was never observed and injection of control antibodies left monopolar arrays unaltered (unpublished data), demonstrating the specificity of bipolarization to dynein inhibition.
The formation of bipolar spindles following inhibition of Eg5 and dynein demonstrates that an additional force drives pole separation. This force could be generated by MT polymerization  and/or additional plus end-directed motors, such as Xklp2 . Residual Eg5 activity, however, is unlikely to contribute to pole separation, given the efficacy of motor inhibition by monastrol . In contrast, incomplete inhibition of dynein following injection of p150-CC1, which interferes with the dynein/dynactin interaction but not dynein’s ATPase activity, could account for the observation that not all monopolar spindles were rescued. Finally, the geometry of MTs and chromosomes in co-inhibited cells may influence the generation of pushing forces that restore spindle bipolarity.
In summary, the observation that dynein inhibition can rescue the monastrol-mediated monopolar phenotype demonstrates that an antagonistic balance between Eg5 and dynein contributes to the establishment of spindle bipolarity. Based on the evidence that Eg5 can slide apart overlapping MTs [4–6], and its antagonistic relationship with dynein , we hypothesize that dynein functions at regions of antiparallel overlap, where it crosslinks and slides antiparallel MTs in opposition to Eg5.
If our hypothesis is valid, then dynein would likely be responsible for monopolar spindle formation in the presence of monastrol, as it would generate an inadequately opposed inward force. Because antiparallel overlap decreases as the distance between centrosomes increases (Figure S3A), and because the magnitude of our postulated dynein-mediated force would depend on the amount of antiparallel overlap, spindles that form in Eg5-inhibited cells should be resistant to collapse above a certain intercentrosomal threshold distance; in other words, spindles should exhibit an intercentrosomal distance-dependent bistability. To examine this, we utilized a nocodazole washout assay , which generates mitotic cells containing centrosomes with widely variable positions (in a manner independent of the inhibitors present) (Figure S4).
In this assay, LLC-Pk1α cells were treated with nocodazole to completely disassemble MTs and then washed 4X with drug free medium to initiate spindle assembly (see Experimental Procedures). Upon removal of drug, MTs assembled at centrosomes and chromosomes . When centrosomal and chromosomal arrays were close enough to interact (proximal centrosomes), these MT populations quickly coalesced, ultimately resulting in bipolar spindles (9 of 13 cells; Figures 1A and 1B; Table S1; Movies S2 and S3); this occurred regardless of the initial spacing between proximal centrosomes. In cells with centrosomal arrays that failed to interact with the chromosomal array (distal centrosomes), acentrosomal bipolar spindles assembled around chromosomes (3 of 5 cells; Figure 1C; Table S1; Movie S4), confirming that mammalian chromosomes alone can organize MTs into bipolar structures [13, 14], even in the continued presence of centrosomes. Examination of cells fixed 60 minutes post-4X washout revealed that ~30% had progressed into or beyond anaphase (Figure 1D), demonstrating that these bipolar spindles are functional, and validating this assay as a tool for studying spindle assembly.
To test the potential bistability of forming spindles in Eg5-inhibited cells, LLC-Pk1α cells were treated first with nocodazole and subsequently with monastrol, and then released into monastrol-containing medium. As predicted, the initial spacing between proximal centrosomes had a profound effect on the resulting MT array. When proximal centrosomes were located close to one another (i.e., < 5.5μm apart), monopolar arrays of MTs formed following release from nocodazole (4 of 5 cells; Figure 2A; Movie S5). In striking contrast, however, when proximal centrosomes were located far from one another (i.e., > 5.5μm apart), bipolar arrays of MTs formed (6 of 7 cells; Figure 2B; Table S1; Movie S6). Furthermore, when centrosomes were distal, chromosomes organized short acentrosomal bipolar arrays in an Eg5-independent manner (3 of 4 cells; Figure 2C; Table S1); chromosomes also organized similar acentrosomal bipolar spindles in a dynein-independent manner (2 of 2 cells; Figure S5).
These data confirm the predicted intercentrosomal distance-dependent bistability and suggest that the requirement for active Eg5 in establishing a bipolar spindle can be bypassed if spindle assembly initiates with well-separated centrosomes (i.e., > 5.5μm apart) or via an exclusively chromosomal pathway. In these cases, we expect the degree of antiparallel MT overlap to be insufficient to mediate dynein-dependent spindle collapse. Furthermore, our data show that Eg5 and dynein are each dispensable for acentrosomal bipolar spindle formation. While chromosome-mediated spindle assembly following dynein inhibition has previously been noted , this is the first demonstration that Eg5 activity is likewise not required for this process.
Though intercentrosomal distance-dependent spindle bistability supports our hypothesis, additional support can be obtained by confirming that dynein is responsible for spindle collapse in the presence of monastrol. To directly test this, LLC-Pk1α cells were treated first with nocodazole and monastrol, then injected with p150-CC1 prior to release into monastrol-containing medium. Consistent with p150-CC1 injections into monastrol-treated monopoles, half of these cells (2 of 4 cells) bipolarized when proximal centrosomes were close to one another (Figure 3A; Table S1; Movie S7); the other half formed monopolar arrays. As expected, when proximal centrosomes were distant, 4 of 5 cells bipolarized (Figure 3B; Table S1). Acentrosomal bipolar arrays additionally formed in Eg5- and dynein-inhibited cells containing distal centrosomes (1 of 1 cell; Figure 3C; Table 1).
These data reveal that monastrol-mediated spindle monopolarity is a dynein-dependent phenotype. Our results are therefore consistent with a model in which Eg5, located on antiparallel MTs, generates an outward sliding force, and that this is resisted by a dynein-generated inward force also acting on antiparallel MTs. This is a novel role for dynein during spindle assembly, which has previously been suggested to exclusively crosslink parallel MTs , and is consistent with dynein’s proposed role during Xenopus extract spindle fusion . We predict specifically that dynein localizes and generates force at or near the plus ends of overlapping MTs, consistent with its plus end localization in fungal systems [18, 19]. Here, dynein could crosslink MTs by binding to one MT via its stalk domain and to a second MT by a non-ATP dependent interaction, mediated, for example, by proteins that bind both dynein and MTs. In strong support of this, recent work has shown that spindle assembly requires the MT binding domain of the p150 subunit of dynactin  and that the MT plus-end binding protein, CLIP-170, which binds to dynein, antagonizes Eg5 .
Although our in vivo data support our hypothesis that dynein localizes to, and generates force at antiparallel MT overlap, confirmation of such necessarily involves visualizing both dynein and antiparallel MTs. In mammalian cells, however, the difficulties associated with genetically tagging and expressing dynein heavy chain preclude the former, while the density of spindle MTs obstructs the latter. We note, however, that dynein has been immunofluorescently localized to mammalian spindle MTs .
Despite these limitations, we can employ a mathematical model of spindle assembly to determine if our in vivo results are consistent with dynein acting on antiparallel MTs. The following assumptions were made while constructing the model: (i) centrosomes nucleate asters consisting of tens to hundreds of MTs undergoing rapid dynamic instability, so that the MT length distribution is exponential ; (ii) a few centrosomal MTs reach chromosome arms and generate a repulsive force (Figure 4A, Force A) either by a polymerization ratchet, or by interacting with chromokinesins; (iii) a few centrosomal MTs reach the spindle equator where Eg5 and dynein motors exert opposite sliding forces at the region of antiparallel overlap (Figure 4A, Force B); and (iv) tension generated at kinetochores pulls chromosomes toward centrosomes (Figure 4A, Force C). Importantly, the precise location of dynein on antiparallel MTs (i.e., whether it’s distributed along the whole overlap length or just at the plus ends) does not make a qualitative difference for the model’s predictions.
These assumptions allow the effective outward force, F, applied to each centrosome to be computed as a function of the half-spindle length, x, assuming that all chromosomes are crowded close together at the spindle equator (Figure 4A). This functional dependence has the form: F(x) = (Ae−x/L − C) − 2Bxe−2x/L, where L is the average MT length, A is the maximal repulsive force on chromosome arms, B is the total motor force per unit length of antiparallel MT overlap, and C is the kinetochore tension force (see Supplemental Text). Using this formula, we found that when parameter B was very small (i.e., when the outward sliding force by Eg5 and the inward pulling force by dynein and possibly other motors canceled each other out or were nonexistent, the total force pushed centrosomes away from the equator when they were close together and toward it when they were far apart (Figures S3B and 4B). In this case, the model predicted a single stable separation between centrosomes where the force balances to zero. With realistic parameters and chromosome distribution in the midplane (see Supplemental Text and Table S2), this stable length was ~11μm when Eg5 and dynein were either both active or inhibited (Figure 4B), a value that matched the spindle length observed in vivo under similar conditions (Table S1).
Less intuitively, the model revealed that when parameter B increased (i.e., when Eg5 alone was inhibited, and there was a significant unopposed inward pulling force by dynein and possibly other motors), the total force on centrosomes exhibited more complex behavior (Figures S3B and 4B). Although the force was still repulsive when centrosomes were close together and attractive when they were far apart, it did not simply decrease monotonically with distance. Rather, it became negative when centrosomes were separated less than ~5μm and positive when centrosomes were separated ~5–11μm. This is because below the ~5μm threshold, antiparallel MT overlap (~2xe−2x/L) is large and the pulling action of dynein is dominant, whereas above the threshold, antiparallel MT overlap becomes smaller and the repulsive action generated by MTs interacting with chromosome arms overcomes the dynein-mediated attraction. As a result, the model predicted a stable separation of ~11μm when the initial centrosomal separation was greater than ~5μm, and collapse when this separation was less than ~5μm. The predicted bistability and length of Eg5-inhibited spindles, as well as the threshold distance, again matched well with the in vivo data (Table S1). Computer simulations of mobile centrosomes and chromosomes were also in agreement with the in vivo observations (Figure S6C and S6D; Movies S8 and S9).
Together, our in silico data accurately simulated our in vivo observations, regarding both the outcome of spindle assembly in the presence of Eg5 and dynein inhibitors and the length of the resulting spindles, and did so with dynein acting on overlapping MTs. Importantly, we varied the model’s assumptions and parameters and established that if dynein were acting from the cell cortex, spindle poles or chromosomes, rather than on antiparallel MTs, the virtual spindle behavior would be incompatible with our observations. Note that some of the modeling assumptions are not crucial: other repulsive interactions than those mediated by chromosome arms, other attractive forces than those brought about by kinetochore tension, and other MT length distributions than the exponential one still predict the same qualitative behavior that we observed. However, the action of dynein specifically on antiparallel MTs is essential.
Our in vivo and in silico results demonstrate that spindle collapse in the absence of functional Eg5 requires dynein activity and an initial intercentrosomal distance of less than ~5μm, supporting our hypothesis that dynein opposes Eg5 by crosslinking and sliding antiparallel MTs. This represents a novel role for dynein during mammalian spindle assembly.
Because centrosome separation in prophase requires dynein, presumably anchored to the nuclear envelope acting on astral MTs, as well as Eg5, acting on antiparallel MTs [2, 23], we propose that as mitosis progresses and centrosomes separate, dynein becomes recruited to newly forming regions of antiparallel overlap where it can antagonize the activity of Eg5 and limit or stabilize centrosome separation, so as to prevent anaphaselike prometaphase . With centrosomes stably separated, the capture of chromosomes by centrosomal MTs may be more efficient, thus enhancing chromosome biorientation and spindle assembly.
All materials for cell culture were obtained from Sigma-Aldrich (St. Louis, MO) with the exception of Opti-MEM, which was obtained from Invitrogen (Carlsbad, CA) and fetal bovine serum, which was obtained from Atlanta Biologicals (Norcross, GA). Unless otherwise noted, all other chemicals were obtained from Sigma-Aldrich (St. Louis, MO).
LLC-Pk1 cells expressing either GFP-tubulin or photoactivatable (PA)-GFP-tubulin were cultured as previously described [7, 25]. Cells were plated on glass coverslips (Corning Inc. Life Sciences, Corning, NY) or etched glass coverslips (Bellco Glass Co., Vineland, NJ) 2 days prior to imaging. For live imaging, cells were mounted in chambers containing non-CO2 MEM supplemented with 0.3U/mL Oxyrase (EC Oxyrase, Oxyrase Inc., Mansfield, OH) and were maintained at ~37°C. Nocodazole treatment and 4X washouts were performed as previously described , except that 5–10min incubations separated each washout.
Monastrol was used at 200μM. p150-CC1 plasmid, a gift of Dr. Tarun Kapoor (The Rockefeller University, New York, NY), was prepared according to protocol  and, following dilution with injection buffer (50nM K-Glu, 1mM MgCl2, pH 7.0), was injected at 25μM. Injection was performed on a Nikon Eclipse TE 300 inverted microscope using either a 60x or 100x phase objective lens and a PV820 Pneumatic PicoPump (World Precision Instruments, Sarasota, FL). Needles were pulled from Omega Dot capillary glass tubes (Friedrich and Dimmock, Inc., Millville, NJ) on a Brown-Flaming P-80 micropipette puller (Sutter Instrument, Co., Novato, CA).
LLC-Pk1 cells were rinsed twice in calcium- and magnesium-free phosphate buffered saline (PBS−/−), and fixed in glutaraldehyde (0.25% glutaraldehyde in PBS−/−), formaldehyde (3.7% formaldehyde in H2O), paraglutaraldehyde (3.7% paraformaldehyde, 0.1% glutaraldehyde and 0.5% Triton X-100 in PBS−/−) or 100% methanol and rehydrated in PBS containing 0.1% Tween and 0.02% sodium azide. The following primary antibodies were used in these experiments: γ-tubulin, used at 1:2000; Mad2, a gift of Dr. Alexey Khodjakov (Wadsworth Center, Albany, NY), used at 1:200; and YL½ (Accurate Chemical, Westbury, NY), used at 1:2. Incubations with primary antibodies were performed overnight at room temperature or for 1hr at 37°C. Cy3- (Jackson ImmunoResearch Laboratories, West Grove, PA) or fluorescein isothiocyanate-labeled (Sigma-Aldrich, St. Louis, MO) secondary antibodies were used at the recommended dilution for 30 or 90min at room temperature, respectively. DNA was visualized with 4′,6-diamidino-2-phenylindole, used at 1:300. Coverslips were mounted in Vectashield (Vector Laboratories, Burlingame, CA) and sealed with nail polish.
Images were acquired using a Nikon Eclipse TE 300 inverted microscope equipped with a 100x phase, NA 1.4 objective lens, a spinning disk confocal scan head (Perkin Elmer, Waltham, MA) and a Hamamatsu Orca ER cooled CCD camera (Hamamatsu, Bridgewater, NJ). All images were taken with a dual wavelength (488/568) filter cube. Image acquisition was controlled by Metamorph software (Molecular Devices, Sunnyvale, CA). Time-lapse sequences were acquired at 5sec-2min intervals using exposure times of 400–800msec. Z-stacks were acquired at 0.2μm steps using similar exposure times. Photoactivation experiments were performed as previously described . Images of fixed cells were acquired by capturing optical sections every 0.2μm using exposure times of 400–600msec (at 488nm) and 600–800msec (at 568nm).
The modeling was based on numerical solutions of the systems of differential equations described and explained in the Supplemental Text. The numerical analysis was done using standard Matlab m-files; simulations were performed on a desktop computer.
p150-CC1 plasmid was a kind gift of Dr. Tarun Kapoor (The Rockefeller University, New York, NY). Antibodies to Mad2 were a kind gift of Dr. Alexy Khodjakov (Wadsworth Center, Albany, NY). We thank members of the Lee and Ross labs (University of Massachusetts, Amherst, MA) for insightful comments. This work was supported by grants from the National Institutes of Health (to P.W. and A.M.).
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.