Roles of Polymerization Dynamics, Opposed Motors, and a Tensile Element in Governing the Length of Xenopus Extract Meiotic Spindles

Abstract

Metaphase spindles assemble to a steady state in length by mechanisms that involve microtubule dynamics and motor proteins, but they are incompletely understood. We found that Xenopus extract spindles recapitulate the length of egg meiosis II spindles, by using mechanisms intrinsic to the spindle. To probe these mechanisms, we perturbed microtubule polymerization dynamics and opposed motor proteins and measured effects on spindle morphology and dynamics. Microtubules were stabilized by hexylene glycol and inhibition of the catastrophe factor mitotic centromere-associated kinesin (MCAK) (a kinesin 13, previously called XKCM) and destabilized by depolymerizing drugs. The opposed motors Eg5 and dynein were inhibited separately and together. Our results are consistent with important roles for polymerization dynamics in regulating spindle length, and for opposed motors in regulating the relative stability of bipolar versus monopolar organization. The response to microtubule destabilization suggests that an unidentified tensile element acts in parallel with these conventional factors, generating spindle shortening force.

Full text

Molecular Biology of the Cell
Vol. 16, 3064–3076, June 2005
Roles of Polymerization Dynamics, Opposed Motors, and a
Tensile Element in Governing the Length of Xenopus
Extract Meiotic Spindles
□
D
□
V
T. J. Mitchison,*
†
P. Maddox,*
‡
J. Gaetz,*
§
A. Groen,*
†
M. Shirasu,*
†
A. Desai,*
储
E. D. Salmon,*
‡
and T. M. Kapoor*
§
*Marine Biological Laboratory, Woods Hole, MA 02543;
†
Department of Systems Biology, Harvard Medical
School, Boston, MA 02115;
‡
Department of Biology, University of North Carolina at Chapel Hill, Chapel Hill,
NC 27599;
§
Laboratory of Chemistry and Cell Biology, Rockefeller University, New York, NY 10021; and
储
Department of Cellular and Molecular Medicine, University of California–San Diego, San Diego, CA 92093-
0685
Submitted March 1, 2005; Accepted March 14, 2005
Monitoring Editor: Tim Stearns
Metaphase spindles assemble to a steady state in length by mechanisms that involve microtubule dynamics and motor
proteins, but they are incompletely understood. We found that Xenopus extract spindles recapitulate the length of egg
meiosis II spindles, by using mechanisms intrinsic to the spindle. To probe these mechanisms, we perturbed microtubule
polymerization dynamics and opposed motor proteins and measured effects on spindle morphology and dynamics.
Microtubules were stabilized by hexylene glycol and inhibition of the catastrophe factor mitotic centromere-associated
kinesin (MCAK) (a kinesin 13, previously called XKCM) and destabilized by depolymerizing drugs. The opposed motors
Eg5 and dynein were inhibited separately and together. Our results are consistent with important roles for polymerization
dynamics in regulating spindle length, and for opposed motors in regulating the relative stability of bipolar versus
monopolar organization. The response to microtubule destabilization suggests that an unidentified tensile element acts
in parallel with these conventional factors, generating spindle shortening force.
INTRODUCTION
A mitotic, or meiotic, spindle at metaphase can maintain a
steady state in size and shape for prolonged periods, despite
rapid turnover of subunits, movement of internal compo-
nents and dissipation of free energy. In this article, we
address the mechanisms that govern metaphase spindle
length. Length is important for spindle function, because it
influences the distance over which chromosomes are segre-
gated. Furthermore, probing the factors that govern length
provide information on assembly and force-producing
mechanisms. Metaphase spindle length tends to be rela-
tively constant within a given cell type, but it varies consid-
erably between species and between cell types in an organ-
ism. Spindle length typically increases with cell size and
genome size, but this relationship can break down in spe-
cialized cells. In large eggs, female meiotic spindles are
typically small compared with the egg cell. This is appro-
priate for meiosis biology; the egg meiotic spindle segre-
gates one set of chromosomes a short distance into a polar
body, retaining the other set near the cortex. In Xenopus
laevis, the egg is ⬃1000
␮
m in diameter, and unfertilized
eggs arrest at metaphase of meiosis II, containing a spindle
⬃25
␮
m in length attached at one pole to the cortex at the
top of the egg (Cha et al., 1998). Spindles assembled in
cytoplasmic extracts made from unfertilized Xenopus eggs
recapitulate meiosis II morphology (Sawin and Mitchison,
1991a), but their length has not been systematically studied.
Spindle length per se has received relatively little atten-
tion, but many models have been proposed for how forces
on chromosomes and poles are generated. These forces are
thought to also govern spindle length, with steady-state
length arising from a balance of pushing and pulling forces.
Force balance models can be divided into those that high-
light the role of microtubule polymerization dynamics (In-
oue and Sato, 1967; Margolis and Wilson, 1981; Mitchison et
al., 1986; Inoue and Salmon, 1995); those that highlight ac-
tion of ATPase motor proteins (McIntosh et al., 1969; Hoyt et
al., 1993; Gaglio et al., 1996; Sharp et al., 2000; Nedelec, 2002;
Cytrynbaum et al., 2003); and those that highlight the role of
the “spindle matrix”, a hypothetical, nonmicrotubule, tensile
element (Pickett-Heaps et al., 1997). Recently, a different type
of model was proposed, in which spindle length is set not by
a balance of forces, but by a concentration gradient of mor-
phogens diffusing from a source at chromatin to a global
sink in the cytoplasm (Karsenti and Vernos, 2001). Most of
these models have in common that they seek to explain
spindle length with the microtubule system, including dy-
namics regulators, motors, and cross-linkers, as the sole
mechanochemical element. Exceptions are the original poly-
merization dynamics model that preceded the discovery of
tubulin (Inoue and Sato, 1967), and spindle matrix models,
that explicitly propose a nonmicrotubule, tensile element
This article was published online ahead of print in MBC in Press
(http://www.molbiolcell.org/cgi/doi/10.1091/mbc.E05–02–0174)
on March 23, 2005.
□
D
□
V
The online version of this article contains supplemental mate
-
rial at MBC Online (http://www.molbiolcell.org).
Address correspondence to: T. J. Mitchison (timothy_mitchison@
hms.harvard.edu).
3064 © 2005 by The American Society for Cell Biology

(Pickett-Heaps et al., 1997). The most widely discussed mod-
els have been those in which spindle length is governed by
some combination of polymerization dynamics and opposed
motor proteins, and the purpose of this study was to criti-
cally evaluate these models in Xenopus extract spindles.
A useful distinction in considering models for spindle
length regulation is between mechanisms that act extrinsic
to the spindle, versus intrinsic mechanisms. Potential extrin-
sic mechanisms include limiting amounts of some subunit,
and forces generated at the cell cortex. Intrinsic mechanisms
include balanced forces within the spindle and a possible
morphogen gradient emanating from chromatin. In mam-
malian tissue culture mitosis, the spindle incorporates ⬃50%
of the cell’s tubulin (Zhai and Borisy, 1994), suggesting
component limitation is a significant factor. Pulling forces
from the cortex acting on astral microtubules are known to
play a significant role in length regulation in several mitotic
systems (Sharp et al., 2000). Thus, extrinsic and intrinsic
factors probably act in concert to govern the length of typical
mitotic spindles. In contrast, extrinsic mechanisms are prob-
ably much less important in egg meiosis. The meiotic spin-
dle is small compared with the egg (⬃10
⫺5
of the egg vol
-
ume in Xenopus) and presumably does not deplete a
significant fraction of the egg’s tubulin. Pulling from the
cortex operates mainly on one spindle pole in egg meiotic
spindles (Lutz et al., 1988) and is probably a minor factor in
governing spindle length. In this study, we formally dem-
onstrate that the length of spindles assembled in Xenopus
egg extracts is governed by intrinsic mechanisms, and we
investigate these mechanisms by perturbation experiments.
MATERIALS AND METHODS
Xenopus egg extract sand spindles with replicated DNA were made by stan-
dard methods (Desai et al. 1999a) and used within 90 min of assembly. Spindle
assembly and imaging were performed at 19 –20°C. For polarization imaging,
we used a Nikon (Melville, NY) TE-300 inverted microscope equipped for
differential interference contrast using de Senarmont compensation, a 20⫻
long working distance objective, a heat reflection filter, and a cooled charge-
coupled device camera. The Wollaston prisms were removed, and the polar-
izer set slightly away from extinction. A chamber was made by drilling a
22-mm-round hole in a 1-mm-thick sheet of stainless steel the shape of a
microscope slide. A 25-mm-round coverslip was cemented under the hole
with valap (33% paraffin wax, 33% beeswax, and 33% lanolin). Three to 10
␮
l
of extract containing spindles was deposited on the coverslip and smeared
out into a disk 5–10 mm in width with the pipette tip. The spread extract was
immediately covered with 250
␮
l of mineral oil. Length measurements were
made by capturing images of many random fields and measuring all spindles,
ignoring multipolar structures. Perturbing reagents were added and mixed
immediately before spreading the extract in the chamber. Microscopes for
wide-field and confocal fluorescence imaging, and fluorescent probes for
tubulin and kinetochores have been described previously (Desai et al. 1999a;
Maddox et al., 2003). Optimal concentrations of X-rhodamine tubulin for
confocal speckle imaging were determined empirically for each extract, with
⬃50 nM typical. Affinity-purified, inhibitory antibody to Xenopus mitotic
centromere-associated kinesin (MCAK) was made and characterized as de-
scribed previously (Walczak et al., 1996). Affinity-purified anti-nuclear mitotic
apparatus (NuMA) IgG was made by immunization with the C-terminal
peptide (C)TAKSPRASNKLFERKQQRNK coupled to keyhole limpet hemo-
cyanin, and affinity purified using the same peptide coupled to agarose, by
using the methods described in Field et al. (1998). Specificity was tested by
immunoprecipitation as described in the legend to Supplemental Figure 1.
Anti-NuMA pulled down primarily a band of the expected molecular weight,
whose identity was confirmed by Western blotting of the immunoprecipitate
(Supplemental Figure 1). Western blotting of whole extract was negative for
NuMA, presumably because the high protein concentration interfered with
transfer to nitrocellulose, a problem we have noticed with other antibodies.
For imaging, the IgG was labeled with Alexa488-NHS ester (Molecular
Probes, Eugene, OR) according the manufacturer’s recommendations. p50
dynamitin was made as described previously (Heald et al., 1997). Immu-
nodepletion with magnetic beads was performed as described previously
(Funabiki and Murray, 2000).
The microtubule-depolymerizing drug N-(2-napthyl)-3-trifluoromethyl-
benzene sulfonamide (105D) was tested for effects on pure tubulin polymer-
ization as described in the legend to Supplemental Figure 2. Its caged deriv-
ative had no effect in this assay. Synthesis of 105D and its caged derivative are
described in the legend for Supplemental Figure 3. Both were ⬎95% pure by
thin layer chromatography (TLC) and liquid chromatography/mass spec-
trometry (LC/MS) and gave the expected molecular ions. On photolysis in
methanol by using a hand-lamp, caged 105D released 105D with ⬃100% yield
by TLC and LC/MS. For experiments in extracts, caged 105D was dissolved
in dimethyl sulfoxide as a 200 mM stock by warming to 60°C and added to
extracts to a final concentration of 400
␮
M. This concentration is approxi-
mately the solubility limit, and it was chosen so that only a fraction of the
compound had to be photocleaved to cause microtubule depolymerization,
thus limiting the UV dose delivered to spindles. The caged drug had no
discernible effect on spindle assembly in the absence of UV light. To measure
the rate of uncaging by using a microscope in extracts, we collected sequential
images with a DAPI filter set. 105D is weakly fluorescent in the UV, whereas
its caged derivative is not, so the field gets brighter as uncaging proceeds
(Supplemental Figure 4). We estimated a half-time for photorelease in extract
of ⬃5 s with 360 nm illumination from a 100W Hg bulb through a 4,6-
diamidino-2-phenylindole (DAPI) filter cube to a 60⫻/1.4 numerical aperture
PlanApo objective in a Nikon 800e upright microscope. For spindle depoly-
merization experiments, caged 105D was added to preformed spindles in
extracts that also contained X-rhodamine tubulin (⬃200 nM) and Alexa488-
anti-CenpA (⬃1
␮
g/ml; Maddox et al., 2003). Spindles were located using dim
rhodamine illumination and time-lapse imaging initiated with a double label
filter cube (wide-field) or no filter cube (confocal). After a few preuncaging
frames had been collected, a DAPI filter cube was brought into the epi path,
and the field illuminated with 360-nm light for 1–2 s. Then, the cube was
changed back and the time-lapse sequence continued. One to2sofUV
illumination in extracts containing 400
␮
M caged 105D generated sufficient
free105D to depolymerize spindles in the field and had no effect on spindles
when the caged drug was not present. The effect of uncaging was remarkably
local. A spindle in the microscope field (⬃200-
␮
m circle) subject to 360-nm
illumination with caged-105D present rapidly disassembled, whereas spin-
dles outside the field were unaffected. Thus, we were able to trigger and
follow depolymerization of several spindles in each slide-coverslip prepara-
tion. To determine why the effects of photoreleasing 105D are local and
persistent, we imaged the drug diffusing away from a UV-illuminated area,
by using its intrinsic fluorescence, and by limiting the observation light to
minimize further photorelease. We observed that 105D partitions into mem-
branes and moves only very slowly through the extract after photorelease
(Supplemental Figure 4), explaining its local effects on minute time scales.
RESULTS
To measure spindle length in a convenient and nonperturb-
ing way, mitotic extract (3–10
␮
l) containing spindles with
replicated chromosomes (“cycled spindles”, Desai et al.
1999a) was spread in a thick (⬃100
␮
m) layer under mineral
oil and imaged by polarization microscopy (Figure 1A).
Spindles were remarkably stable in this preparation, and
retained a constant length for at least 30 min by time-lapse
imaging (our unpublished data). The spindles usually ro-
tated and translated slowly, showing they were free of in-
teractions with the substrate that might influence their
length. To test whether the amount of any extract compo-
nent is limiting for spindle length, we assembled spindles in
parallel at three different concentrations of added sperm and
then measured the distribution of spindle lengths (Figure 1,
A and B). The morphology, birefringence, mean length, and
length distribution were similar in each case. A confocal
fluorescence image of a meiosis II spindle at similar magni-
fication in an unfertilized egg is shown for comparison
(Figure 1C). The average length of extract spindles varied
slightly from preparation to preparation, in part due to
variability in the extent to which spindles fused. Extract
spindles are, on average, a little longer than egg meiosis II
spindles (⬃25
␮
m; Cha et al., 1998), but overall the extract
system does a good job of recapitulating meiosis II morphol-
ogy. Because spindle length was independent of spindle
concentration, and spindles seemed not to interact with the
substrate, we conclude that length is regulated by mecha-
nisms intrinsic to the spindle.
To test whether protein polymerization dynamics plays
an important role in governing spindle length in Xenopus
extracts, as they do in other spindles (Inoue and Sato, 1967),
we first measured the effect of an agent that nonspecifically
Roles of Polymerization Dynamics
Vol. 16, June 2005 3065

promotes protein assembly. We used 2-methypentane-2,4-
diol, called “hexylene glycol,” in the mitosis literature (sug-
gested by Robert Palazzo). This solvent promotes aggrega-
tion of many proteins and is used for this purpose in
crystallography. It also promotes microtubule polymeriza-
tion and stabilizes spindles and asters (Rebhun et al., 1975;
Harris and Clason, 1992). At 3% (vol/vol) and above, hexy-
lene glycol promoted rapid nucleation of microtubule asters
throughout the extract. At 2%, it only slowly promoted
nucleation, but it had a remarkable effect on preformed
spindles, causing them to increase in birefringent retarda-
tion (a measure of spindle microtubule density) and to grow
progressively in length and volume (Figure 2A and Movie
M1). Spindle length increased at a constant rate of ⬃1.7
␮
m/min, at least to the point where length doubled (Figure
2B). To probe the mechanics of elongation in hexylene gly-
col, we imaged fluorescent tubulin at speckle levels by spin-
ning disk confocal microscopy (Figure 2C and Movie M2).
Visual inspection and kymograph analysis (Figure 2D)
showed that speckles throughout the spindle moved pole-
wards at a rate similar to the rate of pole separation. We
conclude that spindles elongate in hexylene glycol by anti-
parallel sliding between the two half-spindles, with little or
no microtubule depolymerization. This situation is reminis-
cent of poleward flux with depolymerization blocked. The
antiparallel sliding component of poleward flux can be
blocked by adenyl-5⬘-yl imidodiphosphate (AMPPNP)
(Sawin and Mitchison, 1991b) and Eg5 inhibitors, including
monastrol (Miyamoto et al., 2004), so we tested the effect of
these agents on spindle elongation. Monastrol caused spin-
dle collapse when added alone, but this collapse could be
suppressed by also adding p50 dynamitin (discussed be-
low). Hexylene glycol-induced elongation was blocked by
both agents (Figure 2B). We conclude that spindle elonga-
tion in hexylene glycol occurs by antiparallel sliding be-
tween the half spindles, most likely driven by Eg5. Consis-
tent with this view, extract spindles also elongate by
antiparallel sliding when pole organization is disrupted
Figure 1. Length of in vitro spindles is independent of spindle density. Cycled spindles were assembled at a final density of 50, 16.7, and
5.5 sperm nuclei/
␮
l and imaged in oil overlay chambers by polarization microscopy 90–120 min after addition of cytostatic factor (CSF)
extract. (A) Representative images at each dilution. Bar, 5
␮
m. (B) Histograms of spindle lengths at each dilution measured from polarization
images. The mean and SD (in micrometers) is written on each histogram. Note they are similar at each dilution. (C) Immunofluorescence
image of the meiosis II spindle in an unfertilized egg, at a comparable magnification, for comparison. The animal cortex of the egg is to the
left. Anti-tubulin staining and laser confocal imaging. Image kindly provided by David Gard (University of Utah) (see Cha et al., 1998).
T. J. Mitchison et al.
Molecular Biology of the Cell3066

(Gaetz and Kapoor, 2004), dependent on Eg5 activity
(Shirasu-Hiza et al., 2004). The rate of spindle expansion in
hexylene glycol (⬃1.7
␮
m/min) is slower than that that rate
of antiparallel sliding during flux in control spindles (⬃4
␮
m/min). Eg5 motors seem to work against an unknown
mechanical load to drive flux (Miyamoto et al., 2004), and we
suspect that hexylene glycol increases this load.
We next tested a more specific microtubule stabilizing
agent, inhibitory antibody to Xenopus MCAK (a kinesin 13,
formerly called XKCM1). This kinesin promotes microtubule
catastrophes in an ATP-dependent reaction (Desai et al.
1999b). It is the most potent known catastrophe factor in
Xenopus egg extract (Tournebize et al., 2000), and removing
or inhibiting it induces massive microtubule polymerization
in M-phase extracts (Walczak et al., 1996). Plus ends in
extract spindles are thought to undergo bounded dynamic
instability, meaning that they do not grow indefinitely, but
rather catastrophe frequently enough to have a defined av-
erage length (Verde et al., 1992). Titrating a catastrophe
inhibitor into extract is predicted to first increase this aver-
age length and then cause a transition to the unbounded
regime, where plus ends grow indefinitely. We titrated anti-
MCAK into extract, adding the inhibitor well before spindle
assembly, and fixing at two time points, to ensure we were
measuring spindle length at steady state. Increasing concen-
trations of anti-MCAK up to 10
␮
g/ml caused a small,
dose-dependent increase in spindle length, and a larger,
dose-dependent increase in total microtubules per spindle
(Figure 3). At 15
␮
g/ml (Figure 3) and above (Figure 4),
microtubules elongated dramatically, forming large asters.
When MCAK was inhibited to this extent before spindle
assembly, disorganized structures formed whose length
could not be defined (Figure 3). We infer that decreasing the
catastrophe rate while staying in the bounded regime mod-
estly increased spindle length. Decreasing catastrophe to the
point of entering the unbounded regime resulted in disor-
ganization, and not a dramatic increase in length.
Figure 2. Hexylene glycol promotes spindle expansion. Hexylene glycol [2% (vol/vol)] was added to cycled spindles from a 20% stock in
water immediately before imaging. (A) Polarization images from a time-lapse sequence. Elapsed time shown in minutes and seconds.
Hexylene glycol was added ⬃1 min before the first time point. Note the growth in spindle length and width. Time elapsed shown as
minutes:seconds. Length of time box, 10
␮
m. See Movie M1 (B) Spindle length as a function of time after addition of hexylene glycol alone
(2%; black circles), hexylene glycol plus AMPPNP (1.5 mM; open squares), or hexylene glycol plus monastrol (400
␮
M) and p50 dynamitin
(0.9 mg/ml; gray diamonds). (C) Spinning disk confocal fluorescence images of a spindle containing speckle level tubulin to which 2%
hexylene glycol has been added. The line in the second panel was used to make the kymograph in D. See Movie M2. Bar, 5
␮
m. (D)
Kymograph through the upper pole of the spindle shown in C. Time is vertical, distance along the line horizontal. Note the movement of
the pole away from the spindle equator, which is to the right in the panel. Note also that many thin lines, which are speckle trajectories,
parallel the movement of the pole (black lines highlight examples). Near the equator, speckles also are moving in the other direction (white
line), in parallel with the opposite pole.
Roles of Polymerization Dynamics
Vol. 16, June 2005 3067

To better understand the effects of unbounding microtu-
bule length, we imaged preformed spindles live after adding
saturating amounts of anti-MCAK. Microtubules rapidly ex-
tended away from the spindle, converting them into large
asters (Figure 4A and Movie M3). During this outgrowth,
the spindle itself did not seem to elongate, and in some cases
the poles even moved slightly closer together. Although the
principle effect of inhibiting MCAK is inhibiting plus end
catastrophes, this protein also has the biochemical capability
to depolymerize minus ends at poles (Desai et al. 1999b), and
other kinesin 13 family members have been implicated in
this activity (Gaetz and Kapoor, 2004; Rogers et al., 2004). To
probe effects on poles, we imaged a fluorescent spindle pole
marker as well as tubulin speckles after inhibiting MCAK
(Figure 4B and Movie M4). The pole marker was affinity-
purified antibody raised to a carboxy-terminal peptide from
Xenopus NuMA, labeled with Alexa488. NuMA accumulates
at the poles of Xenopus extract spindles by a dynein–dynac-
tin-dependent mechanism and has often been used as a pole
marker (Merdes et al., 1996). Our antibody was specific by
immunoprecipitation (Supplemental Figure 1), gave the re-
ported localization for NuMA in live and fixed spindles and
had no discernible effect on spindle assembly or dynamics
when added to extracts or on the response to inhibiting
MCAK as judged by comparing effects in the tubulin chan-
nel with and without the probe. Confocal imaging con-
firmed massive outgrowth of microtubules from the spindle
starting a few minutes after adding anti-MCAK. Anti-paral-
lel microtubules sliding in the center of the spindle that is
characteristic of poleward flux continued (Movie M4). By
tubulin imaging alone, and more informatively by tubulin ⫹
NuMA imaging, we observed progressive disorganization
Figure 3. Titration of anti-MCAK antibody effects. Affinity-purified, inhibitory antibody to MCAK (Walczak et al., 1996) was added to
spindle assembly reactions containing X-rhodamine tubulin 20 min after bring the extract back into M phase after replication, well before
spindle assembly. Samples were squash-fixed at 80 and 120 min after bringing the extract into M phase. (A) Representative spindles fixed
at 80 min as a function of final antibody concentration. Note the massive microtubule growth out of the spindle at 15
␮
g/ml (and higher;
our unpublished data), indicative of transition to unbounded dynamic instability. Bar, 10
␮
m. (B) Quantification of spindle length in this
experiment. Open bars are 80- and black 120-min samples. The error bars indicate one SD above and below the mean, and the number in each
bar the sample size. (C) Quantification of total microtubules per spindle measured as integrated fluorescence intensity minus local
background signal. Labeling as in B.
T. J. Mitchison et al.
Molecular Biology of the Cell3068

of poles. In about half the spindles, this disorganization
takes the form of the pole elongating and curling backward
toward the equator, seeming to track back along the outer
surface of the spindle (Figure 4B and Movie M4). Tubulin
speckles follow this pole movement, moving out from the
main body of the spindle, and curling around with the
moving pole (Figure 4B, blue arrows). Due to the disorga-
nization of pole structure, and lack of a direct assay for
depolymerization, we were unable to quantify microtubule
depolymerization at poles in this experiment.
To further probe the role of microtubule dynamics in
governing spindle length, we rapidly depolymerized micro-
tubules. We added 20
␮
M nocodazole to an aliquot of extract
containing spindles on a microscope slide, mixed them, put
on a coverslip, and initiated imaging by using a dry 40⫻ lens
to facilitate rapid location of spindles. Spindles shortened
and depolymerized completely in 2–3 min with this treat-
ment, and it was necessary to locate them within ⬃10sof
drug addition to obtain useful information on early events.
Spindle shortening in nocodazole was previously argued to
occur by pulling at kinetochores (Cassimeris et al., 1990), so
we used a nonperturbing kinetochore probe (Alexa488-anti-
CenpA IgG; Maddox et al., 2003) to observe possible action
of such forces. As expected from previous work (Inoue and
Sato, 1967; Salmon et al., 1984; Cassimeris et al., 1990), the
microtubule density dropped rapidly, and the pole-to-pole
distance decreased (Figure 5A and Movie M5). Note that the
images in Figure 5 and Movie M5 are normalized to peak
intensity, optimizing visualization of remaining structures,
but giving a misleading impression of microtubule density,
which is quantified as total tubulin fluorescence in Figure
5B. We expected to see stretched kinetochores pulling the
poles inwards in this experiment, and we were surprised to
observe that the distance between sister kinetochores invari-
ably decreased shortly after drug addition, indicating loss of
tension (Figure 5A, red dots; Figure 5B, triangles; and Movie
M5). All kinetochores visualized experienced this relaxation
(⬎50 kinetochore pairs in 13 spindles in 10 sequences). In
some cases, kinetochores seemed to experiencing compres-
sion during spindle collapse. This was evident from lateral
movement away from the spindle axis, twisting of the ki-
netochore pair (Figure 5A, 119 s), and apparent curving or
buckling of kinetochore microtubules (Figure 5A, 119 s, note
microtubule bundle connected to the upper sister of the pair
marked with blue lines; also see Movie M5).
Further investigation of the forces in collapsing spindles
required high-resolution imaging, which was difficult using
nocodazole addition because it took too long to find and
focus on a spindle. Therefore, we synthesized a photochem-
ically “caged” microtubule-depolymerizing drug that al-
lowed us to find and image a spindle before and after
triggering depolymerization (Figure 6A). We started with
105D, a depolymerizing drug that was found by phenotypic
screening of a combinatorial library (Mitchison, 2003). It
arrests cells in mitosis in tissue culture cells with a pheno-
type similar to nocodazole and with an IC
50
of ⬃3
␮
M
(⬃20-fold less potent than nocodazole). It inhibited polymer-
ization of pure tubulin (Supplemental Figure 2) and depo-
lymerized extract spindles with an IC
50
of ⬃20
␮
M, again
⬃20-fold less potent than nocodazole. 105D has several ad-
vantages for making a caged derivative. It is simple to
synthesize and modify (Supplemental Figure 3). The caged
drug had no detectable effect on microtubules or extract
spindles at its solubility limit (⬃400
␮
M). 105D is weakly
fluorescent in the DAPI channel, whereas its caged deriva-
tive is not. Thus, photorelease and subsequent movement of
the drug can be imaged and quantified by fluorescence
microscopy (Supplemental Figure 4).
Time-lapse imaging of spindles before and after photore-
lease of 105D showed effects broadly similar to nocodazole
addition. The intensity of the tubulin signal rapidly de-
creased after photorelease (Figure 6, B and D); the spindle
shortened (Figure 6, B–D, and Movies M6 and M7); and
sister kinetochores moved together and twisted (Figure 6D
and Movie M7). 105D did not depolymerize spindle micro-
tubules as efficiently as nocodazole, and a subset of micro-
tubule bundles was stable for several minutes after photore-
Figure 4. Inhibition of MCAK promotes microtubule outgrow and
pole curling, but not spindle elongation. Affinity-purified, inhibitory
antibody to MCAK was added at 150
␮
g/ml final to cycled spindles
immediately before imaging. (A) Polarization images from a time-
lapse sequence. Note the dramatic outgrowth of microtubules from
the spindle (arrows in second panel). There is some alteration of
spindle shape and pole structure over the sequence, but pole-to-pole
distance remains approximately constant. Elapsed times shown in
minutes:seconds. Antibody was added ⬃1 min before the first time
point. Time bar, 18
␮
m. See Movie M3. (B) Spinning disk confocal
fluorescence images of a spindle containing speckle level X-rhoda-
mine tubulin (green) and Alexa488-labeled antibodies to NUMA at
5
␮
g/ml as a pole marker (red). Note that the upper pole, visualized
by anti-NUMA localization (white arrow), is curled over and ap-
parently attached to the body of the spindle. By the last time point,
the pole has moved toward the spindle equator, and become par-
tially disorganized. In the tubulin channel, the movements of the
microtubules that are associated with this pole movement can be
visualized (blue arrows). In the center of the spindle, microtubules
slide apart in both directions. In the upper part, most of the flow is
upwards. At the top of the spindle, the flow curls over, correlating
with movement of the pole back toward the equator. Bar, 5
␮
m. See
Movie M4.
Roles of Polymerization Dynamics
Vol. 16, June 2005 3069

lease. Most of these stable microtubules terminated at
kinetochores (Figure 6D and Movie M7), implying they are
kinetochore microtubules, which are known to be selectively
resistant to depolymerizing drugs in other systems
(Cassimeris et al., 1990). In nocodazole, kinetochore fibers
eventually disappeared (Figure 5A), but after uncaging
105D they did not, perhaps because 105D partially stabilizes
kinetochore microtubules. Other tubulin drugs are known to
have stabilizing as well as destabilizing effects (Wilson et al.,
1999). Both wide-field (Figure 6C) and confocal (Figure 6D)
sequences suggested that the spindle poles are pulled (or
pushed) together after photorelease, resulting in compres-
sion of attached microtubules. These forces caused buckling
of selectively stable kinetochores fibers, movement of sister
kinetochores toward each other, and twisting or sideways
movement of the sister pair. If the poles were pulled to-
gether by microtubules, we would expect to see straight
bundles of microtubules connecting them. By through-focus
imaging of ⬎10 collapsing and fully collapsed spindles with
wide-field and confocal microscopy, we were unable to find
any straight bundles of microtubules connecting poles or
any sister kinetochore pairs still under tension. Figure 6D,
941 and 978 s, shows two images from a through-focus
confocal image series where all remaining microtubule bun-
dles were either buckled or pushed out sideways from the
spindle axis, and all kinetochore pairs are close together. We
tentatively conclude that when overlap microtubules are
rapidly removed with a drug, the poles are pulled (or
pushed) together by something other than microtubules.
Spindle collapse is unlikely to depend on F-actin, because
our extracts routinely included cytochalasin D (⬃3
␮
M), and
addition of latrunculin B to 30
␮
M (in addition to cytocha-
lasin D) had no effect on the speed, extent, or morphology of
collapse induced by photorelease of 105D (our unpublished
data). The photorelease experiment made it possible to col-
lect before and after data on spindle length and microtubule
density. For seven representative spindles, the peak rate of
shortening (measured pole to pole) averaged 7
␮
m/min
(range 2–13
␮
m/min). Length plateaued after 350 s on av-
erage (n ⫽ 6, range 300 –450 s), when the spindle was 47% on
average of its initial length (range 31–57%). Total tubulin
fluorescence plateaued ⬃300 s after initiating depolymeriza-
tion, at an average of 8% of the initial fluorescence (range
5–11%). The half-time between initiating depolymerization
and reaching the plateau in total fluorescence was 55 s (n ⫽
7, range ⫽ 53– 60 s).
We next probed the role of opposed motor proteins in
governing spindle length, focusing on Eg5 and dynein. Eg5
is essential for bipolarity in extract spindles (Sawin et al.,
1992; Kapoor et al., 2000), where it drives antiparallel sliding
associated with poleward flux and spindle elongation (Fig-
ure 2; Miyamoto et al., 2004; Shirasu-Hiza et al., 2004). Dy-
nein works together with dynactin and NuMA to organize
Figure 5. Rapid microtubule depolymerization by using nocodazole. Cycled spindles containing labeled tubulin (green) and a kinetochore
marker (anti-CenpA; red) were mixed with nocodazole (20
␮
M final) and wide-field time-lapse imaging was initiated ⬃10 s after mixing by
using a 40⫻ dry objective. (A) Imaging of spindle collapse. Elapsed time is shown in seconds relative to the first image; tubulin fluorescence
is normalized to peak brightness in each image, to highlight the organization of remaining microtubules. Insets show 3⫻ magnification of
marked kinetochores. Note the sister kinetochores are initially well separated. The spindle rapidly looses fluorescence, shortens, and the sister
kinetochore move together. Note some evidence of buckling of kinetochore microtubules at 59, 119 s. Bar, 5
␮
m (main panels), 1.7
␮
m (insets).
See Movie M5. (B) Quantification of the sequence in A, showing distance between the marked kinetochores in A (k-k), pole to pole distance
(p-p) and integrated tubulin fluorescence after subtracting local background signal. Note that k-k distance drops faster (in percent terms) than
p-p distance.
T. J. Mitchison et al.
Molecular Biology of the Cell3070

and focus the poles (Merdes et al., 1996) and seems to be the
dominant minus end-directed motor in Xenopus extract spin-
dles on the basis of inhibition experiments (Heald et al.,
1997). Using polarization microscopy, we confirmed the ef-
fects of inhibiting Eg5 with monastrol, and dynactin with
excess p50 dynamitin, (Figure 7, B and C). Unexpectedly,
when both these inhibitors were added together, they coun-
teracted each other (Figure 7D). Almost all spindles were
now bipolar, and their poles were more organized than with
p50 alone. We quantified length and morphology for spin-
dles in the presence of inhibitors, adding them before and
after spindle assembly (Table 1). p50 almost completely
rescued the effect of monastrol on bipolarity, whether it was
added before or after assembly. Monastrol partially rescued
the effect of p50 on poles morphology. When both inhibitors
were added before spindle assembly, length was almost
completely rescued (32 vs. 39
␮
m in controls; Table 1). This
was less true when both inhibitors were added after assem-
bly (23 vs. 37
␮
m in controls; Table 1), probably reflecting
reduced effectiveness of p50 when added after assembly.
Although the double-inhibited spindles had relatively normal
morphology and length by polarization imaging, they were
much more fragile than control spindles. Unlike control spin-
dles, they were easily damaged by squashing between a slide
and coverslip or by touching with microneedles, and their
average length in replicate experiments was more variable than
with controls. We note that the concentrations of p50 we used
was probably insufficient to completely block pole organiza-
tion. Complete inhibition of pole organization tends to increase
spindle length (Gaetz and Kapoor, 2004; Shirasu-Hiza et al.,
2004), perhaps due to displacement of a kinesin 13 depolymer-
ization factor from the poles (Gaetz and Kapoor, 2004).
Figure 6. Microtubule depolymerization using a caged drug. Cycled spindles containing labeled tubulin, kinetochore, and pole markers in
some cases, and caged 105D were imaged by time-lapse wide-field (B and C) or confocal (D) fluorescence. At t ⫽ 0, 105D was photoreleased
by 1- to 2-s illumination with a UV filter set. All times are in seconds relative to photorelease. Bar, 7.5
␮
m (B), 6
␮
m (C), 5
␮
m (D), and 1.25
␮
m (D, insets). (A) Structure of caged 105D and the photochemical reaction that releases active drug. (B) Example of a spindle before and
after photorelease of 105D (wide-field). Tubulin fluorescence is presented with brightness and contrast held constant throughout the sequence
to highlight the rapid decrease in microtubule density. This sequence also contained a fluorescent kinetochore probe; see Movie M6. (C)
Second widefield example. Tubulin fluorescence is normalized to peak brightness in each image, to highlight the organization of remaining
microtubules. Note the decrease in spindle length, and buckling of remaining stable microtubules. See Movie M7. (D) Example of
photorelease of 105D (confocal). Probes are for tubulin (green), kinetochore (anti-CenpA; red), and poles (anti-NUMA; red). Tubulin
fluorescence is normalized to peak brightness in each image. Insets show marked kinetochore pairs at 4⫻ magnification. The focal plane was
changed several times during this sequence, and different kinetochore pairs are shown in each panel. At late time points, most of the
remaining stable microtubules connect to kinetochores. Note that the poles (large red dots) move progressively together. Sister kinetochores
first move together and then twist and move away from the spindle axis, whereas their attached microtubules either buckle or move outward,
suggesting kinetochore fibers experience compression from the collapsing poles. The spindle was optically sectioned twice during this
sequence, and the images at 941 and 978 s are two focal planes from one through-focal series. Note the absence of straight microtubule
bundles directly connecting the poles at any time point or focal plane.
Roles of Polymerization Dynamics
Vol. 16, June 2005 3071

Eg5 is still present in spindles after monastrol treatment
(Kapoor et al., 2001) and might influence the response to
dynein inhibition even when inhibited. We therefore tested
whether p50 also could rescue the effect of Eg5 depletion.
Depletion of Eg5 to ⬍5% of normal resulted in assembly of
almost entirely monopolar spindles as reported previously,
and addition of p50 before assembly rescued this effect
(Figure 8). Eg5 depleted p50 spindles were mostly bipolar,
their poles were more focused than in p50 alone, and they
were approximately the right length. We conclude that al-
Figure 7. Eg5 and dynein/dynactin play an-
tagonistic roles in spindle assembly. Cycled
spindles were assembled in the presence of no
drug, the dynactin inhibitor p50 dynamitin (0.9
mg/ml), the Eg5 inhibitor monastrol (200
␮
M),
or both. Drugs were added at the time of CSF
addition, and spindles were imaged live in oil
overlay chambers by polarization microscopy
90–120 min later. A panel of representative im-
ages is shown for each condition. Note the ex-
pected appearance of p50 spindles with unfo-
cussed poles and monopolar monastrol
spindles. When both drugs were added, bipo-
larity was completely rescued, and pole focus-
ing was partially rescued. Bar, 10
␮
m. See Table
1 for quantitation.
Table 1. Affect of motor protein perturbation on spindle morphology and length scored by polarization microscopy
% Bipolar
spindles
Avg pole focus
index (bipolar
spindles)
Avg pole to pole
length (
␮
m)
SD of pole to
pole length
(
␮
m) No. scored
Agents added before spindle assembly
None 100 1.8 39 4 115
p50 98 0.1 37 9 59
Monastrol 2 * 1 4 100
p50 ⫹ monastrol 100 0.8 32 5 104
Agents added after spindle assembly
None 100 1.9 37 7 107
p50 100 1.0 43 4 65
Monastrol 6 * 1 3 81
p50 ⫹ monastrol 79 1.3 23 2 87
Spindle assembly, drug concentrations, and polarization imaging are as per Figure 7. Motor-perturbing agents were added either before
spindle assembly, at the time of CSF add-back, with images taken 90–120 min later, or 60 min after CSF add-back, when steady-state spindles
were already assembled, with images taken 30–60 min later. Spindles were scored as monopolar if they appeared as a single aster at 20⫻;
otherwise, they were bipolar. Multipolar and aggregated structures represented ⬍20% of the total in all cases and were not counted. Any
bipolar spindles with poles separated by less than ⵑ3
␮
m would be scored as monopolar. To quantify pole focusing, each spindle was
assigned a score of 2 if both poles were well focused, 1 if both poles were partially focused or one pole was focused and the other unfocussed,
and 0 if neither pole was focused. Pole focusing was scored only for bipoles.
* Too few bipoles present to reliably calculate the index. Pole to pole length was averaged for all structures and counted as zero for monpoles.
T. J. Mitchison et al.
Molecular Biology of the Cell3072

though inhibition or removal of Eg5 caused spindles to
collapse to monopoles, coinhibition of dynein reversed that
effect and allowed assembly of spindles that are physically
fragile but nevertheless able to achieve an approximately nor-
mal steady-state length. Rescue of bipolarity in Eg5 inhibited
spindles by p50, discovered here, was a key technical advance
for probing the role of Eg5 in flux (Miyamoto et al., 2004).
DISCUSSION
In this article, we begin a systematic experimental attack on
the mechanisms that govern spindle length in the Xenopus
extract system. Extract spindles, and by implication egg
meiosis II spindles, achieve a steady state in length and mass
by purely intrinsic mechanisms (Figure 1). We performed
perturbation experiments to test standard models for length
regulation based on polymerization dynamics and opposed
motor proteins, finding they can account for some, but not
all, of the results. To account for the response to rapid
microtubule depolymerization, we propose adding a non-
microtubule tensile element. Because egg meiotic spindles
are small relative to the cell than contains them, and their
assembly is largely chromatin driven rather than microtu-
bule organizing center driven (Karsenti and Vernos, 2001),
they may use length-governing mechanisms that are differ-
ent from somatic mitotic spindles.
Polymerization dynamics models predict that increasing
microtubule length by increasing polymerization or decreas-
ing depolymerization should cause spindles to elongate.
They can account for the response of extract spindles to
hexylene glycol (Figure 2) and also for the slight increase in
spindle length observed when the catastrophe factor MCAK
is partially inhibited, but dynamic instability is still bounded
(Figure 3). They fail to account for the response of spindles
to stronger MCAK inhibition, when microtubules go into
unbounded growth and plus ends leave the spindle, but the
spindle poles do not separate further, and rather curl back
toward the equator (Figure 4). This curling phenomenon
might be due to inhibition of minus end depolymerization at
poles by anti-MCAK or simply to disorganization of poles
by misdirected motor activity (discussed below). We cur-
rently lack an assay for measuring depolymerization at
poles that is required to distinguish these possibilities. Ky-
mographs of tubulin speckles do not provide a reliable assay
for depolymerization at poles, because converting sliding
rates into depolymerization rates requires knowing whether
minus ends are static or moving, which has not been mea-
sured in extract spindles. Previous interpretation of kymo-
Figure 8. Antagonizing dynein/dynactin rescues the effects of Eg5 depeletion. Extracts were depleted of Eg5 by using affinity-purified
antibody and magnetic beads. Three rounds of depletion were used to remove all Eg5 that could be detected on Western blots (⬎95%). (A)
Western blot analysis of Eg5 and mock-depleted extracts. Eg5 is removed to below the detection limit in the depleted extract. (B) Quantitation
of spindle morphology. Eg5 and mock-depleted extracts were used to assemble cycled spindles in the presence or absence of p50 dynamitin
(0.9 mg/ml). In Eg5 depletion alone, the majority of spindles were monopolar asters. Addition of p50 promoted assembly of mostly bipolar
spindles. (C) Typical spindle assembled in mock depleted extract. (D) Typical spindle assembled in mock-depleted extract ⫹ p50. Note
bipolar organization with splayed poles. (E) Typical spindle-assembled in Eg5-depleted extract. Note monopolar organization. (F) Typical
spindle assembled in Eg5 depleted extract ⫹ p50. Note rescue of both bipolarity and pole morphology. Bar, 5
␮
m.
Roles of Polymerization Dynamics
Vol. 16, June 2005 3073

graphs assumed static minus ends depolymerizing at poles
during metaphase (Sawin and Mitchison, 1991; Rogers et al.,
2004), and Kinesin 13 family members (previously called
KinI kinesins) were implicated in depolymerization (Gaetz
and Kapoor, 2004; Rogers et al., 2004). However, an alterna-
tive model can be proposed, in which stable minus ends are
distributed throughout the spindle. These ends move pole-
ward at the flux rate without depolymerizing and loose
stability when they reach the poles, leading to loss of the
microtubule from the plus end. To distinguish these models,
we need to localize minus ends in extract spindles and to
measure their dynamic behavior. Perhaps the largest dis-
crepancy from standard polymerization dynamics models
was the response to microtubule-depolymerizing drugs. In
previous work, drug- or pressure-induced spindle shorten-
ing was interpreted as a consequence of pulling forces at
kinetochores generated by microtubule depolymerization
(Cassimeris et al., 1990; Inoue and Salmon, 1995). Instead, we
found that depolymerizing drugs induced a switch from
tension at kinetochores in unperturbed spindles (Maddox et
al., 2003) to compression, arguing that kinetochores were not
pulling the poles together. In fact, our imaging suggested
that the poles are pulled together by something other than
microtubules, as discussed below.
Opposed motor models predict that when the zone of
microtubule overlap in the spindle is increased by microtu-
bule polymerization, more plus end-directed motors are
recruited, increasing the sliding forces between antiparallel
microtubules, and elongating the spindle (Sharp et al., 2000;
Cytrynbaum et al., 2003). Such models predict spindle elon-
gation in hexylene glycol, its sensitivity to AMPPNP and
Eg5 inhibitors, and elongation with partial inhibition of
MCAK. A motor model might also account for the effect of
complete MCAK inhibition on pole morphology (Figure 4B).
Minus end-directed motor complexes at the pole faced with
large numbers of invading microtubules of the wrong po-
larity may track toward the minus ends of the invaders,
causing curling of the poles. We tried to test this by inhib-
iting dynein and MCAK at the same time, but the results
were ambiguous due to disorganization of poles. Opposed
motor models predict opposite effects of inhibiting Eg5 and
dynein/dynactin, and compensating effects when both are
inhibited (Hoyt et al., 1993; Gaglio et al., 1996; Sharp et al.,
2000). In motor inhibition experiments, we tended to see
either relatively normal length bipoles, or monopoles, rather
than intermediate length bipoles (Table 1, inhibitors added
before assembly). This suggests that opposed motor activity
controls the relative stability of monopolar versus bipolar
organization as a switch-like transition, rather than control-
ling spindle length as a continuous variable. More work is
required to address this point experimentally. In a system-
atic investigation of different classes of opposed motor mod-
els, Nedelec (2002) found several stable bipolar solutions. In
all cases, the relative stability of bipolar versus monopolar
organization depended on motor activities, but in most cases
motor activities did not control pole-to-pole distance, which
was instead governed by microtubule polymerization dy-
namics. One of the functions of Eg5 in bipolar spindles is to
drive the sliding components of poleward flux. At the mo-
nastrol concentration used in Figure 7 and Table 1 (200
␮
M),
rate of antiparallel sliding at metaphase is reduced to ⬃10%
of control values (near the detection limit; Miyamoto et al.,
2004), yet average spindle length in spindles treated with
p50 plus monastrol is reduced by ⬍2-fold. Insensitivity of
steady-state length to the rate of antiparallel sliding is sur-
prising, because sliding should tend to increase length, and
spindles indeed elongate when microtubule destabilization
at poles is inhibited (Gaetz and Kapoor, 2004; Rogers et al.,
2004; Shirasu-Hiza et al., 2004). To account for these obser-
vations, either minus end depolymerization rate must de-
crease in concert with sliding rate after Eg5 inhibition/
removal, or spindle length must be governed by a
mechanism that is insensitive to sliding rate. New methods
for probing minus end dynamics are required to distinguish
these hypotheses.
Figure 9. Interpretation of results. See text
for details.
T. J. Mitchison et al.
Molecular Biology of the Cell3074

The most surprising observation in our study was the
response of spindles to microtubule-depolymerizing drugs,
which indicated that poles can be pulled (or pushed) to-
gether by something other than microtubules. It is possible
that our through-focus imaging missed a small number of
straight microtubules running pole to pole that pulled poles
together in response to 105D, but we consider this unlikely
because kinetochore bundles remained buckled for several
minutes after photorelease, implying the pulling factor does
not depolymerize. We also consider it unlikely that 105D
causes microtubules to curve on their own, because it did
not have this action on pure microtubules, and the response
to nocodazole also showed evidence for kinetochore com-
pression and microtubule buckling. We hypothesize that an
unidentified tensile element pulls the poles together and that
this element also opposes elongation in unperturbed spin-
dles. We consider two possible candidates for this element:
external membranes and an internal matrix. Extract spindles
are surrounded by a sheath of membranous organelles, in-
cluding mitochondria and endoplasmic reticulum, that seem
physically connected in thin section electron microscopy
images (Coughlin and Mitchison, unpublished data). Mem-
branes may be important for spindle assembly, because it
failed in high-speed supernatants of Xenopus extracts, unless
they were supplemented with purified membranes (Shirasu-
Hiza and Mitchison, unpublished data). Membranes also
surround meiotic spindles in Drosophila oocytes, where their
importance for spindle assembly was revealed by genetics
(Kramer and Hawley, 2003). Perhaps membranes serve as a
tensile element encapsulating the spindle, tending to oppose
spindle expansion, and driving collapse when microtubules
are depolymerized. Alternatively, spindles may contain
some internal polymer or gel that exerts tension between the
poles, as proposed in the spindle matrix hypothesis (re-
viewed in Pickett-Heaps et al., 1997). Detergent-treated, iso-
lated sea urchin embryo spindles contracted when microtu-
bules were removed using calcium (Salmon and Segall,
1980), an observation more consistent an internal matrix
than tension from membranes. A molecular candidate for an
internal matrix is poly(ADP-ribose), a nonprotein macro-
molecule that is required for bipolar organization of Xenopus
extract spindles, and that seems to turn over much less
rapidly than microtubules (Chang et al., 2004).
In Figure 9, we combine polymerization dynamics, mo-
tors, and a hypothetical tensile element to try and account
for all our data. The model is more explicit in molecular
terms than the data justify, but we hope it provokes discus-
sion and attempts at experimental falsification. We envisage
the matrix as a cross-linked gel that attaches to poles and
plus end-directed motors (Kapoor and Mitchison, 2001), and
thus becomes stretched, storing elastic energy and pulling
the poles inward at steady state. Hexylene glycol, a nonspe-
cific protein-aggregating reagent, promotes recruitment of
more matrix in addition to stabilizing microtubules, result-
ing in balanced growth and increased spindle length while
retaining approximately normal morphology. MCAK inhi-
bition leads to increased tubulin polymerization without a
parallel increase in matrix assembly and to unbounded
growth of plus ends through the poles and out of the spin-
dle. Although plus end-directed motors try to push the
elongated half spindles apart, this is opposed by matrix
stretched between the poles and by curling of poles back
toward the equator driven by dynein-containing complexes
moving on microtubules of the wrong polarity that invaded
the poles (orange arrows). Drug-induced depolymerization
causes microtubules to disassemble faster than the matrix.
Tensile forces focus onto remaining kinetochore microtu-
bules, causing spindle collapse with buckling of kinetochore
fibers. Dynactin inhibition by p50 leads to splaying out of
poles and detachment of matrix. Partial destruction of poles
by p50 (this study) did not increase spindle length, perhaps
because some matrix remains attached. Complete destruc-
tion of poles cause spindles elongation (Gaetz and Kapoor,
2004; Shirasu-Hiza et al., 2004). Eg5 inhibition promotes
movement of the poles together by a combination of matrix
contraction and dynein pulling. Inhibition of both dynactin
and Eg5 results in bipolar spindle that are physically fragile
and lack poleward flux (Miyamoto et al., 2004) but are rela-
tively normal in length (Table 1). We propose these lack
matrix as well as the opposed motor systems and that they
regulate length by dynamic instability alone. Figure 9 does
not address other potentially important processes in spindle
length regulation, including signals diffusing from chroma-
tin and poleward flux, and new experiments are required to
integrate these processes into a complete model. The spindle
matrix hypothesis has long been controversial, but the ex-
periments we report should help in the design of future
experiments to test molecular candidates.
ACKNOWLEDGMENTS
We thank other members of the Marine Biological Laboratory Cell Division
Group and our winter laboratories for comments, and David Gard for infor-
mation on the egg meiosis II spindle. This work was funded by National
Institutes of Health Grants GM-39565 (to T.J.M.), GM-24364 and GM-606780
(to E.D.S.), and GM-65933 (to T.M.K.), and by Marine Biological Laboratory
fellowships from Universal Imaging and Nikon.
REFERENCES
Cassimeris, L., Rieder, C. L., Rupp, G., and Salmon, E. D. (1990). Stability of
microtubule attachment to metaphase kinetochores in PtK1 cells. J. Cell Sci.
96, 9–15.
Cha, B. J., Error, B., and Gard, D. L. (1998). XMAP230 is required for the
assembly and organization of acetylated microtubules and spindles in Xeno-
pus oocytes and eggs. J. Cell Sci. 111, 2315–2327.
Chang, P., Jacobsen, M. K., and Mitchison, T. J. (2004). Poly(ADP-ribose) is
required for spindle assembly and structure. Nature 432, 645–649.
Cytrynbaum, E. N., Scholey, J. M., and Mogilner, A. (2003). A force balance
model of early spindle pole separation in Drosophila embryos. Biophys. J. 84,
757–769.
Desai, A., Murray, A., Mitchison, T. J., and Walczak, C. E. (1999a). The use of
Xenopus egg extracts to study mitotic spindle assembly and function in vitro.
Methods Cell Biol. 61, 385– 412.
Desai, A., Verma, S., Mitchison, T. J., and Walczak, C. E. (1999b). Kin I kinesins
are microtubule-destabilizing enzymes. Cell 96, 69 –78.
Field, C. M., Oegema, K., Zheng, Y., Mitchison, T. J., and Walczak, C. E. (1998).
Purification of cytoskeletal proteins using peptide antibodies. Methods Enzy-
mol. 298, 525–541.
Funabiki, H., and Murray, A. W. (2000). The Xenopus chromokinesin Xkid is
essential for metaphase chromosome alignment and must be degraded to
allow anaphase chromosome movement. Cell 102, 411–424.
Gaetz, J., and Kapoor, T. M. (2004). Dynein/dynactin regulate metaphase
spindle length by targeting depolymerizing activities to spindle poles. J. Cell
Biol. 166, 465–471.
Gaglio, T., Saredi, A., Bingham, J. B., Hasbani, M. J., Gill, S. R., Schroer, T. A.,
and Compton, D. A. (1996). Opposing motor activities are required for the
organization of the mammalian mitotic spindle pole. J. Cell Biol. 135, 399–414.
Harris, P. J., and Clason, E. L. (1992). Conditions for assembly of tubulin-
based structures in unfertilized sea urchin eggs. Spirals, monasters and cytast-
ers. J. Cell Sci. 102, 557–567.
Heald, R., Tournebize, R., Habermann, A., Karsenti, E., and Hyman, A. (1997).
Spindle assembly in Xenopus egg extracts: respective roles of centrosomes and
microtubule self-organization. J. Cell Biol. 138, 615–628.
Hoyt, M. A., He, L., Totis, L., and Saunders, W. S. (1993). Loss of function of
Saccharomyces cerevisiae kinesin-related CIN8 and KIP1 is suppressed by KAR3
motor domain mutations. Genetics 135, 35– 44.
Roles of Polymerization Dynamics
Vol. 16, June 2005 3075

Inoue, S., and Sato, H. (1967). Cell motility by labile association of molecules.
The nature of mitotic spindle fibers and their role in chromosome movement.
J. Gen. Physiol. Suppl. 50, 259 –292.
Inoue, S., and Salmon, E. D. (1995). Force generation by microtubule assem-
bly/disassembly in mitosis and related movements. Mol. Biol. Cell 6, 1619–
1640.
Kapoor, T. M., Mayer, T. U., Coughlin, M. L., and Mitchison, T. J. (2000).
Probing spindle assembly mechanisms with monastrol, a small molecule
inhibitor of the mitotic kinesin, Eg5. J. Cell Biol. 150, 975–988.
Kapoor, T. M., and Mitchison, T. J. (2001). Eg5 is static in bipolar spindles
relative to tubulin: evidence for a static spindle matrix. J. Cell Biol. 154,
1125–1133.
Karsenti, E., and Vernos, I. (2001). The mitotic spindle: a self-made machine.
Science 294, 543–547.
Kramer, J., and Hawley, R. S. (2003). The spindle-associated transmembrane
protein Axs identifies a membranous structure ensheathing the meiotic spin-
dle. Nat. Cell Biol. 5, 261–263.
Lutz, D. A., Hamaguchi, Y., and Inoue, S. (1988). Micromanipulation studies
of the asymmetric positioning of the maturation spindle in Chaetopterus sp.
oocytes: I. Anchorage of the spindle to the cortex and migration of a displaced
spindle. Cell Motil. Cytoskeleton 11, 83–96.
Maddox, P., Straight, A., Coughlin, P., Mitchison, T. J., and Salmon, E. D.
(2003). Direct observation of microtubule dynamics at kinetochores in Xeno-
pus extract spindles: implications for spindle mechanics. J. Cell Biol. 162,
377–382.
Miyamoto, D. T., Perlman, Z. E., Burbank, K. S., Groen, A. C., and Mitchison,
T. J. (2004). The kinesin Eg5 drives poleward microtubule flux in Xenopus
extract spindles. J. Cell Biol. 167, 813–818.
Margolis, R. L., and Wilson, L. (1981). Microtubule treadmills–possible mo-
lecular machinery. Nature 293, 705–711.
McIntosh, J. R., Hepler, R. K., and vanWie, D. G. (1969). Model for mitosis.
Nature 224, 659–663.
Merdes, A., Ramyar, K., Vechio, J. D., and Cleveland, D. W. (1996). A complex
of NuMA and cytoplasmic dynein is essential for mitotic spindle assembly.
Cell 87, 447–458.
Mitchison, T. J. (2004). Probing cell division with “chemical genetics.” In: The
Harvey Lectures, Vol. 98, New York: Wiley–Liss, 19– 41.
Mitchison, T., Evans, L., Schulze, E., and Kirschner, M. (1986). Sites of micro-
tubule assembly and disassembly in the mitotic spindle. Cell 45, 515–527.
Nedelec, F. (2002). Computer simulations reveal motor properties generating
stable antiparallel microtubule interactions. J. Cell Biol. 158, 1005–1015.
Pickett-Heaps, J. D., Forer, A., and Spurck, T. (1997). Traction fibre: toward a
“tensegral” model of the spindle. Cell Motil. Cytoskeleton 37,1–6.
Rebhun, L. I., Jemiolo, D., Ivy, N., Mellon, M., and Nath, J. (1975). Regulation
of the In Vivo mitotic apparatus by glycols and metabolic inhibitors. Ann. N.Y.
Acad. Sci. 253, 362–377.
Rogers, G. C., Rogers, S. L., Schwimmer, T. A., Ems-McClung, S. C., Walczak,
C. E., Vale, R. D., Scholey, J. M., and Sharp, D. J. (2004). Two mitotic kinesins
cooperate to drive sister chromatid separation during anaphase. Nature 427,
364–370.
Salmon, E. D., and Segall, R. R. (1980). Calcium-labile mitotic spindles isolated
from sea urchin eggs (Lytechinus variegatus). J. Cell Biol. 86, 355–365.
Salmon, E. D., McKeel, M., and Hays, T. (1984). Rapid rate of tubulin disso-
ciation from microtubules in the mitotic spindle in vivo measured by blocking
polymerization with colchicine. J. Cell Biol. 99, 1066–1075.
Sawin, K. E., LeGuellec, K., Philippe, M., and Mitchison, T. J. (1992). Mitotic
spindle organization by a plus-end-directed microtubule motor. Nature 359,
540–543.
Sawin, K. E., and Mitchison, T. J. (1991a). Mitotic spindle assembly by two
different pathways in vitro. J. Cell Biol. 112, 925–940.
Sawin, K. E., and Mitchison, T. J. (1991b). Poleward microtubule flux mitotic
spindles assembled in vitro. J. Cell Biol. 112, 941–954.
Sharp, D. J., Rogers, G. C., and Scholey, J. M. (2000). Microtubule motors in
mitosis. Nature 407, 41–47.
Shirasu-Hiza, M., Perlman, Z. E., Wittmann, T., Karsenti, E., and Mitchison,
T. J. (2004). Eg5 causes elongation of meiotic spindles when flux-associated
microtubule depolymerization is blocked. Curr. Biol. 14, 1941–1945.
Tournebize, R., Popov, A., Kinoshita, K., Ashford, A. J., Rybina, S., Pozniak-
ovsky, A., Mayer, T. U., Walczak, C. E., Karsenti, E., and Hyman, A. A. (2000).
Control of microtubule dynamics by the antagonistic activities of XMAP215
and XKCM1 in Xenopus egg extracts. Nat. Cell Biol. 2, 13–19.
Verde, F., Dogterom, M., Stelzer, E., Karsenti, E., and Leibler, S. (1992).
Control of microtubule dynamics and length by cyclin A- and cyclin B-
dependent kinases in Xenopus egg extracts. J. Cell Biol. 118, 1097–1108.
Walczak, C. E., Mitchison, T. J., and Desai, A. (1996). XKCM 1,aXenopus
kinesin-related protein that regulates microtubule dynamics during mitotic
spindle assembly. Cell 84, 37– 47.
Wilson, L., Panda, D., and Jordan, M. A. (1999). Modulation of microtubule
dynamics by drugs: a paradigm for the actions of cellular regulators. Cell
Struct. Funct. 24, 329–335.
Zhai, Y., and Borisy, G. G. (1994). Quantitative determination of the propor-
tion of microtubule polymer present during the mitosis-interphase transition.
J. Cell Sci. 107, 881– 890.
T. J. Mitchison et al.
Molecular Biology of the Cell3076