The Functional Antagonism between Eg5 and Dynein in Spindle Bipolarization Is Not Compatible
with a Simple Push-Pull Model
Stefan Florian
1,2and Thomas U. Mayerl,*
1
Department of Biology and Konstanz Research School Chemical Biology, University of Konstanz, Universitatsstrasse 10, 78457 Konstanz, Germany
2
Present address: Department of Systems Biology, Harvard Medical School, 200 Longwood Avenue, Warren Alpert Building 536, Boston, MA 02115, USA
'Correspondence: thomas.u.mayer@uni-konstanz.de
SUMMARY
During cell division, the molecular motor Eg5 cross- links overlapping anti parallel microtubules and pushes them apart to separate mitotic spindle poles.
Dynein has been proposed as a direct antagonist of Eg5 at the spindle equator, pulling on anti parallel microtubules and favoring spindle collapse. Some of the experiments supporting this hypothesis relied on end point quantifications of spindle pheno- types rather than following individual cell fates over time, Here, we present a mathematical model and proof-of-principle experiments to demonstrate that endpoint quantifications can be fundamentally misleading because they overestimate defective phenotypes. Indeed, live-cell imaging reveals that, while depletion of dyne in or the dynein binding protein Lis1 enables spindle formation in presence of an Eg5 inhibitor, the activities of dynein and Eg5 cannot be titrated against each other, Thus, dynein most likely antagonizes Eg5 indirectly by exerting force at different spindle locations rather than through a simple push-pull mechanism at the spindle equator.
INTRODUCTION
In mitosis, the equal distribution of chromosomes is mediated by the mitotic spindle, a microtubule-based structure whose bipolar shape results from the coordinated activities of microtu- bule-associated proteins and molecular motors (Compton, 2000; Scholey et aI., 2003; Heald and Walczak, 2008; Dumont and Mitchison, 2009). Within the spindle, microtubule minus ends converge at the poles and the plus ends pOint toward the spindle equator where they create a zone of overlapping, anti- parallel microtubules. Due to its homotetrameric structure and plus-end
-directedmotility, Eg5, a conserved member of the kinesin
-5 family, can crosslink these antiparallel microtubulesand push spindle poles apart (Sawin et aI., 1992; Kashina et aI., 1996; Kashina et aI., 1997; Sharp et aI., 1999; Kwok
408
et al., 2004; Kapoor and Mitchison, 2001
;Feren
zet aI., 2010). Consequently, in most systems studied so far, inactiva- tion of Eg5 results in spindle collapse and monopolar spindle formation. Dynein, a minus-end-directed motor, was suggested to be a direct antagonist of Eg5, which pulls spindle poles together and thus promotes spindle collapse (Mitchison et aI., 2005; Tanenbaum et aI., 2008; Ferenz et aI., 2009). The orches- tration of these two antagonistic forces is believed to be essential for bipolar spindle formation
. In allusion to theactin/
myosin-based mechanism of force generation in muscle tissue, this model is called the "push-pull mitotic muscle" model (Mcintosh et aI., 1969; Kapitein et aI., 2005; van den Wildenberg et aI., 2008; Figures 1A and 18). A corollary of this model is that the activities of Eg5 and dynein are titratable, i.e., spindle collapse induced by reduced Eg5 activity should be rescued by lowering dynein activity resulting in the re-equilibration of forces acting within the spindle. Using a fixed-sample-based endpoint quantification, Tanenbaum et al. (2008) showed that (1) depletion of dynein heavy chain (DHC) increases the percentage of bipolar spindles in mitotic cells treated with low doses of
Eg5inhibitor and that (2) the
efficiency of dyneindepletion to rescue spindle bipolarity indeed decreased with increasing inhibition of Eg5, showing no significant effect at saturating inhibitor concentrations. Thus, in line with a simple push-pull model, this study suggested that the activities of Eg5 and dynein at the spindle equator can be titrated against each other.
Here, we develop a mathematical model that explains why fixed
-cellanalyses are often not appropriate for the quanti- fication of cellular phenotypes, and we validate its predictions experimentally. Thus, we show that quantitative statements about the frequency of spindle phenotypes derived from single time point measurements are inaccurate and, without exception, synchronization protocols or live-cell imaging should be used to avoid this problem. Consequently, we apply live-cell imaging to revisit the push-pull antagonism of Eg5 and dynein and find that, in contradiction to previous results from fixed samples, Eg5 is not titratable against the activity of dynein or its binding protein Lis1
.We conclude that the mechanism of bipolar spindle formation is not compatible with a simple push-pull model where Eg5 and dyne in act as direct antagonists.
First publ. in: Cell Reports ; 1 (2012), 5. - S. 408-416 DOI : 10.1016/j.celrep.2012.03.006
Konstanzer Online-Publikations-System (KOPS)
A
c
o
Histone H2B GFP
B
Eg5 phenotype
. 0000 0009 001B 0027 · 0036 0045 0054 ! 01:03
? , x <'t _~ 0 ' ~j
1$ • if f,'f
JI
r~ ,_! /
{f " ~"" ,,# ";#!i" ~
i!$,!!) \~/" I,~ I~.
,:);~
( , ; ) / "01:12 .;," 01.21 (. 0130 0- 0139 !b 01:48 1, 01:57 t 02.06 'Ii 0215
t) t j (>: , \ ~ ?~> ! , ~ ~ ~ t
~
•
~ ~&
$ $ q "> '" ; ' " , ~ ~ ...- ... , /~.... /~ " -
(~';' ~ \~) W® ~~) ~
(,tJi/~ (:~I
%t~:"<'~);.
(\1\ 'ffi1~~!)*1
C' 0224<.
02.33 0242 < 1 0251 " 0300 k 03;09 0 ... 18 0327- t
r
f i ~ I < ~» I'~A '
#1I(J $ " ' . Ir' #' • '11 lIt/>., "
(:\ /"";, :'\ (;'\ /h'" (~;~ tif.) /><.,\
\ , \ \ ~",1 \~!;' \\;~/~~l'
,">
'I?'< !\~_~)"* \~(/ ~~,:! •
\,~)Hoechsl phosptio Histone H3
tflO<~F 15min N itCIVd,·1 00% r~, •
Interphase ~ '( 71 I Interphase
(1'N~".)'1000 /~ ;~.
• death t :135 hTsPllldle assemblydoloci.vo checkpoint,
accumulation ot defective spindles
Figure 1. The Push-Pull Mitotic Muscle Model
(A) Scheme of the role of Eg5 (red) and minus-end- directed motors (blue) in spindle pole separation.
(6) Experimental predictions of the push-pull model, which are tested in this study. Reduced Eg5 activity prevents pole separation and leads to the formation of monopolar spindles. Dynein depletion in cells with reduced (but not absent) Eg5 activity has been suggested to re-enable spindle bipolarization by lowering Eg5-antago- nizing forces.
(C) Representative movie stills and corresponding immunofluorescence images of GFP-H26-ex- pressing HeLa cells treated with VS83 (12,5 I.M), Mitotic cells were defined as featuring a bipolar spindle as soon as they had formed a metaphase plate (orange circle, last frame). The green circle highlights a cell failing to assemble a metaphase plate during the course of the experiment. After 18 hr of imaging, cells were fixed and stained with Hoechst 33258 (DNA) and antibodies against phospho-histone H3 and (J;-tubulin, In all following experiments, the percentage of mitotic cells that formed bipolar spindles over time was calculated from movies (NBc'u.'), In parallel, the percentage of bipolar spindles within the total mitotic population was assessed from fixed samples (Nfix) , Scale bars, 151.m,
(D) The spindle assembly checkpoint causes a delay in mitotic exit for defective spindles, Mitotic spindle lifetimes are values.derived from experiments treating HeLa cells with VS83,
types is still often performed
usingsingle-time-point microscopy on fixed cells. This approach is even more frequent in large-scale screening projects where live-cell imaging,
so-calledhigh- content screening, can become simply
IiiI unfeasible considering the amount of generated data. During experiments analyzing mitotic spindle phenotypes (Figure 1 C), we frequently noticed that there is a
strongdiscrepancy between the prevalence of defective spindle phenotypes
in the mitotic cell populationat the end of
anexperiment and the actual frequency at which the phenotype emerges in mitotic cells during the exper- iment. This happens because normal and defective spindles have significantly
RESULTS different
lifetimes:normal spindles are short-lived, transient
structures that quickly disappear as cells divide and exit mitosis, whereas defective spindles persist for much
longer periods oftime because they activate the
spindle-assemblycheckpoint (SAC), resulting in a mitotic delay (Varetti and Musacchio,
2008; Figure 1 D). Thus, in a single-time point experiment, defec-tive spindles are overrepresented compared to intact spindles.
It
is important
toknow if this analytical bias results
inonly Endpoint Quantifications of Spindle Phenotype
Frequency Are Inherently Inaccurate
Live-cell microscopy has become increasingly popular because
of its potential to provide detailed, time-resolved insights into
cellular processes. However, analysis of the generated data is
time intensive, and, therefore, quantification of cellular pheno-
negligible quantitative or fundamental, even qualitative, errors. It is surprising that, in the context of mitosis research, th
is questionhas not been addressed yet, and fixed-cell analyses are used as a routine method to determine the frequency of phenotypes. We tried to address this issue theoretically and derived a simple mathematical relationship, predicting the fraction of cells within the mitotic population displaying normal spindles at a given time point in fixed samples (N
fix)from the fraction of cells forming normal spindles in mitosis during the experiment
(Nactua/)'We defined the relationship as follows (more details and the exact derivation are provided in the Extended Experimental Procedures):
Nfix(t) = nnormaf-splndles(t)
n mitotic_cells
(t)
nnormaLspindfes (t)
n
n~rmal_splndles(t)
+ ndefective _spindles (t) ,where
nphenotype(t)represents the number of mitotic cells with the indicated phenotype at a given time point. We further assumed (Figure 1 D) that (1) cells enter mitosis at a constant rate, (2) a frac- tion of mitotic cells
(Nactual)forms bipolar spindles and exits mitosis after
tnormalhours in metaphase, and (3) a fraction of 1 -
Nactualmitotic cells develop a defective spindle and exit mitosis after
tdefectlvehours. Under these conditions, if the time of measurement, t, is longer than the times in mitosis of both the defective
(tdefective)and the normal
(tnorma/)mitotic population, the following equation applies:
Nfix(t) = Nactualtnormal . Nactualtnormal
+ (1 -
Nactua/)tdelectiveAccording to this equation, Nfix , i.e., the percentage of mitotic cells displaying normal spindles at a given time point, depends not only on
Nactua"i.e., the fraction of mitotic cells forming bipolar spindles, but also on two other variables:
tnormaland
tdefective.The.se are the average lifetimes of normal and defective spindles, which can both vary independently of
Nactua/.Plotting N"x against
Nactual
for different values of
tnormaland
tdefectivecan help one to understand whether and when fixed-cell analyses are accurate enough to be useful (Figure 2A). To obtain accurate results from fixed samples, N"x should match
Nactualas closely as possible, allowing us to directly estimate the
latterfrom the former. According to the equation, this is the case if
tnormalequals
tdefective
(Figure 2A, green line). Because
tnormal«
tdefective,however, the actual function curve for a typical experiment involving spindle perturbations is highly distorted. Figure 2A shows a plot of the equation for
tnormal= 0.25 hr and
tdefective=
13.5 hr (yellow curve), both average values derived from experi
-ments using the Eg5 inhibitor V883 (published half maximal effective concentration [EC
50),7.27 J.lM; 8arli et aI., 2005). For these values, the whole range of
Nactualfrom 0 to 0.75 corre- sponds to the range from 0 to 0.05 in Nfix , resulting in underesti- mation of changes in
Nactua/'Reciprocally, for
Nactualvalues between 0.8 and 1, small changes in
Nactualcause huge increases in Nfix, thus tending to cause overestimation of effects.
In this situation, it is impossible to determine
Nactualwithout prior knowledge of the lifetimes of individual spindle phenotypes
.Notably, changes in experimental conditions might affect not only Na ctu.1 but also the lifetimes of spindle phenotypes implying that they have to be determined individually for each experi-
mental condition. Thus, in this case, live-cell analysis is the only appropriate approach to quantify cellular phenot ypes.
Experimental Evidence Confirms Inaccuracy of Fixed-Sample Quantifications
For all the experiments presented in this study, the same setup was used: Thirty hours or 54 hr after transfection of HeLa cells stably expressing green fluorescent protein (GFP)-tagged histone H2B with short interfering (si)RNA duplexes, Eg5 inhibi- tors were added and the time-lapse image acquisition was immediately started. At the end of time-lapse acquisition, cells were fixed and stained for immunofluorescence analysis (Fig- ure 1 C). We performed two proof-of-concept experiments to validate the predictions of our model and illustrate the impact of mitotic lifetimes on Nfix' First, we titrated V883 from 3.1
~tMto 100 J.lM in control RNAi cells to gradually lower the fraction of cells forming bipolar spindles and quantified both
Nactualand Nfix' As shown in Figure 2 (indicated with yellow squares in Figure 2A), the resulting data points for
Nactual(x axis) and N"x (y axis) were indeed in close match with the function curve (yellow) defined by the equation using
tnormal '"0.25 hr and
tdefectlve
= 13.5 hr. This confirms that, if
tnorma/« tdefective,values for Nfix and
Nactualdiffer substantially. Next, we wanted to demonstrate how a change in
lifetimesaffects experimental interpretation. To this end, we increased the lifetime of bipolar spindles by depleting 8hugoshin (8g01). Depletion of 8g01 induces premature loss of sister chromatid cohesion, resulting in 8AC activation regardless of spindle function (8alic et aI., 2004). Thus, by prolonging the lifetime of bipolar spindles, 8g01 depletion should result in
tnormal "" tdefectiveand, therefore, in a much better correlation of Nfix with
Nactual(Figure 2A, green line).
Indeed,8g01 depletion caused a marked increase in N"x without actually rescuing bipolar spindle formation
(Nactua,) ,as indicated in Figure 2 by the close-to-vertical connecting lines between control GL2-RNAi (yellow squares) and Sgo1
-RNAi(green triangles) data points. Thus, without knowledge of the function of 8g01, analyses of fixed cells would result in the con-
. elusion that 8g01 antagonizes Eg5 function in spindle assembly,an effect that, as live-cell imaging reveals, is entirely due to changes in mitotic timing. In summary, these proof-of-concept experiments confirm our mathematical model and its prediction that, because perturbations can unpredictably affect both
Nactual
and mitotic
lifetimes, estimationof
Nactualfrom fixed samples is impossible without the knowledge of
tnormaland
tdefective-
Live-Cell Imaging Reveals that the Push-Pull Model Cannot Explain the Antagonism between Dynein and Eg5
Previous studies using fixed samples revealed that depletion of
DHC leads to a huge improvement in bipolar spindle formation
in cells treated with
lowdoses of the Eg5 inhibitor 8-trityl-L-
cysteine (8TLC) (DeBonis et aI., 2004; 8koufias et aI.
, 2006) buthad a negligible effect at maximal Eg5 inhibition (Tanenbaum
et aI., 2008). Thus, residual amounts of Eg5 activity seemed to
be required to rescue spindle bipolarity in DHC-depleted cells,
supporting the idea that Eg5 and dynein are direct antagonists
whose activities can be titrated against each other. We repeated
A a
tnomw,NactUlJ' e uation plots
'C Q) Nh':..-
tOO/fnal N IKIUcl'+ tde1tn;/1vo (1-N actual) tnofma,=O.25 h, tddf6Cliw=13.50 h
:::..
x .9 .00 ¥ .... - . . - " •. -~-• • . ~---< . " " . • • • ~Vl O.25Nac,u.:t1
Q)
r
N,,='0 c .8 0.25N"",,+13.5 (I-N"".)
'0.
3.12,'J.lM
Vl .7
a;
t =t .0 .6 flomt.l/ chIffl(;/'VI'I
0.
:0
?
~/lnilINaCf!ll11.<: .5 N =
.~ I
,,,
.J;;:,,17IJI;N}jclml/+~,mll/(1-~CfUJ)I) I
!!! .4
!
NI~\(=Nac!lMIQ;
0
,
"0 .3
/
c .2
0
J
U ~~~A e2'p~!!..,!,~~!~ __ ... ___ . ______
1" .1 6.25J.1M / / control RNAi "'8g01 RNAi 30h
"ii~ $ .... ~,. .. ,d .. numbers indicate conc. of V883
z 0
. . . . _ ~ ... N ¥
o
.1 .2 .3 .4 .5 .6 .7 .8 .9 1 ~N"",.F fraction of cells assembling bipolar spindles (live)
B c
Vl
.9
Ill! 8go RNAi
o
control RNAi=ai CJ (J) .8
Q)
30h
~~
.7'E ~.6
'019c:
8.
.5:fl~.4
'" c
T '-.§
.3toE
.2z
.1
~O
,,'I- <,<::>
'0' ,,'1-'
,,'I- ",<::>
'0' ,,'1-'
V883 [JJM] V883 [JJM]
Figure 2. A Model Describing the Mathematical Relationship between Phenotype Frequency in Fixed and Live Samples
(A) A plot of the equation describing Nf/x (y axis) as a function of Nectue' (x axis) and mitotic spindle lifetimes (tnon"e' and (defective) based on the variables introduced in Figure 1 D. The green line (45·) would correspond to identical lifetimes for bipolar and defective spindles. The yellow line corresponds to realistic values for tno,mal
(0.25 hr) and (defective (13.5 hr) in an Eg5-inhibitor-treated cell population. Data points represent experimental results for Nf/x and Nac/usl for cells treated with control (GL2) or Sg01 siRNA. Black lines connect data points for GL2-and Sg01-RNAi cells treated with the same dose of VS83. Each data point represents the mean of at least two experiments.
(B and C) Original data for the experiments plotted in (A). HeLa cells were transfected with control (GL2) or Sg01 siRNA and incubated for 30 hr. Next, VS83 was added at the indicated concentrations, and time-lapse movies were acquired for 18 hr, followed by fixation and fixed-cell analysis. Then, Nox (B) and Nactusl
(C) were determined. Bars represent means ± SEM of at least two experiments.
the experiments
under precisely identical conditions but again analyzed inpara
llelboth live-cell and fi
xed samples,i.e., the
samecells studied live were analyzed by immunofluorescence microscopy after the completion of the movie. The efficiency of dyne in depletion was determined by immunoblotting for dyne in
intermediate chain(Figure
3A),w
hich was previously shown tobe degraded upon DHC depletion (Grigoriev et aI., 2007).
Indeed, fixed-cell
analyses confirmed that dynein depletion
significantly increased the percentage of mitotic cells displaying bipolar spi
ndles and, furthermore, that the effect of DHCdepletion (t., difference between DHC-
and control-RNAi cells)is strongest when
EgSis only slightly inhibited (Figure 38).
However,
the situation was different when we
analyzed the cor- respondinglive-cell
movies(Figure 3C). Specifically, while we could confirm that dynein depletion can rescue bipolar spindle
formation in EgS-inhibited cells, we observed that (1) DHC deple- tion had no significant
effecton bipolar spindle formation at the lowest concentration of STLC (t. = 0.02) and (2) the rescue
efficiency of dynein depletion did not decrease with increasingSTLC concentrations but was actually maximal (t.
~0.S6) at high inhibitor concentrations and plateaued up to 80 JlM STLC.
Notably, our live-cell analyses revea
ledthat the EC
sovalue of
STLC was slightly higher than 2
~LM(Figure 3C),
suggestingthat dynein depletion efficiently rescued spindle bipolarization
evenunder conditions
(~40times the
ECso)when EgS was
maxima
lly inhibited. If our mathematical model is correct,the
discrepancy
betweenlive and fixed samples must be caused
by the
effect of dynein depletion on mitotictiming. To
confirmthis, we quantified mitotic
lifetimes for cells forming monopolar and bipolar spindlesin contro
l-or DHC-RNAi cells treated with
A
54 h, 25 nM GL2 RNAi OHC RN AiEg5
[ ::!\!J:m~:~""" . 113o
kOaDynein Intermediate Chain
B
D
LI1!::.:cI!l_iIIW_· . _ _
-1~
kOaHela cells
'l- "
STLC!JlMJ
control RNAi, 2 pM STLC
c
"'·"'"",'=!">'-=.!.-'T-"-'-'="=:;'="T==P"'-=''; III DHCRNAi
.1
~o
'l- "
STLC !JlMJ
OHC RNAi, 2 pM STLC
o control ANAl 54 h
.. rootaphase/
tjpolar
.prophasel
monopolar
10 t5 10 t(h)
Figure 3. Bipolar SpindJe Formation in Eg5-lnhibited and Dynein-Depleted Cells Does Not Fit to.a Push-Pull Mechanism (A) Fifty-four hours after transfection, control (Gl2) and DHC-RNAi cells were analyzed by immunoblotting as indicated.
(B and C) Quantification of N"x (B) and NBe/uB' (C) after transfection with DHC or control siRNA for 54 hr followed by addition of the Eg5 inhibitor STLC and imaging for an additional 18 hr at the indicated doses. Bars represent means ± SEM of triplicate experiments.
(D) Quantification of times in mitosis of bipolar and mono polar spindles at 2 ~IM STLC from one of the experiments shown in (C). Each horizontal bar represents a single cell. The length of the bars represents the time spent in a monopolar (blue) or bipolar (green) state. Note that this representation illustrates in an intuitive manner how N"x and Naetua, are related (see details in the Extended Experimental Procedures): The size of a population on the y axis corresponds to the actual frequency of the phenotype over time (e.g., Nae/ua' for normal spindles), while the total area of the bars of a population corresponds to its relative size in fixed samples (green area for the normal spindle, blue area for the mono polar spindle population). As summarized in the table included in the left panel, DHC depletion significantly changes tprophase. tnorma" and tdefective-
2 J.lM STLC. The results are plotted in Figure 3D, each horizontal bar representing a cell and the length of the bar representing its lifetime (blue = prophase/monopolar, green = bipolar). As shown before (Figure 3C), about 60% of mitotic control-RNAi cells treated with 2 J.lM STLc were able to form bipolar spindles, and, on average, these cells remained in prophase for about 0,93 hr before they formed a bipolar spindle and quickly exited mitosis
(tnormal= 0.36 hr). Notably, consistent with its reported function in checkpoint inactivation (Howell
et aI., 2001),dyne in depletion significantly slowed down metaphase progression of cells forming bipolar spindles from
tnormal= 0.36 hr to 1.15 hr.
Moreover, and unexpectedly,
tdefectivewas dramatically
short-ened from 9.32 hr to 2.9 hr because DHC-depleted cells with
monopolar spindles died much faster than control monopolar
cells, an effect that further aggravated the distortion of N
fix.We
also found that DHC depletion prolongs the time that cells
spend in prophase before forming bipolar spindles (0.93 hr vs
.1.53 hr). Thus, for 2 ,lM STLC, in addition to its effect on
Nactual,DHC depletion induces a
complexchange in lifetimes that
dramatically affects N
fixand explains the differences between
fixed-sample and time-lapse imaging, We conclude that the
previous statement that dynein depletion rescues spindle
bipolarization most efficiently at low Eg5-inhibitor concentrations (Tanenbaum et aI., 2008) is incorrect because of the effect of dynein depletion on the lifetimes of bipolar and monopolar spin- dles. In contrast, our studies relying rigorously on live-cell imaging revealed that dynein depletion rescues spindle bipolarity most efficiently when Eg5 is maximally inhibited, strongly sug- gesting that the functional interrelationship between Eg5 and dynein is more complex than suggested by the simple push- pull model.
Time-Lapse Analysis of Lis1 Depletion Reveals an Effect on Bipolarization Similar to that of DHC1 Depletion
Lis1 (Mesngon et aI., 2006) stimulates the ATPase activity of the dynein complex, and its depletion was shown to promote bipolar spindle formation when Eg5 activity is slightly, but not maximally, inhibited, similar to the depletion of DHC (Tanenbaum et aI., 2008). Our fixed-sample analyses confirmed that Lis1
-depletedcells (Figure 4A) displayed more bipolar spindles than control- RNAi cells over the whole range of STLC concentrations and that the rescue efficiency decreased with increasing inhibition of Eg5 (Figure 4B). Again, this trend could not be observed in time-lapse imaging. As shown in Figure 4C, at 1 /.lM STLC, already 93% of control-RNAi cells formed bipolar spindles and, therefore, Lis1 depletion could not significantly improve the situation, while the maximum rescue effect was consistently achieved at higher doses of Eg5 inhibitor. It was not surprising that the different outcome of fixed- and live-sample analyses was again due to changes in mitotic timing, as Lis1 depletion resulted in an even stronger mitotic delay of bipolar spindles than DHC depletion (Figure 4D, tnorma'). Similar to DHC depletion, Lis1 depletion also resulted in a reduction in
tdefectivefrom 9.51 hr to 7.66 hr due to accelerated death of cells with monopolar spindles. These results confirm that Lis1 depletion rescues spindle bipolarization. The pattern of rescue efficiency, however, strongly argues against the idea that the activities of dynein/Lis1 and Eg5 are titratable and is not compatible with a simple Eg5- dynein push-pull model.
DISCUSSION
The simple push
-pull model where theeffort of Eg5 to push spindle poles apart is continuously antagonized by dynein's inward acting force is intriguingly intuitive. The observation that the activities of Eg5 and dynein seemed to be titratable was an important cornerstone of the idea that Eg5 and dynein are direct antagonists acting both on antiparallel microtubules at the spindle equator. Clearly, our in-depth analyses reveal that dynein/Lis1 and Eg5 are not simply titratable against each other, as we did not observe a negative correlation between the efficiency of spindle bipolarization and the
level of Eg5inhibition in DHC/Lis1
-depleted cells (Figure 4E). From thesedata, we conclude that, while dynein antagonizes Eg5 function in bipolarization at the global cellular level, the molecular details of this antagonism are more complex than anticipated. This conclusion is supported by the fact that dynein localizes to multiple subcellular structures in mitosis where it fulfills diverse functions.
Itis involved in centrosome separation during prophase (Splinter et aI., 2010; Tanenbaum et aI., 2010);
connects astral microtubules to the cell cortex, possibly pulling spindle poles apart (Busson et aI., 1998; Gonczy et aI., 1999;
Grill and Hyman, 2005; Laan et aI., 2012); transports microtubule nucleation factors and spindle assembly checkpoint factors along kinetochore fibers from kinetochores to spindle poles (Ma et aI., 2010; Sivaram et aI., 2009; Chan et aI., 2009 and references therein); and is involved in pole focusing (Gaglio et aI., 1996; Shimamoto et aI., 2011). Taking these diverse func- tions into consideration, it is not surprising that depletion of a multifunctional protein such as dynein results in unexpected and highly complex patterns of spindle formation. In· addition to the study of Tanenbaum et al. (2008), the main evidence for a dynein-Eg5 antagonism in spindle bipolarization comes from Mitchison et al. (2005) and Ferenz et al. (2009). In frog extract, Mitchison et al. (2005) showed that addition of a dynein inhibitor enabled bipolarization
inEg5-inhibited extract. This is in full agreement with our observations, as this study did not address the titrability of Eg5 and dynein. Ferenz et al. (2009) used human cells but relied on a different approach from the one presented here. A dynein inhibitor was injected into Eg5-inhibited cells arrested with nocodazole, and spindle fate was scored depend- ing on the initial position of centrosomes after nocodazole releas· e. It was concluded that, in Eg5-inhibited cells, if the two spindle halves overlap, dynein exerts an inbound force, causing spindle collapse. Considering the very different experi- mental setups, it is hard to directly compare these results to our data. The main conclusion, however, is based on very
low cellnumbers (five dynein-inhibited cells), and most important, this publication observes an artificial spindle formation process after mitotic release from nocodazole, thus lacking a physiolog- ical prophase. Dynein and Eg5, however, are involved in centro- some separation in prophase (Splinter et aI., 2010; Tanenbaum et aI., 2010), and this fu.nction might explain the difference from our results, as we look at mitosis from prophase to telophase.
To definitively understand if dyne
inis
involvedin a push-pull mechanism, direct visualization in living cells will be required.
Until future studies overcome the current technical obstacles to do so, we will have to rely on indirect observations of spindle morphology after molecular manipulations to understand the role of dynein at the spindle equator. However, as shown by our mathematical model and proof-of-concept experiments, these studies have to be performed using
live-cellimaging, as fixed-cell analysi s results in wrong quantifications due to unpre- dictable effects of molecular manipulations on the lifetime of spindle phenotypes.
EXPERIMENTAL PROCEDURES
All experiments in this study were based on the following experimental protocol: RNAi transfection (25 nM final siRNA duplex concentration) was performed and followed by an incubation time of 30 hr (Sg01 RNAi) or 54 hr (all other experiments). Then, either cells were harvested and lysed for immunoblotting, or Eg5 inhibitor was added and time-lapse image acquisition was started immediately and continued for 18 hr. Finally, cells were fixed and stained for immunofluorescence analysis. For quantification of time-lapse data, only cells that entered mitosis at least 10 ·hr before the end of the experiment were included in the analysis to allow accurate determination of cell fate even for very long mitotic lifetimes. See Extended Experimental Procedures for a detailed description of reagents and protocols.
A
B
D
E
Eg5
Us1
n(cells)
.9
'" .B
1l '"
.7~~
6~ ~.
o.!9 .5
.g l.4
~:€ .3
!,JJ~ .2 .1
54 h, 25 nM GL2 RNAi Usl RNAi
1 ,...--... t130
kDa1 ..-
FokDa406312322526404 76B 516711461 593
C
n(cells) 243 235 301 269 214 237 27B 206 225 242o
dyneinl
<], <,
STLC IJJMJ
control RNAi, 21lM STLC
10
R/'I1OIaPlasei bi:x>lar .prophase!
monopolar
151(h)
phenotype fi)(ed!
push-pull
Lis1 Eg5 predictions actual
.9
<], <,
STLC IJJM]
Lis1 RNAi, 21lM STLC
o
10GlUs1 RNAi o control RNAi 54h
151(11)
Figure 4. Bipolar Spindle Formation in Eg5-lnhibited and Lis1-Depleted Cells Cannot Be Explained by a Push-Pull Mechanism (A) Immunoblot analyses for Eg5 and Lis1 of GL2-and Lis1-RNAi cells 54 hr after transfection of siRNAs.
(B and C) Quantification of Nf/x (B) and Nae•ua• (C) after transfection with Lis1 or control siRNA for 54 hr followed by addition of STLC and imaging for an additional 18 hr at the indicated doses. Bars represent means ± SEM of triplicate experiments.
(D) Quantification of times in mitosis of bipolar and monopolar spindles at 2 ~LM STLC from an experiment shown in (B). Mitotic lifetimes were plotted as in Figure 3~. Note the dramatic increase in
tno,ma'
caused by Lis1 depletion.(E) Summary of the inconsistencies between the predictions of the push-pull model and our findings using live· cell imaging.
LICENSING INFORMATION
This is an open-access article distributed under the terms of the Creative Commons Attribution-Noncommercial-No Derivative Works 3.0 Unported License (CC-BY-NC-ND; http://creativecommons.org/licenseslby-nc-nd/3.0/
legalcode).
ACKNOWLEDGMENTS
We thank Anna Brendel for essential help with data analyses, Allon Klein for generous help with the mathematical model, and Tim Mitchison for discus- sions and general support. We are indebted to Lucia Sironi and the Mayer lab for fruitful discussions and to Vasiliki Sarli for VS83. S.F. and T.U.M. were sup- ported by the EU-Research Training Network Fellowship ("Understanding the Dynamic of Cell Division," 512348) and by the CRC-969 "Chemical and Biological Principles of Cellular Proteostasis" of the Deutsche Forschungsge- meinschaft (DFG), respectively.
REFERENCES
Busson, S., Dujardin, D., Moreau, A, Dompierre, J., and De Mey, J.R (1998).
Dynein and dynactin are localized to astral microtubules and at cortical sites in mitotic epithelial cells. Curr. BioI. 8, 541-544.
Chan, y.w., Fava, L.L., Uldschmid, A., Schmitz, M.H., Gerlich, D.w., Nigg, E.A., and Santamaria, A. (2009). Mitotic control of kinetochore-associated dynein and spindle orientation by human Spindly. J. Cell BioI. 185,859-874.
Compton, D.A. (2000). Spindle assembly in animal cells. Annu. Rev. Biochem.
69,95-114.
DeBonis, S., Skoufias, D.A., Lebeau, L., Lopez, R, Robin, G., Margolis, R.L., Wade, RH., and Kozielski, F. (2004). In vitro screening for inhibitors of the human mitotic kinesin Eg5 with antimitotic and antitumor activities. Mol.
Cancer Ther. 3, 1079-1090.
Dumont, S., and Mitchison, T.J. (2009). Force and length in the mitotic spindle.
Curr. BioI. 19, R749-R761.
Ferenz, N.P., Paul, R., Fagerstrom, C., Mogilner, A., and Wadsworth, P. (2009). Dynein antagonizes eg5 by crosslinking and sliding antiparallel microtubules.
Curr. BioI. 19,1833-1838.
Ferenz, N.P., Gable, A., and Wadsworth, P. (2010). Mitotic functions of kinesin- 5. Semin. Cell Dev. BioI. 21, 255-259.
Gaglio, T., Saredi, A, Bin9ham, J.B., Hasbani, M.J., Gill, S.R, Schroer, TA, and Compton, D.A. (1996). Opposing motor activities are required forthe orga- nization of the mammalian mitotic spindle pole. J. Cell BioI. 135,399-414.
Gonczy, P., Pichler, S., Kirkham, M., and Hyman, AA (1999). Cytoplasmic dynein is required for distinct aspects of MTOC positioning, including centro- some separation, in the one cell stage Caenorhabditis elegans embryo. J. Cell BioI. 147, 135-150.
Grigoriev, I., Splinter, D., Keijzer, N., Wulf, P.S., Demmers, J., Ohtsuka, T., Modesti, M., Maly, I.V., Grosveld, F., Hoogenraad, C.C., and Akhmanova, A.
(2007). Rab6 regulates transport and targeting of exocytotic carriers. Dev.
Cell 13, 305-314.
Grill, S.w., and Hyman, AA (2005). Spindle positioning by cortical pulling forces. Dev. Cell 8, 461-465.
Heald, R, and Walczak, C.E. (2008). Mitotic spindle assembly mechanisms. In The Kinetochore: From Molecular Discoveries to Cancer Therapy, P. De Wulf and W.C. Earnshaw, eds. (New York: Springer), pp. 231-268.
Howell, B.J., McEwen, B.F., Canman, J.C., Hoffman, D.B., Farrar, E.M., Rieder, CL, and Salmon, E.D. (2001). Cytoplasmic dynein/dynactin drives kinetochore protein transport to the spindle poles and has a role in mitotic spindle checkpoint inactivation. J. Cell BioI. 155, 1159-1172.
Kapitein, L.C., Peterman, E.J.G., Kwok, B.H., Kim, J.H., Kapoor, T.M., and Schmidt, C.F. (2005). The bipolar mitotic kinesin Eg5 moves on both microtu- bules that it crosslinks. Nature 435, 114-118.
Kapoor, T.M., and Mitchison, T.J. (2001). Eg5 is static in bipolar spindles rela- tive to tubulin: evidence for a static spindle matrix. J. Cell BioI. 154, 1125-1133.
Kashina, A.S., Scholey, J.M., Leszyk, J.D., and Saxton, W.M. (1996). An essen- tial bipolar mitotic motor. Nature 384, 225.
Kashina, A.S., Rogers, G.C., and Scholey, J.M. (1997). The bimC family of kinesins: essential bipolar mitotic motors driving centrosome separation.
Biochim. Biophys. Acta 1357, 257-271.
Kwok, B.H., Yang, J.G., and Kapoor, T.M. (2004). The rate of bipolar spindle assembly depends on the microtubule-gliding velocity of the mitotic kinesin Eg5. Curr. BioI. 14,1783-1788.
Laan, L., Pavin, N., Husson, J., Romet-Lemonne, G., van Duijn, M., L6pez, M.P., Vale, R.D., JOlicher, F., Reck-Peterson, SL, and Dogterom, M. (2012).
Cortical,dynein controls microtubule dynamics to generate pulling forces that position microtubule asters. Cell 148,502-514.
Ma, N., Tulu, U.S., Ferenz, N.P., Fagerstrom, C., Wilde, A., and Wadsworth, P.
(2010). Pcileward transport of TPX2 in the mammalian mitotic spindle requires dynein, Eg5, and microtubule flux. Mol. BioI. Cell 21, 979-988.
Mcintosh, J.R, Hepler, P.K., and Wie, D.G.V. (1969). Model for mitosis. Nature 224, 659-663.
Mesngon, M.T., Tarricone, C., Hebbar, S., Guillotte, AM., Schmitt, E.W., Lanier, L., Musacchio, A., King, S.J., and Smith, D.S. (2006). Regulation of cytoplasmic dynein ATPase by Lis1. J. Neurosci. 26, 2132-2139.
Mitchison, T.J., Maddox, P., Gaetz, J., Groen, A., Shirasu, M., Desai, A, Salmon, E.D., and Kapoor, T.M. (2005). Roles of polymerization dynamics, opposed motors, and a tensile element in governing the length of Xenopus extract meiotic spindles. Mol. BioI. Cell 16, 3064-3076.
Salic, A, Waters, J.C., and Mitchison, T.J, (2004). Vertebrate shugoshin links sister centromere cohesion and kinetochore microtubule stability in mitosis.
Cell 118,567-578.
Sarli, V., Huemmer, S., Sunder-Plassmann, N., Mayer, T.U., and Giannis, A.
(2005). Synthesis and biological evaluation of novel EG5 inhibitors.
ChemBioChem 6, 2005-2013.
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.
Scholey, J.M., Brust-Mascher, I., and Mogilner, A. (2003). Cell division. Nature 422, 746-752.
Sharp, D.J., McDonald, K.L., Brown, H.M., Matthies, H.J., Walczak, C., Vale, R.D., Mitchison, T.J., and Scholey, J.M. (1999). The bipolar kinesin, KLP61 F, cross-links microtubules within interpolar microtubule bundles of Drosophila embryonic mitotic spindles. J. Cell BioI. 144, 125-138.
Shimamoto, Y., Maeda, Y.T., Ishiwata, S., Libchaber, A.J., and Kapoor, T.M.
(2011). Insights into the micromechanical properties of the metaphase spindle.
Cell 145, 1062-1074.
Sivaram, M.V., Wadzinski, TL, Redick, S.D., Manna, T., and Doxsey, S.J.
(2009). Dynein light intermediate chain 1 is required for progress through the spindle assembly checkpoint. EMBO J. 28, 902-914.
Skoufias, D.A., DeBonis, S., Saoudi, Y., Lebeau, L., Crevel, I., Cross, R., Wade, R.H., Hackney, D., and Kozielski, F. (2006). S-trityl-L-cysteine is a reversible, tight binding inhibitor of the human kinesin Eg5 that specifically blocks mitotic progression. J. BioI. Chem. 281,17559-17569.
Open
ACCESS
Splinter, D., Tanenbaum, M.E., Lindqvist, A., Jaarsma, D., Flotho, A., Vu, K.L., Grigoriev, I., Engelsma, D., Haasdijk, E.D., Keijzer, N., et al. (2010). Bicaudal 02, dynein, and kinesin-1 associate with nuclear pore complexes and regulate centrosome and nuclear positioning during mitotic entry. PLoS BioI. 8, e1000350.
Tanenbaum, M.E., Macurek, l., Galjart, N., and Medema, R.H. (2008). Dynein, Lis1 and CLlP-170 counteract Eg5-dependent centrosome separation during bipolar spindle assembly. EMBO J. 27, 3235-3245.
Tanenbaum, M.E., Akhmanova, A., and Medema, R.H. (2010). Dynein at the nuclear envelope. EMBO Rep. 11, 649.
van den Wildenberg, S.M., Tao, l., Kapitein, l.C., Schmidt, C.F., Scholey, J.M., and Peterman, E.J. (2008). The homotetrameric kinesin-5 KLP61 F preferentially crosslinks microtubules into antiparallel orientations. Curr. BioI. 18, 1860-1864.
Varetti, G., and Musacchio, A. (2008). The spindle assembly checkpoint. Curr.
BioI. 18, R591-R595.