Self-assembly and sorting of acentrosomalmicrotubules by TACC3 facilitate kinetochore captureduring the mitotic spindle assemblyWenxiang Fu, Hao Chen1, Gang Wang1, Jia Luo1, Zhaoxuan Deng, Guangwei Xin, Nan Xu, Xiao Guo, Jun Lei, Qing Jiang,and Chuanmao Zhang2

Ministry of Education Key Laboratory of Cell Proliferation and Differentiation and State Key Laboratory of Biomembrane and Membrane Biotechnology,College of Life Sciences, Peking University, Beijing 100871, China

Edited by J. Richard McIntosh, University of Colorado at Boulder, Boulder, CO, and approved August 9, 2013 (received for review July 4, 2013)

Kinetochore capture by dynamic kinetochore microtubule fibers (Kfibers) is essential for proper chromosome alignment and accuratedistribution of the replicated genome during cell division. Althoughthis capture process has been extensively studied, the mechanismsunderlying the initiation of this process and the proper formationof the K fibers remain largely unknown. Here we show that trans-forming acidic coiled-coil–containing protein 3 (TACC3) is essen-tial for kinetochore capture and proper K-fiber formation in HeLacells. To observe the assembly of acentrosomal microtubules moreclearly, the cells were released from higher concentrations of noco-dazole into zero or lower concentrations. We find that small acen-trosomal TACC3–microtubule aster formation near the kinetochoresand binding of the asters with the kinetochores are the initial stepsof the kinetochore capture by the acentrosomal microtubules, andthat the sorting of kinetochore-captured acentrosomal microtu-bules with centrosomal microtubules leads to the capture of kinet-ochore by centrosomal microtubules from both spindle poles. Wedemonstrate that the sorting of the TACC3-associated microtubuleswith the centrosomal microtubules is a crucial process for spindleassembly and chromosome movement. These findings, which arealso supported in the unperturbed mitosis without nocodazole, re-veal a critical TACC3-dependent acentrosomal microtubule nucleationand sorting process to regulate kinetochore–microtubule connectionsand provide deep insight into the mechanisms of mitotic spindleassembly and chromosome alignment.

centrosome | noncentrosomal | cell cycle

To ensure proper segregation of the chromosomes into its twodaughter cells during proliferation, the chromosomes of a

mother cell must be captured by its assembling mitotic spindlethrough attachment of the chromosome kinetochores and thedynamic spindle microtubules (1). A “search-and-capture” modelwas proposed long ago, in which the dynamic spindle micro-tubules nucleated from the centrosomes search for and capturethe chromosome kinetochores (2). Previous studies showed thatthe kinetochores are initially captured by the spindle-pole–nucleated microtubules with their lateral side (3, 4). Once captured,the kinetochores with their chromosomes are transported alongthe microtubules toward a spindle pole, and the microtubulesshrink at their plus ends until the establishment of the end-onattachment (4, 5). However, this model is insufficient to explainthe initial connection of the kinetochore and the spindle micro-tubules in the centrosome-independent spindle assembly process.Recent studies in Xenopus extracts indicated that microtubulesare nucleated near the chromosomes and self-organize into aspindle (6). A new model for acentrosomal spindle assembly hasbeen raised in mouse oocytes, in which self-organized microtubuleorganizing centers (MTOCs) replace the centrosome function (7).The somatic cells may also use the centrosome-independentpathway for their spindle assembly (8–10). In Drosophila cells,the centrosome-independent assembled kinetochore fibers canbe captured by centrosomal microtubules (11–13).

Previous studies have shown that transforming acidic coiled-coil–containing protein 3 (TACC3) is essential for the mitoticspindle assembly and chromosome alignment, but the mecha-nism remains largely unknown (14–18). Here we reveal thatTACC3-dependent small microtubule aster formation and sort-ing near the kinetochores contribute to correct microtubule–kinetochore connections.

Results and DiscussionTACC3 Regulates de Novo Assembly of Acentrosomal Microtubules.Several groups reported recently that TACC3 is essential forchromosome alignment and spindle assembly in mitosis, but theclear mechanism remains unknown (14–18). To test how TACC3regulates the spindle formation, we firstly knocked down TACC3in HeLa cells using small interfering RNA against TACC3(siTACC3) (>95% efficiency) (Fig. 1A). Then we treatedTACC3-knockdown and the irrelevant-knockdown control cellswith 1 μg/mL nocodazole for 5–8 h to abolish microtubule nu-cleation and mitotic spindle assembly, followed by releasingthese cells into medium without nocodazole to allow them toreassemble their microtubules and spindles. As shown (Fig. 1B),in control cells, microtubules were quickly nucleated from boththe separating centrosomes to form two big centrosomal asters;meanwhile, other microtubules were also quickly (within 1.5min) nucleated in the cytoplasm to form many small acen-trosomal asters. Then, these small asters quickly “fused” witheach other and sorted into the big centrosomal asters, and finally

Significance

Mitosis is a highly regulated cell division process in eukaryotes.Assembly of the spindle and segregation of chromosomes inmitosis enable the mother cells to distribute the geneticmaterials equally to their daughter cells. Before chromosomesegregation, the kinetochores at the chromosomes must becorrectly captured by the microtubules. The mechanisms un-derlying the initiation of this process and the proper formationof the kinetochore fibers remain largely unknown. This studyshows that transforming acidic coiled-coil–containing protein 3(TACC3) is essential for proper kinetochore capture and kinet-ochore fiber formation. Our findings reveal a critical TACC3-dependent acentrosomal microtubule nucleation and sortingprocess to regulate kinetochore–microtubule connections.

Author contributions: W.F., Q.J., and C.Z. designed research; W.F., H.C., G.W., J. Luo, Z.D.,G.X., N.X., X.G., and J. Lei performed research; C.Z. contributed new reagents/analytictools; W.F., Q.J., and C.Z. analyzed data; and W.F., Q.J., and C.Z. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

Freely available online through the PNAS open access option.1H.C., G.W., and J. Luo contributed equally to this work.2To whom correspondence should be addressed. E-mail: zhangcm@pku.edu.cn.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1312382110/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1312382110 PNAS | September 17, 2013 | vol. 110 | no. 38 | 15295–15300

CELL

BIOLO

GY

Dow

nloa

ded

by g

uest

on

July

21,

202

1

these big and small microtubule asters assembled into a bi-polar mitotic spindle within 15 min (3.5–15 min). In contrast,in TACC3-knockdown cells, although microtubules were nucle-ated around the separating centrosomes to form the two bigcentrosomal microtubule asters and finally assemble an imma-ture bipolar mitotic spindle, there was nearly no other microtubulenucleation to form the small acentrosomal microtubule asters inthe cytoplasm (Fig. 1 B–D and Fig. S1 A and B). These resultssuggested that TACC3 might contribute to the acentrosomalmicrotubule aster assembly and the microtubule–kinetochoreconnection. To further confirm these, we released these cells intothe medium containing 15 ng/mL nocodazole and also verifiedthat TACC3 is required for acentrosomal microtubule assembly(Fig. 1 C and D and Fig. S1 C and D).To understand why TACC3 is required for the acentrosomal

microtubule aster assembly, we treated the mitotic HeLa cellswith nocodazole or cold temperatures and observed thatTACC3 stayed with the stable kinetochore-connected fibers(K fibers) (Fig. 1E and Fig. S1E). We treated HeLa cells withnocodazole (50 ng/mL) for 16 h to allow the cells to enterinto mitosis without centrosomal microtubule nucleation andverified that TACC3 connected to kinetochore markers, i.e.,Hec1 and Aurora B (Fig. 1F). Meanwhile, a fraction of cla-thrin heavy chain also colocalized with TACC3 on their

fibers, supporting the notion that clathrin and TACC3 co-function as a complex in mitosis (Fig. 1F) (15–18). Moreover,knockdown of Aurora A, which functions upstream ofTACC3 (19), clearly abolished the acentrosomal microtubuleaster formation (Fig. 1G).To further examine the roles of TACC3 in regulating acen-

trosomal microtubule assembly, we performed live-imagingassays in HeLa cells expressing GFP–tubulin to analyze the dy-namic assembly of spindle microtubules. The cells were releasedfrom 500 ng/mL nocodazole into medium with 0 (Fig. 1H andMovies S1–S6) or 15 ng/mL nocodazole (Fig. 1I and MoviesS7–S12). As shown (Fig. 1H and Movies S1 and S2), in controlcells, microtubules were assembled both around the centrosomesand in the acentrosomal regions. In contrast, TACC3 knockdownhad little effect on centrosomal microtubule nucleation whilestrongly inhibiting the formation of acentrosomal microtubules(Fig. 1H and Movies S3 and S4). Moreover, we treated the cellswith MLN8237, a small-molecule inhibitor of Aurora A, andfound it suppressed both the centrosomal and acentrosomalmicrotubule assembly (Fig. 1H and Movies S5 and S6). Tospecifically analyze the formation of acentrosomal micro-tubules, the cells were released into 15 ng/mL nocodazole.Similar as what we observed in fixed samples (Fig. 1 C and G),the live tracking of microtubules also revealed that TACC3

Control

siTACC3-1

siTACC3-2

Num

ber o

f a

cent

roso

mal

m

icro

tubu

le fo

ci

048

121620

No Noc15 ng/ml Noc

chromosome kinetochore fiberunattached acentrosomal MT

kinetochoreTACC3

Tubulin MergeDNA

Si-C

onsi

-TA

CC

3si

-Aur

A

Hec1TACC3 MergeC

ontro

lN

oc(8

min

)C

old

(10

min

)

MLN

8237

02:00 04:20 08:20 22:2015:20 25:40 30:40 35:20 42:00 51:20

02:00 04:00 09:20 12:40 15:40 23:00 28:20 34:00 40:20 44:00

si-T

AC

C3

02:15 08:45 11:15 15:30 17:15 22:00 29:30 49:1504:15 30:45

01:30 03:15 05:30 11:00 15:45 24:30 34:45 45:15 49:4530:00si-C

o

02:30 04:30 11:30 19:00 24:00 29:00 33:00 36:30 40:3016:30

02:30 05:30 09:30 13:30 18:00 22:30 26:30 30:00 35:00 40:00

Noc

rele

ase:

500

ng/

ml

to 0

ng/

ml

02:40 04:20 05:40 09:00 12:20 21:00 23:00 27:00 30:0015:40

01:40 04:00 06:20 08:40 11:20 14:40 20:40 22:20 24:40 29:00

02:20 04:00 07:20 12:40 20:00 24:20 29:00 34:20 38:4016:40

02:40 04:00 06:20 09:20 12:40 17:20 22:00 26:00 30:00 37:40

MLN

8237

si-T

AC

C3

si-C

on02:20 05:20 08:40 11:00 14:40 22:20 26:40 31:20 39:4018:00

03:00 04:20 08:00 10:40 15:00 24:00 27:20 33:40 39:2019:20 Noc

rele

ase:

500

ng/

ml t

o 15

ng/

ml

D

siTA

CC

3-1

siTA

CC

3-2

Con

trol

TAC

C3

α-tu

b

Si-C

onsi

TAC

C3

Noc

: 1 μ

g/m

l to

0

Tubulin Merge Tubulin Merge Tubulin Merge Tubulin Merge Tubulin Merge1.5 min 3.5 min 7.5 min 15 min0.5 min

si-C

onsi

TAC

C3

Noc

: 1 μ

g/m

l to

15

ng/m

l

1.5 min 3.5 min 7.5 min 15 min0.5 minTubulin Merge Tubulin Merge Tubulin Merge Tubulin Merge Tubulin Merge

A B C

E F TACC3 Merge

Cla

thrin

Merge (Tubulin/TACC3/DNA) Merge (Tubulin/TACC3/DNA)

Hec

1A

urB

G

H I

JDNA

Fig. 1. TACC3 is required for de novo assembly of acentrosomal microtubules in mitosis. (A) Detection of TACC3 RNAi depletion efficiency by Westernblotting. HeLa cells were transfected with control and TACC3 siRNAs, respectively. The blots were probed with anti-TACC3 (Upper) and anti–α-tubulin (Lower).(B and C) Representative images of 1 μg/mL nocodazole-arrested normal control or TACC3-knockdown HeLa cells by siRNA followed by release into mediumwithout nocodazole (B) or with 15 ng/mL nocodazole (C) at different time points (0.5, 1.5, 3.5, 7.5, and 15 min). The data are shown as maximum intensityprojections of different z sections. TACC3 is in red, tubulin in green, and DNA in blue. (D) Statistics of numbers of acentrosomal microtubule seeds in controland TACC3 knockdown cells after nocodazole treatment and release for 1.5 min. More than 50 cells for each treatment were counted. (E) Control, noco-dazole-treated (50 ng/mL for 8 min), and cold-treated (10 min on ice) HeLa cells were stained with anti-TACC3 (green) and anti-Hec1 (red) antibodies. (F)Staining of HeLa cells with TACC3 (green) and indicated proteins (Hec1, Aurora B, and clathrin) (red) after nocodazole treatment for 16 h. (G) HeLa cells weretransfected with different siRNAs (control, siTACC3, siAurora A) as indicated. Cells were treated with 50 ng/mL nocodazole for 16 h before fixation. (H and I)Live imaging of HeLa cells expressing GFP–tubulin under indicated different treatments (siRNA control, siTACC3, and Aurora A inhibitor MLN8237). Themitotic cells were arrested with 500 ng/mL nocodazole and released into medium without nocodazole (H) or with 15 ng/mL nocodazole (I). (J) Illustration ofTACC3-dependent acentrosomal microtubule assembly and clustering. During mitosis, TACC3 containing small microtubule seeds/asters were assembledaround the chromosomes and kinetochores, and these small microtubule asters were further assembled and clustered into a complete bipolar spindlestructure. (Scale bar, 10 μm.)

15296 | www.pnas.org/cgi/doi/10.1073/pnas.1312382110 Fu et al.

Dow

nloa

ded

by g

uest

on

July

21,

202

1

and Aurora A is crucial for acentrosomal microtubule for-mation (Fig. 1I and Movies S7–S12). Together, we propose thatTACC3-dependent acentrosomal microtubule nucleation isregulated by Aurora A, and TACC3-containing acentrosomalmicrotubule small aster assembly contributes to microtubule–kinetochore connections (Fig. 1J).

TACC3-Dependent Acentrosomal Microtubule Nucleation Is Required forProper Kinetochore–Microtubule Capture and Correct Mitosis. Becausethe previous conclusions were mainly based on nocodazole-treatedcells, we next investigated the function of TACC3-dependentacentrosomal microtubule nucleation in kinetochore–microtubulecapture during unperturbed mitosis in HeLa cells without noco-dazole (Fig. 2). In normal metaphase cells, proper end-on attach-ments between kinetochores and microtubules were formed,whereas in TACC3-knockdown cells, many kinetochores werenot connected by microtubules or only bound laterally with few

microtubules (Fig. 2 A and B and Fig. S2 A and D). As shown(Fig. 2 C and D and Fig. S2B), TACC3 knockdown resulted inan increase of lagging chromosomes and metaphase plate width.Meanwhile, ablation of TACC3 led to abundant centrosome-nucleated astral microtubules outside the “spindle frame” as in-dicated by arrows, suggesting deficiency of the search-and-captureprocess (Fig. 2A). The microtubule intensity around the chromo-somes in TACC3-depleted cells was much less than control (Fig.2E and Fig. S2C), suggesting insufficient assembly of acentrosomalmicrotubules. Furthermore, we found that TACC3 depletionresulted in marked decrease of distances between paired kinet-ochores (Fig. 2 F and G), indicating that TACC3 is requiredfor maintenance of proper interkinetochore tension. More-over, TACC3 knockdown led to increased BubR1 levels at thekinetochores, suggesting the activation of spindle assembly check-point (Fig. 2 H and I). Because TACC3 localizes to acentrosomalmicrotubules and is not essential for centrosomal microtubulenucleation as discussed above (Fig. 1), we proposed that TACC3-dependent acentrosomal microtubule nucleation and small asterformation mainly contribute to the regulation of kinetochorefunction, including kinetochore–microtubule attachment andmaintenance of interkinetochore tension.TACC3-dependent acentrosomal microtubule nucleation must

be regulated by other mitotic regulators in performing its func-tion, such as clathrin and Aurora A (20). So, here we checked theroles of other mitotic regulators in HeLa cells. Knockdown ofTPX2, which activates Aurora A (21), reduced TACC3 targetingto the spindle, whereas inhibition of Aurora B by small moleculeZM447439 (22) or ablation of centromere protein E had no orlittle effect on the localization of TACC3 to the spindle (Fig.S3A). Moreover, unlike eg5 inhibition by monastral (23), in-hibition of Plk1 by small molecule BI2536 (24) and knockdownof Nedd1 (25) both inhibited the targeting of TACC3 to thespindles (Fig. S3B). Overexpression of HSET, which mainlyfunctions on centrosomal microtubules (26), had no effect on thespindle localization of TACC3 (Fig. S3C). We also knockeddown Hec1 and found the location of TACC3 to the spindles waspartially reduced (Fig. S3D). Taken together, these results in-dicate that inhibiting or knocking down upstream regulators forTACC3 (i.e., TPX2) or interfering with the kinetochore micro-tubule assembly activity (i.e., Plk1, Nedd1, and Hec1) can down-regulate the targeting of TACC3 to the spindles and lead to thedefects in bipolar spindle assembly and chromosome alignmentdue to kinetochore–microtubule attachment failure.

The Sorting of the Small Acentrosomal Microtubule Seeds/Asters intoBig Centrosomal Asters Is a General Mechanism for Spindle Assemblyand Kinetochore Movement. To better illustrate the roles forTACC3 in chromosome alignment and mitotic spindle assembly,we analyzed the microtubule behaviors by live imaging of HeLacells expressing GFP–α-tubulin. We found that the sorting of thesmall acentrosomal asters into the big centrosomal asters is onekey process for the mitotic spindle assembly (Fig. 3 A and B,Figs. S4 and S5, and Movies S13 and S14). During mitosis, boththe centrosomal microtubules and the small acentrosomal astermicrotubules kept growing in the cytoplasm, and when they en-countered, these acentrosomal microtubules were immediatelysorted into the microtubule arrays of the centrosomal asters. Fiveexamples of microtubule sorting from the two movies mentionedabove (Movies S13 and S14) are shown in magnified images(Fig. 3C, boxes a–e). As indicated, during mitosis, acentrosomalmicrotubules can be captured by and sorted into microtubulesfrom one centrosome (Fig. 3C, boxes a, d, and e) or both of thetwo centrosomes (Fig. 3C, boxes b and c). The sorting of acen-trosomal microtubules into centrosomal asters was also con-firmed in other cases (Fig. 3 D and E, Fig. S6 A–C, and MoviesS15 and S16). Together, these data indicate that the microtubulesorting of the small acentrosomal asters with the big centrosomalasters is a general mechanism for spindle assembly.To further characterize the acentrosomal microtubule sorting

process, we analyzed the microtubule behaviors when the HeLa

Inte

r-k

dist

ance

(μm

)

0.0

0.5

1.0

1.5

2.0

Bub

R1

inte

nsity

(a.u

.)

01234567

TubulinCRESTTACC3 DNA Merge

Si-C

onsi

TAC

C3

Si-C

onsi

TAC

C3

BubR1CRESTTACC3 DNA Merge

Si-C

onsi

TAC

C3

A B

C D E

F

0

4

8

12

16

Met

apha

se p

late

wid

th (μ

m)

G

0.00.10.20.30.40.5

si-Control

siTACC3-1

siTACC3-2

0

20

4060

80

100

si-Control siTACC3-1 siTACC3-2

End-on attachment

UnattachedLateral binding

0

20406080

Rel

ativ

e M

T in

tens

ity

arou

nd c

hrom

osom

es

Per

cent

age

of m

itotic

cel

ls w

ith la

ggin

g ch

rom

osom

es (%

)

H

I

si-Control

siTACC3-1

siTACC3-2

si-Control

siTACC3-1

siTACC3-2

si-Control

siTACC3-1

siTACC3-2

si-Control

siTACC3-1

siTACC3-2

Per

cent

age

of

kine

toch

ore

atta

chem

ent t

ype

(%)

BubR1/CRESTHec1 CREST TACC3 DNA Merge

Hec1/CREST

100

Tubulin/CREST

Fig. 2. TACC3 facilitates kinetochore–microtubule capture, interkinetochoretension, and spindle assembly checkpoint. Micrographs are presented asmaximum intensity projections of different z sections. (A) Ablation of TACC3led to defects in chromosome alignment and kinetochore–microtubule at-tachment. (Right) Representative paired kinetochores in single z sections. Theoval lines indicate the spindle frame and the arrows indicate the microtubulesoutside the oval frame. Kinetochores stained with serum from patients withCREST syndrome are in red (CREST), tubulin in green, TACC3 in magenta, andDNA in blue. (B) Quantification of kinetochore attachment types in differentgroups. Percentages of end-on attachment, lateral binding status, andunattached kinetochores are shown. (C) Quantification of the percentages ofmetaphase cells with lagging chromosomes. (D) Measurement of the meta-phase plate width in control and TACC3-depleted cells. (E) Quantification ofthe relative microtubule intensity around chromosomes. (F) HeLa cells werestained with anti-Hec1 (green), CREST (red), and TACC3 (magenta) antibodies.Representative images are shown. (Right) Micrographs in single z sections. (G)Statistics of interkinetochore distances in control and TACC3-ablated cells.The interkinetochore distances were measured according to paired Hec1 andCREST fluorescent signals. The distances between the centers of two pairedHec1 dots were measured with ImageJ. More than 150 kinetochore pairs inthe same z sections in each group were analyzed. Data are presented asmeans plus SEs. (H) Cells were stained with anti-BubR1 (green), CREST (red),and TACC3 (magenta) antibodies. DNA is in blue. (I) Statistics of BubR1fluorescence intensities in control and TACC3-ablated cells. The BubR1 in-tensities were quantified by ImageJ software, and the average cytoplasmicimmunofluorescence intensity was subtracted as background. Data are pre-sented as means plus SEs. (Scale bar, 10 μm.)

Fu et al. PNAS | September 17, 2013 | vol. 110 | no. 38 | 15297

CELL

BIOLO

GY

Dow

nloa

ded

by g

uest

on

July

21,

202

1

cells were released from 500 ng/mL nocodazole to mediumwithout nocodazole (Fig. 1H and Movies S1 and S2) or with 15ng/mL nocodazole (Fig. 1I and Movies S7 and S8). It can beeasily seen in the movies that multiple large microtubule astersand spindle poles were assembled and fused with each other(also see Fig. S6D showing sorting of two spindle poles, derivedfrom Movie S7). Fig. 3 F and G are two cases showing theassembly and sorting of individual acentrosomal foci whennocodazole was removed to observe the de novo assembly ofmicrotubules. With 25 ng/mL nocodazole, the sorting processwas also observed and reorganizations of spindle poles are shown(Fig. S6E and Movie S17). These data further demonstrate thatacentrosomal microtubule sorting is a basic process for con-structing the mitotic spindle and spindle poles during mitosis.Furthermore, by using live cell imaging, we analyzed the

behaviors of acentrosomal microtubules and kinetochores whenthe cells were released from 500 ng/mL nocodazole into 0 ng/mL.As shown (Fig. 3H and Movie S18), acentrosomal micro-tubules were preferentially assembled around and attached to thekinetochores, and sorting of these microtubule structures fur-ther resulted in kinetochore capture and movement. Similarly,when the cells were released into 15 ng/mL nocodazole, theacentrosomal microtubule assembly and sorting process was ac-companied by chromosome kinetochore movement (Fig. 3I andMovie S19). Considering that TACC3 is initially associated with

acentrosomal microtubules and kinetochores (Fig. S7 A–D), wepropose that the TACC3-dependent acentrosomal microtubuleassembly and sorting process facilitate kinetochore movementand capture.A cell normally assembles its mitotic spindle into a bipolar

structure, which is regulated by multiple and complex mechani-cal strengths (27). A mathematical model has been proposed foranastral spindle assembly in Drosophila oocytes (28). To illus-trate the role of the microtubule sorting in establishing thespindle bipolarity, here we hypothesized a simplified mathematicmodel based on our abovementioned results (Fig. S8). Forsorting of microtubules or asters, parallel microtubule motorsmainly function to bundle these microtubules into paralleledbundles and clusters, whereas antiparallel microtubule motorsusually make these microtubules slide and separate (Fig. S8 Aand B). To simply address this, we hypothesized that there existsa minus-end–directed pulling force (F) for generating microtu-bule sorting power (Fig. S8C). For sorting of two microtubulestructures (m, n): If 0° ≤θ <90° (where θ indicates the anglebetween Fm and Fn), fsorting (m,n) function results in the formationof one clustered force: F(m,n) (new) = (Fm + Fn) (jFmj+jFnj)/jFm +Fnj. If 90° ≤θ ≤180°, fsorting (m,n) results in formation of two an-tiparallel forces: Fm (new) = (Fm − Fn)jFmj/jFm − Fnj; Fn (new) =(Fn − Fm)jFnj/jFm − Fnj. For sorting of all of the microtubulestructures: ftotal sorting =

Pðm;nÞfsortingðm;nÞ.

During spindle assembly without centrosomes, the bipolarity isformed through sorting of multiple acentrosomal microtubules(Fig. S8D) (7, 29). In spindle assembly with centrosomes,centrosomes function as the main MTOCs with high sortingforces (Fa and Fb) to generate the bipolarity. When centrosomesare artificially removed in some situations, spindle bipolaritycan still be established through acentrosomal MTOCs (8, 27,30). In nocodazole-treated cells, due to the reduced sortingforce, some acentrosomal forces cannot be sorted, resulting inextra acentrosomal MTOCs, i.e., formation of multipolar mi-crotubule structures (Fig. S8 C and D).

Loading of Preformed Acentrosomal TACC3–Microtubule Seeds/Asters to Kinetochores Followed by Sorting Are Crucial Steps forKinetochore Capture. As indicated above, TACC3-dependentacentrosomal microtubule assembly and sorting facilitate properkinetochore capture. However, how TACC3 regulates microtu-bule and kinetochore behaviors remains unknown. To addressthis, we analyzed the behaviors of TACC3, microtubules, andkinetochores in mitotic HeLa cells in the presence of differentconcentrations (0–1 μg/mL) of nocodazole. With the decrease ofnocodazole concentration, the microtubule nucleation activityand kinetochore capture by the nucleated microtubules wereremarkably increased (Fig. 4A and Fig. S9 A and B). Datashowed that microtubule plus-end tracking protein EB1 spreadthrough the kinetochore-targeted TACC3-containing microtu-bule asters to spindle poles in the presence of nocodazole (Fig.S9C). Next, we confirmed TACC3 dynamically associated withmicrotubules and kinetochores in various treatments usingnocodazole (Fig. S10 A–G). To understand how these TACC3–microtubule seeds/asters and kinetochores get connected, wedetected the behaviors of TACC3–microtubule complexes dur-ing early mitosis in the presence of 100 ng/mL nocodazole (Fig.4B). Before nuclear envelope breakdown, TACC3 and tubulinstayed outside the nucleus and almost no TACC3 or tubulinlocated on the kinetochores. With nuclear envelope break-down in early mitosis, partial TACC3 was loaded onto the shortmicrotubules to form TACC3–microtubule seeds/asters, and theseeds/asters quickly connected with the kinetochores (Fig. 4A–C). We also analyzed the TACC3–microtubule behaviors inHeLa cells with 25 ng/mL nocodazole. There were at least fourkinds of microtubule structures in early mitosis, the microtubuleswithout TACC3, the dissociative TACC3–microtubules, the lat-erally kinetochore-bound TACC–microtubules, and the kineto-chore-attached TACC3–microtubule asters (Fig. 4D). In contrast,most TACC3–microtubule structures attached to kinetochores

00:50 05:00 06:20 07:20 09:10 10:40 12:40 16:0008:0000:00

A

B

C

ab c

de

00:20 01:20 01:50 02:30 03:50 05:47 07:17 08:4103:1000:00

a b c d e D E F 0:451:001:151:301:452:002:152:30

(m:s)

03:43

05:13

02:1302:3302:5303:13

03:5304:1304:3304:53

(m:s) g

03:20

04:10

02:4002:5003:0003:10

03:3003:4003:5004:00

(m:s) f

4:455:005:155:305:456:006:15

4:30

6:306:45

G(m:s)

H

Mer

geH

ec1

Tubu

linM

agni

fyM

erge

Hec

1Tu

bulin

Mag

nify

Noc

: 500

ng/

ml t

o 0

ng/m

lN

oc: 5

00 n

g/m

l to

15 n

g/m

l

I

01:30 03:00 07:30 12:00 14:15 18:00 24:45 40:30 50:1534:30

37:3013:0006:45 18:00 24:00 49:3008:15 29:1500:45 11:15

Fig. 3. Sorting of TACC3-associated acentrosomal microtubules is a generalmechanism for spindle assembly and kinetochore movement. (A–E) Liveimaging of HeLa cells expressing GFP–tubulin. Red arrows indicate cen-trosomal asters and green arrows are acentrosomal microtubules. (A and B)Sorting of acentrosomal microtubules into centrosomal asters. (C) Magnifiedmicrotubule sorting regions (boxes a–e) as shown in A and B. (D and E)Another two examples of microtubule sorting existed in normal mitosis. (Fand G) Assembly and sorting of multiple acentrosomal microtubule seedswhen HeLa cells expressing GFP–tubulin were released from 500 ng/mLnocodazole into medium without nocodazole. (H and I) Live imaging ofHeLa cells coexpressing mcherry–tubulin and GFP–Hec1. The cells were re-leased from 500 ng/mL nocodazole into 0 (H) or 15 ng/mL (I) nocodazole asindicated. (Scale bars in A, B, H, and I, 10 μm and in C–G, 2 μm.)

15298 | www.pnas.org/cgi/doi/10.1073/pnas.1312382110 Fu et al.

Dow

nloa

ded

by g

uest

on

July

21,

202

1

in late mitosis (Fig. 4E), suggesting that the TACC3–microtubuleseeds/asters had been sorted into the kinetochore-attached Kfibers. Thus, we propose that the capture of the kinetochores byspindle microtubules is facilitated through stepwise assembly ofTACC3–microtubule seeds/asters near the kinetochores (Fig.4F) and sorting of the TACC3–microtubule seeds/asters into thebig centrosomal asters during the mitotic spindle assembly.As indicated previously (Fig. 3 H and I and Movies S18 and

S19), the de novo assembly and sorting of microtubules in thecytoplasm facilitate kinetochore association. To further in-vestigate the initial kinetochore capture process, we analyzedthe dynamic behaviors of kinetochores and microtubules duringthe progression from late G2 to mitosis in the presence of 15 ng/mL(Fig. 5A, Fig. S11A, and Movie S20) and 50 ng/mL nocodazole(Fig. 5B, Fig. S11B, and Movie S21) in HeLa cells. In late G2or early mitosis (e.g., before 00:30:00 in Fig. 5A), there were afew polymerized microtubules that were not attached with kinet-ochores. With the time increase, the microtubule seeds/astersbegan to assemble and sort around the kinetochores. In agreementwith what we saw in fixed samples (Fig. 4C), these data in-dicated that loading and self-assembly of microtubules at the

kinetochores facilitate the progression of mitosis and spindleformation.In summary, mainly based on the results observed in noco-

dazole-treated cells, our present work leads us to hypothesizea unique model to illustrate the mechanism of the kinetochorecapture by microtubules (Fig. 5 C and D). First, once the cellenters mitosis and the nuclear envelope breaks down, TACC3and the short microtubules bind together to form small TACC3–microtubule seeds near the kinetochores along with the assemblyof two big centrosomal asters, and the seeds grow gradually intosmall TACC3–microtubule asters through the nucleation of theshort microtubules bound on the seeds. Second, the smallTACC3–microtubule seeds/asters get connected with differentparts of the kinetochores; and meanwhile, the TACC3–micro-tubule seeds/asters keep growing. Third, continuous nucleationand sorting of the TACC3–microtubule seeds/asters enable the

100

ng/m

l Noc

Tubulin/TACC3/ACAEarly mitosis

Late mitosis

100

ng/m

l Noc

TACC3CRESTTubulin MergeB

100

100

050

10Noc

odaz

ole

treat

men

t (ng

/ml)

25

A TACC3CRESTTubulin MergeDNA

DNA

C

Tubulin TACC3 CREST DNA Mergea

bc

d

a

b

c

d

e

e

(-)(+)(-)

(+)

(-)(+)

MT kinetochore TACC3

(+)(-)

D

E FTubulin TACC3 CREST DNA Merge

Late

mito

sis

Ear

ly m

itosi

s

0

Fig. 4. Targeting of preassembled acentrosomal TACC3–microtubule seeds/asters to kinetochore facilitates kinetochore capture by centrosomal micro-tubules. (A) TACC3–microtubule complex dynamically associated withkinetochores upon nocodazole treatment. Staining of mitotic HeLa cellsarrested by different concentrations (1 μg/mL, 100 ng/mL, 50 ng/mL, 25 ng/mL, 10 ng/mL, and 0 ng/mL) of nocodazole. Tubulin is in magenta, CREST inred, TACC3 in green, and DNA in blue. (B) The total of 100 ng/mL nocoda-zole-arrested HeLa cells in early mitosis were stained with tubulin (magenta),CREST (red), TACC3 (green), and DNA (blue). The white dashed lines indicatethe cell boundary. (C) Illustration of acentrosomal TACC3–microtubule (MT)seeds/asters in 100 ng/mL nocodazole-arrested early and late mitotic HeLacells. (D and E) Different types of acentrosomal microtubules assemble themitotic spindle. HeLa cells were treated with 25 ng/mL nocodazole for 5 h.The representative early (D) and late (E) mitotic cells are shown. The maxi-mum intensity projections of 3D images are shown. In early mitosis (D), fourdifferent types of acentrosomal microtubule structures (a–d) are indicatedby projection of selected z stacks followed by magnification. In contrast, onlyone type of acentrosomal microtubules (e in E) was in late mitosis. Thesquare e in E is magnified and illustrated (Below). (F) Illustration of theacentrosomal microtubule nucleation, TACC3–microtubule seeds/asters as-sembly and the end-on capture of the kinetochores by the small aster tubulesduring transition from early mitosis to late mitosis. (Scale bar, 10 μm.)

Mer

geTu

bulin

Hec

1

00:08:00 00:40:00 00:47:00 01:00:00 01:15:0000:25:00 00:33:00 01:32:00A Microtubule assembly in the presence of 15 ng/ml nocodazole

Microtubule assembly in the presence of 50 ng/ml nocodazole00:15:15 00:35:15 00:49:30 01:09:45 01:23:1500:09:0000:00:00 00:19:30

Mer

geTu

bulin

Hec

1

B

Acentrosomal MTsCentrosomal MTs

(-)

(-)

(-)(-)

(-)

(-)TACC3-dependent kinetochore

capture by centrosomal MT through MTOC sorting

Amphitelic

Amphitelic

Syntelic

Syntelic TACC3kinetochore pair

D

C

TACC3 TACC3-MT seedkinetochore pair MT

(1) (2) (3) (4)(-)

(+) (-)

(+)

(+)

(-)(-)(+)

(+)(-)

(-)

(+)

(-)

(-)

(+)

(+)(+)(+)

(-)(-)(-)

(+)(-)

(+)

(+)(+)

(+)(-) (-)

(+)

Fig. 5. A unique model for kinetochore capture by microtubules duringmitotic spindle assembly in somatic cells with centrosomes. (A and B) Liveimaging of HeLa cells coexpressing mcherry–tubulin and GFP–Hec1 in thepresence of 15 ng/mL (A) or 50 ng/mL (B) nocodazole. Magnified images areshown. Tubulin is in red and Hec1 is in green. (Scale bar, 5 μm.) (C) Illus-tration of initial kinetochore capture process by microtubules in a TACC3-dependent way. The small green dots represent TACC3 proteins, the ma-genta lines represent the microtubules (MT), and the large red dots stand forthe kinetochore pair. The initial kinetochore capture by adjacent acen-trosomal microtubules can be divided into four steps: (i) microtubule as-sembly followed by formation of TACC3–microtubule seeds; (ii) binding ofTACC3–microtubule seeds with kinetochores and further nucleation of themicrotubules; (iii) further assembly and sorting of the TACC3–microtubuleseeds lead to the formation of small TACC3–microtubule seeds/asters; and(iv) clustering of TACC3–microtubule seeds/asters and further nucleation ofthe acentrosomal microtubules produces the initial capture of the kinet-ochores by the acentrosomal aster microtubules. (D) Illustration of TACC3-dependent kinetochore capture in establishing bipolar spindle assembly. Theacentrosomal microtubule-captured kinetochores are further captured bycentrosomal microtubules through sorting the small acentrosomal astersinto the big centrosomal microtubule asters to generate amphitelic andsynthelic attachments before final chromosome biorientation.

Fu et al. PNAS | September 17, 2013 | vol. 110 | no. 38 | 15299

CELL

BIOLO

GY

Dow

nloa

ded

by g

uest

on

July

21,

202

1

kinetochores to be captured by the small TACC3–microtubuleasters through lateral binding. Fourth, by skating along themicrotubules of the TACC3–microtubule asters, the kinet-ochores are firmly captured through end-on connection of themicrotubules and the kinetochores. Simultaneously, the smallacentrosomal asters are sorted and join the two big cen-trosomal asters. And finally, depending on the transport ofacentrosomal microtubule asters toward the centrosomes by themicrotubule sorting mechanism, the acentrosomal asters aretotally fused in the centrosomal asters, and a bipolar spindle iseventually established. Noticeably, although the abovementionedresults are mainly based on observations in nocodazole-treatedcells, TACC3 is also required for kinetochore capture in un-perturbed mitosis without nocodazole treatment (Fig. 2). Be-cause acentrosomal microtubule seeds in normal mitosis actsimilarly as in nocodazole-treated samples (Fig. 3 A–E), andTACC3 targets to acentrosomal microtubules (Fig. 1E and Fig.S7) and may regulate acentrosomal microtubule assembly innormal mitosis (Fig. 2E), the acentrosomal TACC3–microtubuleseeds/asters mechanisms may be true in normal mitosis.Efficient chromosome capture requires a bias to complete the

capture process during mitotic spindle assembly (31). Meanwhile,chromosome movements and rotations are required to acceler-ate mitotic spindle assembly according to computer simulations(32). How does a cell speed up the process of kinetochore cap-ture? The lateral surface area of a microtubule is larger than thatof its tip and enables the fast kinetochore capture (3, 4). Theformation of TACC3–microtubule seeds/asters near the kineto-chore may speed up the kinetochore capture process throughforming acentrosomal microtubule-attached kinetochores, thusincreasing the microtubule capture surface of the kinetochoresand accelerating the movement of kinetochores/chromosomesalong the microtubules. Meanwhile, the growth and sorting of

the kinetochore-bound TACC3–microtubule asters also pro-mote the cluster formation and separation of the acentrosomalMTOCs as well as the movement and rotation of chromosomekinetochores. In agreement with the previous study that TACC3depletion delays the mitotic progression (18), our study furtherhighlights the role of the acentrosomal TACC3 in establishingkinetochore capture by microtubules in a fast and accurate way.In mammalian somatic cells, Ran GTPase activity promotes

microtubule nucleation at kinetochores (33), and Ran effectorssuch as microspherule protein 1 and γ-TuRC also function in Kfibers (34, 35). As TACC3 is also a target of the Ran GTPasesystem (20), it may coordinate with Ran to regulate acen-trosomal microtubules for kinetochore capture. Together theseprovide deep insights into the key events during the kinetochorecapture process by the microtubules. However, further analysesat a higher resolution level to elucidate the kinetochore behav-iors and functions are still needed in the future.

Materials and MethodsAll experiments were done in HeLa cells. Living cell imaging data was ac-quired by a DeltaVision system (Applied Precision). For microscopy of thefixed samples, the images were acquired by the DeltaVision or Zeiss LSM710microscope. Details of cell culture and drug treatment, RNA interference,small-molecule inhibitors, antibodies, live cell imaging and microscopy,Western blotting, immunofluorescence, and statistical analyses are providedin SI Materials and Methods.

ACKNOWLEDGMENTS. We thank Xuebiao Yao for HeLa cells stably express-ing mcherry-tubulin and plasmids and Jennifer Deluca, Jun Zhou, XueliangZhu, Michael Lampson, and Wen-Hwa Lee for plasmids and reagents. This workwas supported by funds from the State Key Basic Research and DevelopmentPlan (2010CB833705) and the National Natural Science Foundation of China(31071188, 31030044, and 90913021).

1. Cheeseman IM, Desai A (2008) Molecular architecture of the kinetochore-microtubuleinterface. Nat Rev Mol Cell Biol 9(1):33–46.

2. Kirschner M, Mitchison T (1986) Beyond self-assembly: From microtubules to mor-phogenesis. Cell 45(3):329–342.

3. Hayden JH, Bowser SS, Rieder CL (1990) Kinetochores capture astral microtubulesduring chromosome attachment to the mitotic spindle: Direct visualization in livenewt lung cells. J Cell Biol 111(3):1039–1045.

4. Tanaka K, et al. (2005) Molecular mechanisms of kinetochore capture by spindlemicrotubules. Nature 434(7036):987–994.

5. Gandhi SR, et al. (2011) Kinetochore-dependent microtubule rescue ensures theirefficient and sustained interactions in early mitosis. Dev Cell 21(5):920–933.

6. Heald R, et al. (1996) Self-organization of microtubules into bipolar spindles aroundartificial chromosomes in Xenopus egg extracts. Nature 382(6590):420–425.

7. Schuh M, Ellenberg J (2007) Self-organization of MTOCs replaces centrosome functionduring acentrosomal spindle assembly in live mouse oocytes. Cell 130(3):484–498.

8. Khodjakov A, Cole RW, Oakley BR, Rieder CL (2000) Centrosome-independent mitoticspindle formation in vertebrates. Curr Biol 10(2):59–67.

9. Tulu US, Rusan NM, Wadsworth P (2003) Peripheral, non-centrosome-associated mi-crotubules contribute to spindle formation in centrosome-containing cells. Curr Biol13(21):1894–1899.

10. Tulu US, Fagerstrom C, Ferenz NP, Wadsworth P (2006) Molecular requirements forkinetochore-associated microtubule formation in mammalian cells. Curr Biol 16(5):536–541.

11. Maiato H, Rieder CL, Khodjakov A (2004) Kinetochore-driven formation of kineto-chore fibers contributes to spindle assembly during animal mitosis. J Cell Biol 167(5):831–840.

12. Goshima G, Nédélec F, Vale RD (2005) Mechanisms for focusing mitotic spindle polesby minus end-directed motor proteins. J Cell Biol 171(2):229–240.

13. Goshima G, Vale RD (2003) The roles of microtubule-based motor proteins in mitosis:Comprehensive RNAi analysis in the Drosophila S2 cell line. J Cell Biol 162(6):1003–1016.

14. Schneider L, et al. (2007) The transforming acidic coiled coil 3 protein is essential forspindle-dependent chromosome alignment and mitotic survival. J Biol Chem 282(40):29273–29283.

15. Booth DG, Hood FE, Prior IA, Royle SJ (2011) A TACC3/ch-TOG/clathrin complex sta-bilises kinetochore fibres by inter-microtubule bridging. EMBO J 30(5):906–919.

16. Fu W, et al. (2010) Clathrin recruits phosphorylated TACC3 to spindle poles for bipolarspindle assembly and chromosome alignment. J Cell Sci 123(Pt 21):3645–3651.

17. Hubner NC, et al. (2010) Quantitative proteomics combined with BAC Trans-geneOmics reveals in vivo protein interactions. J Cell Biol 189(4):739–754.

18. Lin CH, Hu CK, Shih HM (2010) Clathrin heavy chain mediates TACC3 targeting tomitotic spindles to ensure spindle stability. J Cell Biol 189(7):1097–1105.

19. Katayama H, Sasai K, Kloc M, Brinkley BR, Sen S (2008) Aurora kinase-A regulateskinetochore/chromatin associated microtubule assembly in human cells. Cell Cycle7(17):2691–2704.

20. Fu W, Jiang Q, Zhang C (2011) Novel functions of endocytic player clathrin in mitosis.Cell Res 21(12):1655–1661.

21. Fu J, Bian M, Liu J, Jiang Q, Zhang C (2009) A single amino acid change convertsAurora-A into Aurora-B-like kinase in terms of partner specificity and cellular func-tion. Proc Natl Acad Sci USA 106(17):6939–6944.

22. Ditchfield C, et al. (2003) Aurora B couples chromosome alignment with anaphase bytargeting BubR1, Mad2, and Cenp-E to kinetochores. J Cell Biol 161(2):267–280.

23. Kapoor TM, Mayer TU, Coughlin ML, Mitchison TJ (2000) Probing spindle assemblymechanisms with monastrol, a small molecule inhibitor of the mitotic kinesin, Eg5.J Cell Biol 150(5):975–988.

24. Steegmaier M, et al. (2007) BI 2536, a potent and selective inhibitor of polo-like ki-nase 1, inhibits tumor growth in vivo. Curr Biol 17(4):316–322.

25. Zhang X, et al. (2009) Sequential phosphorylation of Nedd1 by Cdk1 and Plk1 is re-quired for targeting of the gammaTuRC to the centrosome. J Cell Sci 122(Pt 13):2240–2251.

26. Cai S, Weaver LN, Ems-McClung SC, Walczak CE (2009) Kinesin-14 family proteinsHSET/XCTK2 control spindle length by cross-linking and sliding microtubules.Mol BiolCell 20(5):1348–1359.

27. Dumont S, Mitchison TJ (2009) Force and length in the mitotic spindle. Curr Biol19(17):R749–R761.

28. Hallen MA, Endow SA (2009) Anastral spindle assembly: A mathematical model. Bi-ophys J 97(8):2191–2201.

29. Breuer M, et al. (2010) HURP permits MTOC sorting for robust meiotic spindle bipolarity,similar to extra centrosome clustering in cancer cells. J Cell Biol 191(7):1251–1260.

30. Hornick JE, et al. (2011) Amphiastral mitotic spindle assembly in vertebrate cellslacking centrosomes. Curr Biol 21(7):598–605.

31. Wollman R, et al. (2005) Efficient chromosome capture requires a bias in the ‘search-and-capture’ process during mitotic-spindle assembly. Curr Biol 15(9):828–832.

32. Paul R, et al. (2009) Computer simulations predict that chromosome movements androtations accelerate mitotic spindle assembly without compromising accuracy. ProcNatl Acad Sci USA 106(37):15708–15713.

33. Torosantucci L, De Luca M, Guarguaglini G, Lavia P, Degrassi F (2008) LocalizedRanGTP accumulation promotes microtubule nucleation at kinetochores in somaticmammalian cells. Mol Biol Cell 19(5):1873–1882.

34. Meunier S, Vernos I (2011) K-fibre minus ends are stabilized by a RanGTP-dependent mechanism essential for functional spindle assembly. Nat Cell Biol13(12):1406–1414.

35. Mishra RK, Chakraborty P, Arnaoutov A, Fontoura BM, Dasso M (2010) The Nup107-160 complex and gamma-TuRC regulate microtubule polymerization at kinetochores.Nat Cell Biol 12(2):164–169.

15300 | www.pnas.org/cgi/doi/10.1073/pnas.1312382110 Fu et al.

Dow

nloa

ded

by g

uest

on

July

21,

202

1