centrosome number is controlled by a centrosome- intrinsic

L E T T E R S

NATURE CELL BIOLOGY VOLUME 5 | NUMBER 6 | JUNE 2003 539

Centrosome number is controlled by a centrosome-intrinsic block to reduplicationConnie Wong1 and Tim Stearns1,2

The centrosome duplicates once in S phase. To determinewhether there is a block in centrosome reduplication, we useda cell fusion assay to compare the duplication potential ofunduplicated G1 centrosomes and recently duplicated G2centrosomes. By fusing cells in different cell cycle stages, wefound that G2 centrosomes were unable to reduplicate in acellular environment that supports centrosome duplication.Furthermore, G2 cytoplasm did not inhibit centrosomeduplication in fused cells, indicating that the block toreduplication is intrinsic to the centrosomes rather than thecytoplasm. To test the underlying mechanism, we createdmononucleate G1 cells with two centrosomes by fusing cellswith enucleated cytoplasts. Both centrosomes duplicated,indicating that the block is not controlled bycentrosome:nucleus ratio. We also found that human primarycells have tight control over centrosome number duringprolonged S-phase arrest and that this control is partiallyabrogated in transformed cells. This suggests a link betweenthe control of centrosome duplication and maintenance ofgenomic stability.

The centrosome is the microtubule-organizing centre of animal cells

and is important for organizing the bipolar spindle during mitosis.

Aberrant centrosome number can result in the generation of

monopolar or multipolar spindles1, potentially causing aneuploidy or

cell death. Cells from many types of cancer have multiple centrosomes

and this phenotype has been postulated to function in tumorigenesis

by promoting genomic instability2–4. Centrosome duplication and

DNA replication are initiated at the G1–S transition, and both occur

only once in a single cell cycle. Rao and Johnson used a cell fusion

assay to study the regulation of DNA replication5. They showed that in

a fusion of G1- and S-phase cells, both the G1 and S nuclei replicate;

whereas in a fusion of G2 and S cells, the S nucleus replicates, but the

G2 nucleus does not re-replicate5. This demonstrated the existence of

a block to rereplication of a previously replicated nucleus.

Furthermore, this block is maintained through several mechanisms

and is regulated by the activity of cyclin-dependent kinases (Cdks)6.

Centrosome duplication is under the positive control of Cdk2 and

other factors7,8. However, it is unclear whether duplication is limited

to one round per cell cycle simply by the decline of the positive control

at the end of S phase or by a specific block to reduplication of previ-

ously duplicated centrosomes. We used a cell fusion assay similar to

that of Rao and Johnson5 to address this question. We reasoned that it

should be possible to distinguish these two possibilities by comparing

1Department of Biological Sciences, Stanford University, and 2Department of Genetics, Stanford University Medical School, Stanford, CA 94305-5020, USA.Correspondence should be addressed to T.S. (e-mail: [email protected]).

Percentage of BrdU incorporationPercentage with one centrosomePercentage with two centrosomes

100

80

60

40

20

0

Per

cent

age

of c

ells

Unsync (n = 500) G1 (n = 900) S (n = 700) G2 (n = 700)

a

b

Figure 1 Cell cycle synchronization of fibroblast cells. (a) Fibroblasts werearrested in G1 phase by serum starvation, S phase by double thymidineblock, and G2 phase by timed release of the S-phase cells. The degree ofsynchronization was assessed by BrdU labelling and determination ofcentrosome number by centrin staining (one pair of centrioles is equivalentto one centrosome). Unsynchronized cells were used as a control. (b)Fluorescence microscopy images of cells with one pair (left) or two pairs(right) of centrioles. Centrin, green; DNA, blue. Scale bar represents 1 µm.

© 2003 Nature Publishing Group

L E T T E R S

540 NATURE CELL BIOLOGY VOLUME 5 | NUMBER 6 | JUNE 2003

the duplication potential of unduplicated G1 centrosomes and

recently duplicated G2 centrosomes when placed in S-phase cyto-

plasm, a cellular environment that supports duplication. If duplication

was limited only by presence of a positive signal, then both G1 and G2

centrosomes introduced into the S-phase cytoplasm would be

expected to duplicate. However, if duplication was limited by a centro-

some-intrinsic block to reduplication, then only the G1 centrosomes

would be expected to duplicate. As such a mechanism might be abro-

gated in transformed cells, which often have centrosome abnormali-

ties, we performed all experiments with two cell types that are expected

to have normal cell cycle control: primary human diploid fibroblasts

(HDFs) isolated from infant foreskin and a human diploid fibroblast

cell line immortalized by expression of human telomerase reverse

transcriptase (hTERT-BJ1)9,10. Both cell types gave similar results in all

experiments.

A G1 population was obtained by serum starvation of cells for

approximately 72 h. A G2 population was obtained by arresting cells in

S phase using a double-thymidine block11 followed by a 7.5-h release.

Approximately 95% of cells were undergoing DNA replication at 3 h

after release from the thymidine block, as determined by 5-bromod-

eoxyuridine (BrdU) incorporation (Fig. 1a). In contrast, only approxi-

mately 8% of cells were undergoing DNA replication at 7.5 h after the

release, indicating that most cells had successfully entered G2 phase.

Flow cytometry analysis of the G1 and G2 cells confirmed that approx-

imately 90% of each of the populations had the expected DNA content

(data not shown). We also determined centrosome number in the syn-

chronized cells. As a mature centrosome contains one pair of centri-

oles, we used immunofluorescence microscopy analysis of the

centriolar protein centrin12 as an indicator of centrosome number

(Fig. 1b). We found that 91% of cells in the G1 population had a single

centrosome (one pair of centrioles), whereas 93% of the cells in the G2

population had two centrosomes (two pairs of centrioles; Fig. 1a).

Cells were fused using polyethylene glycol and the desired binucleate

cells were identified by microscopy. Successful analysis of cell fusions

between cells at different cell-cycle stages required the ability to distin-

guish between homophasic (G1–G1, S–S and G2–G2) and heteropha-

sic (G1–S and G2–S) fusions. For this purpose, fluorescent conjugates

of concanavalin A (ConA) — a lectin that binds to cell-surface pro-

teins — were used to label the two cell populations. Immediately

before fusion, one cell population was labelled with ConA–Texas Red

and the other with ConA–Marina Blue. Thus, only heterophasic binu-

cleate cells should contain both Texas Red and Marina Blue. The cell-

surface dyes mixed in fused cells (Fig. 2a), but were not transferable

between adjacent unfused cells (Fig. 2b). Centriole number in the

fused cells could easily be discerned by centrin staining (Fig. 2a). Note

that in several micrographs presented, the ConA–Marina Blue fluores-

cence is obscured by the brightness of the DAPI nuclear stain.

To test whether the fusion process affected the kinetics of centrosome

duplication, cells from two identical G1 populations were fused, allowed

to proceed through the cell cycle for 24 h and then processed for

immunofluorescence microscopy with an anti-centrin antibody. Fusion

of two G1 cells would result in a binucleate cell with two centrosomes;

thus, duplication during the post-fusion incubation would result in four

centrosomes. We found that 42% of binucleate cells had undergone

duplication by 24 h post-fusion. This was similar to the percentage of

untreated G1 cells that underwent duplication 24 h after release (49%)

and to the percentage of treated, but unfused, G1 cells that underwent

duplication 24 h after treatment (47%). Thus, centrosome duplication

occurred with normal kinetics in fused binucleate cells.

Centrosome duplication occurs in S phase, therefore it is important

that the G1–S and G2–S cell fusions remain in S phase for a sufficient

ConA–MB, DAPI ConA–TR Centrin

a

One BrdU-positive nucleusTwo BrdU-positive nuclei

90

80

70

60

50

40

30

20

10

0

Per

cent

age

of c

ells

6 h (n = 66) 12 h (n = 85) 6 h (n = 43) 12 h (n = 56)

G1–S G2–STime after fusion

d

ConA–MB, DAPI ConA–MB, DAPI

ConA–TR

BrdUb c

G2-Sfusion

G1-Sfusion

Figure 2 DNA replication and centrosome number in fused cells. Cells weresynchronized, surface-labelled and fused using polyethylene glycol beforeplating and incubation for the indicated times. DNA replication was assayedby BrdU labelling and centrosomes were counted by staining with an anti-centrin antibody. (a) A fused binucleate cell with mixed ConA–Marina Blue(ConA–MB) and ConA–Texas Red (ConA–TR) surface label. Nucleus isstained with DAPI. This cell has three centrosomes, revealed as three pairsof centrin-staining centrioles, enlarged in insets. (b) Adjacent unfused cellsmaintain discrete ConA surface labelling at 24 h after fusion treatment. (c)A G2–S-fused binucleate cell containing one BrdU-labelled nucleus (top). AG1–S binucleate cell containing two BrdU-labelled nuclei (bottom). One ofthe nuclei, presumably from the S-phase cell, shows greater BrdUincorporation. Scale bar represents 10 µm in a–c. d, Quantification of thenumber of BrdU-labelled nuclei in G1–S and G2–S binucleate cells at timesafter fusion. The number of cells counted at each time is indicated.


L E T T E R S


time after fusion. To test this, we compared the pattern of BrdU incor-

poration in G1–S and G2–S binucleate cells after fusion (Fig. 2d). In

G1–S cells at 6 h after fusion, approximately 60% of the binucleate cells

had one BrdU-labelled nucleus and 30% had two BrdU-labelled

nuclei. At 12 h after fusion, approximately 20% had one BrdU-labelled

nucleus, and more than 70% had two BrdU-labelled nuclei. In cells

with two BrdU-labelled nuclei, one of the nuclei typically showed a

more intense BrdU labelling (Fig. 2c), presumably because the S-phase

nucleus incorporated BrdU over a longer period than the G1 nucleus.

In G2–S cells at 6 h after fusion, approximately 70% of the binucleate

cells had one BrdU-labelled nucleus, whereas only 7% had two BrdU-

labelled nuclei; this distribution was similar at 12 h after fusion

(Fig. 2d). These data indicate that cells in both types of fusion remain

in S phase, that the G2 nucleus does not re-replicate, and importantly,

that the G2 cytoplasm did not inhibit S-phase events, as found by Rao

and Johnson for Hela cells5.

We next compared the behaviour of the unduplicated G1 centro-

somes and newly duplicated G2 centrosomes in G1–S and G2–S

fusions. Fusion of a G1 cell to an S-phase cell results in the generation

of a cell with three centrosomes. At 6 h after fusion, approximately

65% of the G1–S-fused cells had three centrosomes and 30% had four

centrosomes (Fig. 3a). In the remainder, centrosome number could

not be clearly determined. At 12 h after fusion, approximately 25% of

the G1–S-fused cells had three centrosomes, whereas 70% had four

centrosomes, indicating that one of the three had duplicated with the

same timing as DNA replication. Fusion of a G2 cell to an S-phase cell

results in a cell with four centrosomes. At 6 h after fusion greater than

90% of the G2–S binucleate cells had four centrosomes (Fig. 3a). In

contrast to the G1–S fusions, no centrosome duplication was observed

at 12 h (Fig. 3a). In several experiments, cells were analysed for up to

24 h without evidence of centrosome duplication. Thus, previously

duplicated G2 centrosomes were not able to reduplicate in the G2–S

fusions, even though DNA replication could proceed, demonstrating

the existence of a block to centrosome reduplication.

To ensure that the G2–S fusions would have sufficient time in

S phase to carry out centrosome duplication if it were possible, the

fusion experiment was repeated under conditions of S-phase arrest.

G2- and S-phase cells were fused then maintained in S phase by thymi-

dine treatment for 24 h, well in excess of the time needed for centro-

some duplication in the G1–S samples. After 24 h of S-phase arrest,

82% of the G2–S binucleate cells still had four centrosomes, and 18%

still had fewer than four (n = 27 cells), presumably through fusion of

unsynchronized cells. None of the unfused cells had more than four

centrosomes; therefore, extending the S-phase period could not over-

come the block to reduplication.

There are two possible explanations for the inability of previously

duplicated G2 centrosomes to reduplicate in the G2–S fusions. First,

centrosomes contain information about their duplication status that

prevents their reduplication; second, the G2 cytoplasm contains a fac-

tor that functions dominantly to prevent reduplication, regardless of

the duplication status of the centrosome. To rule out the latter expla-

nation, we fused G1 cells to G2 cells and asked if centrosome duplica-

tion could occur in the presence of G2 cytoplasm. In these

experiments, the G1 cells were released from serum starvation block

for 5 h before fusion. Thus, the fused cells were advanced in their cell

cycle relative to the cells above and were assayed at earlier times. This

difference, however, had no effect on the experimental outcome.

Fusion of a G1 cell to a G2 cell results in a cell with three centrosomes.

At 2 h after fusion, 70% of the G1–G2-fused cells had three centro-

somes (Fig. 3b). However, at 7 h after fusion, only 18% of the G1–G2-

fused cells had three centrosomes, whereas 80% had four centrosomes.

Even at 18 h after fusion, the number of G1–G2 binucleate cells with

four centrosomes remained approximately 85% (Fig. 3b). Thus, G2

cytoplasm does not prevent centrosome duplication in fused cells.

Next, we used the Cdk inhibitors Purvalanol A and Roscovitine to

determine whether the centrosome duplication that we observed in

G1–G2 fusions occurs by the normal Cdk2-mediated pathway13–16.

Purvalanol A17 was added to G1–G2-fused cells at 1 h after fusion. At

7 h after fusion, only 30% of the purvalanol A-treated G1–G2 binucle-

ate cells had four centrosomes, a 2.5-fold reduction from the untreated

population (Fig. 3b). Similar results were obtained with another Cdk

inhibitor, Roscovitine18 (data not shown). These results indicate that

centrosome duplication in the binucleate cells occurs through the nor-

mal Cdk-dependent pathway.

We have shown that previously duplicated G2 centrosomes cannot

reduplicate in cell fusions with either G1- or S-phase cells and that the

block to reduplication does not occur at the level of the G2 cytoplasm.

This suggests that centrosome duplication is limited to one round per cell

cycle by a centrosome-intrinsic mechanism, just as DNA replication is

limited by a nuclear-intrinsic mechanism19. We envision two models to

Binucleate cells with three centrosomesBinucleate cells with four centrosomes100

90

80

70

60

50

40

30

20

10

0

Per

cent

age

of c

ells

6 h (n = 40) 12 h (n = 60) 6 h (n = 74) 12 h (n = 43)

G1–S G2–S

Time after fusion

Binucleate cells with three centrosomesBinucleate cells with four centrosomes

100

90

80

70

60

50

40

30

20

10

0

Per

cent

age

of c

ells

2 h (n = 38) 7 h (n = 142) 18 h (n = 144) 7 h (n = 34)

G1–G2 G2–G2 (PurA)

Time after fusion

a

b

Figure 3 Centrosome duplication in G1–S, G2–S and G1–G2 cell fusions.Cell fusions were prepared as in Fig. 2. Quantification of the number ofcentrioles in (a) G1–S- and G2–S-fused binucleate cells and (b) G1–G2-fused binucleate cells at various times after fusion treatment. The Cdkinhibitor Purvalanol A (PurA) blocks centrosome duplication in fused cells.


L E T T E R S


account for the block in reduplication. The first is that duplication

results in a physical alteration of the centrosome such that it cannot be

reduplicated until it passes through mitosis, similar to the licensing

model for DNA replication control19. The second model is that there is a

counting mechanism by which cells assess the ratio of centrosomes to

nuclei and only duplicate the centrosome to the level of two centrosomes

per nucleus. In principle, these two models should be distinguishable by

generating mononucleate G1 cells with two G1 centrosomes. The physi-

cal alteration model predicts that both of these centrosomes would

duplicate, yielding a total of four centrosomes, whereas the counting

model predicts that neither of the centrosomes would duplicate. To test

these models, we created mononucleate G1 cells with two G1 centro-

somes by fusing normal G1 cells to cytoplasts (enucleated cells) obtained

from G1 cells. The cytoplasts were prepared by centrifugation of G1

fibroblasts through a Ficoll gradient; approximately half of the cytoplasts

retained the single G1 centrosome (Fig. 4a). The cells and cytoplasts were

surface-labelled, as described above, to allow identification of the appro-

priate fusions. The G1 cytoplast–cell fusions were released from G1

arrest, allowed to progress to S phase and maintained in S phase with

thymidine treatment so that centrosome number could be determined.

Both unfused cells and cytoplast–cell fusions took approximately 24 h to

enter S phase from G1 arrest, hence the cytoplast–cell fusions were

assayed at 10 h (pre-S phase) and 36 h (S phase) after fusion.

At 10 h after fusion, 46% of the cytoplast–cell fusions had one cen-

trosome, as expected from the fusion of a G1 cell and a cytoplast lack-

ing a centrosome; 44% of the cytoplast-cell fusions had two

centrosomes, as expected from fusion of a G1 cell and a cytoplast con-

taining a centrosome (Fig. 4b, c). In addition, 10% of the cytoplast–cell

fusions contained more than two centrosomes, as expected from

fusion between G1 cells and multiple cytoplasts; note that only 2% of

fusions had four centrosomes. At 36 h after fusion, 8% of the cyto-

plast–cell fusions contained one centrosome, whereas 56% contained

two centrosomes and 25% contained four centrosomes (Fig. 4b, c).

The cytoplast–cell fusions that contained two centrosomes at 36 h

most probably resulted from centrosome duplication in the cells that

contained one centrosome before S phase. Similarly, the cytoplast–cell

fusions that contained four centrosomes most probably resulted from

centrosome duplication in the cells that contained two centrosomes

before S phase. Thus, both centrosomes in a G1 cell with two centro-

somes are able to duplicate, ruling out a centrosome/nucleus counting

mechanism for controlling centrosome number.

We have shown that there is a centrosome-intrinsic block to redupli-

cation in somatic mammalian cells. We wondered how these results

can be reconciled with the observation that embryos and certain

somatic cells can undergo multiple rounds of centrosome duplication

during prolonged S-phase arrest21–23. It is possible that embryos lack a

block to centrosome reduplication and rely on having a limited win-

dow of opportunity for duplication. For example, the mitotic state

inhibits centrosome duplication in both sea urchin and frog

embryos14,24, so the alternation of S phase and M phase in the embry-

onic cell cycle would limit duplication to S phase, which might be long

enough to only allow a single round of duplication. In somatic cells,

centrosome reduplication under conditions of S-phase arrest is only

observed after prolonged incubation, equivalent in length to several

generation times21. It is possible that under these conditions, the block

to reduplication is overcome. Interestingly, centrosome number does

not increase exponentially in S-phase arrest reduplication21, consistent

with the idea that each centrosome has an intrinsic block that has a sto-

chastic chance of being overcome.

If the delay in centrosome reduplication during S-phase arrest was

caused by a block to reduplication, then abrogation of that block

might result in more rapid reduplication. Many types of cancer cells

have been reported to have aberrant centrosome number3,4 and the

tumour suppressor protein p53 has been implicated in the regulation

of centrosome duplication25. Therefore, we compared the kinetics of

centrosome reduplication in primary HDF cells, HCT116 p53+/+

colon cancer cells and HCT116 p53−/− cells generated by somatic

knockout of the p53 gene26 (Fig. 5). Each cell type was subjected to S-

phase arrest by hydroxyurea treatment and assayed for centrosome

number at various times during the arrest; any cell with more than

two centrosomes was counted as aberrant. The anti-centrin antibody

had high background labelling in HCT116 cells; therefore, we used

antibodies against γ-tubulin and pericentrin (components of the

pericentriolar material) to label centrosomes (Fig. 5a). Note that for

these antibodies, a single dot of labelling corresponds to one centro-

some. Before arrest, the normal cells and cancer cells already differed

substantially in centrosome number, with no HDF cells and approxi-

mately 8% of cancer cells having an aberrant centrosome number

(Fig. 5b). This difference increased markedly by day 4 of the S-phase

arrest, with approximately 4% of HDF cells, 24% of the HCT p53+/+

cells and 50% of HCT p53−/− cells having an aberrant centrosome

number (Fig. 5b).

10 h (n = 108)36 h (n = 122)

70

60

50

40

30

20

10

0

Per

cent

age

of c

ells

1 2 4 Other

Number of centrosomes

a b

c

Figure 4 Centrosome duplication in G1–G1 cytoplast–cell fusions. (a) Differential interference contrast (DIC; left) and fluorescence (right)images of a cytoplast adjacent to a normal cell. The cytoplast is labelledwith ConA–Texas Red and contains one centrosome (one pair of centrioles;inset), but no nucleus. The cell is labelled with ConA–Marina Blue andcontains one centrosome (one pair of centrioles; inset) and one nucleus. (b) A G1–G1 cytoplast–cell fusion containing two centrosomes (two pairs ofcentrioles; inset) at 10 h after fusion (left). A G1–G1 cytoplast–cell fusioncontaining four centrosomes (four pairs of centrioles; inset) at 36 h afterfusion (right). Scale bar represents 10 µm. c, Quantification of centriolenumbers in G1–G1 cytoplast–cell fusions at 10 h and 36 h after fusion.


L E T T E R S


These results are consistent with the possibility that cancer cells in

general, and p53−/−cancer cells in particular, have defects in the block to

centrosome reduplication. In addition, they are consistent with previ-

ous results showing that loss of p53 results in deregulation of centro-

some duplication27. However, we note that even in p53−/− cancer cells,

there is a delay in gross reduplication of the centrosomes until four days

of arrest, suggesting that regulation of centrosome number probably

occurs through several overlapping mechanisms and that the effect of

p53 might be indirect. It is also probable that failure of the block to cen-

trosome reduplication is not the only centrosome amplification mecha-

nism; another mechanism is the failure to undergo cytokinesis after

mitosis, resulting in a tetraploid cell with two centrosomes at G1 (refs

28, 29). The control of centrosome duplication is probably critical to

the maintenance of genome integrity and it will be of great interest to

determine the molecular basis for the block to centrosome reduplica-

tion that we have identified.

METHODSTissue culture, cell fusion and cytoplast preparation. HDFs were harvested

from infant foreskin and hTERT-BJ1 cells were purchased from Clontech (Palo

Alto, CA). Both cell types were cultured in DMEM (Invitrogen, Carlsbad, CA)

containing 10% foetal bovine serum. G1 cells were obtained by culturing cells

in DMEM with 0.1% newborn calf serum for approximately three days. G2 cells

were obtained by releasing cells arrested in S phase for 7.5 h. Cells were arrested

in S phase using a double-thymidine block, as previously described11. For the

fusion experiments, 1 × 106 cells of each cell population to be fused were

trypsinized, resuspended in 600 µl of ConA–Texas Red or ConA–Marina Blue

(both at 16.7 µg ml−1) in PBS and incubated at 37 °C for 10 min. Cells were

then washed separately with 10 ml of growth media before being resuspended

and mixed in 600 µl of 50 µM SDS in PBS for 3 min at 37 °C. Next, cells were

centrifuged and resuspended in 600 µl of 50% polyethylene glycol-3350 for 1

min at 37 °C. Serum-free DMEM (1 ml) was added to the cells, mixed gently

and incubated at room temperature for 1 min. This process was repeated until a

total of 5 ml serum-free DMEM had been added. The cells were then washed

once with 10 ml serum-free media before being resuspended in 20 ml of growth

media and seeded onto coverslips. Cytoplasts were prepared by density gradient

centrifugation, as previously described20, except that cytochalasin B (10 µg ml−

1) was added to the cells 20 min before centrifugation. Cytoplasts generated

from 1 × 107 G1 cells were fused with 1 × 107 G1 cells, as described above.

p53+/+ and p53−/− HCT116 cells (a gift of J. Ford, Stanford University, CA) were

cultured in McCoy’s 5A Media (Invitrogen) with 10% foetal bovine serum.

Hydroxyurea (2 mM) was used for the prolonged S-phase arrest of HCT cells.

Cytochemistry. Cells were fixed with methanol at −20 °C then blocked with 3%

bovine serum albumin (w/v) and 0.1% Triton X-100 in PBS for 30 min. For

visualization of centrin, cells were incubated with a monoclonal mouse anti-

centrin antibody (20H5; a gift from J. Salisbury, Mayo Clinic, Rochester, MN

(ref. 30)), followed by Alexa 488-conjugated goat anti-mouse antibody

(Molecular Probes, Eugene, OR). DNA was visualized using 4′,6-diamidino-2-

phenylindole (DAPI; Molecular Probes). Cells were then observed under a Zeiss

Axioskop microscope (Zeiss, Thornwood, NJ) using a 100× oil-immersion

objective. For visualizing replicated DNA, BrdU was added to the cells to a final

concentration of 20 µM and incubated for 30 min at 37 °C. Cells were treated

with 2 M HCl for 30 min at room temperature or with 1 U ml−1 DNase I for 1 h

at 37 °C before treatment with 1000 U ml−1 exonuclease III for 1 h at 37 °C.

Staining was then performed as described above.

ACKNOWLEDGEMENTS

We thank J. Ford for cell lines, J. Salisbury for the anti-centrin antibody, and G.-W.

Fang and P. Jackson for comments on the manuscript. This work was supported by

a grant from the Human Frontier Science Program. C.W. was supported by a

Stanford Graduate Fellowship.

COMPETING FINANCIAL INTERESTS

The authors declare that they have no competing financial interests.

Received 3 January 2003; Accepted 25 March 2003;Published online: 27 May 2003; DOI: 10.1038/ncb993

1. Sluder, G., Thompson, E. A., Miller, F. J., Hayes, J. & Rieder, C. L. The checkpointcontrol for anaphase onset does not monitor excess numbers of spindle poles or bipo-lar spindle symmetry. J. Cell Sci. 110, 421–429 (1997).

2. Brinkley, B. & Goepfert, T. Supernumerary centrosomes and cancer: Boveri’s hypoth-esis resurrected. Cell Motil. Cytoskeleton 41, 281–288 (1998).

3. Lingle, W. L. & Salisbury, J. L. The role of the centrosome in the development ofmalignant tumors. Curr. Top. Dev. Biol. 49, 313–329 (2000).

4. Pihan, G. A. & Doxsey, S. J. The mitotic machinery as a source of genetic instability incancer. Semin. Cancer Biol. 9, 289–302 (1999).

5. Rao, P. N. & Johnson, R. T. Mammalian cell fusion: studies on the regulation of DNAsynthesis and mitosis. Nature 225, 159–164 (1970).

6. Nguyen, V. Q., Co, C. & Li, J. J. Cyclin-dependent kinases prevent DNA re-replicationthrough multiple mechanisms. Nature 411, 1068–1073 (2001).

7. Hinchcliffe, E. H. & Sluder, G. “It takes two to tango”: understanding how centro-some duplication is regulated throughout the cell cycle. Genes Dev. 15, 1167–1181(2001).

8. Stearns, T. Centrosome duplication. A centriolar pas de deux. Cell 105, 417–420(2001).

9. Bodnar, A. G. et al. Extension of life-span by introduction of telomerase into normalhuman cells. Science 279, 349–352 (1998).

10. Vaziri, H. & Benchimol, S. Reconstitution of telomerase activity in normal humancells leads to elongation of telomeres and extended replicative life span. Curr. Biol. 8,279–282 (1998).

11. Spector, D. L., Goldman, R. D. & Leinwand, L. A. Cells a laboratory manual (ColdSpring Harbor Press, Cold Spring Harbor, 1998).

70

60

50

40

30

20

10

0

Per

cent

age

of c

ells

with

two

or m

ore

cent

roso

mes

p53+/+ HDF

p53+/+ HCT116

p53–/– HCT116

Day 0 Day 3 Day 4

a

b

DAPI

Day 3

Day 5

Pericentrin γ-tubulin

Figure 5 Centrosome reduplication during S-phase arrest. (a) HCT116 p53−

/− cells at day 3 and day 5 of hydroxyurea arrest stained for DAPI, γ-tubulinand pericentrin to identify centrosome number. Note that these antibodiesstain the pericentriolar material rather than the centrioles, so that a singledot of staining corresponds to one centrosome. Scale bar represents 10 µm.(b) Percentage of HDF, HCT116 p53+/+ and HCT116 p53−/− cells with anaberrant number of centrosomes (greater than two) before S-phase arrest(day 0) and after three and four days of S-phase arrest.


L E T T E R S


12. Paoletti, A., Moudjou, M., Paintrand, M., Salisbury, J. L. & Bornens, M. Most of cen-trin in animal cells is not centrosome-associated and centrosomal centrin is confinedto the distal lumen of centrioles. J. Cell Sci. 109, 3089–3102 (1996).

13. Hinchcliffe, E. H., Li, C., Thompson, E. A., Maller, J. L. & Sluder, G. Requirement ofCdk2-cyclin E activity for repeated centrosome reproduction in Xenopus egg extracts.Science 283, 851–854 (1999).

14. Lacey, K. R., Jackson, P. K. & Stearns, T. Cyclin-dependent kinase control of centro-some duplication. Proc. Natl Acad. Sci. USA 96, 2817–2822 (1999).

15. Matsumoto, Y., Hayashi, K. & Nishida, E. Cyclin-dependent kinase 2 (Cdk2) isrequired for centrosome duplication in mammalian cells. Curr. Biol. 9, 429–432(1999).

16. Meraldi, P., Lukas, J., Fry, A. M., Bartek, J. & Nigg, E. A. Centrosome duplication inmammalian somatic cells requires E2F and Cdk2–cyclin A. Nature Cell Biol. 1,88–93 (1999).

17. Gray, N. S. et al. Exploiting chemical libraries, structure, and genomics in the searchfor kinase inhibitors. Science 281, 533–538 (1998).

18. De Azevedo, W. F. et al. Inhibition of cyclin-dependent kinases by purine analogues:crystal structure of human cdk2 complexed with roscovitine. Eur. J. Biochem. 243,518–526 (1997).

19. Blow, J. J. & Laskey, R. A. A role for the nuclear envelope in controlling DNA replica-tion within the cell cycle. Nature 332, 546–548 (1988).

20. Wigler, M. H. & Weinstein, I. B. A preparative method for obtaining enucleated mam-malian cells. Biochem. Biophys. Res. Commun. 63, 669–674 (1975).

21. Balczon, R. et al. Dissociation of centrosome replication events from cycles of DNAsynthesis and mitotic division in hydroxyurea-arrested Chinese hamster ovary cells. J.

Cell Biol. 130, 105–115 (1995).22. Gard, D. L., Hafezi, S., Zhang, T. & Doxsey, S. J. Centrosome duplication continues in

cycloheximide-treated Xenopus blastulae in the absence of a detectable cell cycle. J.Cell Biol. 110, 2033–2042 (1990).

23. Sluder, G., Miller, F. J., Cole, R. & Rieder, C. L. Protein synthesis and the cell cycle:centrosome reproduction in sea urchin eggs is not under translational control. J. CellBiol. 110, 2025–2032 (1990).

24. Hinchcliffe, E. H., Cassels, G. O., Rieder, C. L. & Sluder, G. The coordination of cen-trosome reproduction with nuclear events of the cell cycle in the sea urchin zygote. J.Cell Biol. 140, 1417–1426 (1998).

25. Fukasawa, K., Choi, T., Kuriyama, R., Rulong, S. & Vande Woude, G. F. Abnormal cen-trosome amplification in the absence of p53. Science 271, 1744–1777 (1996).

26. Bunz, F. et al. Disruption of p53 in human cancer cells alters the responses to thera-peutic agents. J. Clin. Invest. 104, 263–269 (1999).

27. Tarapore, P. & Fukasawa, K. Loss of p53 and centrosome hyperamplification.Oncogene 21, 6234–6240 (2002).

28. Meraldi, P., Honda, R. & Nigg, E. A. Aurora-A overexpression reveals tetraploidizationas a major route to centrosome amplification in p53−/− cells. EMBO J. 21, 483–492(2002).

29. Borel, F., Lohez, O., Lacroix, F. & Margolis, R. Multiple centrosomes arise fromtetraploidy checkpoint failure and mitotic centrosome clusters in p53 and RB pocketprotein-compromised cells. Proc. Natl Acad. Sci. USA 99, 9819–9824 (2002).

30. Sanders, M. A. & Salisbury, J. L. Centrin plays an essential role in microtubule sever-ing during flagellar excision in Chlamydomonas reinhardtii. J. Cell Biol. 124,795–805 (1994).


centrosome number is controlled by a centrosome- intrinsic

Documents