Aurora A overexpression induces cellular senescence in mammary ...

Oncogene (2008) 27, 4305–4314

& 2008 Macmillan Publishers Limited All rights reserved 0950-9232/08 $30.00 www.nature.com/onc

ORIGINAL ARTICLE

Aurora A overexpression induces cellular senescence in mammary gland hyperplastic tumors developed in p53-deficient mice D Zhang1,6, T Shimizu1,7, N Araki1, T Hirota2, M Yoshie3, K Ogawa3, N Nakagata4, M Takeya5 and H Saya1,7 1

Department of Tumor Genetics and Biology, Graduate School of Medical Sciences, Kumamoto University, Kumamoto, Japan; Department of Experimental Pathology, Cancer Institute of Japanese Foundation for Cancer Research, Tokyo, Japan; 3Department of Pathology, Asahikawa Medical College, Asahikawa, Japan; 4Division of Reproductive Engineering, Center for Animal Resources and Development, Kumamoto University, Kumamoto, Japan and 5Department of Cell Pathology, Graduate School of Medical Sciences, Kumamoto University, Kumamoto, Japan 2

Aurora A mitotic kinase is frequently overexpressed in various human cancers and is widely considered to be an oncoprotein. However, the cellular contexts in which Aurora A induces malignancy in vivo are still unclear. We previously reported a mouse model in which overexpression of human Aurora A in the mammary gland leads to small hyperplastic changes but not malignancy because of the induction of p53-dependent apoptosis. To study the additional factors required for Aurora A-associated tumorigenesis, we generated a new Aurora A overexpression mouse model that lacks p53. We present evidence here that Aurora A overexpression in primary mouse embryonic fibroblasts (MEFs) that lack p53 overrides postmitotic checkpoint and leads to the formation of multinucleated polyploid cells. Induction of Aurora A overexpression in the mammary glands of p53-deficient mice resulted in development of precancerous lesions that were histologically similar to atypical ductal hyperplasia in human mammary tissue and showed increased cellular senescence and p16 expression. We further observed DNA damage in p53-deficient primary MEFs after Aurora A overexpression. Our results suggest that Aurora A overexpression in mammary glands is insufficient for the development of malignant tumors in p53-deficient mice because of the induction of cellular senescence. Both p53 and p16 are critical in preventing mammary gland tumorigenesis in the Aurora A overexpression mouse model. Oncogene (2008) 27, 4305–4314; doi:10.1038/onc.2008.76; published online 31 March 2008

Correspondence: Dr H Saya, Division of Gene Regulation, Institute for Advanced Medical Research, Keio University School of Medicine, 35 Shinano-machi, Shinjuku-ku, Tokyo 160-8582, Japan. E-mail: [email protected] 6 Current address: Breast Cancer Translational Research Laboratory, Department of Stem Cell Transplantation and Cellular Therapy, The University of Texas MD Anderson Cancer Center, Houston, TX, USA. 7 Current address: Division of Gene Regulation, Institute for Advanced Medical Research, Keio University School of Medicine, Tokyo, Japan. Received 18 April 2007; revised 30 January 2008; accepted 20 February 2008; published online 31 March 2008

Keywords: Aurora A; polyploidy; apoptosis; senescence; p16; DNA damage

Introduction Mitotic kinases such as cyclin-dependent kinase 1, pololike kinases, NIMA-related kinases, WARTS/LATS1related kinases and Aurora/Ipl1-related kinases have been widely reported to play crucial roles in regulating cell division and checkpoint functions. Dysfunction of some mitotic kinases leads to chromosome rearrangements and aneuploidy, which are hallmarks of many human cancers, indicating that mitotic kinases are important for maintaining genomic integrity (Nigg, 2001). Aurora A serine/threonine mitotic kinase has crucial roles in various mitotic events such as centrosome maturation and separation, mitotic entry, bipolar spindle assembly, chromosome alignment and cytokinesis (Meraldi et al., 2002; Hirota et al., 2003; Kunitoku et al., 2003; Marumoto et al., 2003, 2005; Giet et al., 2005). Many reports have presented evidence that Aurora A functions as an oncogene product. The human Aurora A gene is located at chromosome 20q13.2, a region frequently amplified in various human cancers (Kallioniemi et al., 1994). Human Aurora A gene amplification has been found in several human tumor cell types, including primary breast and colorectal cancers (Sen et al., 1997; Bischoff et al., 1998; Zhou et al., 1998). Aurora A mRNA and protein are frequently overexpressed in a range of cancer types, including breast, colorectal, gastric, ovarian and hepatocellular carcinoma (Tanaka et al., 1999; Sakakura et al., 2001; Gritsko et al., 2003; Jeng et al., 2004). Aurora A overexpression has been linked with centrosome amplification, aneuploidy and chromosome instability (Zhou et al., 1998; Cahill et al., 1999; Meraldi et al., 2002). Furthermore, ectopic overexpression of Aurora A efficiently transforms immortalized rodent fibroblasts (Bischoff et al., 1998; Zhou et al., 1998).

Aurora A overexpression in p53 KO mice D Zhang et al

4306

Breast cancer is one of the most common cancers and a leading cause of cancer death among women in the Western hemisphere (Howe et al., 2001). Evidence of Aurora A gene amplification and Aurora A overexpression in human breast cancer suggests that Aurora A may be involved in the pathogenesis of that cancer. Increases in endogenous Aurora A expression and centrosome number were also noted in a carcinogen-induced rat mammary tumor model, suggesting that Aurora A has a critical role during mammary tumor development and progression (Goepfert et al., 2002). However, the precise cellular conditions under which Aurora A expression promotes mammary tumorigenesis remain unclear. We previously used the Cre-loxP system to develop a CAG-CAT-Aurora A;Wap-Cre transgenic mouse model that conditionally overexpresses Aurora A in adult mammary glands (Zhang et al., 2004). Elevated Aurora A expression in these mice induces failure of cytokinesis, leading to the formation of binucleated and hyperploid cells in the mammary epithelium. Although small papillary hyperplasias were found in mammary glands overexpressing Aurora A, no tumorous lesions were identified. Given that Aurora A overexpression induced p53-dependent apoptosis in our model, we hypothesized that additional factors such as p53 inactivation may be required for the tumorigenesis of Aurora A-overexpressing mammary epithelium. To test this hypothesis, we generated CAG-CAT-Aurora A;Wap-Cre transgenic mice carrying heterozygous or null alleles for p53 and found that Aurora A overexpression can override the postmitotic checkpoint and lead to the formation of multinucleated polyploid cells in p53-deficient primary mouse embryonic fibroblasts (MEFs). These mice frequently developed tumorous hyperplasia in mammary glands, but lesions were considered nonmalignant and precancerous. Aurora A-expressing p53-deficient mammary gland cells also expressed high levels of senescence markers and p16INK4a (p16). Our findings suggest that p16 plays critical roles for preventing mammary gland malignancy in our p53-deficient Aurora A overexpression mouse model by inducing cellular senescence.

Results Wap-Cre-mediated Aurora A overexpression in the mammary glands of p53-deficient mice We previously reported generating CAG-CAT-Aurora A;Wap-Cre double-transgenic mice that conditionally overexpress human Aurora A in adult mammary glands by using the Cre-loxP recombination system under the control of the Wap promoter (Zhang et al., 2004). Even though Aurora A overexpression induced the formation of binucleated cells and papillary hyperplastic changes in mammary tissues, no malignant tumors were found because Aurora A had induced Oncogene

p53-dependent apoptosis. To further investigate the involvement of p53 in Aurora A-associated tumorigenesis, we generated CAG-CAT-Aurora A;Wap-Cre transgenic mice carrying heterozygous or null alleles for p53 (CAG-CAT-Aurora A;Wap-Cre;p53 þ / mice and CAG-CAT-Aurora A;Wap-Cre;p53/ mice, respectively). Immunohistochemical analyses confirmed the successful Wap-Cre-mediated human Aurora A overexpression in the mammary glands of both the p53-heterozygous and p53-null mice. Aurora A overexpression was not detected in control CAGCAT-Aurora A;p53 þ / or CAG-CAT-Aurora A;p53/ transgenic mice (Supplementary Figure S1). Aurora A overexpression leads to formation of multinucleated polyploid cells in p53-null primary MEFs We previously showed that induction of Aurora A overexpression in p53-wild-type (wt; þ / þ ) mice resulted in failure of cytokinesis, leading to significantly increased numbers of mammary gland cells with two nuclei (Zhang et al., 2004). To investigate the function of Aurora A in p53-deficient cells, we established primary MEFs from CAG-CAT-Aurora A;p53 þ / þ and CAG-CAT-Aurora A;p53/ transgenic mice and induced Aurora A overexpression in those MEFs in vitro by infection with AxCANCre (Kanegae et al., 1995), an adenovirus that encodes the Cre enzyme (Figure 1a). Cre-mediated Aurora A overexpression significantly increased the percentage of cells containing more than three nuclei in CAG-CAT-Aurora A;p53/ MEFs but not in CAG-CAT-Aurora A;p53 þ / þ MEFs (Figures 1b and c). Induction of multinucleated cells by Aurora A expression was further confirmed in vivo. Cells having three or more than three nuclei were frequently detected in mammary glands of p53/ mice, but rarely found in those of p53 þ / þ mice (Supplementary Figure S2). To determine whether those multinucleated polyploid cells could undergo further DNA synthesis, we used a bromodeoxyuridine (BrdU)-incorporation assay and found that the percentage of BrdU-positive polyploid cells was markedly increased when Aurora A was overexpressed in p53/ MEFs (13.6±1.52%), but only a small fraction of the polyploid cells were BrdU positive in p53 þ / þ MEFs (3.5±0.18%) (Figure 1d). These findings demonstrate that p53 deficiency allows further DNA synthesis in postmitotic polyploid MEFs produced by Aurora A overexpression. Therefore, the percentages of cells containing >4 N DNA content (>4 N DNA) (hyperploid cells and aneuploid cells) are significantly increased in Aurora A-overexpressing p53/ MEFs by comparison to the p53 þ / þ MEFs. Markedly increased number of centrosomes and abnormal spindle formation after Aurora A overexpression were also found in p53/ MEFs compared with p53 þ / þ MEFs (Supplementary Figure S3). These findings indicate that Aurora A overexpression on a p53-deficient background leads to the formation of hyperploid cells and aneuploid cells with aberrant cell division.


4307 CAG-CAT-Aurora-A;p53+/+ MEF

(-)

Luc

Cre

Aurora-A

CAG-CAT-Aurora-A;p53-/- MEF (-)

Luc

Cre

Percentage of multinucleated cells (%)

20 CAG-CAT-Aurora-A;p53+/+MEFs CAG-CAT-Aurora-A;p53 - / - MEFs

15

* 10 5

0

n=2

Aurora-A

n= >3

n=2

Luc

CAG-CATAurora-A;p53+/+

(-)

n= >3 Cre

CAG-CAT-Aurora-A;p53+/+

CAG-CATAurora-A;p53-/-

(-)

BrdU

3.55%

Luc

Luc DNA content

CAG-CAT-Aurora-A;p53-/-

Cre

BrdU

13.58%

Cre

DNA content

Figure 1 Aurora A overexpression in p53-deficient mouse embryonic fibroblasts (MEFs) results in the formation of multinucleated polyploid cells. (a) Primary MEFs from CAG-CAT-Aurora A;p53 þ / þ mice (upper panel) and CAG-CAT-Aurora A;p53/ mice (lower panel) were infected with adenoviral luciferase (Luc) or AxCANCre (Cre) virus (which induces Aurora A expression). Western blotting confirmed that Aurora A overexpression appeared 24 h after Cre infection. (), uninfected. (b) Primary MEFs from CAGCAT-Aurora A;p53 þ / þ and CAG-CAT-Aurora A;p53/ mice were infected with adenoviral luciferase or AxCANCre and stained 72 h later with anti-a-tubulin antibodies (green) and propidium iodide (PI) to label the DNA (red). Binucleated cells and multinucleated cells (those with X3 nuclei) induced by Aurora A overexpression are indicated by arrows and arrowheads, respectively. Scale bar, 20 mm. (c) Quantification of nuclei in primary MEFs prepared from CAG-CAT-Aurora A;p53 þ / þ and CAG-CAT-Aurora A;p53/ mice at 72 h after virus infection (Luc or Cre). Values shown are the mean percentages of binucleated or multinucleated cells (with X3 nuclei) from three independent experiments in which >300 cells were counted. *Po0.01. (d) Flow cytometry of MEFs with overexpressed Aurora A. Primary MEFs from CAG-CAT-Aurora A;p53 þ / þ and CAG-CAT-Aurora A;p53/ mice were treated with bromodeoxyuridine (BrdU) for 4 h at 48 h after infection with AxCANCre. The percentages of BrdU-positive polyploid (DNA>4N) cells after Aurora A overexpression were 3.55% in CAG-CAT-Aurora A;p53 þ / þ MEFs and 13.58% in CAG-CATAurora A;p53/ MEFs.

Aurora A overexpression leads to precancerous lesions in p53-deficient mice To investigate whether Aurora A overexpression leads to the formation of mammary tumors, female CAGCAT-Aurora A;Wap-Cre;p53 þ / and CAG-CAT-Aurora A;Wap-Cre;p53/ mice were mated with male C57BL/6 mice and monitored weekly for the development of mammary tumors. In total 5 of 21 CAG-CAT-Aurora A;Wap-Cre;p53 þ / mice and 5 of

11 CAG-CAT-Aurora A;Wap-Cre;p53/ mice developed precancerous lesions that were histologically similar to atypical ductal hyperplasia (ADH) in human mammary tissue (Figure 2b). Hematoxylin and eosin (H&E) staining of the lesions showed increased numbers of abnormal cells within ducts or lobules and disruption of the dual-layer architecture of the mammary gland. However, these lesions lacked characteristics of invasive carcinoma such as cellular and nuclear pleomorphisms Oncogene


4308

CAG-CAT-Aurora-A;Wap-Cre;p53-/-

CAG-CAT-Aurora-A;p53-/-

a

H&E

b

H&E

c

Ki67

d

Aurora-A

e

Aurora-A

Figure 2 Wap-Cre-mediated Aurora A overexpression leads to lesions resembling atypical ductal hyperplasia (ADH) in mammary glands of p53-deficient mice. (a) Hematoxylin and eosin staining (H&E) of a paraffin-embedded mammary gland section from a CAGCAT-Aurora A;p53/ mouse. (b) H&E staining of a mammary gland section from a CAG-CAT-Aurora A;Wap-Cre;p53/ mouse showing ADH-like hyperplastic lesion. (c) Immunohistochemical analysis of Ki67 expression in ADH-like hyperplastic lesion from tissue in (b). (d) Immunohistochemical analysis of an ADH-like lesion in mammary gland of a CAG-CAT-Aurora A;Wap-Cre;p53/ mouse shows high expression of exogenous human Aurora A. (e) Immunohistochemical analysis of a lymphoma in the mammary gland of a CAG-CAT-Aurora A;Wap-Cre;p53/ mouse shows that lymphoma cells do not express human Aurora A. Scale bars, 50 mm.

and disruption of basement membrane. In contrast, no ADH-like lesions were found in 20 CAG-CAT-Aurora A;p53 þ / mice or in 11 CAG-CAT-Aurora A;p53/ mice, which do not express Aurora A (Figure 2a). Although our previous study showed that CAG-CATAurora A;Wap-Cre;p53 þ / þ mice developed small papillary hyperplasia, none of them had ADH-like tumors in their mammary glands (Table 1). Additionally, immunohistochemical analysis revealed that approximately 8% of cells in these ADH-like lesions expressed the proliferation marker Ki67 (Figure 2c). However, in CAG-CAT-Aurora A;Wap-Cre;p53 þ / þ mice, Ki67-positive cells were rarely detected (o1%) (Supplementary Figure S4). These data indicate that Aurora A overexpression induces cell proliferation in p53/ mammary glands more than in p53 þ / þ mammary glands, while not as aggressive as malignant tumors. Exogenous Aurora A levels were high in the ADH-like tumor lesions and adjacent normal mammary epithelium (Figure 2d), but lymphoma and sarcoma lesions (which often develop in p53/ mice) did not express exogenous Aurora A (Figure 2e). Taken together, these results suggest that Aurora A overexpression can induce benign tumors in mammary glands when p53 function are impaired, but it is not sufficient for the development of invasive malignant tumors. Cellular senescence in Aurora A-overexpressing mammary glands Apoptosis and senescence are two cellular responses that prevent the transformation of normal cells into malignant Oncogene

Table 1 Incidence of ADH-like tumor formation in the mammary glands of Aurora A conditional transgenic mice with different p53 background Mouse genotype

CAG-CAT-Aurora CAG-CAT-Aurora CAG-CAT-Aurora CAG-CAT-Aurora CAG-CAT-Aurora a

A;Wap-Cre;p53+/+ A;Wap-Cre;p53+/ A;Wap-Cre;p53/ A;p53+/ A;p53/

Tumor incidence, No. of mice (%)

Age of mice (days±s.d.)

0 of 14 (0) 5 of 21 (24) 5 of 11 (45) 0 of 20 (0) 0 of 11 (0)

512±63.2 354±71.2 137±28.9a 357±96.8 147±38.0a

Most p53/ mice died of lymphoma or sarcoma within 200 days.

cells under oncogenic stimulation. Given our previous observation that overexpression of Aurora A in vivo did not lead to malignancy in p53-deficient mice in which apoptosis was significantly inhibited, we speculated that cellular senescence was important in the prevention of malignant transformation in our mouse model. Thus, we investigated the status of apoptosis and senescence in Aurora A-overexpressing mammary glands. TdTmediated dUTP-biotin nick-end labeling (TUNEL) and the active caspase-3 staining revealed that there are many apoptotic cells in CAG-CAT-Aurora A;Wap-Cre;p53 þ / þ mice, while CAG-CAT-Aurora A;Wap-Cre;p53/ mice had few apoptotic cells (Figure 3a; Supplementary Figure S5), which was consistent with our previous study (Zhang et al., 2004). Endogenous senescence-associated bgalactosidase (SA-b-gal) activity, a well established


4309

p53 þ / þ mice (Figures 3a and b). Senescent cells are also known to exhibit distinct senescence-associated heterochromatic foci containing heterochromatin proteins such as HP1 (Narita et al., 2003); they also upregulated expression of proteins such as Decoy receptor 2 (DcR2), a new oncogene-induced senescence

marker of cellular senescence (Dimri et al., 1995), was found in response to Wap-Cre-mediated Aurora A overexpression in the mammary glands of both p53 þ / þ and p53/ mice, but both the amount of SA-b-gal activity and the numbers of SA-b-gal-positive cells in the p53/ mice were much higher than those in

a

CAG-CAT-Aurora-A; Wap-Cre;p53+/+

CAG-CAT-Aurora-A; Wap-Cre;p53-/-

H&E

H&E

c

CAG-CAT-Aurora-A; Wap-Cre;p53+/+


DcR2

DcR2

PI

PI

Merge

Merge

30

20 TUNEL

TUNEL

10

(%) 20

SA-β-gal

*

d Percentage of DcR2 positive cells

b Percentage of SA-β-gal positive cells

β-gal SA-β

15

10

5

0 CAG-CAT-Aurora-A; Wap-Cre;p53+/+


(%)

*

10 8 6 4 2 0 CAG-CAT-Aurora-A; Wap-Cre;p53+/+


Figure 3 Mammary epithelial cells overexpressing Aurora A are positive for markers of cellular senescence. (a) Hematoxylin and eosin staining (H&E), TdT-mediated dUTP-biotin nick-end labeling (TUNEL) and senescence-associated b-galactosidase (SA-b-gal) assay were performed on frozen sections from mammary glands of CAG-CAT-Aurora A;Wap-Cre;p53 þ / þ (left panels) and CAGCAT-Aurora A;Wap-Cre;p53/ (right panels) mice. Cells from the p53/ mice showed less apoptosis and more SA-b-gal activity than the p53 þ / þ mice. Scale bar, 100 mm. (b) Quantification of SA-b-gal-positive cells in mammary glands of CAG-CAT-Aurora A;Wap-Cre;p53 þ / þ and CAG-CAT-Aurora A;Wap-Cre;p53/ mice. Values shown are the mean percentages of SA-b-gal-positive cells from three animals per experimental group in which over 300 cells were counted. The percentage of SA-b-gal-positive cells in p53/ mice is much higher than those in p53 þ / þ mice (15.7±3.06 vs 2.0±1.00%) (*Po0.05). (c) Immunofluorescent staining of Decoy receptor 2 (DcR2) expression. Mammary glands from a CAG-CAT-Aurora A;Wap-Cre;p53 þ / þ mouse (left panels) and a CAG-CAT-Aurora A;Wap-Cre;p53/ mouse (right panels) were stained with anti-DcR2 antibody (green) and DNA was visualized with propidium iodide (PI) (red). Scale bar, 20 mm. (d) Quantification of DcR2-positive cells in CAG-CAT-Aurora A;Wap-Cre;p53 þ / þ and CAG-CAT-Aurora A;Wap-Cre;p53/ mice. Values shown are the mean percentages of DcR2-positive cells from three animals per experimental group in which over 300 cells were counted. The percentage of DcR2-positive cells in p53/ mice is significantly higher than those in p53 þ / þ mice (7.58±0.42 vs 1.15±0.44%) (*Po0.05). Oncogene


4310

marker (Collado et al., 2005). Thus we performed immunofluorescent staining for HP1-g (Supplementary Figure S6) and DcR2 in Aurora A-overexpressing mammary gland tissues and found that the numbers of positive cells were significantly high in the mammary glands of p53/ mice only when Aurora A was overexpressed (Figures 3c and d). All these findings provide in vivo evidence that Aurora A overexpression induces cellular senescence in a p53-deficient background. Increased p16 expression in Aurora A-overexpressing mammary glands The most compelling link between cellular senescence and tumor suppression is their mutual dependence on tumor suppressor genes such as p16, p21, p53 and retinoblastoma (Rb) (Bringold and Serrano, 2000). The Rb/p16 and p53/p21 tumor suppressor pathways in particular are important regulators of cellular senescence (Ferbeyre et al., 2002). Therefore, we hypothesized that the Rb/p16 pathway may be involved in Aurora A-induced senescence in a p53-deficient background. To test this hypothesis, we performed immunofluorescence analysis of the DcR2 senescence marker and p16 by using serial tissue sections as shown in Figures 3c and 4a. The area having p16-positive cells (Figure 4a, lower panels) contained DcR2-positive cells (Figure 3c, right panels), suggesting that p16 positive cells have potential to become senescent. Immunohistochemical analysis also indicated upregulation of p16 expression in the ADH-like tumor lesions of CAG-CAT-Aurora A;WapCre;p53/ mice (Figure 4b). Aurora A overexpression induces DNA damage in p53-deficient MEFs Senescence is considered a cellular response to stress that limits the proliferation of damaged cells (Campisi, CAG-CAT-Aurora-A; p53 -/p16

PI

PI

Discussion We discovered several new findings in p53/ mice after Aurora A overexpression in the present study in comparison with p53 þ / þ mice in our previous study. Differences in the response of p53 þ / þ and p53/ mice are (1) significantly different induction of cellular senescence, as judged by the staining of three senescence markers, SA-b-gal activity, HP1-g and DcR2. The proportion of SA-b-gal-positive cells in p53/ mice showed a sevenfold increase compared with cells in p53 þ / þ mice, indicating that cellular senescence of the mammary epithelium is hypothesized to be an important contributor to prevent the development of malignant tumors by Aurora A in p53/ mice; (2) different pathological changes between p53 þ / þ and p53/ mice. Induction of Aurora A overexpression in p53/ mice resulted in development of precancerous lesions that were histologically similar to ADH, a precancerous CAG-CAT-Aurora-A; p53-/-

Merge

CAG-CAT-Aurora-A; Wap-Cre;p53 -/p16

2001). It can be provoked by DNA damage, oxidative stress and activated oncogenic ras (Lundberg et al., 2000). To investigate the mechanism that induces cellular senescence after Aurora A overexpression, we assessed DNA damage by monitoring the presence of phosphorylated histone H2AX (g-H2AX). g-H2AXpositive foci were detected in primary MEFs after Aurora A overexpression (Figure 5a); substantially more of those foci were present in CAG-CAT-Aurora A;p53/ MEFs after Aurora A overexpression than in CAG-CAT-Aurora A;p53 þ / þ MEFs (Figure 5b). These findings indicate that Aurora A overexpression elicits DNA damage when p53 functions are impaired, which is one potential mechanism for induction of cellular senescence.


Merge

Figure 4 Expression of p16 is upregulated in p53-deficient Aurora A-overexpressing tissues. (a) Immunofluorescent staining of p16 expression. Mammary glands from the CAG-CAT-Aurora A;p53/ (upper panels) and CAG-CAT-Aurora A;Wap-Cre;p53/ mice (lower panels) shown in Figure 3c were immunostained with anti-p16 antibody (green, left); DNA was visualized with propidium iodide (PI) (red, middle). Scale bar, 20 mm. (b) Immunohistochemical analysis of p16 in the mammary gland of CAG-CAT-Aurora A;p53/ and CAG-CAT-Aurora A;Wap-Cre;p53/ mice. p16 expression was high in hyperplastic lesions of the CAG-CAT-Aurora A;Wap-Cre;p53/ mouse but not in the mammary gland cells of the CAG-CAT-Aurora A;p53/ mouse. Scale bar, 50 mm. Oncogene


4311 (-)

CAG-CAT-Aurora-A;p53 -/- MEFs

CAG-CAT-Aurora-A;p53+/+ MEFs Luc

b

Luc

(-)

Cre

Relative foci number per cell

a

Cre

40 30 20 10 0

(-)

Luc

Cre

CAG-CAT-Aurora-A;p53+/+ MEFs CAG-CAT-Aurora- A ;p53 -/- MEFs Figure 5 Aurora A overexpression induces DNA damage in p53-deficient mouse embryonic fibroblasts (MEFs). (a) Immunofluorescent analysis of primary MEFs from CAG-CAT-Aurora A;p53 þ / þ and CAG-CAT-Aurora A;p53/ mice with an anti-g-H2AX antibody at 48 h after infection with adenoviral luciferase (Luc) or AxCANCre (Cre) virus. (), uninfected. Nuclei are outlined in blue (upper panels). Double staining with anti-g-H2AX and propidium iodide (PI) in representative cells from the upper panels are shown in the lower panels. (b) Quantification of g-H2AX-positive foci per cell after infection with adenoviral Luc or Cre from (a). Values shown are the means from three independent experiments in which 50 interphase cells were counted.

lesion in human mammary tissue. In our previous study, Aurora A overexpression only induced small papillary hyperplasia without ADH in p53 þ / þ mice and (3) upregulated p16 expression in p53/ mice after Aurora A overexpression, indicating that p16 is hypothesized to be critical in preventing mammary gland malignancy by inducing senescence. The presence of Aurora A overexpression in various human cancers suggests that Aurora A may function as an oncogene; however, even though considerable evidence indicates that Aurora A overexpression provokes mitotic abnormalities and chromosome instability, to date no evidence has been presented proving that it leads to malignancy in vivo. Indeed, several groups have reported that Aurora A alone can not induce malignancy. Anand et al. (2003) reported that overexpression of Aurora A could not transform primary cells. Fukuda et al. (2005) demonstrated that a conditional transgenic system for mouse Aurora A showed no abnormalities during mouse development, possibly because the Aurora A protein is unstable and readily degraded in normal mouse tissues. In our present study, Aurora A was stably expressed but did not lead to the development of malignant tumors because of the induction of cellular senescence. In contrast, Wang et al. (2006) reported recently that overexpression of

Aurora A could cause mammary tumor formation through inducing genetic instability and activating AKT. The discrepancy between their results and ours may stem from the use of different mouse strains. The FVB/N mice used by Wang et al. (2006) are highly susceptible to chemically induced squamous cell carcinomas, with a high rate of malignant conversion from papilloma to carcinoma. The different genetic backgrounds of this mouse strain, perhaps, impairment of factors associated with senescence and apoptosis, could result in different outcome of tumorigenesis. Tumorigenesis in humans is a multistep process, and genetic alterations including inactivation of tumor suppressor genes and activation of oncogenes lead to the progressive conversion of normal cells into cancer cells (Hanahan and Weinberg, 2000). The p53 tumor suppressor gene acts to reduce the incidence of cancer by regulating genes involved in cell cycle arrest, apoptosis, and other pathways (Levine, 1997). p53-mediated apoptosis is an important part of the tumor suppressor phenotype (Symonds et al., 1994). Loss of p53 function not only facilitates resistance to apoptosis but also induces genomic instability (Hanahan and Weinberg, 2000). In tetraploid cells resulting from mitotic abnormalities such as failure of mitotic spindle assembly or cytokinesis, the p53-dependent postmitotic checkpoint is Oncogene


4312

activated to induce G1 arrest and apoptosis. However, in p53-deficient cells, tetraploidy caused by mitotic failures leads to the propagation of errors of late mitosis and to the generation of aneuploidy (Andreassen et al., 2001). Tetraploidy enhances the occurrence of chromosomal alterations and promotes tumor development in p53null cells (Fujiwara et al., 2005). In this study, we found that Aurora A overexpression led to mitotic aberrations, including increased numbers of centrosomes and hyperploidy, which consequently generate aneuploid cells, in p53-deficient MEFs. However, in our p53-null model, Aurora A overexpression induced benign lesions but not malignant ones. These findings provide important evidence that overexpression of Aurora A can promote hyperproliferative precancerous lesions when p53 functions are impaired but that additional factors are required for malignant transformation. Cellular senescence is a state of permanent growth arrest provoked by a variety of stresses (Lundberg et al., 2000). Expression of oncogenic ras in primary human or rodent cells results in permanent G1 arrest and provokes premature cell senescence (Serrano et al., 1997). A new study using a mouse model of cancer initiation showed that senescent cells exist in premalignant tumors but not in malignant ones, providing circumstantial evidence that oncogene-induced senescence acts to counter tumorigenesis (Collado et al., 2005). Our mouse model suggests that both senescence and apoptosis seem to act as tumor suppression mechanisms that prevent the further evolution of preneoplastic cells in response to Aurora A dysregulation. Some signaling pathways have been shown to promote apoptosis in one cell type but to induce senescence in others; for example, the transcription factor E2F-1 cooperates with p53 to mediate apoptosis in murine cells but induces senescence when overexpressed in normal human fibroblasts (Wu and Levine, 1994; Dimri et al., 2000). We illustrate here that oncogenic Aurora A can activate both the p53/p21 and the Rb/p16 pathways to induce both apoptosis and senescence programs and that the two programs seem to serve as backups for each other. Cellular senescence can be also triggered by various types of DNA damage or chromatin remodeling in addition to oncogenic forms of ras or raf (Chen et al., 1995; Serrano et al., 1997; Zhu et al., 1998). Much evidence exists supporting a link between DNA damage, p16 expression and cellular senescence. For example, in normal human fibroblasts and tumor cells, DNA double-strand breaks lead to increased p16 expression and senescence (Robles and Adami, 1998; te Poele et al., 2002). To examine whether p16 expression can be a major factor to induce senescence in Aurora A overexpressing cells, we infected the immortalized mouse fibroblasts derived from CAG-CAT-Aurora A mice, which do not express p16, with p16-expressing adenovirus and/or AxCANCre virus, which induces Aurora A expression and tested senescence of those cells with an SA-b-gal assay. While no SA-b-gal-positive cells were detected among immortalized mouse fibroblasts infected with either the p16-expressing adenovirus or the AxCANCre virus, >20% of the fibroblasts co-expresOncogene

sing p16 and Aurora A underwent senescence (data not shown). These findings suggest that deregulated Aurora A promotes cellular senescence in a p16-dependent manner. We speculate that induction of DNA damage by Aurora A could be the mechanism by which Aurora A induces cellular senescence. Possible mechanisms of DNA damage induction might involve the aberrant cell cycle progression mediated by Aurora A overexpression. Further, given that Aurora A is overexpressed not only in mitotic cells but also in interphase cells, it might activate aberrant signaling pathways involved in DNA synthesis and repair. Another possible mechanism is the induction of oxidative stress, a major cause of DNA damage and genetic instability. Recent studies have proposed that activation of oncogenes such as c-Myc and ras alters biochemical pathways to produce reactive oxygen species that lead to DNA damage (Lee et al., 1999; Vafa et al., 2002). Overexpressed Aurora A may have a similar effect. p53 was recently reported to protect the genome from oxidation by reactive oxygen species; downregulation of p53 resulted in excessive oxidation of the DNA (Sablina et al., 2005). We found that the number of g-H2AX-positive foci was significantly higher in p53-deficient MEFs than in p53-wildtype MEFs after Aurora A overexpression. Our results provide new evidence that Aurora A overexpression induces massive DNA damage in p53-deficient cells that can override postmitotic G1 arrest, indicating that p53 might protect the genome from the oxidative stress induced by Aurora A overexpression. Our future studies will focus on illustrating the mechanism by which Aurora A overexpression induces DNA damage in p53-deficient cells. One can speculate that p53 prevents Aurora A-mediated oxidative stress-induced DNA damage that would otherwise trigger p16 expression and cellular senescence. Alternatively, disruption of both p53/p21 and p16/Rb pathways may be necessary for neoplastic transformation by Aurora A. We still do not know by which pathway p16 induces cellular senescence in our model. p16 specifically inhibits cyclin-dependent kinase 4 (cdk4) and cdk6-mediated G1/S progression and therefore limiting cdk4/6-Cyclin D complex formation and phosphorylation of pRb. Narita et al. (2003) reported that the formation of heterochromatic foci depends on the Rb pathway in senescent cells. Given that we found HP1-g expression in p53-deficient tissue in our model, the Rb pathway may be involved in the ability of p16 to promote senescence. The Rb2/p130 protein, which is a member of the Rb family, is also reported to mediate the senescent growth arrest (Helmbold et al., 2006). Similar to pRb, Rb2/p130 is hyperphosphorylated by cyclin D/cdk4 or 6 and cyclin E or A/cdk2 complexes and considered acting as a master regulator of pRb in a sustained senescent arrest. We are working on identifying the status of phosphorylated pRb and Rb2 in our model to confirm whether these factors are responsible for the induction of senescence. Other future studies will include our overexpressing Aurora A in the mammary glands of INK4a/Arf-null mice reported by Serrano et al. (1996). These mice,


4313

which carry a targeted deletion of the INK4a locus that eliminates both p16INK4a and p19ARF, are viable but prone to spontaneous and carcinogen-induced tumors (Serrano et al., 1996); moreover, the introduction of oncogenic ras into INK4a-deficient fibroblasts resulted in neoplastic transformation. We speculate that polyploidy associated with disruption of both p53/p21 and p16/Rb pathways may be necessary for neoplastic transformation by Aurora A.

Materials and methods Generation of transgenic mice We previously generated CAG-CAT-Aurora A transgenic mice (strain C57BL/6) carrying a pCAG-loxP-CAT-loxP-Aurora A construct designed to induce human Aurora A overexpression by Cre-mediated recombination; we also generated CAG-CATAurora A;Wap-Cre double-transgenic mice by mating CAGCAT-Aurora A mice with Wap (whey acidic protein)-Cre mice (Zhang et al., 2004). CAG-CAT-Aurora A;Wap-Cre;p53 þ / transgenic mice, which carry one allele of p53, and CAG-CATAurora A;Wap-Cre;p53/ transgenic mice, which carry neither p53 allele, were typically generated from crosses of CAG-CATAurora A;p53 þ / mice and Wap-Cre;p53 þ / mice. For the study reported here, CAG-CAT-Aurora A;Wap-Cre;p53 þ /, CAG-CAT-Aurora A;Wap-Cre;p53/ and littermate control female mice were mated at 8 weeks of age and housed with male mice thereafter to induce Aurora A expression. Animals were examined weekly for evidence of tumor development. Generation and culture of MEFs and immunofluorescence and flow cytometry analyses Primary MEFs were generated from mouse embryos at E13.5 days and cultured in Dulbecco’s modified Eagle’s medium with 10% fetal bovine serum. Early passage MEFs (p2–3) were used at the initiation of all experiments. Immunofluoresence staining was performed as follows: cells were fixed with 4% paraformaldehyde for 20 min, permeabilized with 0.2% Triton X-100 in Tris-buffered saline (TBS) for 5 min, blocked for 1 h with 5% bovine serum albumin in TBS, and incubated with primary antibodies in 0.3% bovine serum albumin in TBS. Primary antibodies were mouse monoclonal anti-a-tubulin antibody (B-5-1-2; 1:1000, Sigma-Aldrich, St Louis, MO, USA); rabbit polyclonal anti-phosphorylated histone H2AX antibody (1:250; Trevigen, Gaithersburg, MD, USA); and rabbit polyclonal anti-pericentrin antibody (1:1000; Covance, Denver, PA, USA). Fluorescein isothiocyanate (FITC)-conjugated antibodies (Biosource, Carlsbad, CA, USA) were used as secondary antibodies. Cells were counterstained with propidium iodide (PI) before being mounted under glass coverslips and analysed by confocal microscopy (FV300, Olympus). For flow cytometric analysis, MEFs were infected by adenovirus for 48 h, after which BrdU (10 ng ml1) was added in culture medium and the MEFs incubated for further 4 h. Cells were then fixed with 70% ethanol, stained with antiBrdU-FITC and PI, and subjected to flow cytometry according to the manufacturer’s instructions (Becton Dickinson, Franklin Lakes, NJ, USA).

Histologic and immunohistochemical, and immunofluoresence analyses For conventional histologic analysis, mammary glands were fixed in 10% phosphate-buffered formaldehyde for 24 h and embedded in paraffin. Sections (3 mm thick) were stained with H&E using standard techniques. For cryosections, tissues were fixed with 4% paraformaldehyde for 6 h, and then transferred to a 30% sucrose solution for 24 h before 10-mm-thick sections were prepared. Immunohistochemical and immunofluoresent analyses were performed as previously described (Zhang et al., 2004). Primary antibodies were rabbit polyclonal anti-Aurora A (1:100; TransGenic, Kumamoto, Japan); rat monoclonal anti-E-cadherin (ECCD2, 1:1000; Takara, Otsu, Japan); rabbit monoclonal anti-Ki67 (SP6, 1:200; NeoMarkers, Fremont, CA, USA); rabbit polyclonal anti-active caspase 3 (1:100; Chemicon, Temecula, CA, USA); mouse monoclonal antiHP1-g (1:1000; Chemicon); rabbit polyclonal anti-DcR2 (1:250; Stressgen, Ann Arbor, MI, USA) and rabbit polyclonal anti-p16 (M-156, 1:200; Santa Cruz Biotechnology, Santa Cruz, CA, USA). SA-b-gal and TUNEL assays Cryosections (10 mm thick) were subjected to SA-b-gal and TUNEL assays as follows. SA-b-gal activity was detected with a Senescence b-Galactosidase Staining Kit according to the manufacturer’s instructions (Cell Signaling). Apoptotic cells in mammary tissues were detected by TUNEL with an In situ Apoptosis Detection Kit (TaKaRa). Adenovirus infection Primary MEFs were infected with adenovirus at multiplicities of infection of at least 20 for 1 h, after which the virus solution was exchanged for fresh medium. Infection efficiency under these conditions is normally >95%. Statistical analysis Statistical analyses were performed using commercially available software (Statview, version 5.0, SAS Institute, Cary, NC, USA). The data obtained were analysed by t-test. Values represent the mean±s.e. Statistical significance was defined as Po0.05. Acknowledgements We thank Dr Kimi Araki (Kumamoto University) for providing pCAG-CAT-lacZ plasmid; Mr Takenobu Nakagawa (Kumamoto University) for technical assistance; Dr Izumu Saito and Dr Yumi Kanegae (University of Tokyo) for providing adenoviral luciferase, AxCANCre and p16-expressing adenovirus; Mrs Christine F Wogan (The University of Texas MD Anderson Cancer Center) and Dr Sampetrean Oltea (Kumamoto University) for editorial assistance; members of the Saya laboratory for valuable suggestions and members of the Gene Technology Center at Kumamoto University for their technical assistance. This work was supported by a grant for Cancer Research from the Ministry of Education, Culture, Sports, Science, and Technology of Japan (to HS).

References Anand S, Penrhyn-Lowe S, Venkitaraman AR. (2003). AURORA-A amplification overrides the mitotic spindle assembly checkpoint, inducing resistance to Taxol. Cancer Cell 3: 51–62.

Andreassen PR, Lohez OD, Lacroix FB, Margolis RL. (2001). Tetraploid state induces p53-dependent arrest of nontransformed mammalian cells in G1. Mol Biol Cell 12: 1315–1328. Oncogene


4314 Bischoff JR, Anderson L, Zhu Y, Mossie K, Ng L, Souza B et al. (1998). A homologue of Drosophila aurora kinase is oncogenic and amplified in human colorectal cancers. EMBO J 17: 3052–3065. Bringold F, Serrano M. (2000). Tumor suppressors and oncogenes in cellular senescence. Exp Gerontol 35: 317–329. Cahill DP, Kinzler KW, Vogelstein B, Lengauer C. (1999). Genetic instability and darwinian selection in tumours. Trends Cell Biol 9: M57–M60. Campisi J. (2001). Cellular senescence as a tumor-suppressor mechanism. Trends Cell Biol 11: S27–S31. Chen Q, Fischer A, Reagan JD, Yan LJ, Ames BN. (1995). Oxidative DNA damage and senescence of human diploid fibroblast cells. Proc Natl Acad Sci USA 92: 4337–4341. Collado M, Gil J, Efeyan A, Guerra C, Schuhmacher AJ, Barradas M et al. (2005). Tumour biology: senescence in premalignant tumours. Nature 436: 642. Dimri GP, Itahana K, Acosta M, Campisi J. (2000). Regulation of a senescence checkpoint response by the E2F1 transcription factor and p14(ARF) tumor suppressor. Mol Cell Biol 20: 273–285. Dimri GP, Lee X, Basile G, Acosta M, Scott G, Roskelley C et al. (1995). A biomarker that identifies senescent human cells in culture and in aging skin in vivo. Proc Natl Acad Sci USA 92: 9363–9367. Ferbeyre G, de Stanchina E, Lin AW, Querido E, McCurrach ME, Hannon GJ et al. (2002). Oncogenic ras and p53 cooperate to induce cellular senescence. Mol Cell Biol 22: 3497–3508. Fujiwara T, Bandi M, Nitta M, Ivanova EV, Bronson RT, Pellman D. (2005). Cytokinesis failure generating tetraploids promotes tumorigenesis in p53-null cells. Nature 437: 1043–1047. Fukuda T, Mishina Y, Walker MP, DiAugustine RP. (2005). Conditional transgenic system for mouse aurora a kinase: degradation by the ubiquitin proteasome pathway controls the level of the transgenic protein. Mol Cell Biol 25: 5270–5281. Giet R, Petretti C, Prigent C. (2005). Aurora kinases, aneuploidy and cancer, a coincidence or a real link? Trends Cell Biol 15: 241–250. Goepfert TM, Adigun YE, Zhong L, Gay J, Medina D, Brinkley WR. (2002). Centrosome amplification and overexpression of aurora A are early events in rat mammary carcinogenesis. Cancer Res 62: 4115–4122. Gritsko TM, Coppola D, Paciga JE, Yang L, Sun M, Shelley SA et al. (2003). Activation and overexpression of centrosome kinase BTAK/ Aurora A in human ovarian cancer. Clin Cancer Res 9: 1420–1426. Hanahan D, Weinberg RA. (2000). The hallmarks of cancer. Cell 100: 57–70. Helmbold H, Deppert W, Bohn W. (2006). Regulation of cellular senescence by Rb2/p130. Oncogene 25: 5257–5262. Hirota T, Kunitoku N, Sasayama T, Marumoto T, Zhang D, Nitta M et al. (2003). Aurora A and an interacting activator, the LIM protein Ajuba, are required for mitotic commitment in human cells. Cell 114: 585–598. Howe HL, Wingo PA, Thun MJ, Ries LA, Rosenberg HM, Feigal EG et al. (2001). Annual report to the nation on the status of cancer (1973 through 1998), featuring cancers with recent increasing trends. J Natl Cancer Inst 93: 824–842. Jeng YM, Peng SY, Lin CY, Hsu HC. (2004). Overexpression and amplification of Aurora A in hepatocellular carcinoma. Clin Cancer Res 10: 2065–2071. Kallioniemi A, Kallioniemi OP, Piper J, Tanner M, Stokke T, Chen L et al. (1994). Detection and mapping of amplified DNA sequences in breast cancer by comparative genomic hybridization. Proc Natl Acad Sci USA 91: 2156–2160. Kanegae Y, Lee G, Sato Y, Tanaka M, Nakai M, Sakaki T et al. (1995). Efficient gene activation in mammalian cells by using recombinant adenovirus expressing site-specific Cre recombinase. Nucleic Acids Res 23: 3816–3821. Kunitoku N, Sasayama T, Marumoto T, Zhang D, Honda S, Kobayashi O et al. (2003). CENP-A phosphorylation by Aurora A in prophase is required for enrichment of Aurora-B at inner centromeres and for kinetochore function. Dev Cell 5: 853–864.

Lee AC, Fenster BE, Ito H, Takeda K, Bae NS, Hirai T et al. (1999). Ras proteins induce senescence by altering the intracellular levels of reactive oxygen species. J Biol Chem 274: 7936–7940. Levine AJ. (1997). p53, the cellular gatekeeper for growth and division. Cell 88: 323–331. Lundberg AS, Hahn WC, Gupta P, Weinberg RA. (2000). Genes involved in senescence and immortalization. Curr Opin Cell Biol 12: 705–709. Marumoto T, Honda S, Hara T, Nitta M, Hirota T, Kohmura E et al. (2003). Aurora A kinase maintains the fidelity of early and late mitotic events in HeLa cells. J Biol Chem 278: 51786–51795. Marumoto T, Zhang D, Saya H. (2005). Aurora A—a guardian of poles. Nat Rev Cancer 5: 42–50. Meraldi P, Honda R, Nigg EA. (2002). Aurora A overexpression reveals tetraploidization as a major route to centrosome amplification in p53/ cells. EMBO J 21: 483–492. Narita M, Nunez S, Heard E, Lin AW, Hearn SA, Spector DL et al. (2003). Rb-mediated heterochromatin formation and silencing of E2F target genes during cellular senescence. Cell 113: 703–716. Nigg EA. (2001). Mitotic kinases as regulators of cell division and its checkpoints. Nat Rev Mol Cell Biol 2: 21–32. Robles SJ, Adami GR. (1998). Agents that cause DNA double strand breaks lead to p16INK4a enrichment and the premature senescence of normal fibroblasts. Oncogene 16: 1113–1123. Sablina AA, Budanov AV, Ilyinskaya GV, Agapova LS, Kravchenko JE, Chumakov PM. (2005). The antioxidant function of the p53 tumor suppressor. Nat Med 11: 1306–1313. Sakakura C, Hagiwara A, Yasuoka R, Fujita Y, Nakanishi M, Masuda K et al. (2001). Tumour-amplified kinase BTAK is amplified and overexpressed in gastric cancers with possible involvement in aneuploid formation. Br J Cancer 84: 824–831. Sen S, Zhou H, White RA. (1997). A putative serine/threonine kinase encoding gene BTAK on chromosome 20q13 is amplified and overexpressed in human breast cancer cell lines. Oncogene 14: 2195–2200. Serrano M, Lee H, Chin L, Cordon-Cardo C, Beach D, DePinho RA. (1996). Role of the INK4a locus in tumor suppression and cell mortality. Cell 85: 27–37. Serrano M, Lin AW, McCurrach ME, Beach D, Lowe SW. (1997). Oncogenic ras provokes premature cell senescence associated with accumulation of p53 and p16INK4a. Cell 88: 593–602. Symonds H, Krall L, Remington L, Saenz-Robles M, Lowe S, Jacks T et al. (1994). p53-dependent apoptosis suppresses tumor growth and progression in vivo. Cell 78: 703–711. Tanaka T, Kimura M, Matsunaga K, Fukada D, Mori H, Okano Y. (1999). Centrosomal kinase AIK1 is overexpressed in invasive ductal carcinoma of the breast. Cancer Res 59: 2041–2044. te Poele RH, Okorokov AL, Jardine L, Cummings J, Joel SP. (2002). DNA damage is able to induce senescence in tumor cells in vitro and in vivo. Cancer Res 62: 1876–1883. Vafa O, Wade M, Kern S, Beeche M, Pandita TK, Hampton GM et al. (2002). c-Myc can induce DNA damage, increase reactive oxygen species, and mitigate p53 function: a mechanism for oncogeneinduced genetic instability. Mol Cell 9: 1031–1044. Wang X, Zhou YX, Qiao W, Tominaga Y, Ouchi M, Ouchi T et al. (2006). Overexpression of aurora kinase A in mouse mammary epithelium induces genetic instability preceding mammary tumor formation. Oncogene 25: 7148–7158. Wu X, Levine AJ. (1994). p53 and E2F-1 cooperate to mediate apoptosis. Proc Natl Acad Sci USA 91: 3602–3606. Zhang D, Hirota T, Marumoto T, Shimizu M, Kunitoku N, Sasayama T et al. (2004). Cre-loxP-controlled periodic Aurora A overexpression induces mitotic abnormalities and hyperplasia in mammary glands of mouse models. Oncogene 23: 8720–8730. Zhou H, Kuang J, Zhong L, Kuo WL, Gray JW, Sahin A et al. (1998). Tumour amplified kinase STK15/BTAK induces centrosome amplification, aneuploidy and transformation. Nat Genet 20: 189–193. Zhu J, Woods D, McMahon M, Bishop JM. (1998). Senescence of human fibroblasts induced by oncogenic Raf. Genes Dev 12: 2997–3007.

Supplementary Information accompanies the paper on the Oncogene website (http://www.nature.com/onc). Oncogene