Catulin Knockdown Induces Senescence in Cancer Cells

L-C Fan1, W-F Chiang2,3, C-H Liang1, Y-T Tsai4, T-Y Wong4, K-C Chen4, T-M Hong1,5,6 and Y-L Chen1,4

1Institute of Basic Medical Sciences, College of Medicine, National Cheng Kung University, Tainan, Taiwan, ROC; 2Department of Dentistry, Chi-Mei Medical Center, Liouying, Taiwan, ROC; 3School of Dentistry, National Yang-Ming University, Taipei, Taiwan, ROC; 4Institute of Oral Medicine, College of Medicine, National Cheng Kung University, Tainan, Taiwan, ROC; 5Department of Clinical Medical Research, National Cheng Kung Hospital, Tainan, Taiwan, ROC and 6Institute of Clinical Medicine, College of Medicine, National Cheng Kung University, Tainan, Taiwan, ROC

Cellular senescence functions as a tumor suppressor Introduction that protects against cancer progression. a-Catulin, an a-catenin-related protein, is reported to have tumorigenic Cellular senescence is a state of irreversible growth potential because it regulates the nuclear factor-jB arrest (Campisi and d’Adda di Fagagna, 2007) in which (NF-jB) pathway, but little is known about its clinical senescent cells remain metabolically active and display relevance and the mechanism through which it regulates typical phenotype changes, including the absence of cancer progression. Here, we found that a-catulin mRNA mitosis, an enlarged flattened morphology and high levels were significantly upregulated in cancer cell lines activity of senescence-associated b-galactosidase (SA-b- and clinical oral squamous cell carcinomas, which gal) (Dimri et al., 1995). It can be caused by the critical positively correlated with tumor size (P ¼ 0.001) and shortening of telomeres in a process known as replica- American Joint Committee on Cancer (AJCC) stage tive senescence or by various forms of stress such as (P ¼ 0.004). a-Catulin knockdown in the OC2 and A549 oncogene activation and loss of tumor suppressors in cancer cell lines dramatically decreased cell proliferation a process known as premature senescence or stress- and contributed to cellular senescence, and inhibited OC2 induced senescence. Moreover, evidence indicates that xenograft growth. Mechanistic dissection showed that DNA damage is a crucial mediator of cellular senes- a-catulin depletion strongly induced the DNA-damage cence: it triggers the DNA-damage response (DDR) response (DDR) in both cell lines, via a p53/p21-dependent pathway, which enables cells to sense damaged DNA pathway in A549 cells, but a p53/p21-independent path- and to respond by repairing the damage or by arresting way in OC2 cells carrying mutant p53. Global gene cell-cycle progression (Bartkova et al., 2006; Di Micco expression analysis revealed that a-catulin knockdown et al., 2006). Persistent DDR signaling is essential for altered cell-cycle regulation and DDR pathways at the both senescence initiation and maintenance (d’Adda di presenescent stage as well as significantly downregulate Fagagna, 2008). Genetic manipulations such as the several crucial genes related to mitotic chromosome aberrant expression of oncogenes abolishing the senes- condensation, DDR and DNA repair systems, which cence response led to fully developed malignancy (Chen suggests that its depletion-induced cellular senescence et al., 2005; Sarkisian et al., 2007). Cellular senescence is might be caused by chromosome condensation failures, also a proven tumor-suppressive strategy that forces severe DNA damage and impaired DNA repair ability. transformed cancer cells to (re-)enter irreversible cell- Our study provides evidence that a-catulin promotes cycle arrest (Collado et al., 2007). Therefore, triggering tumor growth by preventing cellular senescence and the senescence of tumor cells may contribute to suggests that downregulating a-catulin may be a promis- successful cancer therapy (Lleonart et al., 2009). ing therapeutic approach for cancer treatment. a-Catulin is an 82-kDa vinculin/a-catenin family Oncogene (2011) 30, 2610–2621; doi:10.1038/onc.2010.637; protein encoded by the CTNNAL1 gene, which was published online 31 January 2011 originally identified as a downregulated transcript in sodium butyrate-treated pancreatic cancer cells that Keywords: a-catulin; senescence; p53; DNA damage; were undergoing differentiation and apoptosis (Zhang cDNA microarray; cancer et al., 1998). a-Catulin has similarities to human vinculin and a-catenin, especially in the N-terminal region, which contains binding sites for b-catenin, talin and a-actinin, which suggests that a-catulin may be a cytoskeletal Correspondence: Professor Y-L Chen, Institute of Oral Medicine, College of Medicine, National Cheng Kung University,1 University linker protein (Janssens et al., 1999). Despite the Road, Tainan 70101, Taiwan, ROC. sequence similarity between a-catulin, a-catenin and E-mail: [email protected] or Professor T-M Hong, Institute vinculin, however, a-catulin did not inhibit the Wnt/ of Clinical Medicine, College of Medicine, National Cheng Kung b-catenin signaling pathway (Merdek et al., 2004). The University, 1 University Road, Tainan 70101, Taiwan, ROC. E-mail: [email protected] a-CATULIN/CTNNAL1 gene localizes to chromosome Received 1 May 2010; revised and accepted 20 December 2010; published 9q31-32 (Zhang et al., 1998; Park et al., 2002), a online 31 January 2011 frequent site of allelic loss and a tumor suppressor a-Catulin and senescence L-C Fan et al 2611 marker that has been reported in many human cancers compared with adjacent non-tumor tissue samples (Schultz et al., 1995; Miura et al.,1996),whichimplies (Figure 2a). All 63 patients were enrolled in the that a-catulin may be a tumor suppressor. However, follow-up study, which lasted 77 months (median: 33 other studies (Park et al., 2002; Wiesner et al.,2008) months). The total survival curve for the 63 oral report that a-catulin has tumorigenic potential because it squamous cell carcinoma (OSCC) patients is shown in binds directly to IKK-b and Lbc, a Rho-specific guanine Figure 2b. Although a Kaplan–Meier analysis showed nucleotide exchange factor, which promotes cell migra- no significant correlation between survival and a-catulin tion and increases resistance to apoptosis. The literature expression, we found that OSCC patients with higher contains little evidence of the clinical significance and a-catulin expression levels had a relatively lower function of a-catulin in cancer progression. In this study, cumulative survival rate. Furthermore, we investigated we analyzed the expression of a-catulin in clinical tumor the clinical relevance of differential a-catulin expression specimens to clarify the correlation between a-catulin in paired tumor and non-tumor tissues. a-Catulin expression and tumor progression and to elucidate the overexpression in tumor tissue was not correlated with mechanisms used by a-catulin in cellular senescence. age and gender, cigarette smoking, betel nut chewing, alcohol drinking habits, tumor site or lymph node metastasis but was significantly correlated with tumor Results size (P ¼ 0.001) and AJCC staging (P ¼ 0.004) (Table 1). Although there was no significant correlation between a-Catulin knockdown arrested cell proliferation the grade of tumor cell differentiation and a-catulin and induced senescence in cancer cells expression, a-catulin expression gradually increased First, using quantitative reverse transcription (qRT)– with poorly differentiated grades of tumor (P ¼ 0.057). PCR we measured the mRNA levels of a-catulin in two types of primary cultured healthy oral keratinocytes, the human lung cancer cell line A549 and several human a-Catulin knockdown suppressed tumorigenicity in vivo oral cancer cell lines. a-Catulin mRNA levels were To investigate whether a-catulin depletion impairs tumor- higher in cancer cells than in normal cells (Figure 1a). igenesis in vivo, we subcutaneously injected shLuc- or 6 We generated two lentiviruses, designated shC01 and shC01-infected OC2 cells (2 Â 10 ) into the posterior flank shB01, that produced specific short-hairpin RNA of non-obese diabetic (NOD)-severe combined immuno- (shRNA) targeted to bases 324–344 and 1401–1421 of deficiency (SCID) mice and measured the tumor size on full-length a-catulin mRNA, respectively. To understand days 0, 6, 13, 20, 25, 27, 29, 31 and 33 post-inoculation. the role of a-catulin in tumorigenesis, we infected tumor We found that the tumors became palpable and grew cells with these lentiviruses to silence their a-catulin gene rapidly after day 25 in mice inoculated with shLuc- expression. After 1 week of infection and puromycin infected OC2 cells but grew slowly in mice inoculated with selection, we found that in OC2 and A549 cells a-catulin shC01-infected OC2 cells (Figure 3A). Tumors in the gene expression was silenced and cell proliferation former group were significantly larger on day 33 than dramatically reduced by the shB01 and shC01 viruses those in the latter group (tumor incidence: shLuc: 3 of 3; but not by the control shLuc viruses (Figures 1b–d). shC01: 3 of 4) (Figure 3A, inset). shLuc-induced tumors We also found that a-catulin knockdown in OC2 and were three times heavier than shC01-induced tumors A549 cells induced the accumulation of cell populations (Figure 3B), which indicated that inhibiting a-catulin in S and G2/M cell-cycle phases (Figure 1e) as well expression in cancer cells delayed tumor formation. as decreasing their specific markers cyclin A and cyclin Moreover, a-catulin expression in shC01-infected OC2 B1, respectively, compared with shLuc-infected cells cells was lower than in shLuc-infected OC2 cells (Supplementary Figure 1). Less than 1.5% apoptotic (Figure 3C). Notably, immunohistochemical analysis cells were observed among shLuc- and a-catulin revealed that the percentage of Ki-67-positive tumor cells shRNA-infected cells (Figure 1e). Most of these cells was higher in shLuc infected than in shC01-infected OC2 had senescence-like morphology with an enlarged and cells (Figures 3D and E). In contrast, the percentage of flattened shape (Figure 1f). Notably, endogenous SA-b-gal DcR2-positive tumor cells was higher in shC01-infected activity and the oncogene-induced senescence marker OC2 cells and was morphologically enlarged and flattened DcR2 (Collado et al., 2005) were higher in a-catulin- compared with shLuc-infected cells (Figures 3D and F), silenced cancer cells 6 days after infection with a-catulin strongly indicating tumor cell senescence in vivo driven by shRNA viruses (Figure 1g; Supplementary Figure 1). a-catulin knockdown. Taken together, these results Taken together, these findings suggest that a-catulin suggest that a-catulin knockdown significantly suppressed knockdown induced cellular senescence as the major tumor size by inducing senescence, which inhibited cancer phenotype of cell death in OC2 and A549 cancer cells. cell proliferation in vivo. a-Catulin expression was frequent in oral squamous cell Cellular senescence induced by a-catulin knockdown was carcinoma tissues and positivelycorrelatedwithtumorsize p53 dependent in A549 cells but p53 independent in OC2 We then used qRT–PCR to examine a-catulin expres- cells sion in oral cancer and adjacent non-tumor oral mucosa p53/p21 is the key tumor suppressor pathway; it is trigged tissue samples from 63 oral cancer patients. a-Catulin by DNA damage and has been implicated in senescence levels were higher in 49 of the tumor tissue samples induction (d’Adda di Fagagna, 2008). Therefore, we also

Oncogene a-Catulin and senescence L-C Fan et al 2612

Figure 1 Silencing a-catulin in cancer cells decreased cell proliferation and led to cellular senescence. (a) qRT–PCR of endogenous a-catulin mRNA levels in two NHOK cell primary cultures, oral cancer cell lines (OC2, OECM1 and HSC3) and a lung adenocarcinoma cell line (A549). GAPDH was a loading control. (b) a-Catulin expression in OC2 and A549 cells was knocked down by the stable expression of shRNAs against a-catulin from two lentiviral expression constructs: pLKO.1 shB01 and pLKO.1 shC01. RT–PCR and western blotting were used to determine a-catulin expression. GAPDH and b-actin levels were loading controls. (c) Silencing a-catulin in OC2 and A549 cells markedly inhibited colony formation. Two days after puromycin selection, 104 of shRNA-infected cells were seeded per 10 cm dish. The plates were fixed 14 days after seeding and then stained with crystal violent. Representative photographs of colony formation 14 days after control shLuc and a-catulin shRNA (shB01 and shC01)-infected cells had been cultured. (d) Cell proliferation was measured by MTS assay at the times indicated after shLuc-, shB01- and shC01-infected OC2 and A549 cells. (e) Propidium iodide staining was performed to determine the cell-cycle distribution in OC2 and A549 cells after 6 days infected with shLuc, shB01 and shC01 viruses. a-Catulin shRNA-infected cells showed higher population in S and G2/M cell-cycle phase compared with control shLuc-infected cells. (f) Silencing a-catulin in OC2 cells markedly caused flat and enlarged cell morphology. Morphological features of representative shLuc- and shB01-infected cells. Scale bar, 10 mm. (g) SA-b-gal-positive cells showed markedly greater staining than did control shLuc-infected cells. (Left) Representative micrographs of shB01- and shC01-infected OC2 and A549 cells 6 days after they had been infected with lentivirus. Scale bar, 40 mm. (Right) The percentage of SA-b-gal-positive cells was quantified from three different fields. **Po0.01; #Po0.001.

examined whether DNA damage and the p53/p21 path- as nuclear accumulation (Figures 4a and b). These data way is involved in the senescence induced by a-catulin suggest that the level of a-catulin expression is linked knockdown in OC2 and A549 cells. We found that to changes in the p53/p21 pathway: a low level mediates a-catulin knockdown in OC2 and A549 cells resulted cell-cycle arrest and causes cellular senescence. in increased phosphorylated form of histone H2AX Moreover, we clariﬁed whether p53 activation (g-H2AX, Ser139), which is essential for DDR signaling was required for the senescence caused by a-catulin (Stucki et al., 2005), and p53 and p21 expression as well knockdown. We determined whether p53 knockdown

Oncogene a-Catulin and senescence L-C Fan et al 2613

Figure 2 a-Catulin mRNA expression levels were higher in clinical OSCC tissues. (a) qRT–PCR of a-catulin mRNA levels in 63 paired (N, T) primary OSCC tissues. b2-Microglobulin was an internal control. Error bars represent the mean±s.d. of triplicate experiments. The pie chart illustrates patient distribution by a-catulin expression measured using qRT–PCR. Patients were sorted into two categories based on the expression of a-catulin in tumor versus paired adjacent non-tumor tissues. (b) The survival curves of 63 OSCC patients with and without upregulated a-catulin (calculated using the Kaplan–Meier method).

rescues tumor cells from the deleterious effects of cyclin A and cyclin B1 (Figure 4f, right), suggesting a-catulin knockdown and enables their continued that p53 knockdown directed a-catulin-knockdown- proliferation. To do this, we co-infected A549 and induced senescence to cell proliferation. Unexpectedly, OC2 cells with various combinations of lentiviruses p53 knockdown in OC2 cells failed to abolish a-catulin- expressing a-catulin shRNAs, p53 shRNAs, and several knockdown-induced cellular senescence and proli- different negative control shRNAs. After puromycin ferative inhibition (Figures 4c–e, left). The cells still selection, western blotting confirmed p53 knockdown in accumulated DcR2 and g-H2AX, and cyclin A and pooled clones (Figure 4f). We then determined that the cyclin B1 were not detected (Figure 4f, left). This effect of p53 depletion on SA-b-gal activity was that it suggests that a-catulin-knockdown-induced senescence abrogated a-catulin-knockdown-induced senescence in OC2 cells is not associated with the upregulation of and proliferative inhibition in A549 cells (Figures 4c–e, the p53 pathway. right). p53-mediated senescence induced by a-catulin Because p53 depletion did not affect a-catulin-knock- knockdown was also confirmed in p53-siRNA-trans- down-induced senescence in OC2 cells, we sequenced the fected A549 cells (Supplementary Figure 2). Moreover, complete coding region of TP53, which was PCR p53 knockdown blocked the expression of p21 induced amplified from A549 and OC2 cell cDNA. This analysis by a-catulin silencing in A549 cells but not the active confirmed the wild-type TP53 cDNA sequence for A549 expression of g-H2AX, which suggested that DNA and revealed a point mutation of A to G at position 394 damage accumulated upstream of p53 (Figure 4f, right). in OC2 cells—Lys132 mutated to Glu132 within the p53 knockdown also inhibited DcR2 expression induced DNA-binding domain of p53 (Supplementary Figure 3). by a-catulin silencing, but increased the expression of Moreover, the p16/Rb pathway is crucial for senescence

Oncogene a-Catulin and senescence L-C Fan et al 2614 Table 1 The correlation between a-Catulin mRNA expression and a-Catulin knockdown significantly altered cell-cycle the clinicopathological variables in 63 OSCC patients regulation and DDR-related pathways Clinicopathological Number of a-Catulin P-value To further investigate the mechanism by which a- parameters patients T/N41.5 (%) catulin-knockdown induces senescence in OC2 cells, we used cDNA microarray analysis to examine the genes Age (years) 0.503 regulated as early as the presenescent stage (on day 2 50 or younger 27 15 55.6 Over 50 36 23 63.9 and day 4), during which the expression of SA-b-gal activity and DcR2 was undetectable (Figure 5a). A Tobacco smokinga 0.165 MetaCore software analysis of global genes revealed Yes 53 30 56.6 that cell-cycle regulation and DDR were the two main No 10 8 80 pathways that significantly altered OC2 cells at the Alcohol drinkingb 0.503 presenescent stage (Figure 5b). Notably, the most Yes 32 18 56.3 prominently altered cell-cycle regulation pathway in a- No 31 20 64.5 catulin-knockdown cells was the chromosome conden- Areca nut chewingc 0.452 sation in prometaphase (CCP), which includes the Yes 49 30 61.2 CCNA, CCNB, CDK1, INCENP, AURKA, NCAPD2, No 14 7 50.0 NCAPG2, SMC2, SMC4, TOP2A, H3F3B and H1FX American Joint Committee genes, all of which were markedly downregulated on Cancer (AJCC) stage 0.004** (Figure 5c). An analysis of the time course of each I þ II 24 9 37.5 prominently altered gene expression among these path- III þ IV 39 29 74.4 ways was performed using quantitative real-time PCR Tumor size 0.001*** (Figure 6a). We found that two DNA-damage-induced T1 þ T2 32 13 40.6 genes—growth arrest and DNA-damage-inducible pro- T3 þ T4 31 25 80.6 tein GADD45A and 14-3-3 protein sigma SFN—were markedly upregulated at the presenescent stage. These Differentiation 0.057 Well 31 15 48.4 genes are expressed upon cell-cycle arrest and are Moderate/poor 32 23 71.9 involved in DNA repair. On the other hand, the genes related to CCP, including SMC2, SMC4, NCAPG2, Lymph node metastasis 0.181 TOP2A and AURKA, and the DNA repair-related N(À) 39 21 53.8 N(þ ) 24 17 70.8 genes, including MSH2, MSH6, RAD50 and BRCA1, were strongly downregulated at 2 days and with a Site of tumor 0.884 moderate decrease by 4 days. In addition, the active Tongue 22 13 59.0 expression of g-H2AX was also initially induced at the Buccal mucosa/others 41 25 61.0 presenescent stage (Figure 6b). Indeed, there was a 0 0 Abbreviation: OSCC, oral squamous cell carcinoma. higher percentage of 4 ,6 diamino-2-phenylindole Á 2HCl a10 packs/year tobacco smoking history for over 10 years. (DAPI)-stained nuclei in a-catulin shRNAs-infected bAlcohol drinking 50 cups/year history for over 10 years. OC2 cells with a giant nucleus at the presenescent stage, cAreca nut chewing history for over 10 years. than in shLuc-infected OC2 cells (Figure 6c), suggesting 2 w test, **Po0.01, ***Po0.001. The data shown in bold are that a-catulin depletion-induced chromatin relaxation at statistically significant. presenescence may be primarily due to the loss of the genes involved in chromosome condensation. Taken induction, but our data indicated that p16/Rb was not together, these results suggest that the loss of CCP and activated by a-catulin silencing in OC2 cells (Supple- the impairment of DDR and DNA repair ability during mentary Figure 4). To further confirm the p53- presenescence are the major causative triggers accelerat- dependent and p53-independent functions of a-catulin- ing the onset of cellular senescence (on day 6, Figure 5a) knockdown-induced senescence, we performed experi- by a-catulin knockdown in OC2 cells. ments in two further cell lines, MCF-7 human breast cancer cells and CL1-0 human lung adenocarcinoma cells carrying wild-type and mutant p53, respectively. Discussion Our results also showed that p53 knockdown blocked a-catulin-depletion-induced senescence and proliferative Here, we provide evidence that the expression of inhibition in MCF-7 cells (Supplementary Figures 5a–c, a-catulin is upregulated in most human cancer cells right), which was consistent with the results observed in and clinical oral cancer tissues and positively correlated A549 cells. However, p53 knockdown in CL1-0 cells did with tumor size and cancer stage of OSCC. Knockdown not affect the cellular senescence and proliferative of a-catulin in cancer cells bearing either wild-type or inhibition driven by a-catulin depletion (Supplementary mutant p53 is sufficient to trigger DDR and eventually Figures 5a–c, left), which was also consistent with the induce cellular senescence in vitro as well as suppress observations in OC2 cells. Therefore, these results tumor formation in vivo. These results suggest that suggest that a-catulin-knockdown-induced senescence a-catulin functions as an oncoprotein, sustaining pro- is p53 dependent in p53 wild-type cells but p53 liferation by preventing cellular senescence and promot- independent in mutant p53-expressing cells. ing tumor progression.

Oncogene a-Catulin and senescence L-C Fan et al 2615

Figure 3 The suppressive effect of a-catulin silencing on tumor formation in vivo.(A) Tumor sizes were measured at the times indicated after shLuc- and shC01-infected OC2 cells were injected. The photograph shows representative images of tumor size on day 33 after tumor excision. (B) Tumor weight was measured on day 33 after tumor excision. (C) qRT–PCR of a-catulin was done on each excised tumor. Each data point represents the mean±s.e.m. of three mice. (D) Hematoxylin and eosin, Ki-67 and DcR2 immunohistochemical staining of primary tumors from shLuc- or shC01-infected OC2 cells on day 33 after subcutaneous implantation. Magniﬁcation: panels a–f, Â 100; insert, Â 400. (E, F) The frequencies of Ki-67-positive and DcR2-positive cells were quantiﬁed by using Histoquest software. *Po0.05; **Po0.01.

Ample evidence suggests that p53/p21 is crucial for Our cDNA microarray data revealed that a-catulin- mediating DDR (d’Adda di Fagagna, 2008) and induces knockdown-induced senescence in OC2 cells may be growth arrest to allow DNA repair, or promotes cellular caused primarily by the downregulation of many cell- senescence. Our results show that p53/p21 activation cycle regulation-related genes, especially by CCP. More- mediates a-catulin-knockdown-induced cellular senes- over, an analysis of the time course found that several cence in p53 wild-type cells (A549 and MCF-7). More- CCP-related genes, including SMC2, SMC4, NCAPG2, over, it has been reported that p53 mutations impair the TOP2A and AURKA, were strongly downregulated as DDR pathway (Bartkova et al., 2005; Gorgoulis et al., early as 2 days after a-catulin silencing. Loss of CCP 2005) because mutant p53 is often more stable than genes has been reported to block centrosome separation, wild-type p53 in the nucleus and is a dominant-negative spindle assembly and chromosome segregation during inhibitor against wild-type p53 (de Vries et al., 2002), mitosis. SMC2, SMC4 and NCAPG2 are three of ﬁve which allows cancer cells to develop by bypassing components of the condensing II complex, a complex senescence. However, this is not consistent with our that establishes the mitotic chromosome architecture ﬁnding that OC2 and CL1-0 cells expressing mutant p53 and has a central role in chromosome assembly and also show a-catulin-knockdown-induced senescence, segregation (Hudson et al., 2009). The loss of SMC4 which suggests that pathways other than p53/p21 lead stops the resolution of sister chromatids and cause the to senescence. formation of anaphase chromatin bridges and DNA

Oncogene a-Catulin and senescence L-C Fan et al 2616

Figure 4 The a-catulin-knockdown-induced p53/p21 tumor suppression pathway in cancer cell senescence. (a) Immunofluorescent microscopic analysis of F-actin, g-H2AX (Ser139), p53 and p21 expression on day 6 after lentiviral infection. a-Catulin silencing in shB01- and shC01-infected OC2 cells caused the cells to become larger and flatter. Visualized using rhodamine phalloidin-stained F- actin. g-H2AX (Ser139) was used to monitor DNA damage and the upregulation of p53 and p21. Scale bar, 25 mm. (b) Western blotting of g-H2AX (Ser139), p53 and p21 expression in the cells shown in (a). b-Actin was a loading control. (c) Representative micrographs of SA-b-gal staining in A549 and OC2 cells co-infected with various combinations of lentiviruses encoding shRNAs that targeted a-catulin (shB01 and shC01), p53 (shp53) and negative control luciferase (shLuc) 6 days after lentiviral infection. Scale bar, 40 mm. (d) The percentage of SA-b-gal-positive cells were quantified from three different fields in (c). **Po0.01; #Po0.001. (e) MTS assay was used to detect the cell proliferation at the indicated times following OC2 and A549 cells co-infected with shp53 and a-catulin shRNA (shB01 or shC01) viruses. (f) Western blotting of p53, p21 and g-H2AX (Ser139), cyclin proteins (cyclin A, cyclin B1, cyclin D1 and cyclin E) and DcR2 expression in the cells shown in (c). b-Actin was a loading control.

Oncogene a-Catulin and senescence L-C Fan et al 2617

Figure 5 Pathway analysis of differential global gene expression in a-catulin-knockdown OC2 cells. (a) (Left) Representative micrographs of SA-b-gal staining in shLuc-, a-catulin shRNA (shB01 or shC01)-infected OC2 cells on 2, 4 and 6 days after lentiviral infection. Scale bar, 40 mm. (Right) The percentage of SA-b-gal-positive cells were quantified from three different fields. **Po0.01; #Po0.001. Presenescent cells were defined by the undetectable expression of SA-b-gal activity and DcR2 senescence marker. b-Actin was a loading control. (b) MetaCore software with GeneGo Map Folders was used to analyze the pathways in universal genes differentially expressed in shB01- and shC01-infected OC2 cells on days 2 and 4. Each pathway folder regulated by genes identified from cDNA microarray data is numbered in green, and the number of genes involved in each pathway is numbered in red. (c) CCP is the most statistically significant pathway of the cell-cycle regulation pathway folder. cDNA microarray data showed that several genes were downregulated (red arrows).

Oncogene a-Catulin and senescence L-C Fan et al 2618

Figure 6 The genes involved in DDR and chromosome condensation were apparently altered in presenescent a-catulin knockdown OC2 cells. (a) The expression levels of genes were individually assessed by qRT–PCR and normalized using GAPDH. Data were presented as log2 ratios of the normalized expression levels in a-catulin shRNA-infected (shB01 or shC01) cells over those in shLuc- infected OC2 cells at the indicated time point. (b) Western blotting of g-H2AX (Ser139) in control shLuc- and a-catulin shRNA (shB01 and shC01)-infected OC2 cells 2 and 4 days after viral infection. b-Actin was a loading control. (c) Chromatin relaxation induced by a-catulin silencing in OC2 cells. (Upper) Representative photographs of DAPI-stained nuclei in OC2 cells followed by shLuc, shB01 and shC01 viral infection on day 4. Scale bar, 100 mm. (Lower left) Confocal microscopy was used to examine the DAPI-stained nuclei from shLuc-infected and shC01-infected OC2 cells. Scale bar, 10 mm. (Lower right) Frequencies of giant nucleus cells were quantiﬁed. Cell with a giant nucleus was determined by evaluating the diameter 417 mm. Total cells were count by three representative photographs with Â 400 magniﬁcation.

breakages (Coelho et al., 2003). Loss of TOP2A has DNA repair mechanisms maintain genomic fidelity been demonstrated to drastically disrupt chromatin (Pastink et al., 2001). The loss of these repair genes condensation by preventing the proper formation of usually contributes to genetic defects and eventually the kinetochore (Mikhailov et al., 2002). AURKA has a causes genomic instability (Pastink et al., 2001; van critical role in controlling centrosome separation, Gent et al., 2001), which predisposes the cell to spindle assembly and chromosome segregation during senescence (Suzuki et al., 2002; Walen, 2007). Notably, mitosis (Carmena and Earnshaw, 2003). A recent study our study showed that several DNA repair-related genes reported that the inhibition of ARUKA induces cancer were markedly suppressed in presenescent OC2 cells cell senescence in vitro and in vivo (Huck et al., 2010). induced by a-catulin depletion. MSH2, MSH6 and Although the association between chromatin relaxation PCNA are reported to have key roles in mismatch repair and cellular senescence remains unclear, importantly, (Kunkel and Erie, 2005); ATM, BRCA1 and RAD50 are chromatin relaxation has been hypothesized to be a required in homologous recombination repair and non- response to DNA damage and to be related to homologous end-joining repair (Shrivastav et al., 2008); chromosomal instability (Murr et al., 2006; Ziv et al., and XPC is involved in nucleotide excision repair 2006). Moreover, chromosomal instability induces (Volker et al., 2001; Wakasugi et al., 2002). When the aneuploidy, which eventually leads to cellular senescence expression of these DNA repair genes was inhibited in (Suzuki et al., 2002; Walen, 2007). These findings a-catulin-knockdown cells, this was powerfully reflected strongly suggest that the inhibition of a-catulin expres- in the high incidence of genomic instability. On the other sion may lead to the absence of chromosome condensa- hand, we found that DNA damage upregulated the tion, caused by the inhibition of these genes, result in expression of the GADD45A and SFN (14-3-3 sigma) chromosomal instability and cause senescence. genes. GADD45A is known to have a central role as a

Oncogene a-Catulin and senescence L-C Fan et al 2619 cellular stress sensor (Li et al., 2009), and its expression cDNA. Primer sequences are listed in Supplementary Table 1. increased following stressful growth arrest conditions Gene expression levels of a-catulin and a housekeeping gene and treatment with DNA-damaging agents (Rosemary (glyceraldehyde 3-phosphate dehydrogenase (GAPDH) or Siafakas and Richardson, 2009). Moreover, upregulated b2-microglobulin) were determined for each sample. Relative SFN gene expression inhibits Akt-mediated cell growth, quantities of a-catulin mRNA were calculated using the transformation and tumorigenesis, all of which cause comparative Ct method, subtracting Ct (a-catulin) from Ct (housekeeping gene) to derive DCt for each sample. Analyses DNA damage (Yang et al., 2006). Taken together with were done in triplicate to confirm the data. these findings, our study strongly suggests that one major driving force behind a-catulin-knockdown-in- Lentivirus production and infection duced senescence in OC2 cells is insufficiency of DDR shRNA targeting a-catulin, p53 and luciferase were expressed and the impairment of DNA repair ability, which causes from the pLKO.1 hairpin vector, which harbors an expression severe DNA damage and genomic instability, both of cassette for a puromycin resistance gene driven by the human which eventually result in irreversible senescence. phosphoglycerate kinase promoter. The shRNA sequences Cellular senescence is thought to be important for used in this study are shown in Supplementary Table S1. preventing unregulated growth and malignant transfor- shRNA lentiviruses were generated from HEK293T packaging mation (Campisi and d’Adda di Fagagna, 2007). cells. Cell culture supernatant containing lentivirus was Cellular senescence induced by loss of function of harvested every 24 h until 72 h after transfection. Viral titers oncogenes such as Cdk4 (Zou et al., 2002) and CIP2A were determined by transmuting HEK293 T cells using diluted (Li et al., 2008) has been considered as a therapeutic culture supernatants and tested by counting the number of viable cells after 2 days of culture in the presence or absence of approach and has been verified by previous studies puromycin (Sigma-Aldrich, St Louis, MO, USA). Viral (Lleonart et al., 2009). Knockdown of a-catulin is supernatants were stored at À801C. Cells were infected with crucial for ensuring the irreversibility of the senescence shRNA or control lentiviruses in the presence of 8 mg/ml of arrest even in p53 mutant cells. Our findings not only polybrene (Sigma-Aldrich). After 48 h of infection, the cells offer new perspectives in the modulation of senescence were treated with 1 mg/ml of puromycin for selection. by a-catulin but also suggest a novel therapeutic target Puromycin-resistant cells were pooled for subsequent analysis. for cancer treatment. Colony formation assay Approximately 105 cells were seeded per 10 cm dish and cultured Materials and methods in growth medium with puromycin (1 mg/ml) for 14 days. Clonogenic survival was determined by staining the colonies Cell culture with crystal violet and visualizing them with a digital camera. OC2 and OECM1 human oral cancer cell lines established from Taiwanese men with a history of betel quid chewing were MTS proliferation assay obtained from Dr RC Chang (Veterans General Hospital, Cell proliferation was determined by using CellTiter 96 AQueous Taipei, Taiwan) and Dr SY Liu (Chi-Mei Medical Center, One Solution Cell Proliferation Assay (MTS) kit (Promega, Tainan, Taiwan), respectively. An HSC3 cell line derived from Madison, WI, USA). Approximately 103 cells were seeded in human tongue carcinoma with lymph node metastasis was 96-well plate and cultured in growth medium with puromycin from the JCRB cell bank of Japan. OC2, OECM1 and HSC3 (1 mg/ml). Following phosphate-buffered saline washing and cells were maintained in RPMI 1640 medium, Dulbecco’s addition of MTS reagent, absorbance at 490 nm was recorded. modified Eagle’s medium. A549 human lung cancer cells (American Type Culture Collection, Washington, DC, USA) were maintained in Dulbecco’s modified Eagle’s medium. Cell-cycle distribution Medium was supplemented with 10% fetal bovine serum and For the analysis of cell-cycle distribution, both floating and 100 units/ml of penicillin and streptomycin (Invitrogen, attached cells were collected and centrifuged before washed Carlsbad, CA, USA). Human normal oral keratinocytes with cold phosphate-buffered saline, and then fixed in 70% (NHOK), taken with informed consent, were isolated from cold ethanol overnight at À201C. Propidium iodide staining the gingival tissue of healthy patients in the Department of was performed after incubation of the cells with 50 mg/ml Dentistry, National Cheng Kung University Hospital. The propidium iodide and 20 mg/ml RNase A in the dark at room NHOK were cultured in keratinocyte-serum-free medium temperature for 30 min, which were then analyzed by flow supplied with L-glutamine, epidermal growth factor (EGF) cytometry (BD Biosciences, San Jose, CA, USA). and bovine pituitary extract (BPE) (Invitrogen) and then incubated at 371C in a humidified 5% CO2 atmosphere. b-Galactosidase staining SA-b-gal activity was measured using a kit (Senescence RNA extraction and qRT–PCR b-Galactosidase Staining Kit; Cell Signaling Technology, Total RNA was isolated using a reagent (Trizol; Invitrogen) Beverly, MA, USA) according to the manufacturer’s instruc- according to the manufacturer’s protocol. First-strand cDNA tions. Light microscopy was used to identify and count was synthesized from 1 mg of total RNA using reverse senescent (blue stained) cells. transcription (ImProm-II Reverse Transcription System; Promega, Madison, WI, USA). qRT–PCR was done (Light- Immunofluorescent staining Cycler 480 Instrument; Roche Applied Sciences, Indianapolis, Cells seeded on a coverslip were fixed in 4% paraformaldehyde IN, USA) using SYBR Green I dye. Each 10 ml reaction for 15 min. After they had been permeabilized with 0.25% contained 4 mmol/l of MgCl2, 200 nmol/l each of forward and Triton X-100, the cells were blocked in 5% bovine serum reverse primers, 1 ml of dye (LightCycler Fast Start DNA albumin (BSA) for 30 min at room temperature and then Master SYBR Green I; Roche) and 2 ml of 10-fold diluted incubated with antibodies against g-H2AX (Ser 139) (#2577;

Oncogene a-Catulin and senescence L-C Fan et al 2620 Cell Signaling Technology), p53 (sc-126) or p21 (sc-817) (Santa was done using primary antibody for Ki-67 (550609; BD Cruz Biotechnology, Santa Cruz, CA, USA). Rhodamine Pharmingen, San Diego, CA, USA) and DcR2 (ab2019; phalloidin (Alexa Fluor 633-conjugated phalloidin; Invitro- Abcam); the sections were then incubated with anti-mouse/ gen) and DAPI were used for F-actin and nuclear staining. rabbit immunoglobulin G-horseradish peroxidase-conjugated secondary antibody. The signal was detected using a kit Western blotting (Aminoethyl Carbazole Substrate Kit; Zymed Laboratories Whole-cell lysates were resolved by sodium dodecyl sulfate Inc., San Francisco, CA, USA). The sections were counter- polyacrylamide gel electrophoresis. Proteins were transferred stained with hematoxylin. Image analysis of Ki-67- and DcR2- onto a polyvinylidene difluoride membrane (Millipore, positive cell was further quantified by using Histoquest Billerica, MA, USA) and incubated with the primary antibody, software (Tissue Gnostics, Vienna, Austria). and then incubated with horseradish peroxidase-conjugated secondary antibodies. Specific proteins were detected using Microarray analysis and pathway analysis enhanced enhanced chemiluminescence (ECL) chemilumines- Microarray hybridization was done at the Phalanx Biotech cence reagent (Amersham Biosciences, Piscataway, NJ, USA). Service Center (Hsinchu, Taiwan). RNA integrity and quality The primary antibodies used for western blotting were a- were assessed using a kit (RNA 6000 Nano Assay; Agilent catulin (B01P; Abnova, Taipei, Taiwan), g-H2AX (Ser 139, Technologies, Inc., Santa Clara, CA, USA), a spectrophot- #2577; Cell Signaling Technology), p53 (sc-126) and p21 (sc- ometer (NanoDrop ND-1000; Thermo Fisher Scientific Inc., 817) (Santa Cruz Biotechnology); DcR2 (ab2019; Abcam, Wilmington, DE, USA) and agarose gel electrophoresis. RNA Cambridge, UK), the Cyclin Antibody Sampler kit (#9869; was labeled with Cy5 using aminoallyl RNA labeling and then Cell Signaling Technology) and b-actin (Sigma-Aldrich). hybridized in duplicate on full human genome arrays (Whole Genome Human OneArray HOA 4.3; Phalanx Biotech) Patients and tissue samples containing 30 968 human genome probes and 1082 experi- The study was approved by the institutional review boards of mental control probes in a one-block array format, and with Chi-Mei Hospital and National Cheng Kung University each probe a 60 mer oligonucleotide designed in the sense Hospital, Tainan, Taiwan. Sixty-three paired primary OSCC direction. Filtered data were log2 transformed and corrected samples were included in this study. Immediately after surgery, using quantile normalization before their average ratios and cancerous and adjacent non-cancerous tissues were stored in significance values were calculated. Genes that showed a liquid nitrogen. Pathologists determined the histological grade significant (Po0.01) expression difference between a-Catulin of each specimen. The OSCC samples were staged according to silenced (shB01 and shC01 infected) and non-silenced (shLuc) the American Joint Committee on Cancer (5th AJCC) 1997 were analyzed for pathways (GeneGo Map Folders of cancer staging guidelines. MetaCore software suite; GeneGo, St Joseph, MI, USA). Significance probability was calculated using the hypergeome- trical distribution based on gene ontology terms. Because Survival and statistical analysis 2 one gene is frequently involved in multiple pathways, A w test was used to analyze the clinicopathological variables all pathways corresponding to the genes with significance of qRT–PCR results. Survival rates were calculated using the probability were listed. Kaplan–Meier method, and statistical significance was determined using the log-rank test. Statistical significance was set at Po0.05. All statistical analyses were performed using SPSS 13.0 for Windows (SPSS Inc., Chicago, IL, USA). Conflict of interest

Tumorigenicity assays in NOD/SCID mice The authors declare no conflict of interest. After lentiviral infection and puromycin selection (1 mg/ml), shLuc- or shC01-infected OC2 cells (2 Â 106)in50mlof phosphate-buffered saline were subcutaneously injected into Acknowledgements the posterior flank of 6-week-old NOD–SCID mice. Tumor 3 size (mm ) was measured for each mouse on days 0, 6, 13, 20, This work was supported by grants NSC 96-2311-B-006-005- 25, 27, 29, 31 and 33 post-inoculation. On day 33, primary MY3, NSC 97-2314-B-384-003-MY3, NSC 99-3112-B-006-011 tumors were excised and weighed. Tumor size was monitored and NSC 99-2627-B-006-003 from the National Science by measuring the length (L), width (W) and height (H), and Council, and DOH99-TD-C-111-003 from the Department of calculated with the formula (L Â W Â H). Health, Taiwan. RNAi reagents were obtained from the National RNAi Core Facility located at the Institute of Histopathology and immunohistochemistry Molecular Biology/Genomic Research Center, Academia Specimens fixed in 10% buffered formalin were embedded in Sinica, supported by the National Research Program for paraffin and then sectioned and stained with hematoxylin and Genomic Medicine Grants of National Science Council, eosin. Immunohistochemical analysis of the paraffin sections Taiwan (NSC 97-3112-B-001-016).

References

Bartkova J, Horejsi Z, Koed K, Kramer A, Tort F, Zieger K et al. Campisi J, d’Adda di Fagagna F. (2007). Cellular senescence: when (2005). DNA damage response as a candidate anti-cancer barrier in bad things happen to good cells. Nat Rev Mol Cell Biol 8: 729–740. early human tumorigenesis. Nature 434: 864–870. Carmena M, Earnshaw WC. (2003). The cellular geography of aurora Bartkova J, Rezaei N, Liontos M, Karakaidos P, Kletsas D, Issaeva N kinases. Nat Rev Mol Cell Biol 4: 842–854. et al. (2006). Oncogene-induced senescence is part of the tumorigenesis Chen Z, Trotman LC, Shaffer D, Lin HK, Dotan ZA, Niki M barrier imposed by DNA damage checkpoints. Nature 444: 633–637. et al. (2005). Crucial role of p53-dependent cellular senescence

Oncogene a-Catulin and senescence L-C Fan et al 2621 in suppression of Pten deficient tumorigenesis. Nature 436: Park B, Nguyen NT, Dutt P, Merdek KD, Bashar M, Sterpetti P et al. 725–730. (2002). Association of Lbc Rho guanine nucleotide exchange Coelho PA, Queiroz-Machado J, Sunkel CE. (2003). Condensin- factor with alpha-catenin-related protein, alpha-catulin/CTNNAL1, dependent localisation of topoisomerase II to an axial chromosomal supports serum response factor activation. J Biol Chem 277: structure is required for sister chromatid resolution during mitosis. 45361–45370. J Cell Sci 116: 4763–4776. Pastink A, Eeken JC, Lohman PH. (2001). Genomic integrity and Collado M, Blasco MA, Serrano M. (2007). Cellular senescence in the repair of double-strand DNA breaks. Mutat Res 480–481: cancer and aging. Cell 130: 223–233. 37–50. Collado M, Gil J, Efeyan A, Guerra C, Schuhmacher AJ, Barradas M Rosemary Siafakas A, Richardson DR. (2009). Growth arrest and et al. (2005). Tumour biology: senescence in premalignant tumours. DNA damage-45 alpha (GADD45alpha). Int J Biochem Cell Biol Nature 436: 642. 41: 986–989. d’Adda di Fagagna F. (2008). Living on a break: cellular senescence as Sarkisian CJ, Keister BA, Stairs DB, Boxer RB, Moody SE, Chodosh a DNA-damage response. Nat Rev Cancer 8: 512–522. LA. (2007). Dose-dependent oncogene-induced senescence in vivo de Vries A, Flores ER, Miranda B, Hsieh HM, van Oostrom CT, Sage and its evasion during mammary tumorigenesis. Nat Cell Biol 9: J et al. (2002). Targeted point mutations of p53 lead to dominant- 493–505. negative inhibition of wild-type p53 function. Proc Natl Acad Sci Schultz DC, Vanderveer L, Buetow KH, Boente MP, Ozols RF, USA 99: 2948–2953. Hamilton TC et al. (1995). Characterization of chromosome 9 in Di Micco R, Fumagalli M, Cicalese A, Piccinin S, Gasparini P, Luise human ovarian neoplasia identifies frequent genetic imbalance on C et al. (2006). Oncogene-induced senescence is a DNA damage 9q and rare alterations involving 9p, including CDKN2. Cancer Res response triggered by DNA hyper-replication. Nature 444: 638–642. 55: 2150–2157. Dimri GP, Lee X, Basile G, Acosta M, Scott G, Roskelley C et al. Shrivastav M, De Haro LP, Nickoloff JA. (2008). Regulation (1995). A biomarker that identifies senescent human cells in culture of DNA double-strand break repair pathway choice. Cell Res 18: and in aging skin in vivo. Proc Natl Acad Sci USA 92: 9363–9367. 134–147. Gorgoulis VG, Vassiliou LV, Karakaidos P, Zacharatos P, Kotsinas Stucki M, Clapperton JA, Mohammad D, Yaffe MB, Smerdon SJ, A, Liloglou T et al. (2005). Activation of the DNA damage Jackson SP. (2005). MDC1 directly binds phosphorylated histone checkpoint and genomic instability in human precancerous lesions. H2AX to regulate cellular responses to DNA double-strand breaks. Nature 434: 907–913. Cell 123: 1213–1226. Huck JJ, Zhang M, McDonald A, Bowman D, Hoar KM, Stringer B Suzuki T, Fujii M, Ayusawa D. (2002). Demethylation of classical et al. (2010). MLN8054, an inhibitor of Aurora A kinase, induces satellite 2 and 3 DNA with chromosomal instability in senescent senescence in human tumor cells both in vitro and in vivo. Mol human fibroblasts. Exp Gerontol 37: 1005–1014. Cancer Res 8: 373–384. van Gent DC, Hoeijmakers JH, Kanaar R. (2001). Chromosomal Hudson DF, Marshall KM, Earnshaw WC. (2009). Condensin: stability and the DNA double-stranded break connection. Nat Rev architect of mitotic chromosomes. Chromosome Res 17: 131–144. Genet 2: 196–206. Janssens B, Staes K, van Roy F. (1999). Human alpha-catulin, a novel Volker M, Mone MJ, Karmakar P, van Hoffen A, Schul W, alpha-catenin-like molecule with conserved genomic structure, but Vermeulen W et al. (2001). Sequential assembly of the nucleotide deviating alternative splicing. Biochim Biophys Acta 1447: 341–347. excision repair factors in vivo. Mol Cell 8: 213–224. Kunkel TA, Erie DA. (2005). DNA mismatch repair. Annu Rev Wakasugi M, Kawashima A, Morioka H, Linn S, Sancar A, Mori T Biochem 74: 681–710. et al. (2002). DDB accumulates at DNA damage sites immediately Li W, Ge Z, Liu C, Liu Z, Bjorkholm M, Jia J et al. (2008). CIP2A is after UV irradiation and directly stimulates nucleotide excision overexpressed in gastric cancer and its depletion leads to impaired repair. J Biol Chem 277: 1637–1640. clonogenicity, senescence, or differentiation of tumor cells. Clin Walen KH. (2007). Origin of diplochromosomal polyploidy in near- Cancer Res 14: 3722–3728. senescent fibroblast cultures: heterochromatin, telomeres and Li Y, Qian H, Li X, Wang H, Yu J, Liu Y et al. (2009). Adenoviral- chromosomal instability (CIN). Cell Biol Int 31: 1447–1455. mediated gene transfer of Gadd45a results in suppression by Wiesner C, Winsauer G, Resch U, Hoeth M, Schmid JA, van Hengel J inducing apoptosis and cell cycle arrest in pancreatic cancer cell. J et al. (2008). Alpha-catulin, a Rho signalling component, can Gene Med 11: 3–13. regulate NF-kappaB through binding to IKK-beta, and confers Lleonart ME, Artero-Castro A, Kondoh H. (2009). Senescence resistance to apoptosis. Oncogene 27: 2159–2169. induction; a possible cancer therapy. Mol Cancer 8:3. Yang H, Wen YY, Zhao R, Lin YL, Fournier K, Yang HY et al. Merdek KD, Nguyen NT, Toksoz D. (2004). Distinct activities of the (2006). DNA damage-induced protein 14-3-3 sigma inhibits protein alpha-catenin family, alpha-catulin and alpha-catenin, on beta- kinase B/Akt activation and suppresses Akt-activated cancer. catenin-mediated signaling. Mol Cell Biol 24: 2410–2422. Cancer Res 66: 3096–3105. Mikhailov A, Cole RW, Rieder CL. (2002). DNA damage during Zhang JS, Nelson M, Wang L, Liu W, Qian CP, Shridhar V et al. mitosis in human cells delays the metaphase/anaphase transition via (1998). Identification and chromosomal localization of CTNNAL1, the spindle-assembly checkpoint. Curr Biol 12: 1797–1806. a novel protein homologous to alpha-catenin. Genomics 54: Miura K, Suzuki K, Tokino T, Isomura M, Inazawa J, Matsuno S 149–154. et al. (1996). Detailed deletion mapping in squamous cell Ziv Y, Bielopolski D, Galanty Y, Lukas C, Taya Y, Schultz DC et al. carcinomas of the esophagus narrows a region containing a putative (2006). Chromatin relaxation in response to DNA double-strand tumor suppressor gene to about 200 kilobases on distal chromosome breaks is modulated by a novel ATM- and KAP-1 dependent 9q. Cancer Res 56: 1629–1634. pathway. Nat Cell Biol 8: 870–876. Murr R, Loizou JI, Yang YG, Cuenin C, Li H, Wang ZQ et al. (2006). Zou X, Ray D, Aziyu A, Christov K, Boiko AD, Gudkov AV et al. Histone acetylation by Trrap-Tip60 modulates loading of repair (2002). Cdk4 disruption renders primary mouse cells resistant to proteins and repair of DNA double-strand breaks. Nat Cell Biol 8: oncogenic transformation, leading to Arf/p53-independent senes- 91–99. cence. Genes Dev 16: 2923–2934.

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

Oncogene