Oncogene (2008) 27, 1961–1970 & 2008 Nature Publishing Group All rights reserved 0950-9232/08 $30.00 www.nature.com/onc ORIGINAL ARTICLE Cellular senescence bypass screen identifies new putative tumor suppressor

JFM Leal1, J Fominaya1, A Casco´ n2, MV Guijarro1, C Blanco-Aparicio1, M Lleonart3, ME Castro1, SRamon y Cajal 3, M Robledo2, DH Beach4 and A Carnero1

1Experimental Therapeutics Programme, Centro Nacional de Investigaciones Oncolo´gicas (CNIO), Madrid, Spain; 2Molecular Pathology Programme, Centro Nacional de Investigaciones Oncolo´gicas (CNIO), Madrid, Spain; 3Departamento de Patologı´a, Hospital Vall d’Hebron, Barcelona, Spain and 4Institute for Cell and Molecular Sciences, Center for Cutaneous Biology, London, UK

Senescence is a mechanism that limits cellular lifespan and Introduction constitutes a barrier against cellular immortalization. To identify new senescence regulatory genes that might play a Replicative senescence is characterized by a progressive role in tumorigenesis, we have designed and performed a loss of proliferative potential with the increase of popu- large-scale antisense-based genetic screen in primary lation doublings, resulting in a permanent and irrever- mouse embryo fibroblasts (MEFs). Out of this screen, sible cell-cycle arrest. Although the process of senescence we have identified five different genes through which loss occurs both in vitro and in vivo (Dimri et al., 1995; of function partially bypasses senescence. These genes Schmitt et al., 2002; Shay and Roninson, 2004; Braig belong to very different biochemical families: csn2 et al., 2005; Collado et al., 2005; Michaloglou et al., (component of the Cop9 signalosome), aldose reductase 2005), the transition to the senescent phenotype is (a metabolic enzyme) and brf1 (subunit of the RNA commonly studied in culture where a cell population polymerase II complex), S-adenosyl homocysteine hydro- can be grown and monitored. Replicative senescence can lase and Bub1. Inactivation, at least partial, of these genes be a consequence of shortening and dysfunction of the confers resistance to bothp53- and p16INK4a-induced telomeres, due to the inability of most human cells to proliferation arrest. Furthermore, such inactivation in- replicate the chromosomal ends (Serrano and Blasco, hibits p53 but not E2F1 transcriptional activity and 2001; Greider and Blackburn, 2004). Nevertheless, impairs DNA-damage-induced of p21. Since cellular senescence (also including accelerated senes- the aim of the screen was to identify new regulators of cence) is a more general process not only triggered tumorigenesis, we have tested their inactivation in human by accumulation of cell divisions, but also by activation tumors. We have found, either by northern blot or of oncogenes (Serrano et al., 1997), DNA damage, epi- quantitative reverse transcriptase–PCR analysis, that genetic changes and oxidative stress (Campisi, 2001; the expression of three genes, Csn2, Aldose reductase Shay and Roninson, 2004). Senescent cells cannot be and Brf1, is lost at different ratios in tumors of different stimulated to re-enter the cell cycle by physiological origins. These genes are located at common positions of mitogens, become resistant to apoptosis and acquire loss of heterogeneity (15q21.2, 7q35 and 14q32.33); altered differentiated characteristics. Moreover, accele- therefore,we have measured genomic losses of these rated senescence driven by any of these stresses is inde- specific genes in different tumors. We have found that pendent of the telomeric length (Serrano and Blasco, Csn2 and Brf1 also show genomic losses of one allele in 2001; Suzuki et al., 2001) and is not prevented by expres- different tumors. Our data suggest that the three genes sion of telomerase (Morales et al., 1999). Regardless of identified in the genome-wide loss-of-function genetic the mechanism triggering cellular senescence, the signal- screen are putative tumor suppressors located at ing cascade leading to growth arrest seems to be similar 15q21.2; 7q35 and 14q32.33. and it involves the p53 and pRB pathways (Sherr and Oncogene (2008) 27, 1961–1970; doi:10.1038/sj.onc.1210846; McCormick, 2002). Overcoming the restriction that published online 29 October 2007 cellular senescence poses to immortalization may be a primary step during the transformation to malignancy Keywords: senescence; genetic screening; tumor sup- (Hanahan and Weinberg, 2000). pressor; p53; BRF1; Csn2 During immortalization, cells acquire genetic altera- tions that override the normal mechanisms controlling senescence (For review see Hanahan and Weinberg, 2000; Sherr and McCormick, 2002; Shay and Roninson, Correspondence: Dr A Carnero, Experimental Therapeutics 2004; Campisi, 2005). Among the alterations that Programme, Centro Nacional de Investigaciones Oncologicas (CNIO), immortalize cells, those that inactivate tumor suppressor c/Melchor Fernandez Almagro no3, Madrid 28029, Spain. E-mail: [email protected] genes in the p53 and pRB pathways are very frequent Received 20 June 2007; revised 31 August 2007; accepted 7 September in cancer. Tumor suppressor genes may get inactivated 2007; published online 29 October 2007 by either deletion of one or both alleles, promoter Cellular senescence bypass screen JFM Leal et al 1962 methylation, splice-site mutations and nonsense muta- fragments, we measured their efficacy at reducing the tions that induce premature translational termination levels of their respective targets. NIH3T3 cells and destabilize mRNA transcripts (or a combination carrying either of Csn2, ARase and BRF1 HA-tagged thereof). Such alterations result in a complete absence or were infected with the correspondent antisense partial reduction of the tumor suppressor protein in the in the pMARXIV vector, and the effect of the antisense affected cells, conferring them selective advantage in was quantitated by western blotting. As shown in clonal selection for tumor progression. Figure 1a, the protein level of the different HA-tagged In this work, we designed and performed a genome- genes was reduced by average of 50%. Expression of wide loss-of-function genetic screen to identify addi- these antisense constructs partially bypasses growth tional putative tumor suppressor genes controlling arrest in senescence and is able to produce a moderate senescence that might act as tumor suppressors. increase in the lifespan of MEFs (Figure 1b and data not shown). In 3T3 experiments, the antisense fragments increased MEFs lifespan approximately four population doublings (Figure 1c), although the cells finally entered Results senescence. Moreover, the expression of the different antisense fragments also provides an early escape from Pre-senescent MEFs were infected with retrovirus senescence of the cultures. However, downregulation of carrying a library of senescent MEF’s transcriptome in these genes do not alter the p53 levels nor phosphoryla- antisense orientation (Carnero et al., 2000). MEFs were tion status in MEFs (Supplementary Materials 2). seeded at low concentration at doubling time of 10–12 In IMR90 human diploid fibroblasts, active shRNA and left to senesce. Clones that were able to grow were against Csn2 and ARase do not produce viable cell lines, identified, provirus recovered and sequenced. while active shRNA against BRF1 enhances the lifespan Out of our genome-wide loss-of-function screen, we in 6–8 population doublings causing cells to abruptly identified five different antisense fragments that targeted enter apoptosis (data not shown). proteins with very different biochemical properties (Table 1). Among them, an antisense against Csn2 (a component of the Cop9 signalosome) was recovered Effect of antisense on pRB and p53-dependent cell growth eight independent times. Antisense fragments against arrest Aldose reductase and S-adenosyl homocysteine hydro- To explore the possible mechanisms through which lase (two metabolic enzymes) and against BRF1 (a these genes might affect senescent arrest, we studied the subunit of the RNA polymerase III complex) were effect of the different antisense fragments on the p53 and recovered twice. Finally, an antisense against Bub1 was pRb checkpoints. First, we measured the ability of these recovered once. The size of the fragments varies from antisense constructs to bypass p53-induced arrest. To 130 to 210 nucleotides and they map to different regions this end, we expressed the different antisense fragments in the mRNAs, including the 30 untranslated region, as in a cell line derived from p53-null MEFs expressing the in the case of Csn2 and Aldose reductase. temperature-sensitive mutant of p53 (val135). These S-adenosyl homocystein hydrolase (SAHH) was cells grew actively at 39 1C but arrested at 32 1C, when identified previously in large-scale loss-of-function the p53 protein adopts the active conformation (Car- screen for genes involved in p53-dependent arrest and nero et al., 2000). The antisense fragments inhibited the their role in cancer has been published previously in cell cycle arrest dependent on p53 with rates ranging Berns et al.(2004). Suppression of SAHH by ShRNA between 20 and 40% (Figure 2a). confers resistance to both p53-dependent and p19ARF- dependent proliferation arrest; it also abolishes a DNA- The antisenses affect the transcriptional activity of p53 damage-induced G1 cell-cycle arrest (Berns et al., 2004) We examined the effect of the antisense fragments and is significantly downregulated in colon and lung on the p53-dependent transcriptional activation. A carcinomas (Lleonart et al., 2006). p21waf1-luciferase reporter construct was deployed as Here, we focused on the fragments against Csn2, a p53 transcriptional target. All of the antisense frag- Arase and BRF1, since they have been related to LoH. ments greatly reduced the transcriptional activity of p53 (Supplementary Materials 1). To validate the antisense (Figure 2b). To check whether the p53 transcriptional

Table 1 Antisense fragments identified in the large scale loss of function screening Antisense no Times recovered Length (nts) CDS Location (from ATG)

8 COPs2 (Csn2) 8 209 1323 1261–1468 45 HBRF1 (BRF1) 2 134 2031 100–229 312 Aldose reductase 2 238 951 1072–1306 B1 (Arase) 328 S-adenosyl 2 139 1593 283–414 homocysteine Hydrolase 1 (SAHH) 435 Bub1 1 202 3177 1619–1436

Oncogene Cellular senescence bypass screen JFM Leal et al 1963 activity was compromised by the antisense in a more p53 stabilization and activation produced an increase physiological setting, we transduced NIH3T3 cells with of p21waf1 protein, clearly visible 6 h after exposure to each one of the antisense fragments, selected mass doxorubicin or cisplatin and maintained 18 h later (24 h cultures carrying the retroviral vectors and analysed if after exposure). In the mass cultures of NIH3T3 cells DNA damage-induced p53 stabilization was followed by expressing the antisenses, we still observed the p53 an increase in the p21waf1 gene expression (Figure 2c). stabilization indicating that the different genes do not In NIH3T3 cells as well as in the mass cultures expres- connect upstream of the p53 pathway. The stabiliza- sing each one of the antisense fragments, p53 protein tion of p53 was not translated into an increase of p21 levels were increased 6 h after exposure to 0.2 mgml-1 of protein, suggesting that the transcriptional activity of doxorubicin or 100 mM of cisplatin, decreasing 24 h after p53 was compromised by the antisenses (Figure 2c). exposure in the case of doxorubicin and being main- tained in the cisplatin-treated cells. In NIH3T3 cells, this The antisense fragments help evading arrest induced by the pRB/E2F pathway Alteration of the pRb/E2F checkpoint has also been described to alter the replicative senescence in MEFs a s b 25 α p53[273H] (Yamasaki et al., 1998; Sage et al., 2000). Therefore, 20 we have determined whether the overexpression of the

v ARase- different antisense fragments has any effect on E2F- 15 induced transactivation. To that end, we co-transfected α-HA HEK293 cells with human E2F, a synthetic E2F 10 ARase α s promoter-luciferase reporter, and each one of the anti- α -tub 5 sense fragments (Figure 2d). None of the antisense frag- Proliferation index vector ments significantly affected the expression of luciferase, 0 indicating that the proteins they shut down do not ARase 0481026 interfere with the transcriptional activity of E2F1. Then, 18 we investigated whether the antisense fragments were 16 p53[273H] s able to bypass p16-induced cell cycle arrest by perform- α 14 ing a colony formation assay. We first transduced wild- v 12 BRF- type MEFs with the antisense fragments and selected 10 mass cultures, expressing them with the appropriate 8 BRF αs α -HA antibiotic. Then, each mass culture was transduced with 6 p16INK4a at low viral titer and cultured for 2 weeks α 4 -tub Proliferation index under the appropriate culture selection. Cells rapidly 2 vector entered growth arrest, as can be seen in the control 0 HBRF 06824 plate (Figure 2e), but in those plates where each one of 30 the antisenses were expressed, many cells escaped arrest p53[273H] and formed colonies. In this case, the most effective s 25 α antisense was against Aldose reductase. 20 v Csn2- 15 Human tumors frequently lose the expression of the α-HA Csn2α s candidate tumor suppressor genes 10 We have shown that Csn2, Aldose reductase and BRF1 α-tub 5 protein loss-of-function can partially extend MEF Proliferation index vector lifespan, probably by regulating p53 transcription and 0 Csn2 0681024 Time (d) Figure 1 Antisense fragments obtained in the screening increase the proliferation rate of MEFs. (a) Each one of the antisense c 25 fragments and empty vectors as controls, were transduced into NIH3T3 previously selected for the expression of the correspond- ing HA-tagged target gene to check their efficiency to reduce the 20 p53(273H) expression of their targets. After selection, total protein was extracted and specific protein levels identified by HA antibodies. ARase αs 15 (b) Transduced antisense fragments were able to increase the proliferation index of CD1 MEFs over the vector alone. Mutant Csn2 αs p53 (273 H) was used as a positive control. Presenescent MEFs 10 HBRF αs were infected in passage 2 with viruses carrying mutant p53(273 H), the indicated antisense fragment or vector alone. Cells were selected for positive infection with antibiotic and seeded at low density at Relative cell number 5 passage 5 to perform a time course curve. In these conditions cell wt growth respond only to senescence bypass. (c) Cell proliferation of wild-type MEFs after infection with the indicated molecules. The 0 standard 3T3 passage protocol was followed as indicated in 56 7 8 9 10 11 12 13 Materials and methods.

Oncogene Cellular senescence bypass screen JFM Leal et al 1964 by controlling p16ink4a-induced arrest, supporting we have directly analysed the involvement of these genes a molecular mechanism for their biological role in in human tumors. senescence. Since loss of function might regulate the First, we checked the expression of the different genes effectors of senescence and the p53 and pRB pathways, of interest in several arrays of paired normal/tumoral tissue samples from the same patients. cDNA probes from each one of the candidate genes were radioactively 60 7 labeled and hybridized with the array of tumoral/normal ) 50 6 6 RNA samples. As an expression control, we hybridized the arrays with an ubiquitin-specific probe. We normal- 5 40 ized the signals of the specific probes against the signal 4 30 of ubiquitin and quantitated the signal in tumor and 3 normal samples. Finally, we identified those tumor 20 2 samples in which the signal was decreased by at least 10 50% in tumors with respect to normal tissue (Figure 3a). 1 We find that all three of our putative tumor suppressor 0 Luciferase activity (x10 0 genes had a marked reduction of their expression in s s s s  s s s Cp c v c some tumoral samples. An especially high percentage of Colony formation (% control) Csn2 BRF1 csn2 BRF1 the tumoral samples showed at least a 50% reduction of ARase ARase p53 the expression of Csn2 in thyroid tumors (Figure 3a). waf1p High percentages of Csn2 mRNA loss were observed in Vector ARase tumors of pancreas, breast, ovary, kidney, uterus and Doxo cis Doxo cis rectum. BRF1 mRNA was at least 50% lost in some 02406246 0 6 24 0 6 24 ovary, kidney, uterus and rectum tumors (Figure 3a). p53 p53 Aldose reductase was lost in a few tumor types, mainly p21 p21 in thyroid, uterus, lung and stomach and at a lower percentage than the other two genes (Figure 3a).  -tub -tub To complement these studies, we have performed Csn2 BRF1 quantitative analysis of the expression of Csn2, ARase Doxo cis Doxo cis and BRF1 in three different types of tumors: lung, colon 0 6 24 0 6 24 0 6 24 0 6 24 and prostate. In each case, we quantified the mRNA in p53 p53 the tumoral sample versus the normal sample of the same p21 p21 tissue of the same patient (Figure 3b and Supplementary -tub -tub Materials 3). Our data show that all of the genes are reduced by at least 50% in the three types of tumors 10 (Figure 3b). When we averaged the total level of mRNAs of the different genes in normal tissue and compared

) 8 6 these levels with the same in the tumoral samples, we 6

4 Figure 2 The antisense fragments partially bypassed both p53- or Luciferase activity (x10 2 p16-induced growth arrests. (a) p53-null MEFs stably transfected to express the temperature sensitive mutant p53val135 were grown at 0 39 1C and transduced with each of the antisenses. Cultures were then s s s s transferred to 32 1C and colonies arising from cells with bypassed Cp     c v c p53-induced growth arrest were counted. Data are given as the csn2 relative percentage of colonies arising under the previous conditions ARase BRF1 against the colonies arising from the same number of cells seeded E2F1 and cultured at 39 1C(b) HEK293 cells were transfected with either E2Fp pGL3-p21promoter alone, pGL3-p21promoter plus p53wt or pGL3-p21promoter plus p53wt and each one of the antisense vector ARase s fragments. Luciferase activity was measured after 48 h and is expressed relative to protein content. (c) NIH3T3 cells were transduced with vector control or each one of the antisense fragments and mass cultures were selected expressing the antisenses. The cells where then treated with 0.2 mg/ml of doxorubicin or 100 mM of cisplatin and analysed by western blot for p53 and p21 protein content 6 and 24 h after treatment. (d) MEFs transduced with each one of the antisense fragments and mass culture-selected were then transduced with p16Ink4a at low viral titter and cultured under antibiotic selection for 2 weeks. Cells that bypassed p16Ink4a- induced growth arrest and formed colonies were visualized with crystal violet. (e) HEK293 cells were cotransfected either with pGL3-E2Fprom, pGL3-E2Fprom and E2F1 or pGL3-E2Fprom, E2F1 and each of the antisense fragments. Luciferase activity was   Csn2 s BRF s measured 48 h after transfection, relative to protein content.

Oncogene Cellular senescence bypass screen JFM Leal et al 1965 * 50 45 12 40 Csn 2 10 * 35 * 30 8 25 6 * * 20 * 4 * * 15 10 * 2 * 5 Relative levels mRNA

* 0 % tumor samples mRNA lost 0 NTNTNT Prostate Lung Colon 7.5 Csn2

ARase Skin (10) ARase Pancreas (7) BRF1 Testis (10) 5.0 Thyroid (10) Cervix (11) Rectum (17) Uterus (17) Lung (13) Prostate (7) Kidney (25)

Colon (21) 2.5 Ovary (13) Stomach (18)

Breast (19) tumor(nºsamples) Relative levels mRNA *** 0.0 50 NTNTNT *** 45 Prostate Lung Colon *** ** ** 40 50 ** BRF1 35 ** Lung 40 30 Colon 25 Prostate 30 * 20 20 15 10 10 5 Relative levels mRNA 0 NT N T N T

% tumor samples mRNA is >50% 0 decreased over non tumoral tissue* Csn2 ARase BRF1 Prostate Lung Colon Figure 3 Percentage of tumor samples showing decreased levels of mRNA. Cancer Profiling Array membranes (BD Biosciences) with paired non tumoral/tumoral samples from the same patient were hybridized with P32-labeled probes of the genes of interest to determine the respective amount of mRNAs. (a) Relative amount of repressed samples (>50%) for each type of tumor in relation to their paired sample of normal tissue. (b) Expression of the genes of interest is reduced in tumoral samples. Average amount of Csn2, ARase and BRF1 mRNA present in tumors versus non tumoral samples from the same patient. Individual data is presented in the Supplementary material 2. (c) Average of the mRNA quantitative measurements in tumoral vs non tumoral samples for each one of the genes analysed. Statistical analysis was performed by unpaired t-test. *Po0.05; **Po0.005; ***Po0.001. found a consistent reduction of the Csn2, ARase and BRF1 showed single allelic genomic losses in all tumor BRF1 mRNAs in the tumoral samples (Figure 3c). types analysed (Figure 4a) with variations amongst tumor types. The analysis of Aldose Reductase was hampered by the existence of several pseudogenes that complicated Candidate tumor suppressor genes are located in the interpretation of the data (data not shown). chromosomal regions frequently lost in cancer Csn2 showed specific allelic losses in 38.7% of thyroid Finally, we wanted to investigate whether the genes tumors correlating with the high levels of mRNA losses identified in our loss-of-function screening were detected previously in Figure 3a. Cns2 was also lost in affected by allelic loss in cancer, since they are located 33.3% of clear cell renal carcinomas, 6.25% of larynx in chromosomal regions significantly lost (>20%) in tumors, 23% rectum and 30% in colon carcinomas different tumors, according to data obtained in a CGH (Figure 4b). Also HBRF1 was lost in 50% CCRC and in tumor database (see Supplementary Information 3). 23% of rectal tumors (Figure 4a). Nevertheless, the Thus, Csn2 is located at 15q21.2, Aldose reductase percentage of HBRF1 loss in other tumors was smaller, at 7q35, and BRF1 at 14q32.33. Therefore, we have accounting for 6.25% in larynx, 7% in thyroid and explored whether these genes could be the putative 7.7% in colon. tumor suppressors localized in these loci. To that end, we analysed the allelic losses of these genes in DNA from tumor samples of different origins. We Csn2, BRF1 and ARase reduced the foci induced by analysed samples from thyroid (n ¼ 31), kidney (n ¼ 9, all oncogenic ras clear-cell renal carcinoma), larynx (n ¼ 16), colon (n ¼ 13) To finally assess the capability of these genes to act as and rectum (n ¼ 12) tumors (Ref 4). Analysis of Csn2 and tumor suppressors we analysed the alteration of the

Oncogene Cellular senescence bypass screen JFM Leal et al 1966 a 300 Csn2 +Ras 250 Exon 1 Control 2 200 Exon 13 Exon 5 Exon 10 150 Control 1 Control 3

Control 4 100

Number of colonies 50

0 V Csn2 ARase BRF1 BRF1 10 b +Ras Exon 2 Control 1 Exon 5 Control 2 Exon 8 Control 3 Control 4 Exon 17 5 Number of foci

*** Thyroid (n=31) 50% 0 Kidney (n=9) V Csn2 ARase BRF1 45% *** Larynx (n=16) *** Figure 5 (a) Antisense fragments obtained in the screening 40% Colon (n=13) ** increase the colony number induced by Ras in soft agar. NIH3T3 35% ** Rectum (n=12) stably expressing either of the antisense fragments against Csn2, 30% BRF1 or Arase were infected with viruses carrying oncogenic Hras ** 4 25% (Val12). After selection, 10 Cells were seeded in soft agar and cultured following the procedure described in Materials and 20% methods, 4 weeks later they were fixed, pictures taken. Data shows genomic loss 15% the average number of colonies of one representative experiment 10% performed in triplicate. (b) Antisense fragments obtained in the 5% screening reduced the foci formation by oncogenic ras in NIH3T3. NIH3T3 seeded at 50% confluence were co-transfected with a 0% plasmid carrying rasval12 under LTR promoter and either of the Csn2 BRF1 Csn2, Arase or BRF1 full length cDNAs. Cells were cultivated for Figure 4 Analysis of genomic loses for Csn2 and BRF1. 15 days without selection changing medium every 3 days. At this (a) Example of allelic loss of Csn2 and BRF1 in one tumoral point plates were stained with crystal violet and washed extensively. sample. (b) Percentage of tumors showing allelic loss in each type Results show the average number of foci per 2.5 cm dish in two of tumor. Statistical analysis was performed by unpaired t-test. independent experiments performed in triplicate. *Po0.05; **Po0.005; ***Po0.001.

oncogenic susceptibility induced in the absence or indicate that these genes might be potential tumor presence of the genes identified. To such extent we suppressor candidates. expressed oncogenic ras in NIH3T3 cells previously expressing each of the antisense fragment identified. After selection, cells were cultured in soft agar and the Discussion ability to grow in semisolid media assessed. We observed that the different antisenses increased the growth Exploiting the entry of MEFs into culture stress-induced potential in soft agar in around 50% (Figure 5a). A senescence, we have identified three genes through which further prerequisite to be considered as tumor suppres- loss of function, at least partially, might downregulate sor candidate is the ability of the genes to reduce the growth arrest. These three genes belong to very different oncogenic potential. To that end, we coexpressed biochemical families and have not been associated oncogenic ras and either of the full-length cDNAs of before with this phenotype. Furthermore, the analysis the genes identified in a classical foci formation assay of the genetic status of these genes shows that they are (Carnero and Beach, 2004). Ras induced a number lost through different mechanisms in tumor samples of foci in NIH3T3 (Figure 5b). However, the ectopic from several tissues. These results suggest that these overexpression of Csn2, BRF1 or Arase almost com- genes might function as tumor suppressors and their loss pletely abolish foci formation (Figure 5b). The fact that could confer some advantage to tumoral cells. the absence of the genes increases transforming proper- Csn2 is a subunit of the COP9 signalosome, a ties and its overexpression reduced these properties multimeric protein complex conserved among most

Oncogene Cellular senescence bypass screen JFM Leal et al 1967 eukaryotes (Wei et al., 1998; Bech-Otschir et al., 2002). of tumors of different origins and with a higher Disruption of csn2 in mice produced p53 accumulation incidence in lung, breast, brain and colorectal carcino- and early embryonic death (Lykke-Andersen et al., mas. The average incidence of csn2 allelic loss across 2003). It was shown that COP9 signalosome-associated different tumors (26.99%) is in the range observed for kinase activity phosphorylates p53 in the core domain well known tumor suppressor genes, such as pRB and targets it for proteasomal degradation (Bech- (38.64%), PTEN (30.68%) or BRCA2 (24.01%). This Otschir et al., 2001). Recently, it has been proposed observation is in agreement with the existence of a that the Cop9 signalosome is essential for the activity of candidate tumor suppressor gene in the 15q21.2 . MLF1, a negative regulator of cell-cycle progression Our hypothesis is further supported with detailed functioning upstream the tumor suppressor p53 cytogenetic studies using microsatellite probes to detect (Yoneda-Kato et al., 2005). Disruption of Csn2 by specific loss of heterogeneity Frequency of loss ranges anti-Csn2 antibodies inhibits signalosome-dependent from 27% in cervical carcinomas to 62% in head and p27 stabilization and G1/Sarrest (Yang et al., 2002). neck squamous cell carcinomas (HNSCC) (See Table 2 Knockdown of the Csn3 protein by siRNA interferes in Supplementary Information). The aldose reductase with the activation of p53 by genotoxic stress, implying gene is located at 7q35. This region has a lower inci- that the COP9 signalosome is an essential factor in dence of loss in tumors, still being significant in the chain of events leading to p53 activation after DNA some tumors such as HNSCC (22.22%) (See Table 1 damage (Yoneda-Kato et al., 2005). Our data support in Supplementary Information). LOH analysis shows the notion that the COP9 signalosome is essential for a higher impact, especially in oropharyngeal epithe- p53 activity, since depletion of csn2 with antisense lial carcinoma, with a 50% allelic loss (see Table 2, noticeably affects the ability of p53 to induce cell cycle in Supplementary Information). A similar situation arrest. Furthermore, the same antisense partially accounts for BRF1, at 14q32.33. CGH analysis shows abolishes p53 transactivation of the p21 promoter and a moderate loss, with a higher incidence (20.69%) in impairs the increase of p21 protein after treatment with colorectal carcinomas. LOH data confirms this observa- DNA damaging agents, indicating that the COP9 tion, providing higher values of allelic loss (up to 66% in signalosome acts somehow downstream p53. metastatic renal carcinoma) and affecting a variety of BRF1 is a subunit of the RNA polymerase III tumor types (See Supplementary Information). transcription factor IIIB. RNA polymerase III synthe- The data presented in this paper support the notion sizes several small RNA species, such as the tRNAs and that the genes obtained in the screening could partici- 5 SrRNA. In cell cycle-arrested cells, BRF1 levels pate as tumor suppressors in the cancer-safeguarding are markedly reduced due to a decrease in protein mechanism of senescence. Further research will be stability (Eichhorn and Jackson, 2001). Furthermore, necessary to fully understand the mechanism of action pRB regulates RNA polymerase III somehow binding of all three genes, and to add new insight into the possi- to BRF1 (Larminie et al., 1997). We do not have any ble prognostic value of the loss or mutation of these mechanistic insight about the mode of action of Brf1 on genes in cancer. the p53 pathway impairing p21 transcription. Aldose reductase (1.1.1.21) is an enzyme that parti- Materials and methods cipates in glucose metabolism, plays a protective role against toxic aldehydes and is important for osmo- Cell culture regulation in the kidney. There is some evidence that MEFs were generated from 13.5 day CD1 mouse embryos. expression of AKR1B1 is decreased in adrenocortical The head and blood organs were eliminated and the rest of the cancer (Lefrancois-Martinez et al., 2004). embryo was minced and dispersed in 0.1% trypsin (60 min at The expression of the three genes obtained in our 37 1C). Adherent cells growing in Dulbecco’s modified Eagle’s screening was altered in tumoral tissue when compared medium (Gibco, Invitrogen SA, Barcelona, Spain) supplemen- with matched non-tumoral samples from the same tissue ted with 10% FBS(Sigma)were cultured during 2 population doublings and then frozen. MEFs were subcultured 1:4 when and patient. We observed that Csn2 expression was they reached confluence; each passage was considered two reduced in 18.5% of the tumors analysed, with 50% population doublings (PDs). For colony-formation assays, of thyroid tumors showing complete disappearance (5 cells were plated at 103 cells per 10 cm plate. After 10 days, of 10 samples analysed). The expression of the other colonies were stained with crystal violet and counted. two genes was lost with lower frequencies, indicating 3T3 protocol. Every 3 days, cells were trypsinized, that their activity can probably be lost through other counted and 106 cells were plated per 10-cm plate. The relative mechanisms, such as mutations or post-translational number of cells is considered as a measure of the number of modifications. We did not investigate whether each one cells per passage related to the initial number of cells seeded of the genes behave as tumor suppressors in a specific per plate. tissue, but the high specificity of the loss of expression of each gene in very unique tumoral tissues strongly Growth in soft agar To measure the anchorage-independent growth, 104 cells were suspended in 1.4% agarose D-1 Low suggests this possibility. EEO (Pronadisa) growth medium containing 10% FBS, The three genes identified in our loss-of-function disposed onto a solidified base of growth medium containing screening are located in regions with signi- 2.8%. agar (agarose D-1 Low EEO, Pronadisa), and overlaid ficant (>20%) allelic loss in different tumors. Csn2 is with 1 ml of growth medium. After 24 h, media containing located at 15q21.2, a region frequently lost in a variety 10% FBSwere added to each 35 mm dish and renewed two

Oncogene Cellular senescence bypass screen JFM Leal et al 1968 times weekly. Colonies were scored 3 weeks after, and all Q RT–PCR values were determined in triplicate. Normal tissue and tumor tissue from 20 patients with colon carcinoma, 20 patients with prostate carcinoma and 20 Foci formation In all 50% confluent NIH3T3cells were patients with lung carcinoma were randomly chosen from transfected with 0.1 ı` g of pbabepuro-ras (val12) and 0.5 ı` gof the tumor bank at the Pathology Department of Vall pwzl-hygro alone or carrying the full length cDNA of Csn2, d’Hebro´ n Hospital (Barcelona, Spain). Biopsy samples are Arase or BRH1 using Jet-Pei as indicated by the vendor. Cells routinely collected, quickly frozen and stored at –801C imme- were cultured 15 days, then fixed and stained with crystal diately after surgery. All tumors were histologically examined violet. to confirm the diagnosis of carcinoma. All procedures of the present study have been approved by the Ethics Committee Library generation, transduction and recovery of the active of the Hospital Vall d’Hebron. proviruses Total RNA was extracted from normal and tumor tissue As a source of the mRNA, we used MEFs terminally arrested with the RNAeasy mini kit (Qiagen, Hilden, Germany). The at replicative senescence. Aliquots of 2 mg of total mRNA were RNA nano Lab Chip kit (Agilent, Palo Alto, CA, USA) was used for the generation of the library. Randomly primed used to quantify and determine the integrity of the isolated cDNA fragments of the polyA þ mRNA were synthesized, total RNA. cDNA synthesis was done using random primers selected by size (50–500 nucleotides) on a S400 column with SuperScriptTM II reverse transcriptase (Invitrogen, (Pharmacia and Upjohn SA, Barcelona, Spain) and cloned Carlsbad, CA, USA) and aliquots were stored at À20 1C. into the EcoRI and XhoI sites of pMARXIVpuro in the Quantitative real-time TaqMan reverse transcriptase–PCR antisense orientation. technology (Applied Biosystems, Foster City, CA, USA) was Transduction of the library and recovery of the proviruses used to determine the differential expression of the selected was performed as described previously by Carnero and co- genes. Relative quantification analysis was performed with the workers (Carnero et al., 2000). ABI PRISM 7700 instrument (Applied Biosystems). Data were analysed with sequence detection software. The PCR cycling Western blot analysis program consisted of denaturing at 95 1C for 10 min and 50 Cells were washed twice with ice-cold phosphate-buffered cycles at 95 1C for 15 s, and annealing and elongation at 60 1C saline (PBS) and lysed by sonication in lysis -buffer (50 mM for 1 min. Tris–HCl pH 7.5, 1% NP-40, 10% glycerol, 150 mM NaCl, Cyclophilin (ref. 4326316E), an endogenous control, was 2mM Complete protease inhibitor cocktail Roche Diagnostics, used to normalize variations in cDNA quantities from Barcelona, Spain). The protein content of the lysates was different samples. Each reaction was performed in triplicate determined by the modified method of Bradford using bovine with cDNA from normal and tumor tissue from each patient serum albumin (BSA) as a standard. The appropriate quantity studied. A blank sample (no DNA) was included in each of protein was then dissolved in Laemmli buffer (62.5 mM experiment. A new RNA extraction was randomly performed Tris–HCl pH 6.8, 10% (v/v) glycerol, 1% (w/v) SDS, 5% (w/v) from the original tissue of some samples and reproducible 2-mercaptoethanol, 0.0025% (w/v) bromphenol blue -Sigma-) quantitative real-time PCR results were obtained (data not and these samples were then separated by SDS–PAGE, shown). transferred on to Immobilon-P membrane (Millipore, Iberica SA, Madrid, Spain), and immunostained. The following primary antibodies were used: PAb FL-393 anti-p53 (Santa Multiplexed genomic loss detection Cruz Biotechnology Inc., Santa Cruz, CA, USA; sc-6243), To investigate the presence of deletions affecting the BRF1 MAb anti-p21 (Oncogene Research Products, EMD Bio- and Csn2 genes, we designed a technique based on specific sciences Inc., San Diego, CA, USA; OP64), MAb anti-HA multiplex amplification of each gene. We first designed and (Sigma-Aldrich, St Louis, MO, USA; clone HA7), MAb anti-a labeled (50 6-FAM) a pair of primers for each exon. The primer tubulin (Sigma T9026), Ab anti-p53 P-ser18 (Cell Signalling pairs for amplification were designed on the basis of genomic Technology, Danvers, MA, USA), and horseradish peroxi- sequences of BRF1 and Csn2 ( Gene ID: 2972 and 9318, dase-labeled rabbit anti-mouse (Promega Corp., Madison, WI, respectively), and were as follows: TRIP15-1F (50-AGCTGAG USA; 17331001), goat anti-rabbit (Calbiochem, EMD Bio- AGTGACGCCTCTG-30); TRIP15-1R (50-ATCACCACCCT sciences Inc., San Diego, CA, USA; 401315) and rabbit anti- CAGAGTTCC-30); TRIP15-6F (50-CAGGTACAAATTTGC goat (Santa Cruz sc-2768) as secondary antibodies. Proteins TGTTTACAA-30); TRIP15-6R (50-AGCTACCTGGCACGA were visualized using the ECL detection system (Amersham CTGAT-30); TRIP15-10F (50-TTGAGAAGCAGGGTCTTT Biosciences, Fairfield, CT, USA). GG-30); TRIP15-10R (50-TGGCATTTCTTTGACAGTGG-30); TRIP15-13F (50-TTTAATGTAGGCAGCTGCTTTTT-30); RNA array hybridization TRIP15-13R (50-GCACCACCCCTCTTCTGAT-30); BRF1-2F Cancer profiling array membranes (BD Biosciences, Franklin (50-CCCGCTTTGTTCTCACTTTG-30); BRF1-2R (50-CGGT Lanes, NJ, USA) were pre-hybridized with ExpressHyb GGGAAATACCATTCTG-30); BRF-5F (50-ACCAATGCTG hybridization solution for 4 h at 65 1C. The appropriate probe TTTCTGGTTG-30); BRF-5R (50-GCACGCTGGTCCCTAC was then labeled by PCR with 50 1C of redivue dCTP32 AG-30); BRF-8F (50-AGCTCCATTTCCCATGTCTG-30); (Amersham), using the primers previously described for the BRF-8R (50-GCTCTGCTACAACCCCAGTC-30); BRF-17F reverse transcriptase–PCR. The labeled probe was then (50-GGTCTGAGGGGTCTGTTGAG-30); BRF-17R (50-TGA purified from free hot nucleotides with a sepharose G-50 AGCTAACAGGCCCAAGT-30). We used control fragments column Nick (Amersham). The purified probe was then from 1, 3 and 11 as internal controls of the denatured for 3 min at 100 1C and added to the hybridization assay. We amplified genomic DNA of tumors by means of a solution. The hybridization was performed overnight at 65 1C. multiplex PCR kit, following standard recommendations Then the membrane was washed at 65 1C two times with 2 Â (Qiagen GmbH). Briefly, multiplex amplifications were per- SSPE, 0.1% SDS; once with 1 Â SSPE, 0.1% SDS and once formed in 25 ml of a mixture containing 1 Â multiplex PCR more with 0.1 Â SSPE, 0.1% SDS. The membrane was then master mix, 0.2 mM of each primer, and 100–200 ng of genomic exposed to a Biomax MSfilm (Kodak). DNA. The PCR program started with an initial heat-activating

Oncogene Cellular senescence bypass screen JFM Leal et al 1969 step at 95 1C for 15 min to activate HotStartTaq DNA frozen, thawed and centrifuged 2 min at 13 200 r p m , 4 1C. polymerase. Amplification was for 20 cycles: 30 s at 94 1C, Twenty microliters of each sample were then transferred to 90 s at 60 1C and 90 s at 72 1C, followed by 10 min of final a 96 -well plate and luciferase activity measured with the extension at 721C. PCR amplification products obtained from automated addition of 100 ml of luciferase substrate (Promega) the deletion tests were used for fragment analysis on an ABI by a Victor II (Wallac Oy) reader. Firefly luciferase activity was PRISMTM 310 capillary sequencer (Applied Biosystems, then normalized with Renilla luciferase activity. Perkin Elmer), and analysed using GeneScan v3.1 (Applied Biosystems). Normalization was performed by overlapping Statistical analysis tumor samples with a control sample, determining the peak t- surface of all fragments and calculating the normal peak Statistical analysis was performed by unpaired test compar- fractions (Schouten et al., 2002). ing against the percentage of samples that show >50% increase in normal samples vs tumoral samples of the same patient (Figure 3), against the frequency of allelic loss Luciferase assays in non tumoral samples (Figure 4). *Po0.05; **Po0.005; Cells were seeded in 6-well plates and cultured for 24 h at ***Po0.001. 37 1C. Then the medium was changed and 2 h later, the cells were transiently transfected overnight with the appropriate DNAs (and additionally with phRG-TK vector, a renilla Acknowledgements luciferase reporter) using the calcium phosphate protocol. The cells were then washed from the calcium phosphate crystals We thank the CNIO Tumor Bank for providing the tumor with PBSand a glycerol shock (DMEM supplemented with samples. This work has been partly supported by grants from 10% glycerol) was performed during 1 min. Fresh medium was Spanish Ministry of Health (FIS-02/0126), Fundacio´ n Mutua added to the transfected cells and cultured at 37 1C. Cells were Madrilen˜ a and the Spanish Ministry of Education and Science harvested 48 h after transfection and lysed with 200 mlof1Â (SAF2005-00944) to AC and MRC and Wellcome Trust (to Reporter Lysis Buffer (Promega). The lysates where then DHB).

References

Bech-Otschir D, Kraft R, Huang X, Henklein P, Kapelari B, Pollmann with malignancy in human sporadic adrenocortical tumors. J Clin C et al. (2001). COP9 signalosome-specific phosphorylation targets Endocrinol Metab 89: 3010–3019. p53 to degradation by the ubiquitin system. Embo J 20: 1630–1639. Lleonart ME, Vidal F, Gallardo D, Diaz-Fuertes M, Rojo F, Bech-Otschir D, Seeger M, Dubiel W. (2002). The COP9 signalosome: Cuatrecasas M et al. (2006). New p53 related genes in human at the interface between signal transduction and ubiquitin-depen- tumors: significant downregulation in colon and lung carcinomas. dent proteolysis. J Cell Sci 115: 467–473. Oncol Rep 16: 603–608. Berns K, Hijmans E M, Mullenders J, Brummelkamp TR, Velds A, Lykke-Andersen K, Schaefer L, Menon S, Deng XW, Miller J B, Heimerikx M et al. (2004). A large-scale RNAi screen in human cells Wei N. (2003). Disruption of the COP9 signalosome Csn2 subunit identifies new components of the p53 pathway. Nature 428: 431–437. in mice causes deficient cell proliferation, accumulation of p53 Braig M, Lee S, Loddenkemper C, Rudolph C, Peters AH, and cyclin E, and early embryonic death. Mol Cell Biol 23: Schlegelberger B et al. (2005). Oncogene-induced senescence as an 6790–6797. initial barrier in lymphoma development. Nature 436: 660–665. Michaloglou C, Vredeveld LC, Soengas MS, Denoyelle C, Kuilman T, Campisi J. (2001). Cellular senescence as a tumor-suppressor mechan- van der Horst CM et al. (2005). BRAFE600-associated senescence- ism. Trends Cell Biol 11: S27–S31. like cell cycle arrest of human naevi. Nature 436: 720–724. Campisi J. (2005). Suppressing cancer: the importance of being Morales CP, Holt SE, Ouellette M, Kaur KJ, Yan Y, Wilson KS et al. senescent. Science 309: 886–887. (1999). Absence of cancer-associated changes in human fibroblasts Carnero A, Beach DH. (2004). Absence of p21WAF1 cooperates with immortalized with telomerase. Nat Genet 21: 115–118. c-myc in bypassing Ras-induced senescence and enhances oncogenic Sage J, Mulligan GJ, Attardi LD, Miller A, Chen S, Williams B et al. cooperation. Oncogene 23: 6006–6011. (2000). Targeted disruption of the three Rb-related genes leads to Carnero A, Hudson JD, Hannon GJ, Beach DH. (2000). Loss-of- loss of G(1) control and immortalization. Genes Dev 14: 3037–3050. function genetics in mammalian cells: the p53 tumor suppressor Schmitt CA, Fridman JS, Yang M, Lee S, Baranov E, Hoffman RM model. Nucleic Acids Res 28: 2234–2241. et al. (2002). A senescence program controlled by p53 and Collado M, Gil J, Efeyan A, Guerra C, Schuhmacher AJ, Barradas M p16INK4a contributes to the outcome of cancer therapy. Cell 109: et al. (2005). Tumour biology: senescence in premalignant tumours. 335–346. Nature 436: 642. Schouten JP, McElgunn CJ, Waaijer R, Zwijnenburg D, Diepvens F, Dimri GP, Lee X, Basile G, Acosta M, Scott G, Roskelley C et al. Pals G. (2002). Relative quantification of 40 nucleic acid sequences (1995). A biomarker that identifies senescent human cells in culture by multiplex ligation-dependent probe amplification. Nucleic Acids and in aging skin in vivo. Proc Natl Acad Sci USA 92: 9363–9367. Res 30: e57. Eichhorn K, Jackson SP. (2001). A role for TAF3B2 in the repression Serrano M, Blasco MA. (2001). Putting the stress on senescence. Curr of human RNA polymerase III transcription in nonproliferating Opin Cell Biol 13: 748–753. cells. J Biol Chem 276: 21158–21165. Serrano M, Lin W, McCurrach ME, Beach D, Lowe SW. (1997). Greider CW, Blackburn EH. (2004). Tracking telomerase. Cell 116: Oncogenic ras provokes premature cell senescence associated with S83–S86, 1 p following S86. accumulation of p53 and p16INK4a. Cell 88: 593–602. Hanahan D, Weinberg RA. (2000). The hallmarks of cancer. Cell 100: Shay JW, Roninson IB. (2004). Hallmarks of senescence in carcino- 57–70. genesis and cancer therapy. Oncogene 23: 2919–2933. Larminie CG, Cairns CA, Mital R, Martin K, Kouzarides T, Jackson Sherr CJ, McCormick F. (2002). The RB and p53 pathways in cancer. SPet al. (1997). Mechanistic analysis of RNA polymerase III Cancer Cell 2: 103–112. regulation by the . Embo J 16: 2061–2071. Suzuki K, Mori I, Nakayama Y, Miyakoda M, Kodama S, Watanabe Lefrancois-Martinez AM, Bertherat J, Val P, Tournaire C, Gallo- M. (2001). Radiation-induced senescence-like growth arrest requires Payet N et al. (2004). Decreased expression of cyclic adenosine TP53 function but not telomere shortening. Radiat Res 155: monophosphate-regulated aldose reductase (AKR1B1) is associated 248–253.

Oncogene Cellular senescence bypass screen JFM Leal et al 1970 Wei N, Tsuge T, Serino G, Dohmae N, Takio K, Matsui M et al. Yang X, Menon S, Lykke-Andersen K, Tsuge T, Di X, Wang X et al. (1998). The COP9 complex is conserved between plants and (2002). The COP9 signalosome inhibits p27(kip1) degradation and mammals and is related to the 26Sproteasome regulatory complex. impedes G1-Sphase progression via deneddylation of SCFCul1. Curr Biol 8: 919–922. Curr Biol 12: 667–672. Yamasaki L, Bronson R, Williams BO, Dyson N J, Harlow E, Jacks T. Yoneda-Kato N, Tomoda K, Umehara M, Arata Y, Kato JY. (2005). (1998). Loss of E2F-1 reduces tumorigenesis and extends the Myeloid leukemia factor 1 regulates p53 by suppressing COP1 via lifespan of Rb1(+/-)mice. Nat Genet 18: 360–364. COP9 signalosome subunit 3. Embo J 24: 1739–1749.

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

Oncogene