Oncogene (2008) 27, 2084–2090 & 2008 Nature Publishing Group All rights reserved 0950-9232/08 $30.00 www.nature.com/onc ORIGINAL ARTICLE Array CGH and -expression profiling reveals distinct genomic instability patterns associated with DNA repair and cell-cycle checkpoint pathways in Ewing’s sarcoma

BI Ferreira1, J Alonso2,7, J Carrillo2, F Acquadro1, C Largo1, J Suela1, MR Teixeira3, N Cerveira3, A Molares4, G Gome´ z-Lo´ pez4,5,A´ Pestan˜ a2, A Sastre6, P Garcia-Miguel6 and JC Cigudosa1,7

1Molecular Cytogenetics Group, Centro Nacional de Investigaciones Oncolo´gicas (CNIO), and CIBER on Rare Diseases (CIBERER), Madrid, Spain; 2Departamento de Biologı´a Molecular y Celular del Ca´ncer, Instituto de Investigaciones Biome´dicas ‘A. Sols’ (CSIC-UAM), Madrid, Spain; 3Department of Genetics, Portuguese Oncology Institute and Biomedical Sciences Institute (ICBAS), Porto, Portugal; 4Fundacio´n Biome´dica del CHUVI, Hospital do Rebullo´n, Vigo, Pontevedra, Spain; 5Unidad de Bioinforma´tica, Centro Nacional de Investigaciones Oncolo´gicas (CNIO), Madrid, Spain and 6Unidad Oncohematologı´a Infantil, Hospital Infantil La Paz, Madrid, Spain

Ewing’s sarcoma (ES) is characterized by specific Oncogene (2008) 27, 2084–2090; doi:10.1038/sj.onc.1210845; translocations, the most common being t(11;22)(q24;q12). published online 22 October 2007 Additionally, other type of genetic abnormalities may occur and be relevant for explaining the variable tumour biology and Keywords: Ewing’s sarcoma; arrayCGH; expression clinical outcome. We have carried out a high-resolution array profile; genomic instability CGH and expression profiling on 25 ES tumour samples to characterize the DNA copy number aberrations (CNA) occurring in these tumours and determine their association with gene-expression profiles and clinical outcome. CNA were observed in 84% of the cases. We observed a median number Introduction of three aberrations per case. Besides numerical chromosome changes, smaller aberrations were found and defined at Ewing’s sarcoma (ES) is the second most common 5p, 7q and 9p. All CNA were compiled to malignant bone tumour in children and young adults define the smallest overlapping regions of imbalance (SORI). (Kovar, 1998), and belongs to a wider group of A total of 35 SORI were delimited. Bioinformatics analyses neuroectodermal tumours known as ES Family of were conducted to identify subgroups according to the pattern Tumours (Ushigome et al., 2002). Their genetic hallmark of genomic instability. Unsupervised and supervised clustering is the presence of the reciprocal t(11;22)(q24;q12) translo- analysis (using SORI as variables) segregated the tumours in cation (Aurias et al., 1984; Mugneret et al., 1988), or other two distinct groups: one genomically stable (p3 CNA) and less frequent variant translocations t(21;22)(q12;q12) and other genomically unstable (>3 CNA). The genomic unstable t(7;22)(p22;q12), and their morphology falls in the small- group showed a statistically significant shorter overall survival blue-round-cells category. The t(11;22) translocation leads and was more refractory to chemotherapy. Expression profile to the fusion of the EWS gene on 22q12 with the FLI1 analysis revealed significant differences between both groups. gene on 11q24, resulting in the formation and expression related with chromosome segregation, DNA repair of the chimaeric transcript EWS-FLI1 (Delattre et al., pathways and cell-cycle control were upregulated in the 1992) that is thought to contribute to the pathogenesis of genomically unstable group. This report elucidates, for the first this tumour by modulating the expression of target genes. time, data about genomic instability in ES, based on CNA and Secondary chromosomal aberrations acquired in addi- expression profiling, and shows that a genomically unstable tion to the EWS rearrangements have been reported for group of Ewing’s tumours is correlated with a significant poor many years based on conventional cytogenetic approaches. prognosis. The most common correspond to trisomies of chromosome 8 (Mugneret et al., 1988), 12 and gain of DNA sequences in Correspondence: Dr JC Cigudosa, Molecular Cytogenetics Group, 1q (Armengol et al., 1997). The relation between these Centro Nacional de Investigaciones Oncolo´ gicas (CNIO), 3 C/Melchor secondary changes and clinical outcome is quite contro- Ferna´ ndez Almagro, Madrid 28089, Spain. E-mail: [email protected] (or) versial. Some authors refer that these aberrations may Dr J Alonso, Departamento de Biologı´ a Molecular y Celular del contribute to tumour progression and may serve as criteria Ca´ ncer, Instituto de Investigaciones Biome´ dicas ‘A Sols’, CSIC-UAM, for the aggressiveness of the disease (Zielenska et al., 2001), C/ Arturo Duperier 4, Madrid 28029, Spain. although this relationship is not clearly established. E-mail: [email protected] Development of new techniques as microarrays provides 7These two authors contributed equally to this work. Received 12 March 2007; revised 4 July 2007; accepted 10 September an opportunity to perform genome-wide analysis at 2007; published online 22 October 2007 different levels. Previous works have focused their research Genomic instability in Ewing’s sarcoma BI Ferreira et al 2085 on array expression profiles of ES cell lines and primary and chromosome-based CGH have been previously used tumours in order to identify tumour classification profiles, to study genomic alterations in ES (Brisset et al., 2001; new EWS/FLI1 targets or to reveal distinct expression Ozaki et al., 2001; Hattinger et al., 2002). Recurrent signatures associated to different outcomes (Khan et al., changes involving whole chromosome numeric aberrations 2001; Ohali et al., 2004; Bandres et al., 2005; Mendiola were all confirmed in our series but aCGH also unveiled et al., 2006). A more recent genomic approach is the use of frequent chromosomal changes such as partial losses at array-based comparative genomic hybridization (aCGH), 7q11.2 (5/25) or total/partial 5p gains (5/25) not previously which has been successfully used for the detection of described. genomic imbalances in human tumours (Pinkel and Application of aCGH facilitates the definition and Albertson, 2005) in a most precise manner than chromo- limits of the aberrant regions that are present in a some-based conventional techniques. The patterns of copy tumour sample, but it also provides a unique view of the number alterations identified by aCGH have been useful overall stage of genomic instability. Our study shows to differentiate tumours in more biologically and clinically that ES is a genomically heterogeneous disease at relevant subtypes and the higher resolution has led to diagnosis. Data obtained by aCGH provided a global precise mapping of the boundaries of amplified and profile for each tumour that was further analysed in deleted regions indicating candidate genes relevant to terms of genomic instability assessed by different cellular control pathways. parameters. First, we studied the extent of the genome Considering that our current knowledge of clinical affected by this instability, defined as the fraction (%) of and biological features is still far from depicting a the genome altered (FGA) by CNA. The median of complete scenario of ES, we combined different FGA for all samples was 6% (with a wide range between approaches to search for novel genomic changes that 0 and 52%), similarly to what has been described for occur during the development of this disease. In this bladder carcinoma (Blaveri et al., 2005). Other measures report, we describe for the first time the results of an of genomic instability were taken into account such as integrated genomic analysis using both array-based the number of whole chromosome aberrations, the CGH and gene-expression profiling (GEP) in ES number of copy number transitions (CNT) within a primary tumours. chromosome, the presence and number of high-level amplifications or homozygous deletions (HD), and the number of chromosomes containing such transitions. The genomic instability profile of ES is summarized in Results and discussion Figure 2a and all studied parameters are provided in the Supplementary Table S1. Chromosome numerical Genomic imbalance profiling changes were seen in 17 cases (68%) and 15 cases A total of 23 primary and 2 recurrent ES tumours were (60%) showed CNT. analysed by whole genome array CGH (aCGH). A total of Of interest, four HD were detected in four cases: case 21 tumours (84%) showed DNA copy number aberrations number 22 showed two located at 7p22 and 7q36, case (CNA). The most frequent gains involved entire chromo- numbers 24 and 21 showed one at 9p21 and case number somes 8 (56%), 12 (20%), 18 (12%), 20 (12%) and the 7 showed one at 1p36 (Table 1, validation by fluorescent short arm of chromosome 5 (5p) (20%). Most frequent in situ hybridization (FISH) is shown in Supplementary losses affected entire chromosomes 10 (16%) and 19 Figure S1). The HD region at 9p21 included the (16%), and partial regions of chromosome arms 16q CDKN2A/B loci, confirming the role of this tumour (16%) and 7q (25%). The summary of the CNA detected suppressor gene in ES pathogenesis (Kovar, 1998). by aCGH is shown in Figure 1. Conventional cytogenetics Among the other HD, the 1p36 region contains a

Figure 1 Ideogram of DNA copy number aberrations. All aberrant regions are represented along the chromosomes: black lines on the right of the chromosomes represent regions of gain and gray on the left side of the chromosomes indicate regions of losses.

Oncogene Genomic instability in Ewing’s sarcoma BI Ferreira et al 2086

Figure 2 Genomic analysis of Ewing’s sarcoma (ES) tumours reveals two subtypes. (a) Genomic instability graphic report. Each tumour is quantitatively characterized by its number of chromosome transitions (CNT) (yellow bars), number of chromosomes with structural changes (blue bars) and the number of aneuploid chromosomes (green bars). (b) Hierarchical unsupervised clustering. Purple squares refers to highly instability group of tumours (>3SORI) and blue squares refers to low instability group of tumours (0–3 SORI). Presence of trisomy of chromosome 8 is represented by ( þ ) and absence by (À). Most significant SORI are signed with an asterisk. GEPAS Suite software (http://gepas.bioinfo.cipf.es/) was used for supervised clustering and gene annotation analysis of CGH data.

substantial numbers of putative tumour suppressor (Figure 2b), two main branches were segregated. One of genes (Table 1) and it has been shown to be deleted in them, named genomic unstable group, is characterized other tumours, such as neuroblastoma (Thompson by a high number of aberrations (median number of 8 et al., 2001). On the other hand, high-level amplifica- SORI per case, ranging from 4 to 18) while the other, the tions were not observed, in accordance with previous genomic stable group, only showed a small number of chromosome-based CGH studies. SORI (median number of 1 SORI per case, ranging To disclose genomic groups within the series, an from 0 to 3). We observed that trisomy 8, the most unsupervised clustering analysis, using the presence of common secondary aberration in ES, was equally smallest overlapping regions of imbalance (SORI) as distributed in both branches, ruling out its potential variable, was performed. According to this analysis role in the genomic instability segregation. We then

Oncogene Genomic instability in Ewing’s sarcoma BI Ferreira et al 2087 decided to run a supervised clustering study of the Correlation of genomic instability with clinical outcome genomic profiling in our ES series, using the genomic Clinical data and follow-up were available from 20 instability groups previously mentioned. A total of six patients with tumour samples at diagnosis (Supplementary regions, mainly located in chromosomes 12 and 19, Table S2). We first analysed the relation between primary showed a statistically significant association (Po0.05) site location (axial vs peripheral), stage (localized vs with the highly unstable genomic group. The proposed metastatic), gain of chromosome 8 and overall genetic genomic stability groups also differed in their genomic status (stable vs unstable) with clinical outcome by instability parameters, as defined previously (Supple- univariate survival analysis. As shown in Figure 3, both mentary Table S1). The group with a higher degree of stage and overall genetic status at diagnosis were negative genomic instability showed a higher number of CNT, of predictors of overall survival (Po0.015 and Po0.017 chromosomes with structural changes, of whole chromo- respectively), while gain of chromosome 8 (the second some changes and a higher proportion of FGA than the more frequent chromosome alteration in ES) and primary group with lower number of changes. The latter is site location were not associated with survival. By multi- mainly composed of cases with the complete absence of variate analysis the genomic instability status remained as aberrations or a small percentage of cases that presented the most significant independent prognostic factor (Fig- numerical changes. Interestingly, the two recurrent ure 3). In addition, the patients from the genomic stable tumours included in our series (numbers 9 and 14) group showed a higher tendency to achieve complete belonged to the genetically unstable group, harbouring remission during or after treatment than those from an elevated number of chromosome alterations. genomic unstable group (100 vs 62%, respectively,

Table 1 Regions of homozygous deletions Tumour sample Chromosome Start (bp) End (bp) Size (bp) No probes Candidate genes

7 1 9 778 925 10 706 826 927 901 31 CLSTN1, CTNNBIP1, THC2037606, LZIC, NMNAT1, RBP7, UBE4B, KIF1B, PGD, APITD1, CORT, DFFA, PEX14, FLJ20321, ENST00000344008

22 7 1 349 463 1 622 358 272 895 8 MAFK, MGC9712, MGC10911, KIAA1908, chr7:001622299-001622358

22 7 99 885 360 99 923 768 38 408 5 Chr7:099885360-099885419, AK055267, GNB2, PERQ1

21;24 9 21 795 270 21 999 029 203 759 11 MTAP, C9orf53, CDKN2A, CDKN2B

Figure 3 Correlation of clinical and genomic data with overall survival. Kaplan–Meier plots of overall survival according to (a) primary site location, (b) stage, (c) genetic stability status and (d) chromosome 8 gain. Univariate analysis was performed using the log- rank test. Multivariate analysis was performed using Cox proportional-hazards regression. MedCalc software (Mariakerke, Belgium) was used for statistical computations.

Oncogene Genomic instability in Ewing’s sarcoma BI Ferreira et al 2088 P ¼ 0.027, w2-square test). These data showed that genomic GSEA revealed statistically significant associations unstable group are more refractory to chemotherapy and, (false-discovery rate (FDR) o0.25) between the un- as a consequence, are associated to a poor prognosis. stable genomic status and several Biocarta Pathways (www.biocarta.com). We found that most of differen- Gene-expression profiling and genomic aberrations tially expressed genes in the unstable group were correlation associated with DNA repair and cell-cycle control The definition of these two separated groups, based on pathways such as ATRBRCA or RB (see Table 2 for the presence of CNA, led us to wonder about possible details), association supported by the qRT–PCR results associations of CNA patterns with alterations in cellular (Figure 4). Thus, our data show that genomic instability pathways or genes involved in maintenance of genome in ES is associated with cell-cycle checkpoint pathways instability. For that, we analysed the gene-expression profiles and a superior activity in the machinery of chromosome of both groups of tumours (unstable vs stable) to look for segregation and mitosis. If this particular expression genes potentially related with the genomic instability profile. profile is the cause or the consequence of the genomic A group of 290 probes, corresponding to 265 unique genes, instability is at the moment largely unknown. were shown to be differentially expressed between both In summary, we have shown for the first time a study combining the use of high-resolution genome-wide groups of tumours (Po0.01) (Supplementary Table S3). A total of 10 of these genes that are known to be involved in screening tools such as aCGH and GEP, in a series of DNA repair and cell-cycle control pathways (BLM, ES primary tumours. Array-CGH allowed the definition CCNB1, CDKN1C, CENPE, CHEK1, DBF4, FANCM, of minor aberrations but also elucidated about the size FGFR10P2, RGS1 and TOP2A) were additionally analysed and gene content of previously described alterations. by quantitative real-time RT–PCR in order to validate With this study, we were able to segregate a homo- microarray data. As shown in Figure 4, all genes analysed, genous tumour series into two groups based on genomic except CDKN1C, were upregulated in the genomic unstable instability that are characterized by different patterns of group and these differences were highly significant in six of gene-expression profiles and clinical outcome. these genes (Figure 4). A functional analysis of this group of genes was conducted using their annotation to terms Materials and methods (see Supplementary methods for details). Genes impli- cated in mitosis and nuclear division and associated to Tumour samples and clinical data pericentromeric chromosome regions and to the kinesin Genetic and clinical data were collected from a series of 26 complex were significantly overrepresented in this group patients provided from Pathology Departments and Oncology of genes (Table 2 and Supplementary Figures S2 and S3). Units of several Spanish Children’s Hospitals and from Portuguese Institute of Oncology (IPO-Porto). A total of 24 A second approach was conducted by using the Gene Set samples were primary tumours obtained before any treatment Enrichment Analysis (GSEA) (Subramanian et al., 2005). and two of them were recurrent classified as ES harbouring the This bioinformatics analysis compares experiments of t(11;22) or t(21;22) based on immunohistologic examination and genome-wide expression profiles from samples belonging molecular detection (RT–PCR/cytogenetics) of fusion transcripts. to two classes (see Supplementary methods for details). All samples were processed for RNA and DNA extraction.

Figure 4 Validation of genes differentially expressed between genomic stable and unstable groups by quantitative real-time PCR. RNA expression levels of 10 genes differentially expressed between both groups of tumours, according to microarray studies, were analysed by quantitative real-time RT–PCR in 10 stable and 12 unstable tumours. Data shown represent the relative expression of each gene vs the control gene TATA binding (TBP) (mean7s.e.). Statistical significance (two-tailed Student’s t-test) is also shown.

Oncogene Genomic instability in Ewing’s sarcoma BI Ferreira et al 2089 Table 2 Functional and biological comparison of genes differentially considered for the study after filtering for quality of expresseda between stable and unstable genomic subgroups hybridization according to quality control parameter (DLRS) provided by the CGH Analytics v3.2.25 Software (Agilent Analysis based on gene ontology annotations Technologies). Microarray data were extracted with Feature GO ID GO term Fisher’s exact Extraction Software v8.1. Data analysis and chromosome test segmentation were performed with InSilicoArray CGH soft- ware—smoothing method (Eilers and de Menezes, 2005), Biological process categorizing changes as 0, 1 or À1 indicating no change, gain Go:0007049 Cell cycle 8.35EÀ08 or loss, respectively. Alterations in copy number with less than Go:0000278 Mitotic cell cycle 1.17EÀ07 5 consecutive probes and known copy number polymorphisms Go:0000279 M phase 2.45EÀ06 did not score as aberrations. ArrayCGH data have been Go:0000280 Nuclear division 4.86EÀ06 Go:0007067 Mitosis 7.28EÀ06 deposited in NCBIs Gene-Expression Omnibus (GEO, http:// www.ncbi.nlm.nih.gov/geo/) and they are accessible through Cellular component GEO Series temporary accession number 15303288 or freely at Go:0005871 Kinesin complex 1.27EÀ04 our ftp repository (www.cnio.es). Go:0000775 Chromosome, pericentric region 2.80EÀ04 Go:0005634 Nucleus 3.61EÀ04 Go:0005694 Chromosome 4.54EÀ04 Expression array Go:0005819 Spindle 2.32EÀ03 RNA was extracted from all samples. A total of 14 samples provided RNA suitable for hybridization (assessed with Molecular function Agilent 2100 Bioanalyzer, Agilent). Microarray experiments Go:0008094 DNA-dependent ATPase activity 8.23EÀ04 were performed with CodeLink Human 20K (GE Healthcare, Go:0004003 ATP-dependent DNA helicase 1.58EÀ04 formerly Amersham Biosciences, Piscataway, NJ, USA) activity Go:0003684 Damage DNA binding 3.33EÀ03 according to the manufactures’ protocol and scanned at Genomic Go:0003678 DNA helicase activity 8.83EÀ03 core facility of the Instituto de Investigaciones Biome´ dicas (see Go:0003676 Nucleic acid binding 9.63EÀ03 Supplementary methods for more details). Microarray data have been deposited in NCBIs Gene Expression Omnibus (GEO, Analysis based on gene enrichment GSEA (FDRo0.25) http://www.ncbi.nlm.nih.gov/geo/) and are accessible through GEO Series accession number GSE8303. Pathway name Brief description FDR q-val

G2PATHWAY Activated Cdc2-cyclin B kinase 0.020966584 Definition of smallest overlapping regions of imbalance and regulates the G2/M transition; FISH validation DNA damage stimulates the Aberrant recurrent regions were defined as a sequence of DNA-PK/ATM/ATR kinases, which inactivate Cdc2 altered probes common to a set of CGH profiles and SORI as ATRBRCA- BRCA1 and two block cell-cycle 0.011020134 a recurrent region that contains no smaller recurrent region PATHWAY progression in response to DNA (Rouveirol et al., 2006). FISH assays were performed to damage and promote double- validate the HD. UCSC genome browser (http://genome.ucsc. stranded break repair; mutations edu) was used to select the four BAC clones covering the 1p36 induce breast cancer susceptibility region: RP11-624A15, RP11-199O1, RP11-709H9 and RP11- FIBRINOLY- Thrombin cleavage of fibrinogen 0.041925494 473K14, and three BAC clones covering the 9p21 region: SISPATHWAY results in rapid formation of fibrin RP11-454C15, RP11-478M20 and RP11-615P15. The BACs threads that form a mesh to were obtained from BACPAC Resource Center at the capture platelets and other blood cells into a clot Children’s Hospital Oakland Research (Institute in Oakland, RBPATHWAY The ATM protein kinase recog- 0.048073806 CA, USA). Commercial centromeric probes for chromosomes nizes DNA damage and blocks cell 1 and 9 (Vysis Inc., DownersGrove, IL, USA) were used as cycle progression by phosphory- controls. FISH assays were carried out as according to the lating chk1 and p53, which manufacturer’s instructions. FISH scoring of 1p36 and 9p21 normally inhibits Rb to allow G1/S regions fluorescence signals was carried out in each sample by transitions counting the number of single copy gene and control probe TNFR2PATH- Tumour necrosis factor-, produced 0.19711672 signals in an average of 100 nuclei. HD status was considered WAY by activated lymphocytes, binds to as positive for a sample when the ratio test region/control was its receptor TNFR2 to induce activation in immune cells and below 0.25 in >50% of tumour cells. apoptosis in many other cells Bioinformatics and statistical analyses Abbreviations: FDR, false-discovery rate; GO, gene ontology; GSEA, GEPAS Suite software (http://gepas.bioinfo.cipf.es/) was used Gene Set Enrichment Analysis. aSee Supplementary methods for details. for supervised clustering and gene annotation analysis of CGH data. Associations between genomic instability and other clinical parameters were assessed using Fisher’s exact test for qualitative variables and Mann–Whitney test was used for CGH array continuous variables. All expression array data were analysed All samples were hybridized against CGH using CodeLink Expression Analysis Software (GE Health- 44k microarrays (Agilent Technologies, Palo Alto, CA, USA), care). Expression data were normalized in R by the cyclic-loss spanning the entire human genome at a median resolution method (Wu et al., 2005), differential expression analysis was B75 kb. Hybridizations were done according to the manufac- carried out using MeV (TIGR, Rockville, MD, USA) and turer’s protocols. A total of 25 out of 26 samples (tumour Acuity 4.0 (Molecular Devices) softwares. Functional genomic sample number 11 was excluded from all the studies) were analysis was performed using GSEA.

Oncogene Genomic instability in Ewing’s sarcoma BI Ferreira et al 2090 Real-time quantitative RT–PCR hazards regression. MedCalc software (Mariakerke, Belgium) First-strand cDNA was synthesized with the High capacity and GraphPad Prism statistical software version 4.0 (GraphPad cDNA archive kit (Applied Biosystems, Foster city, CA, USA) Software, San Diego, CA, USA) were used for statistical from 1.5 mg. of total RNA in a 30 ml reaction. Real-time computations. Data are presented as mean7s.e. quantitative PCR was performed in a 10 ml reaction with 2.5 ml from 1/10 reverse transcription dilution using the Taqman Universal PCR Master Mix (Applied Biosystems) containing Acknowledgements 0.2 mM of mRNA-specific primers and 0.1 mM Universal Probe Library probes (Roche Applied Science, Basel, Switzerland). We thank to the Spanish Society of Pediatric Oncology Duplicate reactions were run on a 7900HT Fast Real-Time (SEOP) and particularly to Drs Ma I Pintor, J Dura´ n, A PCR System (Applied Biosystems). Cycle threshold (Ct) values Jime´ nez, A Navajas, JL Vivanco, T Acha, A Cantalejo, P for each gene and for TATA binding protein (used as an Galaro´ n, A Mun˜ oz, M Torrent, N Pardo, A Carbone´ ,C internal housekeeping control) were calculated and exported to Calvo, A Nieto, M Guibelalde and J Molina for providing us Excel spreadsheets for additional analysis. Relative expression with the tumour samples and clinical data used in this ÀDCt was calculated as 2 , where DCt ¼ CtgeneÀCtTBP. study.We also thank Ministerio de Educacio´ n y Ciencia (SAF2003-01068, SAF2005-04340 and SAF2006-07586), Min- Statistical analysis isterio de Sanidad (G03/089, C03/10, PI050197), Fundacio´ n For a single comparison of two groups, the two-tailed Student’s Inocente Inocente and Fundacio´ n Enriqueta Villavecchia, for t-test was used. Univariate analysis between clinical or genetic their financial support. BIF has a Marie Curie PhD Earyl parameters and outcome was performed using the log-rank test. Stage Research Training Fellowship. J Carrillo is supported Multivariate analysis was performed using Cox proportional- for a postdoctoral contract from the FGUAM.

References

Armengol G, Tarkkanen M, Virolainen M, Forus A, Valle J, up-regulated by the EWS/FLI1 oncoprotein and is highly Bohling T et al. (1997). Recurrent gains of 1q, 8 and 12 in the expressed in Ewing tumors. Int J Cancer 118: 1381–1389. Ewing family of tumours by comparative genomic hybridi- Mugneret F, Lizard S, Aurias A, Turc-Carel C. (1988). zation. Br J Cancer 75: 1403–1409. Chromosomes in Ewing’s sarcoma. II. Nonrandom addi- Aurias A, Rimbaut C, Buffe D, Zucker JM, Mazabraud A. tional changes, trisomy 8 and der(16)t(1;16). Cancer Genet (1984). Translocation involving chromosome 22 in Ewing’s Cytogenet 32: 239–245. sarcoma. A cytogenetic study of four fresh tumors. Cancer Ohali A, Avigad S, Zaizov R, Ophir R, Horn-Saban S, Cohen IJ Genet Cytogenet 12: 21–25. et al. (2004). Prediction of high risk Ewing’s sarcoma by gene Bandres E, Malumbres R, Escalada A, Cubedo E, Gonzalez I, expression profiling. Oncogene 23: 8997–9006. Honorato B et al. (2005). Gene expression profile of ewing Ozaki T, Paulussen M, Poremba C, Brinkschmidt C, Rerin J, sarcoma cell lines differing in their EWS-FLI1 fusion type. Ahrens S et al. (2001). Genetic imbalances revealed by J Pediatr Hematol Oncol 27: 537–542. comparative genomic hybridization in Ewing tumors. Genes Blaveri E, Brewer JL, Roydasgupta R, Fridlyand J, DeVries S, Chromosomes Cancer 32: 164–171. Koppie T et al. (2005). Bladder cancer stage and outcome by Pinkel D, Albertson DG. (2005). Comparative genomic array-based comparative genomic hybridization. Clin Cancer hybridization. Annu Rev Genomics Hum Genet 6: 331–354. Res 11: 7012–7022. Rouveirol C, Stransky N, Hupe P, Rosa PL, Viara E, Barillot E Brisset S, Schleiermacher G, Peter M, Mairal A, Oberlin O, et al. (2006). Computation of recurrent minimal genomic Delattre O et al. (2001). CGH analysis of secondary alterations from array-CGH data. Bioinformatics 22: 849–856. genetic changes in Ewing tumors: correlation with metastatic Subramanian A, Tamayo P, Mootha VK, Mukherjee S, Ebert disease in a series of 43 cases. Cancer Genet Cytogenet 130: BL, Gillette MA et al. (2005). Gene set enrichment analysis: a 57–61. knowledge-based approach for interpreting genome-wide Delattre O, Zucman J, Plougastel B, Desmaze C, Melot T, expression profiles. Proc Natl Acad Sci USA 102: 15545–15550. Peter M et al. (1992). Gene fusion with an ETS DNA- Thompson PM, Maris JM, Hogarty MD, Seeger RC, binding domain caused by chromosome translocation in Reynolds CP, Brodeur GM et al. (2001). Homozygous human tumours. Nature 359: 162–165. deletion of CDKN2A (p16INK4a/p14ARF) but not within Eilers PH, de Menezes RX. (2005). Quantile smoothing of 1p36 or at other tumor suppressor loci in neuroblastoma. array CGH data. Bioinformatics 21: 1146–1153. Cancer Res 61: 679–686. Hattinger CM, Potschger U, Tarkkanen M, Squire J, Zielenska M, Ushigome U, Machinami R, Sorensen PH. (2002). Ewing Kiuru-Kuhlefelt S et al. (2002). Prognostic impact of sarcoma/primitive neuroectodermal tumor (PNET). In: Fletcher chromosomal aberrations in Ewing tumours. Br J Cancer 86: CDM, Unni KK, Mertens F (eds). Pathology and Genetics of 1763–1769. Tumors of Soft Tissue and Bone. Wold Health Organization Khan J, Wei JS, Ringner M, Saal LH, Ladanyi M, Classification of Tumors. IARC Press: France. pp 298–300. Westermann F et al. (2001). Classification and diagnostic Wu W, Dave N, Tseng GC, Richards T, Xing EP, Kaminski N. prediction of cancers using gene expression profiling and (2005). Comparison of normalization methods for CodeLink artificial neural networks. Nat Med 7: 673–679. Bioarray data. BMC Bioinformatics 6: 309. Kovar H. (1998). Ewing’s sarcoma and peripheral primitive Zielenska M, Zhang ZM, Ng K, Marrano P, Bayani J, neuroectodermal tumors after their genetic union. Curr Opin Ramirez OC et al. (2001). Acquisition of secondary Oncol 10: 334–342. structural chromosomal changes in pediatric Ewing sarcoma Mendiola M, Carrillo J, Garcia E, Lalli E, Hernandez T, de is a probable prognostic factor for tumor response and Alava E et al. (2006). The orphan nuclear receptor DAX1 is clinical outcome. Cancer 91: 2156–2164.

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

Oncogene