Oncogene (2006) 25, 7324–7332 & 2006 Nature Publishing Group All rights reserved 0950-9232/06 $30.00 www.nature.com/onc ONCOGENOMICS DNA copy number amplification profiling of human neoplasms S Myllykangas1, J Himberg2,TBo¨ hling1, B Nagy1,3, J Hollme´ n2 and S Knuutila1 1Department of Pathology, Haartman Institute and HUSLAB, University of Helsinki and Helsinki University Central Hospital, Helsinki, Finland; 2Laboratory of Computer and Information Science, Helsinki University of Technology, Espoo, Finland and 31st Department of Obstetrics and Gynecology, Scmmclwcis University, Budapest, Hungary DNA copy number amplifications activate oncogenes and been shown to increase significantly in breast and are hallmarks of nearly all advanced tumors. Amplified prostate cell lines and in primary tumors, but the genes represent attractive targets for therapy, diagnostics number of overexpressed genes in the amplified chro- and prognostics. To investigate DNA amplifications in mosomal areas varies upon cancer tissue type (Pollack different neoplasms, we performed a bibliomics survey et al., 1999, 2002; Hyman et al., 2002). Amplification- using 838 published chromosomal comparative genomic activated human oncogenes include AKT2 in ovarian hybridization studies and collected amplification data at cancer, ERBB2 in breast and ovarian tumors, MYCL1 chromosome band resolution from more than 4500 cases. in small-cell lung cancer, MYCN in neuroblastoma, Amplification profiles were determined for 73 distinct REL in Hodgkin’s lymphoma, EGFR in glioma and neoplasms. Neoplasms were clustered according to the non-small-cell lung cancer and MYC in various cancers amplification profiles, and frequently amplified chromo- (Futreal et al., 2004). Amplifications are frequently somal loci (amplification hot spots) were identified using observed in advanced cancers, which have lost the p53- computational modeling. To investigate the site specificity mediated maintenance of genomic integrity (Livingstone and mechanisms of gene amplifications, colocalization of et al., 1992; Yin et al., 1992). Amplifications of various amplification hot spots, cancer genes, fragile sites, virus drug target genes have been observed to confer drug integration sites and gene size cohorts were tested in a resistance in several in vitro experiments as well as in statistical framework. Amplification-based clustering clinical studies (Wahl et al., 1979; Johnston et al., 1983; demonstrated that cancers with similar etiology, cell-of- Goker et al., 1995; Shannon, 2002). origin or topographical location have a tendency to obtain DNA copy number amplifications arise as multi- convergent amplification profiles. The identified amplifi- plication of intra-chromosomal regions of 0.5–10 Mb in cation hot spots were colocalized with the known fragile length. In contrast to the definition of amplification, sites, cancer genes and virus integration sites, but DNA copy number increase in larger chromosomal global statistical significance could not be ascertained. areas or intact chromosomes, owing to translocations or Large genes were significantly overrepresented on the aneuploidy, is defined as a gain (Lengauer et al., 1998). fragile sites and the reported amplification hot spots. Amplifications produce excess gene copies in homo- These findings indicate that amplifications are selected in geneously staining regions (HSRs) and extrachromo- the cancer tissue environment according to the qualitative somal acentric DNA fragments (double minutes and traits and localization of cancer genes. episomes) (Albertson et al., 2003). HSRs manifest as a Oncogene(2006) 25, 7324–7332. doi:10.1038/sj.onc.1209717; ladder-like structure of inverted repeats within chromo- published online 5 June 2006 somes (Schwab, 1998). Extrachromosomal amplicons appear as double minutes, which can be seen using Keywords: cancer; gene amplification; fragile site; bio- conventional cytogenetics (Hahn, 1993), and episomes informatics; data mining; molecular pathology (B250 bp in length), which are only detectable using molecular biology methods (Maurer et al., 1987; Graux ONCOGENOMICS et al., 2004). HSRs, double minutes and episomes may contain fused genetic material from different chromo- Introduction somal loci (Guan et al., 1994; Graux et al., 2004). Models for pathways that result in gene amplification Gene amplifications cause an increase in the gene copy have been reviewed in detail by Myllykangas and number and subsequently elevate the expression of the Knuutila (2006). Amplification pathway models predict amplified genes. The expression of amplified genes has that two independent DNA double-stranded breaks at the margins of an amplified region are required to occur Correspondence: S Myllykangas, Department of Pathology, Haartman to enable the amplification of the intra-chromosomal Institute, University of Helsinki, PO Box 21, FI-00014 Helsinki, DNA segment. The human genome is perturbed by Finland. specific labile domains, such as fragile sites, virus E-mail: samuel.myllykangas@helsinki.fi Received 1 February 2006; revised 13 April 2006; accepted 20 April 2006; integration sites and large genes, which are more published online 5 June 2006 sensitive to DNA damage than other loci. Fragile sites Amplification profiling of human neoplasms S Myllykangas et al 7325 are chromosomal regions, in which chromosomal breaks Therefore, amplification hot spots, the genomic loci can be induced using replication-stalling chemicals. that are prone to amplify in human neoplasms, were Based on their prevalence in the population, the fragile identified using computational modeling, specifically the sites (119) are classified as common or rare. Eighty-eight independent component analysis (ICA). Chromosome common fragile sites are found in all individuals and 31 bands with values >0.5 were included in the genome- rare fragile sites in less than 5% of the population. wide ICA factors representing the amplification hot Susceptibility to damage varies between the fragile sites, spots. Values between 0.5 and 1 within the amplification as 75% of the DNA damage lesions have been located to hot spots are shown using gray to black scaling. 22 common fragile sites (out of the 139 chromosomal Figure 1b presents the top 30 identified amplification loci with identified lesions) (Glover et al., 1984). hot spots (10% of all identified independent factors) in Especially, common fragile sites are colocalized with the order of decreasing stability index, which was amplifications (Hellman et al., 2002). WWOX (1.1 Mb) evaluated by bootstrapping the ICA factors. Besides and FHIT (0.8 Mb) are particularly large genes and factor 19, which embodies the whole chromosome 21, located in two of the most damage-prone fragile sites the longest coherent amplification hot spots encom- (16q23.1 and 3p14.2, respectively). WWOX and FHIT passed single chromosome arms. Note that three factors are frequently involved in DNA breaks. Besides fragile are composed of two disjoint regions (factors 9, 12 and site expression, retroviral DNA insertion at specific 27). The disjoint hot spots spanned across the centro- genomic sites is believed to induce DNA breaks. meres of chromosomes 8, 17 and 5, respectively. Amplification hot spots did not cross chromosome boundaries, although this would have been conceivable in the continuous vector presentation of amplifications. Results The amplification hot spots are also presented in Table 1, which lists the fragile sites and cancer genes that DNA copy number amplification profiles of human colocalize with the amplification hot spots. The ampli- neoplasms fication hot spots covered 178 chromosome bands DNA copy number amplifications in human neoplasms (45% of the genome). were extracted from a publicly available data collection. A data matrix containing binary chromosome sub- band-specific information of DNA amplifications was Amplification profile-based clustering reveals relations generated (downloadable at http://www.helsinki.fi/cmg/ between human neoplasms cgh_data.html). The number of examined loci was 393 DNA amplification profiling (Figure 1a) showed that (list of all analysed chromosomal loci in Appendix 1 in amplifications were not randomly distributed in the the Supplementary information). The assembled data human genome, but specific amplifications were ob- matrix contained 4590 cases with 5740 coherent, served in distinct neoplasms. Hierarchical clustering was amplified chromosomal regions. In all, 1 803 870 chro- performed to investigate whether there were amplifica- mosome bands were analysed and 26 527 of them were tion-based similarities between neoplasms. Figure 2 labeled amplified. Appendix 2 in the Supplementary shows the correlation matrix and the dendrogram of information presents an overview of the studied the neoplasms. Four main neoplasm clusters are marked neoplasms (n ¼ 73) and DNA amplifications. The in correlation matrix. majority of the neoplasms were malignant tumors or hematologic malignancies, but non-malignant or pre- Associations between amplification hot spots, cancer malignant lesions were also included in this study. All 73 genes, fragile sites, virus integration sites and gene size neoplasm categories included cases with amplifications. cohorts Amplification profiling was carried out by calculating The identification of amplification hot spots (Figure 1b) chromosome sub-band-specific amplification frequen- suggested that some structural genomic properties or cies. Amplification profiles of all studied neoplasms are labile genomic features might explain the site specificity presented in Figure 1a,