Microarray Analyses Reveal Strong Influence of DNA Copy Number
Total Page:16
File Type:pdf, Size:1020Kb
Oncogene (2005) 24, 1794–1801 & 2005 Nature Publishing Group All rights reserved 0950-9232/05 $30.00 www.nature.com/onc Microarray analyses reveal strong influence of DNA copy number alterations on the transcriptional patterns in pancreatic cancer: implications for the interpretation of genomic amplifications Markus Heidenblad*,1, David Lindgren1, Joris A Veltman2, Tord Jonson1, Eija H Mahlama¨ ki3, Ludmila Gorunova1, Ad Geurts van Kessel2, Eric FPM Schoenmakers2 and Mattias Ho¨ glund1 1Department of Clinical Genetics, Lund University Hospital, SE-221 85 Lund, Sweden; 2Department of Human Genetics, University Medical Center Nijmegen, PO Box 9101, 6500 HB Nijmegen, The Netherlands; 3Laboratory of Cancer Genetics, University of Tampere and Tampere University Hospital, Tampere, Finland DNA copy number alterations are believed to play a major 1997). The consistent presence of these genomic altera- role in the development and progression of human tions indicates that an apparently tumor-type indepen- neoplasms. Although most of these genomic imbalances dent selective advantage is added to the transformed have been associated with dysregulation of individual cell, most probably through modified expression levels genes, their large-scale transcriptional consequences of genes in the affected segments. This is supported by remain unclear. Pancreatic carcinomas frequently display studies of breast cancer, where parallel microarray gene copy number variation of entire chromosomes as well analyses of mRNA (expression profiling) and DNA as of chromosomal subregions. These changes range from (genomic profiling) have shown a consistent association homozygous deletions to high-level amplifications and are between gene copy numbers and expression levels believed to constitute key genetic alterations in the (Hyman et al., 2002; Pollack et al., 2002). These studies cellular transformation of this tumor type. To investigate indicate that DNA copy number has a major impact on the transcriptional consequences of the most drastic transcription, especially in high-level amplifications genomic changes, that is, genomic amplifications, and to where as many as 44–62% of the affected genes also analyse the genome-wide transcriptional effects of DNA show overexpression. However, similar data from colon copy number changes, we performed expression profiling cancer indicate that only 4% of the amplified genes of 29 pancreatic carcinoma cell lines and compared the respond in terms of increased expression (Platzer et al., results with matching genomic profiling data. We show 2002). In the present study, we set out to determine that a strong association between DNA copy numbers and whether the strong gene dosage/expression correspon- mRNA expression levels is present in pancreatic cancer, dence is restricted to breast cancer, or reflects a more and demonstrate that as much as 60% of the genes within common phenomenon that also applies to other tumor highly amplified genomic regions display associated over- types. Pancreatic carcinomas frequently show imbal- expression. Consequently, we identified 67 recurrently ances of entire chromosomes as well as of smaller overexpressed genes located in seven precisely mapped genomic segments. The DNA copy number variations, commonly amplified regions. The presented findings which range from homozygous deletions to high-level indicate that more than one putative target gene may be amplifications, are believed to constitute key genetic of importance in most pancreatic cancer amplicons. alterations in the development and progression of this Oncogene (2005) 24, 1794–1801 doi:10.1038/sj.onc.1208383 tumor type. To investigate the genome-wide transcrip- Published online 31 January 2005 tional effects of gene copy number alterations, in particular of genomic amplifications, we performed Keywords: expression profiling; genomic profiling; mi- expression profiling analyses of 29 pancreatic carcinoma croarray; pancreatic cancer cell lines using cDNA microarrays, and compared the results with matching genomic profiling data (Heiden- blad et al., 2004). The analysis of the combined genomic and expression data sets allowed us to (i) determine the A common feature in human malignancies is the genome-wide association between gene copy number presence of genomic imbalances. Even though some of and expression levels in pancreatic cancer, (ii) identify 67 these are tumor specific, the majority can be found in a recurrently overexpressed genes in seven precisely large number of different tumor types (Mertens et al., mapped commonly amplified regions, and (iii) analyse the genome-wide mRNA effects two of the most frequently amplified regions in pancreatic cancer, *Correspondence: M Heidenblad; E-mail: [email protected] 8q23-24 and 12p11-12. Received 26 July 2004; revised 15 November 2004; accepted 18 To evaluate the global transcriptional consequences November 2004; published online 31 January 2005 of gene copy number alterations, paired gene expression Transcriptional patterns in pancreatic cancer M Heidenblad et al 1795 and DNA copy number ratios from 10 of the cell lines respectively, did not result in any major changes of were analysed. A total of 79 927 paired observations, the expression levels. However, when the fold-changes representing 13 720 different cDNA clones, were in- exceeded 2.0, a large effect was seen, both on the cluded. Based on gene copy number ratios, these were number of genes affected and on the maximum subdivided into seven categories with fold-changes expression levels. This pattern was even more o0.7, 0.7–0.9, 0.9–1.1, 1.1–1.5, 1.5–2.0, 2.0–2.5, pronounced in the category with >2.5-fold DNA and >2.5. In Figure 1a, all expression levels amplification. within the respective categories are plotted. Gene copy To investigate specifically the relationship between fold-changes below 0.9, and between 1.1 and 2.0, gene copy numbers and proportion of over-/under- corresponding to gains or losses of 1–2 gene copies, expressed genes, these were defined as the upper and lower 7th percentiles (Hyman et al., 2002) of the B80 000 expression measurements. The fraction of over- and underexpressed genes within each gene copy a 8 number category revealed that the former showed a 7 strong positive association with DNA copy numbers, 6 and that the latter showed a less pronounced negative 5 association with the number of genomic copies present (Figure 1b). The category with gene copy fold-changes 4 o0.7 showed over- and underexpression of 5.1 and 3 12%, respectively, of the genes in this group, whereas 2 the equivalent fractions in the highest copy number Expression ratio 1 category were 53 and 3.4%. In the normal copy number group (ratios 0.9–1.1), 6.2% of the genes were over- 0 expressed and 6.5% underexpressed. To evaluate the -1 effects of the most highly amplified genomic sequences < 0.7 0.7-0.9 0.9-1.1 1.1-1.5 1.5-2.0 2.0-2.5 > 2.5 Copy number ratio in more detail, we raised the threshold ratio of the highest DNA copy number category to 3.0 and 3.5. This b 60 analysis showed that 62% of the genes in the former and Overexpressed 60% of the genes in the latter were overexpressed, Underexpressed 50 suggesting that a maximum fraction of susceptible genes seems to be affected when fold-changes exceed 3.0, 40 equivalent to approximately 10 gene copies in the present microarray platform. A consequence of these 30 findings is that a change from a low- to a high-level amplification not only results in further upregulation of 20 genes already overexpressed at low gene copy numbers Proportion ESTs (%) but also in additional genes within a given segment 10 becoming upregulated. Hence, the biological outcome of 0 high-level amplifications may be very different from that < 0.7 0.7-0.9 0.9-1.1 1.1-1.5 1.5-2.0 2.0-2.5 > 2.5 of gains and low-level amplifications of the same Copy number ratio segment. The maximum of 60% affected genes in high- Figure 1 Association between gene copy number and expression level amplifications is in agreement with observations in levels. To obtain a reliable and representative assembly of paired breast cancer using array-based comparative genomic genomic and transcriptional observations for the association hybridization (CGH) for amplicon mapping (Hyman analysis, we used the following selection criteria. First, based on et al., 2002; Pollack et al., 2002), but differs substantially the genomic aberration spectra and hybridization qualities of both hybridizations, 10 tumor cell lines were selected; AsPC-1, DANG, from findings in colon cancer based on traditional CGH HupT3, LPC3p, LPC4p, LPC5 m, LPC10 m, LPC11p, LPC13p, for amplicon delineation (Platzer et al., 2002). A partial and LPC14p. Second, to enrich high-quality observations within explanation for this discrepancy could be that CGH this subset of tumor cell lines, we used a signal-to-noise filter of 3.0 analyses on metaphase chromosomes generally over- for both dyes, and for both experiments. Only cDNA clones with at least one paired observation remaining after this filtration were estimate the size of amplified segments, potentially included in the association analysis (a total of 13 720 of 25 468). In leading to analysis of a large number of unamplified total, 79 927 paired ratios from the 10 cases were analysed. For genes, which in turn would reduce the frequency of detailed information on cell lines and hybridizations, see legends to affected genes. Figure 2. (a) Box-Whisker plots illustrating the transcriptional We previously determined the exact genomic bound- distributions of 79 927 observations divided into seven different copy number categories: o0.7 (N ¼ 2943), 0.7–0.9 (N ¼ 21 325), aries of 60 amplicons in the investigated cell lines 0.9–1.1 (N ¼ 31 885), 1.1–1.5 (N ¼ 21 427), 1.5–2.0 (N ¼ 1971), 2.0– (Heidenblad et al., 2004).