Identification of candidate genes involved in neuroblastoma ...

7 downloads 246 Views 652KB Size Report
May 28, 2007 - Correspondence: Dr MS Jackson, Institute of Human Genetics,. International Centre for Life, Central Parkway, Newcastle upon Tyne. NE1 3BZ ...
Oncogene (2007) 26, 7432–7444

& 2007 Nature Publishing Group All rights reserved 0950-9232/07 $30.00 www.nature.com/onc

ORIGINAL ARTICLE

Identification of candidate genes involved in neuroblastoma progression by combining genomic and expression microarrays with survival data M Łastowska1, V Viprey2, M Santibanez-Koref1, I Wappler1, H Peters1, C Cullinane3, P Roberts4, AG Hall5, DA Tweddle5, ADJ Pearson6, I Lewis7, SA Burchill2 and MS Jackson1 1 Institute of Human Genetics, University of Newcastle upon Tyne, Newcastle upon Tyne, UK; 2Children’s Cancer Research Laboratory, Cancer Research UK Clinical Centre, St. James’s University Hospital, Leeds, UK; 3Department of Pathology, St. James’s University Hospital, Leeds, UK; 4Department of Cytogenetics, St. James’s University Hospital, Leeds, UK; 5Northern Institute for Cancer Research, University of Newcastle upon Tyne, Newcastle upon Tyne, UK; 6Institute of Cancer Research, Sutton, Surrey, UK and 7Department of Paediatric Oncology and Haematology, St. James’s University Hospital, Leeds, UK

Identifying genes, whose expression is consistently altered by chromosomal gains or losses, is an important step in defining genes of biological relevance in a wide variety of tumour types. However, additional criteria are needed to discriminate further among the large number of candidate genes identified. This is particularly true for neuroblastoma, where multiple genomic copy number changes of proven prognostic value exist. We have used Affymetrix microarrays and a combination of fluorescent in situ hybridization and single nucleotide polymorphism (SNP) microarrays to establish expression profiles and delineate copy number alterations in 30 primary neuroblastomas. Correlation of microarray data with patient survival and analysis of expression within rodent neuroblastoma cell lines were then used to define further genes likely to be involved in the disease process. Using this approach, we identify >1000 genes within eight recurrent genomic alterations (loss of 1p, 3p, 4p, 10q and 11q, 2p gain, 17q gain, and the MYCN amplicon) whose expression is consistently altered by copy number change. Of these, 84 correlate with patient survival, with the minimal regions of 17q gain and 4p loss being enriched significantly for such genes. These include genes involved in RNA and DNA metabolism, and apoptosis. Orthologues of all but one of these genes on 17q are overexpressed in rodent neuroblastoma cell lines. A significant excess of SNPs whose copy number correlates with survival is also observed on proximal 4p in stage 4 tumours, and we find that deletion of 4p is associated with improved outcome in an extended cohort of tumours. These results define the major impact of genomic copy number alterations upon transcription within neuroblastoma, and highlight genes on distal 17q and proximal 4p for downstream analyses. They also suggest that integration of discriminators, such as survival and comparative gene expression, with microarray data

Correspondence: Dr MS Jackson, Institute of Human Genetics, International Centre for Life, Central Parkway, Newcastle upon Tyne NE1 3BZ, UK. E-mail: [email protected] Received 14 December 2006; revised 22 March 2007; accepted 19 April 2007; published online 28 May 2007

may be useful in the identification of critical genes within regions of loss or gain in many human cancers. Oncogene (2007) 26, 7432–7444; doi:10.1038/sj.onc.1210552; published online 28 May 2007 Keywords: neuroblastoma; arrays; candidate genes

expression

arrays;

SNP

Introduction The presence of recurrent chromosomal copy number alterations in many solid tumours which correlate with disease outcome suggests that changes in the expression of specific genes within these regions are critical to the disease process (Popescu and Zimonjic, 1997). Recently, microarray gene expression profiling has allowed the relationship between copy number and gene expression in such regions to be analysed in detail in a variety of tumours including prostate, glioblastoma, multiple myeloma and colon carcinoma (Wolf et al., 2004; Nigro et al., 2005; Tsafrir et al., 2006; Walker et al., 2006). All these analyses indicate that expression of a significant fraction of genes within gained or deleted regions are altered in a manner consistent with the underlying genomic alteration, with increased expression in regions of gain and decreased expression in regions of loss. Although gene ontology and expression profiles within tissues of interest can be used to further define candidate genes, the extensive copy number dependence of gene expression suggests that additional information will be required to identify those genes which are likely to be critical for tumour biology (Bussey et al., 2006; Walker et al., 2006). The potential value of additional criteria to assess candidate genes in regions of loss and gain is particularly clear in neuroblastoma, the most common extracranial childhood solid tumour, where multiple genomic alterations of proven prognostic value have been identified. These alterations, which include MYCN amplification (Seeger et al., 1985), 1p deletion (del)

Genomic and expression microarray analysis in neuroblastoma M Łastowska et al

7433

(Caron et al., 1996a), 11q del (Attiyeh et al., 2005) and 17q gain (Bown et al., 1999), help to define three nonoverlapping clinicogenetic subtypes of neuroblastoma: The first is characterized by the presence of numerical chromosomal abnormalities and low stage of disease; the second by the presence of 17q gain, 11q and 3p del; the third by the presence of 17q gain, 1p del and MYCN amplification (Lastowska et al., 2001; Chen et al., 2004; Vandesompele et al., 2005). Among these alterations, 17q gain is the single most important indicator of poor prognosis (Lastowska et al., 2001; Vandesompele et al., 2005). The relevance of genes within these regions of recurring copy number alteration is supported further by the fact that many chromosomal gains and losses observed in a transgenic mouse model of neuroblastoma are syntenic to regions altered in the human disease (Hackett et al., 2003). Recently, two analyses have integrated genomic copy number estimates obtained from BAC arrays or PCR with microarray expression analyses to specifically investigate 1p del (JanoueixLerosey et al., 2004), and 1p del, 11q del and 17q gain (Wang et al., 2006). These investigations generated lists of genes whose expression is significantly altered by copy number changes, and established that the expression of between 15 and 61% of all genes in these regions are copy number dependent, highlighting further the need to use additional criteria for candidate gene appraisal. Here, we present the results of an integrated single nucleotide polymorphism (SNP) and expression microarray analysis of 30 primary neuroblastomas, which define deregulated genes within eight recurrent chromosomal copy number alterations. Collectively, the genes identified represent over 3.5% of all genes analysed. Since it has been shown previously that gains of chromosomal regions syntenic to human 17q occur in rodent neuroblastoma (Lastowska et al., 2004), we have used gene expression data from mouse and rat neuroblastoma cell lines to further pinpoint candidates from this genomic region. Furthermore, since many of the genomic alterations analysed have been shown to be of prognostic importance, we have also used the extent to which gene expression and SNP copy number estimates correlate with patient survival as an additional criterion to prioritize differentially expressed genes in these regions. This greatly reduced the number of candidate genes for downstream analysis from over 1000 to 84. Genes in two regions stand out in these analyses; distal 17q gain, which is of known prognostic importance, and distal 4p which, although deleted in B20% of neuroblastomas, has not been associated previously with survival. Results Generation of gene expression and DNA copy number data We have used the Affymetrix U133 Plus 2 chip to assess gene expression in neuroblastomas from 30 patients, 10 with stages 1, 2, 3 or 4S disease, and 20 with stage 4

disease. To correlate gene expression and DNA copy number the status of MYCN, 1p and 17q was established in all tumours using fluorescence in situ hybridization (FISH). The breakpoint positions of unbalanced gains and losses were delineated further at high resolution using Affymetrix 50 K HindIII SNP arrays and validated through comparison with data from constitutional DNA samples and 50K XbaI arrays (see Materials and methods). Since low-stage tumours are characterized by numerical changes in whole chromosomes (Lastowska et al., 2001; Chen et al., 2004; Vandesompele et al., 2005), this analysis was confined to stage 4 tumours where DNA was available (n ¼ 19). The SNP results were consistent with the FISH data in all cases, except that the SNP analysis identified an additional 17q gain in one tumour (NB29), where the region gained (64.41Mb-qter) lies distal to the probe used in FISH analyses (MPO, 53.7Mb). Results showing breakpoint delineation on chromosome 1 in three tumours, and confirmation of 2p gain using direct comparison between tumour and constitutional DNA, are shown in Supplementary Figure S1. In total, the SNP analysis delineated 132 gains and losses, 71 of which (B54%) were accounted for by eight recurring alterations each of which was present in at least 25% (5) of the tumours analysed: MYCN amplification, gains of 17q and 2p, and losses of 1p, 3p, 4p, 10q and 11q (Table 1). We therefore restricted our analysis of the relationship between gene expression and copy number to these eight genomic alterations. Identification of genes coamplified and overexpressed with MYCN Expression and DNA copy number data for distal 2p in the 10 MYCN amplified tumours are shown in Figure 1. Diploid copy number estimates (black) for the MYCN gene range from 30 to 200 (position B16 Mb), with the amplicon extending proximally for 6–8 Mb in three tumours (NB6, 25 and 19), and additional peaks of amplification clearly visible in NB26 (B2 Mb) and NB10 (28–31 Mb). Excluding MYCN and its antisense NCYM, a total of 59 genes have a DNA copy number estimate of >8 in one or more tumour samples (Table 2) and of these, 43 (73%) show a >4-fold increase in gene expression in one or more tumours where there is an increase in DNA copy number. Of these, 30 genes show a significant (Po0.05) and positive (r2>0.7) correlation between gene expression and copy number across all samples. However, the correlation between copy number and expression is weak for MYCN (r2 ¼ 0.1668), and for both DDX1 and NAG, the genes most commonly co-amplified with MYCN (r2 ¼ 0.0791 and 0.01095). Identification of genes differentially expressed in regions of genomic gain and loss To identify genes differentially expressed as a result of low copy number gains and losses, the dataset was systematically dichotomized with respect to each of the eight genomic alteration present in at least 25% of the tumours analysed, and genes were identified where Oncogene

Genomic and expression microarray analysis in neuroblastoma M Łastowska et al

7434 Table 1 Genomic regions altered in five or more samples 1p NB2 NB3 NB5 NB6 NB7 NB9 NB10 NB13 NB18 NB19 NB20 NB24 NB25 NB26 NB27 NB28 NB29 NB31 NB34

2p+

MYCN

pter-56.54 pter-51.67

AMP

pter-43.37 pter-28.99

AMP

3p

4p

10q

11q

17q+

pter-55.49

pter-27.61

71.43-qter

pter-59.65

pter-68.26

30.69-qter 15.83-qter 64.95-qter

37.93-qter 32.20-qter 36.4-qter 35.1-qter 43.73-qter 38.17-qter 40.52-qter

pter-45.02

73.56-qter

pter-27.89 80.46-qter

pter-115.44

AMP

pter-23.39 pter-87.04 pter-52.53

AMP AMP AMP

76.57-qter 80.08-qter

pter-49.77

pter-40.40 pter-qter pter-68.28

pter-66.9 pter-26.23 pter-116.66

pter-49.19

AMP AMP AMP

pter-79.14 pter-103.28 pter-10.22

AMP pter-49.54 pter-188.28

pter-41.1

62.97-qter 43.83-qter

pter-60.47 pter-38.70 pter-5.65

69.83-qter 73.74-qter

34.2 -qter 39.2-qter 46.85-qter 43.81-qter 55.27-qter 27.98-qter 39.96-qter 32.63-qter 64.4-qter 41.72-qter 33.66-qter

The extent of unbalanced copy number changes (in megabases), as defined by the closest unaffected SNP, is given to the nearest 10 kb.

Figure 1 Copy number and gene expression within the MYCN amplicon. Genomic copy number estimates derived from flanking SNPs for each gene are shown in black, fold change in gene expression levels relative to the median value for each gene are shown in red. The most distal 280 genes on 2p are shown in linear order, left to right (lower scale). The physical scale, in megabases, is therefore only an approximation (upper scale). One probe per gene is shown, except for MYCN which is represented by three probes. Only one gene shows significant variation in expression which is independent of copy number (SLC30A3, visible at B27 Mb in NB25 and NB27). For details see Materials and methods and Table 2. SNP, single nucleotide polymorphism. Oncogene

expression levels differed according to the presence or absence of each alteration (Wilcoxon test, FDR ¼ 0.05). A list of all genes identified in each of these analyses is provided as Supplementary information (Supplementary Tables S2 and S3). Two of the eight genomic alterations involved gain of chromosomal material; distal 17q and proximal 2p. Because gain of an entire chromosome 17 is associated with good prognosis, but unbalanced partial gain of the long arm is associated with poor prognosis (Lastowska et al., 2001), gene expression across this entire chromosome was analysed. Out of 1656 gene probes, 239 (14%) showed statistically significant differential expression (Figure 2a). Tumours with 17q gain have marked underexpression of differentially expressed genes on 17p and proximal 17q, which is not gained, but marked overexpression of genes within distal 17q, which is gained (Figure 2a). The commonly gained region contains 40 differentially expressed genes (49 probes), all but two of which are upregulated. The difference in mean expression levels between 17q gained and not gained tumours is plotted in Figure 2b, and illustrates that the vast majority of genes on distal 17q show an increase in log2 ratio of between 0.2 and 0.9 (mean ¼ 0.471), consistent with a dosage effect (a log2 ratio of 0.59 represents a fold increase of B1.5). Interestingly, genes on 17p and proximal 17q are underexpressed in tumours with 17q gain relative to 17q normal tumours (mean ¼ 0.434). Gain of 2p was observed in seven tumours, with the commonly gained region encompassing the distal 45.02 Mb (Table 1). Of the 886 genes on 2p, only 17 were found to be significantly altered by the Wilcoxon test. All map between 3.0 and 39.1 Mb on 2p and were upregulated in 2p gained tumours (Supplementary Table S2). Expression data from the analyses of all five commonly observed regions of loss are presented in Figure 3. Owing to the huge variation in breakpoint

Genomic and expression microarray analysis in neuroblastoma M Łastowska et al

7435 Table 2 Affymetrix ID

Gene name/ID

241535_at 220487_at 210342_s_at 216941_s_at 239046_at 235878_at 230948_at 222830_at 1553137_s_at 209773_s_at 242556_at 212552_at 200790_at 244728_at 1565577_s_at 222684_s_at 208638_at 208639_x_at 216640_s_at 210263_at 225579_at 211504_x_at 205862_at 206899_at 234250_at 212274_at 202479_s_at 1566716_at 234331_s_at 231439_at 229546_at 240579_at 207478_at 207477_at 201241_at 207028_at 211377_x_at 230276_at 226324_s_at 218843_at 225050_at 220321_s_at 209313_at 201838_s_at 218682_s_at 203781_at 57540_at 1562111_at 225262_at 235703_at 1553310_at 231036_at 228222_at 221229_s_at 214662_at 238186_at 242710_at 226425_at 242964_at 208211_s_at 222408_s_at

LOC388920 SNTG2 TPO TAF1B Anon FLJ42198 yj20d11.s1mRNA LBP-32 TIEG2 RRM2 Anon HPCAL1 ODC1 Anon FLJ34486 FLJ14075 ATP6V1C2 P5 FLJ23273 KCNF1 MGC33602 ROCK2 GREB1 NTSR2 FLJ20410 LPIN1 TRB2 DKFZp566F0224 NSE1 ZD76G03 cDNA LOC400944 NAG Anon Anon DDX1 NCYM MYCN DKFZP566A1524 SLB FRCP1 ZNF512 FLJ13646 XAB1 STAF65 SLC4A1AP MRPL33 RBSK MAGE:5285600 FOSL2 PLB Anon Anon PPP1CB FLJ20628 KIAA0007 Anon Anon FLJ21069 Anon ALK CGI-127

Genes co-amplified within MYCN

n CN>8

n EXP>4X

Mean CN

Mean EXP XP

Correlation

P-value

1 1 1 2 2 2 2 3 3 3 3 3 3 3 2 2 2 2 2 2 2 2 3 3 3 2 3 3 3 3 9 9 9 9 9 10 10 7 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1

1 1 1 2 2 2 2 2 1 3 0 3 3 0 2 2 2 2 2 0 0 2 3 1 1 2 2 0 2 2 3 4 0 1 5 9 9 0 0 0 1 1 1 1 1 1 1 1 0 1 1 1 0 0 0 0 1 0 1 1 0

50 99 99 11.3 11.3 11.3 11.3 13.5 13.5 18 19 19 19 18 19 19 19 19 19 19 34 34 31 31 44 19 15 26.5 26.5 26.5 33.1 33.1 38 38 41 35 35 17.4 13 17 17 18 18 18 18 18 17 13 43 43 46 46 16 16 16 16 42 42 42 42 19

64 86 21.7 13.8 7.7 10.8 8.8 7 4.16 7.8 — 7.75 6.2 — 10.6 11.6 5.2 6.8 9.1 — — 10.3 12.5 374* 8.5 19.2 15 — 21.7 6.45 97* 45.8 — 6.1 6.6 38.9 33.7 — — — 8.5 20.8 20 20.6 22.9 10.1 14.8 6.9 — 7.3 34.8 83.8 — — — — 4.8 — 20.9 24.9 —

0.9942 0.9995 0.9965 0.7173 0.5421 0.7689 0.6960 0.8563 0.7963 0.8278 0.7540 0.9765 0.8856 0.8472 0.9910 0.9740 0.8986 0.8720 0.9068 0.2528 0.1702 0.9187 0.7809 0.9515 0.8528 0.8888 0.5062 0.5915 0.1630 0.7525 0.7604 0.1095 0.1966 0.4958 0.0791 0.3067 0.1668 0.5437 0.1363 0.2611 0.0229 0.0206 0.0353 0.0010 0.0261 0.1578 0.0896 0.0332 0.0370 0.6513 0.6647 0.6569 0.2694 0.1656 0.7171 0.4237 0.9214 0.8820 0.9732 0.9891 0.4318

o0.0001 o0.0001 o0.0001 0.0195 0.1055 0.0093 0.0254 0.0016 0.0058 0.0031 0.0118 o0.0001 0.0007 0.0020 o0.0001 o0.0001 0.0004 0.0010 0.0003 0.4811 0.6382 0.0002 0.0077 o0.0001 0.0017 0.0006 0.1355 0.0717 0.6528 0.0120 0.0107 0.7633 0.5861 0.1450 0.8280 0.3886 0.6452 0.1043 0.7074 0.4662 0.9533 0.9581 0.9280 0.9800 0.9470 0.6633 0.8056 0.9274 0.9192 0.0574 0.0508 0.0546 0.4834 0.6475 0.0196 0.2224 0.0002 0.0007 o0.0001 o0.0001 0.2128

Columns 3–6 show the number of tumours in which DNA amplification is observed (n CN>8), the number of tumours in which gene expression is increased more than fourfold (n EXP>4X), the mean copy number estimate within amplified tumours (mean CN), and mean fold change in expression within amplified tumours (mean EXP). Correlations between expression and copy number across all samples, together with associated P-values are given. Fold change estimates marked with an asterisk indicate that gene expression levels were too low to be accurately measured in some samples. In column 2, known genes are shown in bold. In columns 7 and 8 correlations which are significant (Po0.05) are shown in bold. A copy number of >8 was chosen to define MYCN amplified tumours as this is comparable to existing measures of MYCN amplification (Ambros and Ambros, 2001) and provided total discrimination between tumours with 2p gain and MYCN amplification for SNPs within the MYCN gene.

Oncogene

Genomic and expression microarray analysis in neuroblastoma M Łastowska et al

7436

17q not gained

a

p

17q gained

b

Change in mean Log2R -2 -1 0 +1 +2 0 Mb

10 Mb

20 Mb

cen 30 Mb

40 Mb

q 50 Mb

60 Mb

64.4 Mb 70 Mb

Figure 2 Differential expression of genes on chromosome 17 with respect to 17q status. Gene expression was analysed in 28 tumours, dichotomized with respect to 17q gain (two tumours with mixed 17 and 17q gain were excluded). (a) Heat map of genes differentially expressed between 17q gained and not gained groups, shown in linear order with respect to chromosomal position (17pter to 17qter). Red indicates overexpression relative to median value and blue indicates underexpression. (b) Difference in mean expression level of each gene observed in the 17q gained group relative to the not gained group, shown with respect to physical position on the chromosome. For ease of presentation, only genes with differences in log2 ratio between 2 and þ 2 are shown. Four genes with more extreme variation are excluded as a result: ATP2A3 (3.31464), 232887 at (2.88767) and MGC87631 (4.732143) on 17p and ANKFN1 (2.56757) on 17q.

position on 1p, the comparison of expression level of genes in tumours with and without 1p del was performed in a stepwise manner, partitioning the data in genomic intervals according to breakpoint position. Using this approach, 520 probes from chromosome arm 1p (31%) showed significant changes in expression, with 516 of these (99%) showing downregulation in 1p deleted tumours. In some tumours, the expression differences were consistent enough to allow breakpoint positions to be inferred directly (for example, NB6, 7 and 26, Figure 3a). The data from chromosomes 3, 4, 10 and 11 all show similar patterns, with chromosomal loss consistently associated with reduced gene expression. On 3p, 176 probes (19%) were significantly altered, with 100% being downregulated in 3p deleted tumours (Figure 3b). On this chromosome, the small terminal deletion in tumour 31 defines a 5.6 Mb region of common loss which contains six downregulated genes: CHL1, CNTN4, CRBN, LRRN1, SETMAR and ARL8B. On 4p, 81 probes (22%) were significantly altered, all but two being downregulated in 4p deleted tumours (Figure 3c); on 10q, 215 probes (25%) showed significant expression differences, all downregulated in Oncogene

10q deleted tumours (Figure 3d); and on 11q, 291 probes (27.7%) were altered, with 274 probes (94%) being downregulated in 11q deleted tumours (Figure 3e). To illustrate the uniform nature of the expression changes associated with chromosomal loss, the difference in mean expression levels between tumours with and without 11q loss is plotted in Figure 3f. Because the changes in gene expression are often relatively small, particularly in regions of gain, it was desirable to confirm the gene expression results in a number of genes of interest using an independent method. We therefore reanalysed the expression of four genes which showed significant copy number dependent expression (PMP22, WSB1, BRCA1 and BIRC5) using real-time PCR. These genes were chosen as they map to distinct regions of chromosome 17 (17p12, 17q11, 17q21 and 17q25, respectively), and because they have been implicated previously in neuroblastoma or other cancers (Chen et al., 2006; Deng, 2006; Miller et al., 2006). Significant correlations were observed between Affymetrix and real-time PCR data for all four genes, r2 varying from 0.683 for BRCA1 to 0.961 for PMP22 (Supplementary Figure 2S).

Genomic and expression microarray analysis in neuroblastoma M Łastowska et al

7437

a

1p

10.2 bM

b

3p

4p

5.6Mb

27.6Mb

c

e

54.4Mb

58.9Mb

10q

11q

d

f 2 1 0 -1 -2

55

65

75

85

95

105

115

125Mb

Figure 3 Differential expression of genes in regions of chromosomal loss. Heat maps of genes differentially expressed between tumours dichotomized with respect to the region of loss indicated. In each case, the genes are shown in linear order with respect to chromosomal position. Red indicates overexpression relative to median value and blue indicates underexpression.: (a) Chr 1; pter-cen, (b) Chr 3; pter-cen, (c) Chr 4; pter-cen (d) Chr 10; cen-qter, (e) Chr 11; cen-qter and (f) difference in mean expression level of differentially expressed genes on 11q in the 11q deleted group relative to the not deleted group, shown with respect to physical position on the chromosome. For each gene, the mean change in log2 ratio between 11q deleted and 11q normal tumours is shown. Mean change in log2 ratio across all genes in deleted region is 0.5112, var 0.0364.

Oncogene

Genomic and expression microarray analysis in neuroblastoma M Łastowska et al

7438

Correlation with survival identifies candidate genes on 4p and 17q Based on the hypothesis that the expression of genes in regions of loss or gain which are involved in tumour development or progression should correlate more strongly with disease outcome than others, we have analysed the relationship between copy number/expression and patient survival to define further candidate genes. We first analysed the impact of 17q gain, MYCN amplification and 1p del (as detected by FISH) upon overall survival within our dataset using the log-rank test. All three cytogenetic alterations were significantly associated with poor survival, in line with prior expectation, P-values ranging from 0.013 to 0.001 (data not shown). We then ranked all SNPs with respect to their impact upon overall survival in stage 4 tumours (see Materials and methods). The fold over-representation of high-ranking SNPs on each autosomal arm, compared to expectation, is shown in Figure 4a. Chromosome arms 1p and 2p show a greater than 10fold excess of SNPs at one or more of the percentile ranges analysed, consistent with the survival analyses of FISH data. However, chromosome arm 4p, and to a lesser extent 22q, also shows significant over-representation in this analysis. The physical position of SNPs on chromosomes 1, 2, 4, and 22, which rank within the top 5% are shown in Figure 4b. On chromosome 1, these are distributed throughout the proximal short arm, whereas on chromosome 2, they map precisely to MYCN (B16 Mb) or to ALK (B30 Mb, co-amplified with MYCN in NB10). On chromosome 4, all of the SNPs cluster within the distal 40 Mb of the short arm. The SNPs on chromosome 22 are distributed throughout the long arm of this chromosome. Because some of these copy number changes have not been associated previously with survival, we then used the log-rank test to analyse the impact of each alteration under study upon survival within stage 4 tumours. Only 4p gave a significant result (P ¼ 0.003). To extend this observation into a larger dataset, we combined results from the current series with data from all stage 4 tumours of Lastowska et al. (2001) and analysed the impact of 4p loss on overall survival using the log-rank test. Although copy number in Lastowska et al. (2001) was analysed using cytogenetic methods as opposed to SNPs, the combined data identifies a significant impact of 4p loss on survival, with deletion being associated with better response to treatment (Figure 4c, P ¼ 0.013, 10 tumours with 4p del compared to 38 tumours without 4p del). As chromosome 22 was not altered at high frequency within our dataset, it was not studied further. To directly assess the association of gene expression within the genomic regions under study upon disease outcome, we then ranked the expression of all genes with respect to impact upon overall survival, and established if each minimal region of gain or loss contained more genes whose expression is associated with survival than expected by chance. This was achieved by comparing the rank of all genes present within each region to the rank of all genes in the whole Oncogene

Figure 4 SNP copy number and survival in stage 4 tumours. (a) Fold over-representation of SNPs on each autosome arm, relative to expectation within the top 5, 1 and 0.1% of SNPs ranked according to impact upon overall survival. This analysis is uninformative for 17q gain as all but two stage 4 tumours possess this alteration. (b) Physical distribution of SNPs ranked within top 5% from chromosomes 1, 2, 4 and 22. The position of the centromeres is shown by vertical grey bars. X axis, position in megabase; Y axis, inverse of score test P-value for each SNP. (c) Kapan–Meier survival curves for 4p deleted and 4p not deleted groups within stage 4 tumours (including data from Lastowska et al., 2001). SNP, single nucleotide polymorphism

genome using a Wilcoxon test. Only the minimal regions of chromosomes 2p gain, 4p loss and 17q gain showed significant enrichment for genes whose expression correlates with overall survival (P ¼ 0.01, P ¼ 1  107 and P ¼ 0.032, respectively). Over 45% of all differentially expressed genes from the commonly altered regions on 4p and 17q had expression levels associated with overall survival (Po0.005). However, none of the nine genes on 2p which was differentially expressed with respect to 2p gain correlated with overall survival,

Genomic and expression microarray analysis in neuroblastoma M Łastowska et al

7439

indicating that the enrichment on 2p is not due to the genomic alteration under study. The frequency of differentially expressed genes, which correlated with overall survival in the other regions under study (1p, 3p, 10p and 11q) was also analysed and found to be low, varying between 0 and 4%. The results for three genomic regions (1p, 4p and 17q) are shown in Figure 5. The results for chromosome arm 1p (Figure 5a) are presented as an example of a genomic region, where few of the differentially expressed genes correlate with overall survival. Only two of the 19 differentially expressed genes on the chromosome arm which correlate with survival (Po0.005) lie within the minimal region of loss (PRKCZ and CAMTA1). In sharp contrast, 25 out of 48 differentially expressed genes within the shortest region of 4p loss (27.6 Mb) are associated with survival (Po0.005, Figure 5b), including six genes represented by more than one probe. A similar result is obtained for

17q. Although genes associated with survival are present on both arms of this chromosome (Supplementary Table S2), 18 out of 40 differentially expressed probes within the 14.2 Mb minimal region of gain are associated with survival (Po0.005, Figure 5c). Furthermore, since mouse and rat neuroblastoma cell lines are known to possess gains of chromosomal regions which are syntenic to human 17q (mouse chromosome 11 and rat chromosome 10), it was also possible to assess gene expression in these regions relative to rodent neuronal controls (see Materials and methods). Strikingly, all but one of the genes on distal 17q associated with survival in humans was also expressed at high levels in rodent cell lines (SLC25A19, Figure 5c), further supporting their candidature for involvement in neuroblastoma progression. This analysis of expression with respect to survival is complicated by the fact that expression levels are not

Figure 5 Expression and survival in critical regions on 1p, 4p and 17q. The left panel in each case shows the position of breakpoints observed in tumours (black bars), the position of genes which correlate with survival (Po0.005, pink bars), and the minimally altered region on each arm (grey shading). The right panel in each case lists probes which correlate with survival (Po0.005) and their ranking in the analysis. For chromosomes 1 and 4 only genes associated with survival are shown due to space limitations. Genes ranked within the top 500 are highlighted further with darker shading. In addition, the expression of genes marked in violet correlates with survival when the data is stratified for each genomic alteration (see results). For genes on 17q with known orthologues, the fold change of expression in rodent neuroblastoma cell lines relative to appropriate rodent brain controls is also shown. Increase in expression (over twofold) is highlighted in green, an asterisk indicating an absent call in brain tissue (ND: not determined due to absence of orthologous probe on rodent array). Oncogene

Genomic and expression microarray analysis in neuroblastoma M Łastowska et al

7440

independent of the genomic alteration on each chromosome. To address this, we recalculated the correlation with overall survival of all significant genes stratified by genomic alteration (for example, for 17q–17q gained and not gained, for 1p–1p loss and no loss). A total of 34 probes show a significant correlation in this analysis (Po0.05, highlighted in purple in Figure 5) including six represented by two or more probes (MACF1, WSHC1, HCAP-G, TSEN54, PTDSR and SFRS2). Discussion We have used the combination of FISH, SNP copy number analysis and microarray expression analysis to identify genes which map within eight recurrent genomic copy number alterations in neuroblastoma, and whose expression is significantly changed by these alterations. The total number of genes identified represents over 3.5% of all genes analysed, with >1000 genes represented by >1300 probes being significantly changed. This highlights both the significant impact of copy number alterations upon the transcriptional programme of this tumour, and the need for additional criteria to further define candidate genes within such regions. As a first step to address this problem, we have used a combination of clinical and biological features (survival and expression of orthologues in relevant rodent cell lines), which have enabled us to define a list of 84 genes in four genomic regions (1p, 2p, 4p and 17q) as being of particular interest. Analysis of gene ontology indicates that this list is significantly enriched (Po0.05) for genes with biological functions of potential relevance to cancer, including RNA processing (CROP, DDX1, DHX15, DUS1L, EIF4A3, NOL14, SFRS1, SFRS2, SLBP, TRSPAP1 and TSEN54), DNA metabolism (BRCA1, BRIP1, EME1, KPNA2, RBPSUH, RRM2, TK1 and WHSC1) and apoptosis (BIRC5, BRCA1, CROP, MAEA, NME1, PRKCZ, PTDSR and SH3GLB1). Of these, more than half have been associated previously with other cancers, and eight have already been defined as candidate genes for involvement in neuroblastoma (Table 3). Approximately 20% of these genes have more than one Affymetrix probe associated with survival at po0.005, confirming the consistency of the expression results. Furthermore, the analysis of expression pinpoints the commonly altered regions on 4p and 17q as being significantly enriched for genes whose expression correlates with survival, while the analysis of SNP data within stage 4 tumours highlights 1p, 2p and 4p as regions where copy number correlates with survival (17q is gained in all but two stage 4 tumours, so this analysis is uninformative for this region). Proximal 1p is frequently deleted in progressing neuroblastomas, but to date no tumour suppressor gene in this region has been clearly implicated in disease progression. In our study, approximately 330 genes from 1p were downregulated in 1p deleted tumours. These results both confirm the results of two earlier analyses of the 1p35–36 region (Janoueix-Lerosey et al., Oncogene

2004; Wang et al., 2006), as B70% of downregulated genes found in these investigations are also downregulated in the present study, and extend them due to the higher resolution chip used in the current analysis (Affymetrix U133 Plus 2). More importantly, our survival analysis has shortlisted 19 genes on this chromosome arm whose expression level is associated with patient survival. Two, PRKCZ and CAMTA1, are located within a commonly deleted region, and expression of the latter has been linked recently to survival of patients with neuroblastoma in two independent studies (Henrich et al., 2006; Asgharzadeh et al., 2006). However, 10 of these genes are located proximal to 39 Mb, indicating that genes from the proximal region of 1p may have additional impact on survival. Among these, MACF1 is of particular interest because its open reading frame has been found to be disrupted by a translocation breakpoint in a neuroblastoma cell line (Schleiermacher et al., 2005). The combination of expression and SNP microarray data allows gene expression within MYCN amplicons to be analysed at the resolution of a single gene for the first time. It confirms the coamplification and overexpression of previously defined amplified genes (Chen et al., 2004; Seltzer et al., 2005) and the poor correlation between MYCN expression and copy number (Nisen et al., 1988; Slavc et al., 1990). While the extensive variation in amplicon size and complexity we observe is consistent with previous reports (Figure 1; Seltzer et al., 2005; Stallings et al., 2006), the present analysis defines for the first time over 30 genes as being co-amplified at both the DNA and expression level, several of which have been implicated in other neoplasias (Table 3). This complexity suggests that the relationship between gene expression/copy number within the MYCN amplicon and disease outcome, which has produced conflicting results using lower resolution methodologies (Weber et al., 2004; De Preter et al., 2005), may warrant renewed investigation. This is particularly relevant since in our dataset the expression level of MYCN correlated poorly with survival, whereas expression of genes co-amplified with MYCN (GREB1, LPIN1, ODC1, RRM2 and TAF1B) showed stronger correlation. The distal long arm of chromosome 17 is of particular interest as gain of this region is strongly associated with poor prognosis (Bown et al., 1999). Our investigation revealed 96 upregulated genes from this region, 22 being represented by two or more probes. More importantly, we identified 17 genes as prime candidates as these are located within the common region of gain, are associated with survival, and have orthologues which are also overexpressed in mouse and rat neuroblastoma cell lines relative to control neuronal tissue (Figure 5). Several of these genes have been implicated previously in neuroblastoma or other cancers including BIRC5, PTDSR, PYCR1 and TK1 (Table 3). Our analyses have also identified a high concentration of both genes and SNPs on distal 4p whose expression/ copy number correlates with survival. Although deletions of 4p, which occur in B20% of all neuroblastomas (Caron et al., 1996b), have not been associated

Genomic and expression microarray analysis in neuroblastoma M Łastowska et al

7441 Table 3 Genes identified in this study previously implicated in cancer Affymertix ID

Gene Symbol

Involvement in cancer

Referencesa

1p arm 202178_at 1555370_a_at 205277_at 231110_at 208634_s_at 208924_at 227123_at

PRKCZ CAMTA1 PRDM2 FABP3 MACF1 RNF11 RAB3B

Glioblastoma; critical for proliferation Neuroblastoma; downregulation linked to poor survival Pheochromocytoma, AML; downregulated Breast cancer; candidate tumour suppressor gene Neuroblastoma; disrupted by translocation Breast cancer; highly expressed Megakaryoblastic leukaemia, pancreatic carcinoma; upregulated During differentiation Ovarian cancer; downregulated HeLa cells; proapoptotic, loss may contribute to tumorigenesis

1 2 and 3 4 and 5 6 7 8 9 and 10

202741_at 209090_s_at 4p arm 209492_x_at 40225_at 213980_s_at 206052_s_at 218308_at

PRKACB SH3GLB1

222778_s_at 224416_s_at 218662_s_at 235554_x_at 223414_s_at 201385_at 225014_at 207785_s_at 203343_at 230896_at 17q arm 212281_s_at 204531_s_at 211603_s_at 208835_s_at 201577_at 221703_at 211762_s_at 64.4 Mb to qtel 204868_at 217932_at 212723_at 200753_x_at 1554408_a_at 202095_s_at 201303_at 203931_s_at 200656_s_at 226414_s_at 202148_s_at MYCN coamplified 201241_at 205862_at 212274_at 200790_at 209773_s_at 216941_s_at

ATP5I GAK CTBP1 SLBP TACC3

11 12

WHSC1 MED28 HCAP-G MGC29898 LYAR DHX15 LOC389203 RBPSUH UGDH CCDC4

Hepatocellular carcinoma, ovarian cancer; upregulated Prostate cancer hormone refractory; upregulated May act as an oncogene Required for progression through the cell cycle Multiple myeloma; co-ordinately upregulated with WHSC1 Ovarian cancer; upregulated in chemoresistant tumours Multiple myeloma; upregulation linked to poor prognosis Various cancers; stimulates cellular proliferation Pancreatic cancer; upregulation contributes to cells motility Glioblastoma; amplified Leukaemia; upregulated, may regulate cell growth MPNSTs and Barrett’s carcinoma; copy number gains Glioblastoma; amplified and overexpressed Gastrointestinal carcinoid cells; activation of Notch signalling Ovarian carcinoma; associated with CDDP resistance Neuroblastoma; upregulation linked to poor survival

10 and 13 14 15 16 17 and 18 17 19 20 21 22 23 and 24 21 25 26 27

MAC30 BRCA1 ETV4 CROP NME1 BRIP1 KPNA2

Meningioma and other cancers; upregulated Breast and ovarian cancers; mutated Ewing sarcoma, prostate cancer; fusion gene Cloned from cisplatin-resistant cell lines Neuroblastoma; upregulation in advanced stages; poor prognosis Interacts with BRCA1 Breast cancer, melanoma; poor prognostic marker

28, 29 30 31 and 32 33 34 and 35 36 37 and 38

ICT1 MRPS7 PTDSR SFRS2 TK1 BIRC5 DDX48 MRPL12 P4HB ANAPC11 PYCR1

Colon carcinoma; downregulated during the in vitro differentiation Osteosarcoma; enriched mRNAs compared to normal osteoblasts Glioblastoma; poor prognostic marker Increased expression during mouse fibroblasts transformation Proliferation marker in a number of tumours Neuroblastoma; overexpression linked to poor prognosis Variety of cancers; autoantibodies to DDX48 Ovarian cancer; upregulated HER-2/neu-positive breast cancer, melanoma; upregulated Osteosarcoma; enriched mRNAs compared to normal osteoblasts Ovarian cancer; upregulated; antiapoptotic

39 40 41 42 43 and 44 45 46 11 47 and 48 40 10 and 49

DDX1 GREB1 LPIN1 ODC1 RRM2 TAF1B

Neuroblastoma, coamplified with MYCN Breast cancer; suppression blocks estrogen-induced growth Neuroblastoma, coamplified with MYCN Neuroblastoma, coamplified with MYCN Pancreatic cancer; therapeutic target MSH-I colorectal carcinoma; mutated in 82% of tumours

50 51 52 53 54 55

Genes that are underlined were identified by two or more probes. aReferences are listed in Supplementary Material.

previously with survival, the possible impact of dosage of genes on 4p upon progression in neuroblastoma is supported by gain of the distal 2 Mb of 4p in a complex unbalanced translocation [t(1;4;17)] in the CLB-Bar neuroblastoma cell line. This line was obtained from a relapsed tumour, and the translocation resulted in gain of 4p and 17q and deletion of 1p arm (Schleiermacher et al., 2005). It is also noteworthy that an analysis of rare inherited neuroblastomas identified linkage to 4p16

with a logarithm of the odds score which peaked in this region (Perri et al., 2002). The distal 2 Mb region of 4p includes 11 genes whose overexpression is associated with poor survival in this study, including four (ATP5I, SLBP, TACC3 and WHSC1) where the correlation is independent of 4p del status. We have defined a shortlist of candidate genes within regions of known prognostic importance using gene expression profiling, and employed both overall survival Oncogene

Genomic and expression microarray analysis in neuroblastoma M Łastowska et al

7442

and comparative gene expression as additional criteria to define genes of particular interest. Given the complexities which underpin tumorigenesis, patient survival and species-specific expression differences, it is clear that no single criterion will be ideal and that additional discriminators could be used to further enhance the power of such analyses. Recently, for example, Walker et al. (2006) used a combination of SNP and expression microarrays to identify 3041 genes whose expression was altered in loss and gained regions in multiple myeloma, and used expression change during the transition from normal to myeloma plasma cells as a further criterion for candidate gene selection, reducing the number of candidate genes to just 47. The combination of methodologies employed here has identified regions of known prognostic importance in neuroblastoma (1p, 2p and 17q) and identified specific candidate genes within them for further analysis. Since these candidate genes were identified in metastatic tumours from patients with a poor prognosis, it is likely that they play a role in tumour progression or response to treatment. Our analysis has also uncovered a novel region, distal 4p, with a high density of genes whose expression correlates with both DNA copy number and survival. This is particularly striking, as these correlations are observed within metastatic tumours (stage 4). Furthermore, over 50% of candidate genes identified by this combined approach within the minimal regions of recurrent alteration on 17q and 4p have been implicated previously in cancer. These results strongly suggest that, despite the complexities, the integration of clinical or biological data with gene expression and high resolution estimates of DNA copy number may be a powerful tool with which to pinpoint candidate genes involved in tumorigenesis in a wide variety of cancers.

Materials and methods Patient samples and tissue processing Patients attended the Leeds (n ¼ 22) and Newcastle (n ¼ 8) NHS Trusts between January 1995 and September 2003, with patient age ranging from 1 month to 10 years. Diagnosis and staging were according to the International Neuroblastoma Staging System (Brodeur et al., 1993). Therapy was administered according to protocols of the United Kingdom Children’s Cancer Study Group (UKCCSG, now named Children’s Cancer and Leukaemia Group, CCLG), European Neuroblastoma Study Group and Localised Neuroblastoma European Study Group, with similar treatment being administered in both contributing centres. Informed consent was obtained to use tumour material for research at both centres. Primary tumours were snap frozen and stored at 801C. Cryosections (10 m) were stained with haemotoxylin and eosin and examined by a pathologist to identify tumours with >90% of neuroblastoma cells for subsequent RNA and DNA extraction. Total RNA was extracted using the RNeasy Micro kit (Qiagen, Hilden, Germany) and quality was checked using an Agilent 2100 Bioanalyser (Agilent Technologies, Palo Alto, CA, USA). DNA was extracted from the same tumour material using a phenol/chloroform/isoamyl alcohol extraction after proteinase K treatment. The quantity of RNA and DNA Oncogene

was established using the NanoDrop ND-1000 spectrophotometer (NanoDrop, Rockland, DE, USA). Expression microarrays Expression within human tumours was assessed using the Human Genome U133 Plus 2.0 Array (Affymetrix UK Ltd., High Wycombe, UK). Gene expression analysis of mouse and rat neuroblastoma cell lines, which harbour gain of regions syntenic to human 17q, was performed using GeneChip Mouse Expression Set 430 and GeneChip Rat Genome 230 2.0 Array. Cell lines were purchased from ATCC (USA), DSMZ (Germany) and the Interlab Cell Line Collection (Italy), and cells were cultured in RPMI medium with 10% foetal calf serum. RNA from whole brain preparations of mouse and rat were used as reference samples for tissues of neuronal origin. Preparation of in vitro transcription products, hybridization and scanning using the GeneChip scanner 3000 were performed according to Affymetrix protocols using a minimum of 1 mg of total RNA to prepare antisense biotinylated RNA without a second round of amplification. The quality of hybridization was assessed in all samples following the manufacturer’s recommendations. Data were analysed with Affymetrix GCOS 1.1.1 software using global scaling to a target signal of 500. Data were then imported into GeneSpring GX (Agilent Technologies) for subsequent analysis. Expression data for each probe was normalized with a ‘per chip normalization’ to the 50th percentile of all values on the chip, and a ‘per gene normalization’ to the median expression level of the gene across all samples. Probes which gave a present call signal in less than 10% of samples were excluded from analyses. Rodent gene expression data were processed in the same way as the human data with the exception that per gene normalization to the median was replaced by normalization to the reference sample (brain) and gene expression levels were presented as a fold change. To identify genes that were differentially expressed between two groups of tumours, with and without specific chromosomal alterations, a Wilcoxon test was applied with a Benjamini and Hochberg false discovery rate (FDR) of 0.05. All genes were also ranked for their association between gene expression and overall survival using a Cox proportional hazard model with a score test (Rao, 1973). To establish if minimal regions of gain or loss contained more genes whose expression is associated with survival than expected by chance, the rank of all genes present within each minimal regions of gain or loss was compared to the rank of all genes in the whole genome using a Wilcoxon test. Gene ontologies were analysed using the GO Ontology Browser within the Genespring GX analysis platform (Agilent Technologies). SNP Microarrays SNP analyses were performed on 19 stage 4 tumours where DNA was available, and five constitutional DNA patient samples obtained from peripheral blood, using Affymetrix GeneChip Mapping 50 K arrays (Affymetrix UK Ltd.) according to the manufacturer’s recommendations. All samples were analysed using the HindIII 50K array with one exception, NB20, which was analysed using the XbaI 50K array. Where DNA quantity was limited, a whole genome amplification step (multiple displacement amplification) was performed before digestion using F29 DNA polymerase according to the manufacturer’s instructions (Quiagen Inc., Valencia, CA, USA). SNP genotypes were established with GDAS3.0 software using default settings, resulting in an average of 95% positive SNP calls per sample. Copy number

Genomic and expression microarray analysis in neuroblastoma M Łastowska et al

7443 analysis was performed using both the Affymetrix gene chip Copy Number Analysis Tool (CNAT) version 2.0 (Huang et al., 2004) with a 500 kb Genome Smoothed Average (GSA) and using CNAG software (Nannya et al., 2005) with a 10 SNP sliding window. Only copy number changes identified by both methods were recorded, except in three polyploid tumours where the data from CNAG was preferred, as it was found to be more robust (as copy number can be defined relative to one or more regions of known ploidy, defined by cytogenetic analysis). No discrepancies involving gains or losses of >2 Mb were observed between the two methods. The consistency of copy number estimation was assessed in four cases through additional hybridization to the XbaI 50K mapping array (Affymetrix), and in five cases through direct comparison of tumour and constitution samples. All copy number changes >2 Mb in size were confirmed by these analyses. Genome smoothed copy number estimates from all HindIII arrays (n ¼ 18) were also ranked to investigate the association between copy number and overall survival using a Cox proportional hazard model with a score test (Rao, 1973). Real-time reverse transcriptase–PCR Real-time PCR for BRCA1, PPIA, WSB1, PMP22 and PPIA was performed using the 50 nuclease assay on the ABI PRISMt 7700 Sequence detector (Perkin-Elmer, Applied Biosystems, Foster City, CA, USA). Oligonucleotides were designed using Primer Express software (v.3.0, PE Biosystems, Foster City, CA, USA); forward and reverse primers were selected in adjacent exons to avoid amplification of genomic

DNA. Primers, TaqMan TAMRAt probes and specific conditions are listed in Supplementary Table S1. The mean Ct value was normalized against that of the endogenous control gene PPIA (Fischer et al., 2005). Relative gene expression was calculated with the 2DDCt method (Livak and Schmittgen, 2001), using as a reference for each gene the tumour RNA which showed the median expression level for that gene as measured by microarray analysis (tumour 24 for PMP22, tumour 10 for WSB1, tumour 3 for BRCA1 and BIRC5). Molecular cytogenetic analysis Presence of MYCN amplification and status of chromosomes 1 and 17 were analysed in all 30 tumours using FISH. Shortterm culture of primary tumour cells, harvesting and slide preparation were performed according to published protocols (Bown et al., 1994). Conditions for FISH hybridization, washing, detection and the probes used for 17q, 1p and MYCN analyses have been described elsewhere (Lastowska et al., 1997, 2002). Acknowledgements The financial support of the Neuroblastoma Society UK, Newcastle Healthcare Charity, the Candlelighters Trust Leeds and Cancer Research UK is gratefully acknowledged. The authors also thank the CCLG for access to constitutional DNA samples.

References Ambros PF, Ambros IM. (2001). SIOP Europe Neuroblastoma Pathology, Biology, and Bone Marrow Group. Pathology and biology guidelines for resectable and unresectable neuroblastic tumors and bone marrow examination guidelines. Med Pediatr Oncol 37: 492–504. Asgharzadeh S, Pigue-Regi R, Sposto R, Wang H, Yang Y, Shimada H et al. (2006). Prognostic significance of gene expression profiles of metastatic neuroblastomas lacking MYCN gene amplification. J Natl Cancer Inst 98: 1193–1203. Attiyeh EF, London WB, Mosse´ YP, Wang O, Winter C, Khazi D et al. (2005). Chromosome 1p and 11q deletions and outcome in neuroblastoma. New Eng J Med 353: 2243–2253. Bown N, Cotterill S, Lastowska M, O’Neill S, Pearson ADJ, Nicholson J et al. (1999). Gain of chromosome arm 17q and adverse outcome in patients with neuroblastoma. New Eng J Med 340: 1954–1961. Bown N, Reid MM, Malcolm A, Davison EV, Craft AW, Pearson ADJ. (1994). Cytogenetic abnormalities of small round cell tumours. Med Pediatr Oncol 23: 124–129. Brodeur GM, Pritchard J, Berthold F, Carlsen NL, Castel V, Castelberry RP et al. (1993). Revision of the international criteria for neuroblastoma diagnosis, staging, and response to treatment. J Clin Oncol 11: 1466–1477. Bussey KJ, Chin K, Lababidi S, Reimers M, Reinhold WC, Kuo WL et al. (2006). Integrating data on DNA copy number with gene expression levels and drug sensitivities in the NCI-60 cell line panel. Mol Cancer Ther 5: 853–867. Caron H, Sluis P, Kraker J, Bokkerink J, Egeler M, Laureys G et al. (1996a). Allelic loss of chromosome 1p as predictor of unfavourable outcome in patients with neuroblastoma. New Engl J Med 334: 225–230.

Caron H, van Sluis P, Buschman R, Pereira do Tanque R, Beks L, de Kraker J et al. (1996b). Allelic loss of the short arm of chromosome 4 in neuroblastoma suggests a novel tumour suppressor gene locus. Hum Genet 97: 834–837. Chen QR, Bilke S, Wei JS, Greer BT, Steinberg SM, Westermann F et al. (2006). Increased WSB1 copy number correlates with its over-expression which associates with increased survival in neuroblastoma. Genes Chromosomes Cancer 45: 856–862. Chen QR, Bilke S, Wei JS, Whiteford CC, Cenacchi N, Krasnoselsky AL et al. (2004). cDNA array-CGH profiling identifies genomic alterations specific to stage and MYCNamplification in neuroblastoma. BMC Genomics 5: 70. De Preter K, Speleman F, Combaret V, Lunec J, Board J, Pearson A et al. (2005). No evidence for correlation of DDX1 gene amplification with improved survival probability in patients with MYCN-amplified neuroblastomas. J Clin Oncol 23: 3167–3168. Deng CX. (2006). BRCA1: cell cycle checkpoint, genetic instability, DNA damage response and cancer evolution. Nucleic Acids Res 6: 1416–1426. Fischer M, Skowron M, Berthold F. (2005). Reliable transcript quantification by real-time reverse transcriptase–polymerase chain reaction in primary neuroblastoma using normalization averaged expression levels of the control genes HPRT1 and SDHA. J Mol Diagn 7: 89–96. Hackett CS, Hodgson JG, Law ME, Fridlyand J, Osoegawa K, de Jong PJ et al. (2003). Genome-wide array CGH analysis of murine neuroblastoma reveals distinct genomic aberrations which parallel those in human tumors. Cancer Res 63: 5266–5273. Henrich KO, Fisher M, Mertens D, Benner A, Wiedemeyer R, Brors B et al. (2006). Reduced expression of CAMTA1 Oncogene

Genomic and expression microarray analysis in neuroblastoma M Łastowska et al

7444 correlates with adverse outcome in neuroblastoma patients. Clin Cancer Res 12: 131–138. Huang J, Wei W, Zhang J, Liu G, Bignell GR, Stratton MR et al. (2004). Whole genome DNA copy number changes identified by high density oligonucleotide arrays. Hum Genomics 1: 287–299. Janoueix-Lerosey I, Novikov E, Monteiro M, Gruel N, Schleiermacher G, Loriod B et al. (2004). Gene expression profiling of 1p35–36 genes in neuroblastoma. Oncogene 23: 5912–5922. Lastowska M, Cotterill S, Bown N, Cullinane C, Variend S, Lunec J et al. (2002). Breakpoint position on 17q identifies the most aggressive neuroblastoma tumours. Genes Chromosomes Cancer 34: 428–436. Lastowska M, Cullinane C, Variend S, Cotterill S, Bown N, O’Neill S et al. (2001). Comprehensive genetic and histopathologic study reveals three types of neuroblastoma tumours. J Clin Oncol 12: 3080–3090. Lastowska M, Chung Y-J, Cheng Ching N, Haber M, Norrin MD, Kees UR et al. (2004). Regions syntenic to human 17q are gained in mouse and rat neuroblastoma. Genes Chromosomes Cancer 40: 158–163. Lastowska M, Nacheva E, McGuckin A, Curtis A, Grace C, Pearson A et al. (1997). Comparative genomic hybridization study of primary neuroblastoma tumors. Genes Chromosomes Cancer 18: 162–169. Livak KJ, Schmittgen TD. (2001). Analysis of relative gene expression data using real-time quantitative PCR and the 2(-Delta Delta C(T)) Method. Methods 4: 402–408. Miller MA, Ohashi K, Zhu X, McGrady P, London W, Hogarty M et al. (2006). Survivin mRNA levels are associated with biology of disease and patient survival in neuroblastoma: a report from the Children’s Oncology Group. J Pediatr Hematol Oncol 28: 412–417. Nannya Y, Sanada M, Nakazaki K, Hosoya N, Wang L, Hangaishi A et al. (2005). A robust algorithm for copy number detection using high-density oligonucleotide single nucleotide polymorphism genotyping arrays. Cancer Res 65: 6071–6079. Nigro JM, Misra A, Zhang L, Smornov I, Colman H, Griffin C et al. (2005). Integrated array-comparative genomic hybridization and expression array profiles identify clinically relevant molecular subtypes of glioblastoma. Cancer Res 65: 1678–1686. Nisen PD, Waber PG, Rich MA, Pierce S, Garvin Jr JR, Gilbert F et al. (1988). N-myc oncogene RNA expression in neuroblastoma. J Natl Cancer Inst 80: 1633–1637. Perri P, Longo L, Cusano R, McConville CM, Rees SA, Devoto M et al. (2002). Weak linkage at 4p16 to predisposition for human neuroblastoma. Oncogene 28: 8356–8360. Popescu NC, Zimonjic DB. (1997). Molecular cytogenetic characterization of cancer cell alterations. Cancer Genet Cytogenet 93: 10–21.

Rao CR (ed). (1973). Linear Statistical Interference and its Applications (Wiley Series in Probability & Mathematical Statistics) 1st edn. John Wiley & Sons Inc: New York. Schleiermacher G, Bourdeaut F, Combaret V, Pierron G, Raynal V, Aurias A et al. (2005). Stepwise occurrence of a complex unbalanced translocation in neuroblastoma leading to insertion of a telomere sequence and late chromosome 17q gain. Oncogene 24: 3377–3384. Seeger RC, Brodeur GM, Sather H, Dalton A, Siegel SE, Wong KY et al. (1985). Association of multiple copies of the N-myc oncogene with rapid progression of neuroblastoma. New Engl J Med 313: 1111–1116. Seltzer R, Richmond TA, Pofahl NJ, Green RD, Eis PS, Nair P et al. (2005). Analysis of chromosome breakpoints in neuroblastoma at sub-kilobase resolution using fine-tilling oligonucleotide array CGH. Genes Chromosomes Cancer 44: 305–319. Slavc I, Ellenbogen R, Jung WH, Vawter GF, Kretschmar C, Grier H et al. (1990). MYC gene amplification and expression in primary human neuroblastoma. Cancer Res 50: 1459–1463. Stallings RL, Nair P, Maris JM, Catchpoole D, McDermott M, O’Meara A et al. (2006). High-Resolution analysis of chromosomal breakpoints and genomic instability identifies PTPRD as a candidate tumor suppressor gene in neuroblastoma. Cancer Res 66: 3673–3680. Vandesompele J, Baudis M, De Preter K, Van Roy N, Ambros P, Bown N et al. (2005). Unequivocal delineation of clinicogenetic subgroups and development of a new model for improved outcome prediction in neuroblastoma. J Clin Oncol 10: 22280–22299. Tsafrir D, Bacolod M, Selvanayagam Z, Tsafrir I, Shia J, Zeng Z et al. (2006). Relationship of gene expression and chromosomal abnormalities in colon cancer. Cancer Res 66: 2129–2137. Walker BA, Leone PE, Jenner MW, Li C, Gonzalez D, Johnson DC et al. (2006). Integration of global SNP-based mapping and expression arrays reveals key regions, mechanisms and genes important in the pathogenesis of multiple myeloma. Blood 108: 1733–1743. Wang Q, Diskin S, Rappaport E, Attiyeh E, Mosse Y, Shue D et al. (2006). Integrative genomics identifies distinct molecular classes of neuroblastoma and shows that multiple genes are targeted by regional alterations in DNA copy number. Cancer Res 66: 6050–6062. Weber A, Imisch P, Bergmann E, Christiansen H. (2004). Coamplification of DDX1 correlates with an improved survival probability in children with MYCN-amplified human neuroblastoma. J Clin Oncol 22: 2681–2690. Wolf M, Mousses S, Hautanieni S, Karhu R, Huusko P, Allinen M et al. (2004). High-resolution analysis of gene copy number alterations in human prostate cancer using CGH on cDNA microarrays: impact of copy number on gene expression. Neoplasia 6: 240–246.

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

Oncogene