Integrative Genomics Identifies Distinct Molecular Classes of Neuroblastoma and Shows That Multiple Genes Are Targeted by Regional Alterations in DNA Copy Number

Research Article

Qun Wang,1 Sharon Diskin,1,4 Eric Rappaport,1 Edward Attiyeh,1 Yael Mosse,1 Daniel Shue,1 Eric Seiser,1 Jayanti Jagannathan,1 Suzanne Shusterman,5 Manisha Bansal,1 Deepa Khazi,1 Cynthia Winter,1 Erin Okawa,1 Gregory Grant,4 Avital Cnaan,2 Huaqing Zhao,2 Nai-Kong Cheung,6 William Gerald,6 Wendy London,7 Katherine K. Matthay,8 Garrett M. Brodeur,1 and John M. Maris1,3

Divisions of 1Oncology and 2Biostatistics, Children’s Hospital of Philadelphia and Department of Pediatrics and 3Abramson Family Cancer Research Institute, University of Pennsylvania School of Medicine; 4Center for Bioinformatics, University of Pennsylvania, Philadelphia, Pennsylvania; 5Department of Pediatric Oncology, Dana-Farber Cancer Institute and Department of Pediatrics, Harvard Medical School, Boston, Massachusetts; 6Memorial Sloan-Kettering Cancer Center, New York, New York; 7Department of Statistics, University of Florida and Children’s Oncology Group, Gainesville, Florida; and 8University of California at San Francisco School of Medicine, San Francisco, California

Abstract also differentially expressed in the independent validation data set, providing a prioritized list of candidate neuro- Neuroblastoma is remarkable for its clinical heterogeneity blastoma suppressor genes. Taken together, these data are and is characterized by genomic alterations that are strongly consistent with the hypotheses that the neuroblastoma tran- correlated with tumor behavior. The specific genes that scriptome is a sensitive marker of underlying tumor biology influence neuroblastoma biology and are targeted by genomic and that chromosomal deletion events in this cancer likely alterations remain largely unknown. We quantified mRNA target multiple genes through alteration in mRNA dosage. expression in a highly annotated series of 101 prospectively Lead positional candidates for neuroblastoma suppressor collected diagnostic neuroblastoma primary tumors using an genes can be inferred from these data, but the potential oligonucleotide-based microarray. Genomic copy number multiplicity of transcripts involved has significant implica- status at the prognostically relevant loci 1p36, 2p24 (MYCN), tions for ongoing gene discovery strategies. 11q23, and 17q23 was determined by PCR and was aberrant in (Cancer Res 2006; 66(12): 6050-62) 26, 20, 40, and 38 cases, respectively. In addition, 72 diagnostic neuroblastoma primary tumors assayed in a different labora- tory were used as an independent validation set. Unsupervised Introduction hierarchical clustering showed that gene expression was Neuroblastoma is an important childhood cancer, as it accounts highly correlated with genomic alterations and clinical for f15% of all pediatric oncology deaths (1). The clinical hallmark markers of tumor behavior. The vast majority of samples of neuroblastoma is heterogeneity, with the likelihood of tumor MYCN with amplification and 1p36 loss of heterozygosity progression varying widely according to age and disease burden at (LOH) clustered together on a terminal node of the sample diagnosis. A large body of data support the hypothesis that the dendrogram, whereas the majority of samples with 11q clinical behavior of human neuroblastoma may be reliably deletion clustered separately and both of these were largely predicted based on the analysis of a panel of prognostic variables distinct from the copy number neutral group of tumors. Genes (reviewed in ref. 2). The Children’s Oncology Group (COG) currently involved in neurodevelopment were broadly overrepresented stratifies patients into low-, intermediate-, or high-risk categories in the more benign tumors, whereas genes involved in RNA based on analysis of well-defined prognostic factors, including processing and cellular proliferation were highly represented patient age at diagnosis (3), International Neuroblastoma Staging in the most malignant cases. By combining transcriptomic System (INSS) stage (4), tumor histopathology (5), DNA index (6), and genomic data, we showed that LOH at 1p and 11q was and MYCN amplification status (7, 8). This stratification system is associated with significantly decreased expression of 122 currently used to assign therapeutic intensity, mandating that (61%) and 88 (27%) of the genes mapping to 1p35-36 and the algorithm be as precise as possible. Clinical experience with all of 11q, respectively, suggesting that multiple genes may be this system suggests that the algorithm is useful, but misclassifi- targeted by LOH events. A total of 71of the 1p35-36genes were cations almost certainly occur resulting in overtreatment or undertreatment (2). Additional tumor-specific prognostic markers may be required to achieve maximal predictive power. A large number of genomic aberrations have been defined in Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/). neuroblastoma, and the pattern of these somatically acquired Presented in part at the 96th Annual Meeting of the AACR, Anaheim, CA, April 16- changes correlate with tumor behavior (9). Amplification of 20, 2005 and the 11th Conference of the Advances in Neuroblastoma, Genoa, Italy, June 16-19, 2004. the MYCN oncogene at 2p24 provides a paradigm for the clinical Requests for reprints: John M. Maris, Division of Oncology, Children’s Hospital of utility of a tumor-specific DNA rearrangement. MYCN amplification Philadelphia, 34th Street and Civic Center Boulevard, Philadelphia, PA 19104-4399. is present in f20% of newly diagnosed tumors and remains a Phone: 215-590-2821; Fax: 215-590-3770; E-mail: [email protected]. I2006 American Association for Cancer Research. powerful biomarker for an aggressive phenotype and poor survival doi:10.1158/0008-5472.CAN-05-4618 probability even in light of modern dose-intensive therapies

Cancer Res 2006; 66: (12). June 15, 2006 6050 www.aacrjournals.org

(10, 11). Hemizygous deletions and/or loss of heterozygosity (LOH) assessed using Affymetrix GeneChip Test 3 arrays. Statistical modeling of have been defined at multiple genomic regions, including probe set behavior was conducted using Probe Profiler (Corimbia, Berkeley, chromosomal arms 1p, 3p, 4p, 11q, 14q, 16p, and 19q at frequencies CA) as described by the manufacturer. The software implements a model- between 20% and 45% of cases using a variety of techniques based approach for data extraction from probe sets and incorporates quality-control features to detect and correct for the contribution of (2, 12). Likewise, unbalanced gain of 17q material may be the most nonspecific cross-hybridization effects. The model weights probe pairs common genomic aberration in neuroblastoma primary tumors, as based on consistency of performance. Normalization to remove chip- it is present in 50% to 75% of cases. Perhaps most importantly, to-chip variation is accomplished by subtracting the second percentile these events are nonrandom and are clinically relevant, with LOH intensity value and dividing by the interquartile range. A quantitative at 1p36 and 11q being independently prognostic for disease expression score (e-score) is then calculated for each probe set. outcome (13). Despite the clear clinical relevance of these genomic Data from this experiment, including sample annotations, have been aberrations, no definitive neuroblastoma suppressor gene has been deposited in Gene Expression Omnibus database and are available for free identified. at http://www.ncbi.nlm.nih.gov/geo/. Genome-scale profiling technology allows for detailed reassess- Unsupervised clustering. Unsupervised two-way agglomerative hierar- ment of the relationship between DNA aberrations with regional chical clustering with Pearson’s correlation coefficient as a distance metric gene expression (14, 15). Microarray studies in neuroblastoma to was done using GeneSpring 7.0 (Silicon Genetics, Redwood City, CA). Probe sets with low average expression or low variance were excluded in date have mainly focused on defining prognostic signatures or an effort to minimize noise. Specifically, probe sets exhibiting an average identifying genes associated with oncogenic MYCN (16–18). In e-score <50 or an estimated SD <75 were excluded, leaving 2,177 probe sets addition, McArdle et al. correlated global gene expression profiles remaining for cluster analysis. with 11q deletions in a limited number of samples but showed Identification of enriched gene sets. Gene Ontology (GO) annotation the power of this combined approach for gene discovery (19). We was used to characterize the functional relationship of differentially therefore sought to determine regional and global gene expression expressed gene classes between the varying sample classes (23). EASE differences between a large number of samples with and without version 2.0 was selected to examine the GO categories of molecular clinically relevant chromosomal aberrations in primary neuroblas- function, biological process, and cellular component (24). EASE functions tomas. These data were used to further our understanding of the to determine the overrepresentation of GO terms within a given gene list. genetic basis for neuroblastoma heterogeneity and to prioritize For each of the three GO categories, EASE establishes the total number of annotated terms within the input list and population as well as the regional candidate genes for further analyses. abundance of a specific term within the input list and population. Each gene class was processed against the population list of 2,177 probe sets and overrepresented GO terms were statistically determined by the one-sided Materials and Methods Fisher’s exact probability. GO terms with P V 0.05 were selected as Patient samples and molecular characterization. A total of 101 significant. In addition, a bootstrap false discovery rate (FDR) was diagnostic neuroblastoma primary tumor samples were selected from the calculated to ensure that the P of the GO terms remained significant over COG (n = 91 prospectively collected) or Children’s Hospital of Philadelphia multiple tests. This multiple testing correction creates random gene lists (n = 10) neuroblastoma tumor banks. All specimens were snap frozen in the from the population whose size is equal to that of the input list. These institutional pathology departments immediately on arrival according to random lists allow for the determination of the frequency of false-positive standard COG neuroblastoma biology protocols. Specimens were examined terms in the input list with Ps greater than or equal to the term of interest. for percent tumor content before RNA extraction and excluded if <70% In addition to having a P V 0.05, GO terms were required to have a neuroblastic content were estimated. Each tumor specimen was also bootstrap FDR V0.05. assayed for MYCN copy number by fluorescence in situ hybridization, Differential gene expression. Differential gene expression was DNA index by flow cytometry, histopathology according to the system of measured with both Patterns from Gene Expression (PaGE; ref. 25) and Shimada, and allelic status at chromosome bands 1p36, 11q23, and 17q23 Significance Analysis of Microarray (SAM; ref. 26) algorithms. Samples using standard techniques as described previously (refs. 5, 6, 20, 21). We were first segregated into those with and without a detectable genomic differentiate 11q LOH into two categories based on our prior data (13): alteration at the region being studied. Binary comparisons were then balanced 11q LOH = LOH at both 11q and 11p polymorphic loci and done for each region and results were visualized with heat maps. For unbalanced 11q LOH = loss of 11q loci with no loss at 11p loci. RNA was visualization purposes, e-scores were standardized by subtracting the isolated from f50 mg snap-frozen tumor tissue using TRIzol reagent average and dividing by the SD estimated from those samples without a (Invitrogen, Carlsbad, CA) and purified with the RNeasy System (Qiagen, detectable genomic alteration. Standardized e-scores were then binned Valencia, CA) according to the manufacturer’s recommendations and into five levels based on the number of SD from the mean (>1.5 SD below digested with DNase I (Ambion, Houston, TX). RNA quantity and quality the mean, 1-1.5 SD below the mean, within 1 SD of the mean, 1-1.5 above were determined by spectrophotometry, 1% agarose gel, and microfluidics- the mean, and >1.5 SD above the mean). Heat maps of the standardized based electrophoresis (Agilent 2100 Bioanalyzer, Agilent, Palo Alto, CA). and binned e-scores were then generated using Spotfire (Spotfire Fetal brain RNA (Clontech, Palo Alto, CA) was used as a control sample. The DecisionSite 8.0, Somerville, MA). Transcriptome maps were generated Children’s Hospital of Philadelphia Institutional Review Board approved this by placing the Affymetrix U95Av2 GeneChip probe sets located on 1p, 11q, study and informed consent was obtained before collection of all samples. 2p24-25, and 17q in linear order according to chromosomal map position An independent set of 72 primary neuroblastomas obtained at diagnosis using the May 2004 assembly of the human genome according to the for patients treated at Memorial Sloan-Kettering Cancer Center (MSKCC) University of California at Santa Cruz Genome Browser (hg 17; http:// was also made available for this study (16, 22). Specimens were procured genome.ucsc.edu; ref. 27). and processed in a manner identical to that described above. Samples were Reverse transcription-PCR. Real-time quantitative reverse transcrip- analyzed for LOH at 1p36 and amplification of MYCN similar to as tion-PCR (RT-PCR) was done as described previously (28). All primer/probe described above. Samples showing discordance of MYCN copy number and sets spanned exon boundaries to assure specificity for cDNA and primer expression values were excluded from this sample set (22). sequence data are available upon request. Relative expression of target Expression profiling. Total RNA (5-15 Ag) was used to generate cRNA gene was determined by normalization to glyceraldehyde-3-phosphate targets that were hybridized to Affymetrix U95Av2 oligonucleotide arrays dehydrogenase (GAPDH) using a standard curve method with 10 serial containing 12,625 probe sets according to the manufacturer’s instructions dilutions according to the manufacturer’s instruction. All RT-PCR experi- (Affymetrix, Santa Clara, CA). The quality of labeled cRNA target was ments included a no template control and were done in triplicate. The www.aacrjournals.org 6051 Cancer Res 2006; 66: (12). June 15, 2006

Wilcoxon’s rank-sum test was used to test for statistically significant imposing further restrictions on sample selection. For example, differences in expression based on class label. Correlations between RT-PCR although low-risk tumors can be INSS stages I, IIA, IIB, or IVS and microarray data were tested with the Spearman’s rank correlation. from patients of a variety of ages, we only selected INSS stage I cases for patients ages <18 months for this cohort because this is Results the most common and unambiguous representation of low-risk Sample selection strategy and quality control. The patient disease. Region-specific allelotyping data are also shown in Table 1 and tumor characteristics for the 101 samples analyzed on the and are consistent with past reports in terms of correlations with 12,625 oligonucleotide probe set Affymetrix U95Av2 array are high-risk disease features (2, 13, 29). Relationships between shown in Table 1. Samples were selected so that a minimum of expression data and patient survival were not a focus of this 20 cases would be available in four clinically and biologically report because the median duration of follow-up for the 80 patients distinct subsets of neuroblastoma: low-risk, intermediate-risk, who had not died was only 2.3 years. high-risk, and high-risk with MYCN amplification. In addition, we Quality control was assured by using the standard Affymetrix attempted to limit biological heterogeneity within each subset by algorithms as well as metrics available in Probe Profiler. Unscaled brightness was within 2.5 SD of the mean across all experiments. Four primary neuroblastomas (one from each class) were randomly selected from the set of 101 neuroblastomas and assayed twice on Table 1. Patient and tumor characteristics for 101 test and the Affymetrix array to assure consistency of results. Pairwise 72 validation cases Spearman’s rank correlation coefficients for each replicate comparison were 0.97 to 0.99 (data not shown). The standard Characteristics Test cases Validation cases intensity values from the two housekeeping genes GAPDH and b-actin (ACTB) showed very tight interexperimental consistency n 101 72 and the vast majority of signals were within 2.5 SD of the mean Age (y) across all experiments. <1 41 21 z16051Class discovery. To characterize the molecular phenotype of INSS stage human neuroblastoma, we first identified 2,177 probe sets that I274showed the highest degree of variability across the entire sample II 1 13 set. Unsupervised two-way agglomerative hierarchical clustering III 23 8 results are shown in Fig. 1. There were five main sample classes IV 50 41 discovered in this analysis and these were clearly distinguished IVS 0 6 from the control RNA sample derived from fetal brain. Sample class MYCN 1 (S1) was composed almost exclusively of cases with genomic Not amplified 81 60 amplification of MYCN (17 of 18) and 1p36 LOH (15 of 18). The only Amplified 20 12 sample in this class that did not have amplification of MYCN Histology Favorable 48 32 showed LOH at 1p36. All of these samples would be classified as Unfavorable 50 33 high-risk in the current COG classification schema due to MYCN Not available 3 7 amplification status or, in the one discordant sample, metastatic DNA index disease in a patient older than 1 year at diagnosis. Sample class S4, Hyperdiploid 66 26 on the other hand, was composed of 28 cases that were almost Diploid 26 12 exclusively without metastatic disease and few or no genomic Not available 9 34 alterations at the loci assayed. All but one of these samples would 1p36 status be classified as low or intermediate-risk by the current COG No LOH 74 39 system. Nine of the 12 cases in the S2 cluster (n = 12) were older LOH 26 18 than 1 year with metastatic disease at diagnosis, but all showed Not available 1 15 11q23 status normal MYCN and 1p36 status. Sample classes S3 (n = 28) and S5 No loss 59 ND (n = 15) were the most heterogeneous in terms of known clinical LOH 40 ND and genomic covariates, as 17 of 28 and 7 of 15 would be classified Unbalanced LOH* 17 ND as high-risk, respectively. Whereas there was only one case with Not available 2 72 MYCN amplification in each of these sample clusters, the majority 17q23 status showed genomic alterations at 1p36, 11q23, and/or 17q23. No gain 51 ND Examination of the gene dendrogram showed three distinct Gain 38 ND classes, and these correlated strongly with genomic aberrations Not available 12 72 (Fig. 1). Gene cluster 1 (G1) contained 907 probe sets that show COG risk group overexpression in the S1 sample group with MYCN amplification. Low 28 22 Intermediate 21 8 Genes located within or nearby the core region of the MYCN High 32 30 amplicon all were located in G1 (MYCN, NCYM, DDX1, and High + MYCN 20 12 NAG), and this cluster included multiple bona fide MYCN transcriptional target genes, such as ODC1 and MCM7 (30). This cluster was enriched for genes involved in cellular *See text for definition (unbalanced 11q LOH is a subset of cases with metabolism, nucleic acid, protein and macromolecule biosynthe- 11q LOH). sis, RNA processing, ribosomal turnover, and RNA processing (Supplementary Table S1).

Cancer Res 2006; 66: (12). June 15, 2006 6052 www.aacrjournals.org

Figure 1. Unsupervised hierarchical clustering of 101 primary neuroblastomas and human fetal brain. Unsupervised two-way agglomerative hierarchical clustering results show five main sample groups, S1 (red), S2 (yellow), S3 (blue), S4 (pink), and S5 (green), in addition to fetal brain control (light blue). Clinical and genomic covariates: Filled box (black), age = older than 1 year at diagnosis; INSS stage = 4; 1p36 = LOH; MYCN = amplified; 11q23 = LOH; 11q23 UNB = LOH; 17q = unbalanced gain. Gray, the condition is not known for that sample. Three main gene clusters are noted as G1 (blue), G2 (yellow), and G3 (red). Genes of interest are noted (with chromosome number in parentheses for those mapping to regions studied for DNA copy number as well), and those shown more than once reflect different probe sets for the same gene on the U95Av2 chip. www.aacrjournals.org 6053 Cancer Res 2006; 66: (12). June 15, 2006

A second cluster (G2; Supplementary Table S2) contained 437 expressed probe sets) or the comparison of unbalanced 11q genes and was enriched for those involved in host immune LOH versus balanced 11q LOH versus normal 11q status (452 response and antigen processing. High expression of these genes differentially expressed probe sets). was associated with sample cluster 2, composed mainly of high-risk We next focused only on the genes mapping within the genomic cases, but without MYCN amplification. A tight subcluster region that defined the class labels of aberrant versus normal in the containing several MHC class II genes showed nearly identical differential expression analyses. Of the 201 genes mapping to 1p35- expression patterns across the sample set (Fig. 1). 36, 122 (61%) showed significantly lower expression in the 26 cases The final cluster (G3; Supplementary Table S3) of 741 genes was with 1p36 LOH compared with only 5 1p35-36 genes that showed enriched for those highly expressed in the more benign subsets of significantly higher expression. Similar results were seen at each of neuroblastoma. GO terms overrepresented included cellular the other loci: 27% of the 11q mapped genes showed significantly communication and processing, vesicle-mediated transport, signal decreased expression in the 40 cases with 11q LOH, 79% of transduction, neurogenesis, and neurophysiologic processes. Genes the 2p24-25 mapped genes showed significantly increased expres- known to be involved in normal sympathetic nervous system sion in the 20 cases with MYCN amplification, and 40% of the 17q development, such as TH, DBH, PHOX2B, GATA3, and NTRK1, mapped genes showed significantly increased expression in the all showed higher expression in the nonmetastatic cases within 38 cases with unbalanced 17q gain and very few genes this cluster. This cluster was also enriched for genes mapping to showed significant differential expression in the opposite direction chromosome arm 1p, such as STX12, PTP4A2, GNB1, CLSTN1, (Table 2). PUM1, RERE, CHD5, and BAI2, and these showed lower expression To prove that the mRNA copy number estimates from the in the metastatic cases with 1p36 LOH. The nonreceptor tyrosine microarray accurately reflected true mRNA copy number in kinase FYN also showed differential expression in a similar pattern. the tumor cells, we did quantitative RT-PCR for 18 genes shown Region-specific alterations in gene expression. We created to be differentially expressed in these analyses. All Spearman’s rank transcriptome maps to visualize gene expression pattern based correlation coefficient Ps comparing microarray and RT-PCR data on chromosomal map position (Fig. 2). All genes present on the were <0.0001 (data not shown). A representative sampling of these Affymetrix U95Av2 GeneChip located on 1p, 2p24-25, 11q, and 17q data segregated by genomic alterations is shown in Fig. 3. were displayed in linear order according to genome map position. To prioritize regional candidate neuroblastoma suppressor genes Samples were then segregated into those with and without a at 1p and 11q, we arranged the probe sets in rank order by their detectable genomic aberration at the region being studied. There t statistic for differential expression based on allelic status. were a large number of genes showing differential expression at Figure 4A and B shows the top 50 differentially expressed genes each of the loci under study. For examples, the presence of 1p36 mapping to 1p35-36 and 11q, respectively. This visualization tool LOH was associated with low or absent expression of hundreds of allows one to identify samples that show mRNA expression genes mapping telomeric to cytoband 1p21 (Fig. 2A). Samples patterns that are discordant with their class assignment. For with MYCN amplification showed a very tight region of high example, samples 1,009, 1,043, and 1,068 showed no evidence for overexpression noted at 2p24, with MYCN, NCYM, DDX1, and NAG 1p36 LOH yet showed expression patterns more similar to the showing striking differential expression based on the presence or samples with 1p36 LOH. Each of these samples also showed absence of the amplification event (Fig. 2B). Samples with MYCN amplification. In contrast, several of the samples coded as unbalanced 11q LOH showed the most marked differential having 1p36 LOH but not MYCN amplification showed expression expression of 11q14-q25 mapped transcripts, but many of the patterns more similar to the cases without 1p36 LOH (e.g., 58, 205, samples with balanced 11q LOH also showed a similar pattern 1,118, and 1,204). (Fig. 2C). A similar pattern in the opposite direction is noted at To determine if our methods reproducibly identify transcripts 17q, but this is less robust (Fig. 2D). that can be considered as regional candidate tumor suppressor Identification of genes differentially expressed based on genes or oncogenes, we used an independent data set annotated genomic alterations. We next determined which genes were with 1p36 and MYCN genomic data (Table 1). Of the 122 genes differentially expressed throughout the genome in each binary showing significant differential expression in the expected comparison of genomic status at 1p36, 2p24, 11q23, and 17q23 direction at 1p36 in our data set, 71 (58%) also showed significant using both PaGE and SAM algorithms. We found no difference differential expression in the same direction in the independent between these two techniques (data not shown) and only consider validation data set (Table 2). Not surprisingly, concordance was the PaGE data here (full output from this analysis can be visualized highest for the most significantly differentially expressed genes at http://stokes.chop.edu/programs/marislab/data). Summary (Table 3). Table 3 shows the 1p35-36 transcripts concordant in both PaGE statistics are shown in Table 2. On a genome scale, the data sets as ranked by t statistic. Figure 5 compares data for the binary comparisons of normal versus aberrant at 1p36, MYCN, 11q, most differentially expressed genes at the MYCN locus at 2p24-25 in and 17q showed differential expression of 4,386, 6,269, 293, and both data sets, again showing a high degree of concordance. 3,360 probe sets, respectively. For each analysis, the majority of the most statistically significantly differentially expressed genes in the expected direction (lower in LOH cases for 1p and 11q and higher Discussion in gain cases for 2p24/MYCN region and 17q) mapped to the region Neuroblastoma has served as a paradigm for the clinical being studied. Many of the genes showing decreased expression importance of tumor genomic data (2, 13, 29). MYCN oncogene in the 11q analyses also mapped to 3p, a region known to be amplification is reliably associated with an aggressive clinical coordinately deleted in many cases with 11q deletions (31). phenotype and remains a powerful prognostic maker even in the The much lower number of differentially expressed genes in the era of modern dose-intensive therapy. However, despite a wealth chromosome 11 analysis did not change substantively if we of data on the clinical relevance of other chromosomal aberrations considered unbalanced 11q LOH versus all others (305 differentially in this disease, no bona fide neuroblastoma suppressor gene has

Cancer Res 2006; 66: (12). June 15, 2006 6054 www.aacrjournals.org

Figure 2. Transcriptome maps. All genes present on the Affymetrix U95Av2 GeneChip located on chromosomes cytobands 1p11-1p36 (A), 2p24-25 (B), 11q11-11q25 (C), and 17q11-17q25 (D) are shown in linear order according to map position. Patient samples (columns) were grouped according to their genomic status at the locus of interest. Expression levels were first normalized and then binned into five levels (green, >1.5 SD below the mean; dark green, 1-1.5 SD below the mean; black, within 1 SD of the mean; dark red, 1-1.5 above the mean; red, >1.5 SD above the mean). Genes within or near a typical MYCN amplicon (37, 79) are annotated in (B). www.aacrjournals.org 6055 Cancer Res 2006; 66: (12). June 15, 2006

Downloaded from cancerres.aacrjournals.org on September 27, 2021. © 2006 American Association for Cancer Research. Cancer Research been unequivocally identified. The fact that the majority of (Fig. 1, sample class S2). These were all locally invasive (INSS stage hemizygous deletions mapped in neuroblastoma cells are typically III) or metastatic tumors, and most had evidence for copy number dozens of megabases in size led us to speculate that multiple alterations at 11q and/or 17q but not 1p. Although the duration of genes are likely targeted by these acquired genomic alterations. follow-up is short, it is intriguing that only 2 of 12 patients in this This has important implications for gene discovery efforts and subset have died from tumor progression, although 9 of these cases for ongoing studies directed toward identifying genetic biomarkers would be classified as high-risk using the current COG system with of disease outcome and/or new therapeutic targets. Thus, the expected survival probabilities of <40% (44). Although it is certainly primary aim of this study was to combine DNA copy number and possible that this signature is due to tumor-infiltrating lympho- mRNA expression data to understand the magnitude of transcrip- cytes rather than the tumor tissue itself, significant lymphocytic tional alterations at candidate tumor suppressor gene and infiltration of neuroblastomas have generally been observed in low- oncogene loci. stage cases with the associated opsoclonus-myoclonus syndrome Class discovery methods have been used to differentiate distinct (45–47). We speculate that evidence for immune activation may human cancers (32, 33) and more recently have been applied have an influence on tumor biology, but this very preliminary to subclassify within a given histopathology (34–36). It is well observation requires careful follow-up. In contrast, low-risk tumors known that human neuroblastomas show dramatic clinical and had an expression signature enriched in genes involved in normal biological heterogeneity. Using unsupervised class discovery sympatheticoadrenal developmental pathways, such as PHOX2B, methods, we now show that there are distinct expression profiles GATA3, NTRK1, TH, DBH, and FYN, suggesting that loss of a that correlate with clinically and biologically relevant subsets of neuronal differentiation/maturation pathway may be critical for neuroblastoma. The most striking mRNA signature is embedded developing a more malignant tumor phenotype. in the cases with MYCN amplification (Fig. 1). These cases almost Our gene expression profiling data suggest three distinct classes all cluster together. One case that clustered elsewhere in the based on mRNA signatures (S2, S3, and S5) with no clear sample dendrogram showed significantly lower MYCN mRNA copy distinguishing features in terms of known clinical and biological number, whereas the other two were the only two infants with covariates. It will be important to determine, as the data set MYCN amplified tumors in this study. In general, MYCN-amplified matures, whether there is a correlation with tumor clinical samples showed overexpression not only of genes mapping within behavior and patient outcome. It is tantalizing to speculate that or near the typical core MYCN amplicon (37), such as MYCN, the S2 class samples with the strong immune signature might be NCYM, DDX1, and NAG, but also of known transcriptional targets, correlated with improved antitumor control and high survival such as ODC1 and MCM7 (30, 38). Genes mapping to chromosome probability, but the relatively small subset sample sizes and short 17q were also enriched, and this is likely due to the strong follow-up duration preclude these analyses at the present time. correlation of 17q gain with MYCN amplification (39) and the Deletions of chromosome arms 1p and 11q are observed in fact that the NME1 and PPM1D genes likely have an oncogenic neuroblastomas with a more malignant phenotype and detection role during neuroblastoma progression (40–42). In contrast, of these genomic alterations is being applied in the clinic as these samples showed low or absent expression of a large number prognostic biomarkers (13). Despite enormous effort to define of genes mapping to chromosome 1p, a genomic region that is the genes targeted by these deletions, no definitive 1p or 11q deleted in the vast majority of samples also with MYCN neuroblastoma suppressor gene has been identified. Our data are amplification (13, 43). consistent with the hypothesis that perhaps dozens of genes A novel finding was the highly enriched immunity signature, mapping to 1p35-36 and 11q14-25 contribute to the malignant including much higher expression of HLA molecules in a distinct phenotype of high-risk neuroblastomas. By evaluating a completely subset of high-risk neuroblastomas without MYCN amplification independent group of samples for which 1p allelic status was

Table 2. Region-specific alterations in gene expression

Region studied (Mb) No. differentially expressed No. genes No. genes No. differentially expressed No. validated differentially c b genes-genome* in region on U95Av2 genes-region expressed genes-region

1p35-36 (0-34) 2,034 # 563 201 122 # 71 # 2,352 " 5 " 2p24-25 (0-24) 3,033 " 130 47 37 " 22 " 3,236 # 3 # 11q (53-134) 203 # 1,079 329 88 # Not done 90 " 3 " 17q (22-80) 1,580 " 931 372 150 " Not done 1,780 # 58 #

*Number of differentially expressed genes (probe sets) with PaGE confidence levels z0.8 in the entire data set. Arrows indicate direction of change compared with the samples without the genomic alteration being analyzed. cNumber of differentially expressed genes with PaGE confidence levels z0.8 in the region under consideration. bNumber of differentially expressed genes with PaGE confidence levels z0.8 in the validation data set that were also present in the study data set for each region under consideration.

Cancer Res 2006; 66: (12). June 15, 2006 6056 www.aacrjournals.org

Figure 3. Quantitative RT-PCR validation of oligonucleotide-based gene expression. Representative results for five genes located within genomic regions of interest. Log-transformed data were normalized to GAPDH and represented as box and whisker plots for both microarray (A) and RT-PCR (B) experiments. Statistical significance was determined with the Wilcoxon’s test statistic.

known, we could show tight concordance in the 1p genes discovered in this study implies a role in neuroblastoma differentially expressed in both data sets (Table 3). It is also suppression as well. reassuring that the vast majority of 1p genes identified here were Classic positional cloning efforts may be streamlined by also seen by Janoueix-Lerosey et al. as candidate neuroblastoma considering genome-scale mRNA copy number data. We recently suppressors using a similar study design in a smaller sample set mapped a 729-kb consensus region of deletion at 1p36.3 in a large (48). Specifically, 10 of the 11 1p genes in their study (GNB1, series of primary neuroblastomas (53). This region overlapped and CLSTN1, CDC42, MFN2, SLC35E2, CAPZB, TCEB3, VAMP3, RBAF600, was partially defined by constitutional deletions in two children and STMN1) that were also represented on our microarray with developmental abnormalities who subsequently developed platform were significantly differentially expressed in both our neuroblastoma, suggesting that a gene within this region is primary and validation data sets. Caren et al. recently identified involved in neuroblastoma initiation. Of the 24 known transcripts several genes at 1p36.22 that show lower expression in neuroblas- within this interval, CHD5, ICMT, TNFRSF25, KIAA0469, MGC33488, tomas with a poor prognosis, and two of these genes (UBE4B and and CAMTA1 each showed highly significant decreased expression CORT) were also identified as differentially expressed based on in both primary and validation data sets. Likewise, several other 1p36 allelic status in this study (49). CASP9 was differentially positional candidate 1p neuroblastoma suppressor genes have expressed in both data sets, but its role in neuroblastoma biology been postulated, and the CDC2L2 (PITSLRE), PRDM2 (RIZ1), has been controversial (50, 51). Lastly, CAMTA1 has recently been CDC42, and CHD5 genes were each highly differentially expressed postulated to function as an oligodendroglioma suppressor gene in both data sets (41, 54–56), supporting a role for each gene in through haploinsufficiency (52), and its differential expression tumor evolution. Taken together, these data support the hypothesis www.aacrjournals.org 6057 Cancer Res 2006; 66: (12). June 15, 2006

Figure 4. Heat map representation of most differentially expressed genes at 1p35-36 (A) and 11q (B). Top 50 genes discovered to be differentially expressed in each binary comparison of LOH versus no LOH for chromosome bands 1p36 and 11q23. Genes are sorted by their PaGE generated t statistic (available at http://stokes.chop.edu/programs/marislab/data).

Cancer Res 2006; 66: (12). June 15, 2006 6058 www.aacrjournals.org

Table 3. Genes mapping to distal 1p with significantly Table 3. Genes mapping to distal 1p with significantly lower lower expression in both test and validation data sets expression in both test and validation data sets (Cont’d)

c c Gene* Locus t statistics Confidence Gene* Locus t statistics Confidence

Penn MSKCC Penn MSKCC Penn MSKCC Penn MSKCC

STX12 1p35.3 9.09 2.66 0.996 0.980 ARHGEF16 1p36.3 2.55 1.31 0.986 0.867 PTP4A2 1p35 7.62 1.66 0.996 0.894 CASP9 1p36.21 2.49 1.54 0.985 0.889 GNB1 1p36.33 7.21 1.75 0.996 0.904 KIAA1037 1p36.11 2.47 2.96 0.984 0.980 CLSTN1 1p36.22 7.05 2.47 0.996 0.973 HTR6 1p36.13 2.44 1.57 0.983 0.889 PUM1 1p35.2 6.73 1.83 0.996 0.915 STMN1 1p36.11 2.34 1.66 0.98 0.894 RERE 1p36.23 6.32 2.48 0.996 0.973 LYPLA2 1p36.11 2.22 1.28 0.973 0.866 DNAJC8 1p35.2 6.22 2.51 0.996 0.974 TAF12 1p35.2 2.12 1.54 0.967 0.889 CDC42 1p36.1 6.22 1.57 0.996 0.889 ZNF151 1p36.13 2.07 2.72 0.963 0.980 KIAA0962 1p36.1 5.95 1.77 0.996 0.909 KIAA0478 1p36.12 1.98 1.39 0.958 0.881 MFN2 1p36.21 5.90 2.06 0.996 0.942 LAPTM5 1p35.2 1.89 2.90 0.951 0.980 KIAA0842 1p36.13 5.90 2.39 0.996 0.971 SCNN1D 1p36.33 1.84 1.23 0.947 0.858 BAI2 1p35 5.87 2.54 0.996 0.974 G1P2 1p36.33 1.73 1.45 0.935 0.889 ICMT 1p36.21 5.72 1.12 0.996 0.835 G1P2 1p36.33 1.66 1.33 0.926 0.868 PRDM2 1p36 5.65 1.47 0.996 0.889 CLIC4 1p36.11 1.57 1.24 0.918 0.859 SLC35E2 1p36.33 5.64 2.69 0.996 0.980 RAP1GA1 1p36.12 1.47 1.37 0.904 0.879 P29 1p36.11 5.52 1.90 0.996 0.920 CDW52 1p36 1.47 1.35 0.904 0.873 PIK3CD 1p36.2 5.51 1.99 0.996 0.938 CORT 1p36.22 1.30 1.63 0.874 0.889 CHD5 1p36.23 5.45 1.07 0.996 0.824 SDR1 1p36.1 1.18 1.50 0.85 0.889 CAMTA1 1p36.23 5.19 1.99 0.996 0.938 LCK 1p35.2 1.15 1.41 0.845 0.882 CLCN6 1p36 5.04 1.44 0.996 0.889 HMGN2 1p36.1 1.06 1.84 0.835 0.915 FRAP1 1p36.2 5.02 0.99 0.996 0.805 GJB5 1p35.1 0.99 1.52 0.821 0.889 SDC3 1p35.2 4.94 3.23 0.996 0.983 PMSCL2 1p36.22 4.88 2.22 0.996 0.957 CTNNBIP1 1p36.22 4.86 2.73 0.996 0.980 *Of the 146 probe sets representing 122 genes with confidence levels CDC2L1 1p36 4.79 1.53 0.996 0.889 >0.8 in the Penn data set, 111 probe sets representing 98 genes were RER1 1p36.33 4.78 1.57 0.996 0.889 also shown to be significantly differentially expressed with confidence KIAA0453 1p36.21 4.77 2.03 0.996 0.940 levels >0.8 in the MSKCC data set. Genes listed more than once in DJ159A19.3 1p36.13 4.70 1.77 0.996 0.909 cases where more than one probe set for the given gene on the SRRM1 1p36.11 4.60 3.29 0.996 0.985 Affymetrix U95Av2 array showed differential expression. Genes in SFRS4 1p35.2 4.51 2.66 0.996 0.980 boldface have been postulated to be neuroblastoma suppressor genes UBE4B 1p36.3 4.49 1.69 0.996 0.895 previously (see text) and/or were also found to be differentially CAPZB 1p36.1 4.47 3.10 0.996 0.980 expressed based on 1p LOH status by Janoueix-Lerosey et al. (48) or TCEB3 1p36.1 4.47 1.54 0.996 0.889 Caren et al. (49). EIF4G3 1p36.12 4.39 1.77 0.996 0.909 cThe t statistic and confidence levels as computed with the PaGE VAMP3 1p36.23 4.35 2.03 0.996 0.940 algorithm. Negative numbers denote decreased expression in samples CASP9 1p36.21 4.35 0.97 0.996 0.805 with 1p LOH. FUCA1 1p36.11 4.10 2.02 0.996 0.940 RER1 1p36.33 4.07 1.30 0.996 0.866 AKR7A2 1p36.13 3.86 1.51 0.996 0.889 ECE1 1p36.1 3.80 2.09 0.996 0.945 CDC2L2 1p36.3 3.73 1.99 0.996 0.938 that transcriptional alterations in multiple genes located at the CDC2L2 1p36.3 3.73 2.54 0.996 0.974 distal short arm of chromosome 1 cooperate during neuroblastoma DJ465N24.2.1 1p36.11 3.60 3.57 0.995 0.985 initiation and/or progression. MGC33488 1p36.23 3.51 0.97 0.995 0.805 There was remarkable concordance between the two data NUDC 1p36.11 3.49 1.44 0.995 0.889 sets regarding transcripts mapping to 2p24-25 that are overex- PTP4A2 1p35 3.48 2.41 0.995 0.971 pressed in samples with MYCN amplification (Fig. 5). These data SHARP 1p36.13 3.38 1.53 0.995 0.889 are reassuring for ongoing efforts to map DNA copy number KIAA0495 1p36.32 3.29 1.31 0.995 0.867 aberrations using mRNA expression profiles (15). Whereas the KIAA0469 1p36.23 3.26 2.40 0.994 0.971 alterations in mRNA transcription secondary to high-level genomic VAMP3 1p36.23 3.13 1.08 0.994 0.824 amplification at a particular locus seem very reproducible despite KIAA0450 1p36.32 3.12 1.44 0.994 0.889 TNFRSF25 1p36.2 3.12 2.72 0.994 0.980 variations in amplicon size, the downstream consequences of such PRDM2 1p36 3.09 1.88 0.994 0.918 an event are myriad. Over half of the transcriptomes under study TNFRSF1B 1p36.22 2.92 2.67 0.992 0.980 (6,269 probe sets; Table 2) showed significant differential RBAF600 1p36.13 2.91 1.47 0.992 0.889 expression in the binary comparison of samples dichotomized on TNFRSF1B 1p36.22 2.67 2.69 0.988 0.980 MYCN amplification status. The distribution was approximately RUNX3 1p36 2.61 1.54 0.987 0.889 equal between genes showing higher or lower expression in the HMGCL 1p36.11 2.60 1.07 0.987 0.824 MYCN-amplified cases (3,033 versus 3,236 probe sets, respectively). These data further support the notion that the Myc proteins are www.aacrjournals.org 6059 Cancer Res 2006; 66: (12). June 15, 2006

Downloaded from cancerres.aacrjournals.org on September 27, 2021. © 2006 American Association for Cancer Research. Cancer Research promiscuous transcriptional activators and repressors with pleio- our primary data allow for prioritization of potential 11q tropic and diverse effects on cell function. neuroblastoma suppressor genes. Known tumor suppressor genes, Only 40% of high-risk neuroblastomas show MYCN amplifica- such as IGSF4 (63), SDHD (64), MLL (65), and PPP2R1B (66), tion, suggesting that alternative pathways to a malignant showed low or absent expression in the majority of cases with 11q phenotype exist. Deletion of chromosome arm 11q has recently LOH. A similar pattern was seen for genes involved in response to emerged as a critical genomic event in the evolution of high-risk DNA damage (ATM and H2AFX; ref. 67), cell adhesion (NCAM1; neuroblastomas that do not have MYCN amplification (13, 57). ref. 68) and nonreplicative senescence (LOH11CR2A; ref. 69). However, no 11q neuroblastoma suppressor gene has been Each could be considered candidate neuroblastoma suppressor identified. Like 1p deletions, 11q deletions are often large spanning genes based on our expression data, but only the SDHD gene has over 60 Mb from 11q13 to 11qter (58), and the only known been studied in detail in this disease (70). constitutional deletions at 11q are likewise very large (59, 60). Epigenetic modification of DNA is an important somatically Whereas a critical region has been mapped to 11q23 (61), several acquired mechanism to silence gene expression in human cancer. tumors show deletions that map outside of this region (13), and Several examples exist in neuroblastoma, including dense hyper- functional studies have suggested other chromosome 11 loci methylation of promoter sequences for several genes, including critical for suppression of differentiation (62). Whereas our CASP8 (71), RASSF1A (72), TSP1 (73), NR1I2 (74), etc. It has independent data set was not annotated for 11q allelic status, also been shown that a CpG island hypermethylation phenotype

Figure 5. Heat map representation of most differentially expressed genes at 2p24-25 in the Penn (A) and MSKCC (B) data sets. Genes are sorted by their PaGE generated t statistic (available at http://stokes.chop.edu/programs/marislab/data) and show complete concordance for the genes most significantly differentially expressed between MYCN-amplified and single-copy cases.

Cancer Res 2006; 66: (12). June 15, 2006 6060 www.aacrjournals.org

Downloaded from cancerres.aacrjournals.org on September 27, 2021. © 2006 American Association for Cancer Research. Integrative Genomics of Neuroblastoma is associated with more aggressive clinical behavior and poorer In conclusion, this study shows that the genomic data embedded survival probability (75, 76). It is certainly possible that many of the in neuroblastoma diagnostic biopsies can be used to subcategorize genes showing low expression in regions of hemizygous deletion in the disease into molecular subsets. Our data also clearly show this study are subjected to epigenetic silencing of the retained that the regional copy number alterations are correlated with a allele. Ongoing studies will clarify the frequency of epigenetic broad array of transcriptional alterations genome-wide. For LOH versus genetic (mutation) events to result in biallelic inactivation and low copy number gain events that are generally large enough as well as to define if the second event is not required for some to be visible at the karyotype level, the number of genes residing genes and that a phenotypic effect can be manifested simply from within these regions that show significant mRNA copy number haploinsufficiency of the deleted allele. alterations is daunting. Our data suggest that multiple genes Oncogenic activation of key signaling pathways may provide from several discrete regions of the human genome cooperate to the most realistic target for pharmacologic manipulation. Although suppress neuroblastoma tumorigenesis and progression. DNA the MYCN oncogene is an obvious drug target, the profound degree alterations likely target multiple genes; therefore, detection of copy of transcriptional deregulation makes complete inhibition challeng- number alterations may provide more prognostically relevant ing, and this abnormality is found in only 20% of all neuroblastoma information than mRNA expression signatures. Ongoing studies cases. Unbalanced gain of 17q material is nearly ubiquitous in will test whether DNA or RNA alterations are more useful in a high-risk neuroblastomas (39, 57, 77), and we postulate that clinical setting or whether the two will provide complementary combining genomic and transcriptomic data may help prioritize information. potential molecular targets for rationale drug design. Our data are consistent with previous reports that NME1 and PPRM1D are Acknowledgments overexpressed in neuroblastomas with 17q gain (40–42) but also Received 1/2/2006; revised 3/22/2006; accepted 3/30/2006. identify multiple other potential oncogenes, including some that are Grant support: NIH grants R01-CA78545 and R01-CA87847 (J.M. Maris), potentially drugable tyrosine kinases (78), such as MAP3K3 and P01-CA97323 (G.M. Brodeur, S. Shusterman, and J.M. Maris), P01-CA106450 (N-K. Cheung and W. Gerald), R01-CA39771 (G.M. Brodeur), and U10-CA98543 (COG), Alex’s TLK2. Combining array comparative genomic hybridization and Lemonade Stand Foundation (J.M. Maris), Hope Street Kids Foundation (Q. Wang and expression profiling data sets at higher resolution genome-wide will J.M. Maris), and Abramson Family Cancer Research Institute (J.M. Maris). The costs of publication of this article were defrayed in part by the payment of page provide a much more refined list of molecular targets that may be charges. This article must therefore be hereby marked advertisement in accordance exploitable. with 18 U.S.C. Section 1734 solely to indicate this fact.

References 13. Attiyeh EF, London WB, Mosse YP, et al. Chromo- gene expression data by assigning confidence to differ- some 1p and 11q deletions and outcome in neuroblas- entially expressed genes. Bioinformatics 2000;16:685–98. 1. Brodeur GM, Maris JM. Neuroblastoma. In: Pizzo PA, toma. N Engl J Med 2005;353:2243–53. 26. Tusher VG, Tibshirani R, Chu G. Significance analysis Poplack DG, editors. Principles and practice of pediatric 14. Nigro JM, Misra A, Zhang L, et al. Integrated array- of microarrays applied to the ionizing radiation oncology. 4th ed. Philadelphia: J.B. Lippincott Co.; 2002. comparative genomic hybridization and expression response. Proc Natl Acad Sci U S A 2001;98:5116–21. p.895–938. array profiles identify clinically relevant molecular 27. Kent WJ, Sugnet CW, Furey TS, et al. The 2. Maris JM. The biologic basis for neuroblastoma subtypes of glioblastoma. Cancer Res 2005;65:1678–86. human genome browser at UCSC. Genome Res 2002;12: heterogeneity and risk stratification. Curr Opin Pediatr 15. Yi Y, Mirosevich J, Shyr Y, Matusik R, George AL, Jr. 996–1006. 2005;17:7–13. Coupled analysis of gene expression and chromosomal 28. Wang Q, Hii G, Shusterman S, et al. ID2 expression is 3. Breslow N, McCann B. Statistical estimation of location. Genomics 2005;85:401–12. not associated with MYCN amplification or expression prognosis for children with neuroblastoma. Cancer Res 16. Alaminos M, Mora J, Cheung NK, et al. Genome-wide in human neuroblastomas. Cancer Res 2003;63:1631–5. 1971;31:2098–103. analysis of gene expression associated with MYCN in 29. Vandesompele J, Baudis M, De Preter K, et al. 4. Brodeur GM, Pritchard J, Berthold F, et al. Revisions of human neuroblastoma. Cancer Res 2003;63:4538–46. Unequivocal delineation of clinicogenetic subgroups the international criteria for neuroblastoma diagnosis, 17. Wei JS, Greer BT, Westermann F, et al. Prediction of and development of a new model for improved out- staging, and response to treatment. J Clin Oncol 1993;11: clinical outcome using gene expression profiling and come prediction in neuroblastoma. J Clin Oncol 2005;23: 1466–77. artificial neural networks for patients with neuroblas- 2280–99. 5. Shimada H, Ambros IM, Dehner LP, et al. The toma. Cancer Res 2004;64:6883–91. 30. Shohet JM, Hicks MJ, Plon SE, et al. Minichromosome International Neuroblastoma Pathology Classification 18. Ohira M, Oba S, Nakamura Y, Hirata T, Ishii S, maintenance protein MCM7 is a direct target of the (the Shimada system). Cancer 1999;86:364–72. Nakagawara A. A review of DNA microarray analysis of MYCN transcription factor in neuroblastoma. Cancer 6. Look AT, Hayes FA, Shuster JJ, Douglass EC, human neuroblastomas. Cancer Lett 2005;228:5–11. Res 2002;62:1123–8. Castleberry RP, Brodeur GM. Clinical relevance of tumor 19. McArdle L, McDermott M, Purcell R, et al. Oligonu- 31. Breen CJ, O’Meara A, McDermott M, Mullarkey M, cell ploidy and N-myc gene amplification in childhood cleotide microarray analysis of gene expression in Stallings RL. Coordinate deletion of chromosome 3p and neuroblastoma: a Pediatric Oncology Group study. neuroblastoma displaying loss of chromosome 11q. 11q in neuroblastoma detected by comparative genomic J Clin Oncol 1991;9:581–91. Carcinogenesis 2004;25:1599–609. hybridization. Cancer Genet Cytogenet 2000;120:44–9. 7. Brodeur GM, Seeger RC, Schwab M, Varmus HE, 20. Mathew P, Valentine MB, Bowman LC, et al. 32. Golub TR, Slonim DK, Tamayo P, et al. Molecular Bishop JM. Amplification of N-myc in untreated human Detection of MYCN gene amplification in neuroblasto- classification of cancer: class discovery and class neuroblastomas correlates with advanced disease stage. ma by fluorescence in situ hybridization: a pediatric prediction by gene expression monitoring. Science Science 1984;224:1121–4. oncology group study. Neoplasia 2001;3:105–9. 1999;286:531–7. 8. Seeger RC, Brodeur GM, Sather H, et al. Association of 21. Morowitz M, Shusterman S, Mosse Y, et al. Detection 33. Khan J, Wei JS, Ringner M, et al. Classification and multiple copies of the N-myc oncogene with rapid of single-copy chromosome 17q gain in human neuro- diagnostic prediction of cancers using gene expression progression of neuroblastomas. N Engl J Med 1985;313: blastomas using real-time quantitative polymerase profiling and artificial neural networks. Nat Med 2001;7: 1111–6. chain reaction. Mod Pathol 2003;16:1248–56. 673–9. 9. Brodeur GM. Neuroblastoma: biological insights into a 22. Alaminos M, Gerald WL, Cheung NK. Prognostic 34. Yang XJ, Tan MH, Kim HL, et al. A molecular clinical enigma. Nat Rev Cancer 2003;3:203–16. value of MYCN and ID2 overexpression in neuroblasto- classification of papillary renal cell carcinoma. Cancer 10. Schmidt ML, Lukens JN, Seeger RC, et al. Biologic ma. Pediatr Blood Cancer 2005;45:909–15. Res 2005;65:5628–37. factors determine prognosis in infants with stage IV 23. Ashburner M, Ball CA, Blake JA, et al. Gene Ontology: 35. Haven CJ, Howell VM, Eilers PH, et al. Gene neuroblastoma: a prospective Children’s Cancer Group tool for the unification of biology. The Gene Ontology expression of parathyroid tumors: molecular subclassi- study. J Clin Oncol 2000;18:1260–8. Consortium. Nat Genet 2000;25:25–9. fication and identification of the potential malignant 11. Cohn SL, Tweddle DA. MYCN amplification remains 24. Hosack DA, Dennis G, Jr., Sherman BT, Lane HC, phenotype. Cancer Res 2004;64:7405–11. prognostically strong 20 years after its ‘‘clinical debut.’’ Lempicki RA. Identifying biological themes within lists 36. Onken MD, Worley LA, Ehlers JP, Harbour JW. Gene Eur J Cancer 2004;40:2639–42. of genes with EASE. Genome Biol 2003;4:R70. expression profiling in uveal melanoma reveals two 12. Maris JM, Matthay KK. Molecular biology of neuro- 25. Manduchi E, Grant GR, McKenzie SE, Overton GC, molecular classes and predicts metastatic death. Cancer blastoma. J Clin Oncol 1999;17:2264–79. Surrey S, Stoeckert CJ, Jr. Generation of patterns from Res 2004;64:7205–9. www.aacrjournals.org 6061 Cancer Res 2006; 66: (12). June 15, 2006

37. Reiter JL, Brodeur GM. High-resolution mapping of a apoptotic mediators in neuroblastoma tumours with kinase inhibitors. Proc Natl Acad Sci U S A 2005;102: 130-kb core region of the MYCN amplicon in neuro- unfavourable biology. Eur J Cancer 2005;41:635–46. 749–54. blastomas. Genomics 1996;32:97–103. 52. Barbashina V, Salazar P, Holland EC, Rosenblum MK, 66. Wang SS, Esplin ED, Li JL, et al. Alterations of the 38. Vasudevan SA, Nuchtern JG. Gene profiling of high Ladanyi M. Allelic losses at 1p36 and 19q13 in gliomas: PPP2R1B gene in human lung and colon cancer. Science risk neuroblastoma. World J Surg 2005;29:317–24. correlation with histologic classification, definition of a 1998;282:284–7. 39. Bown N, Cotterill S, Lastowska M, et al. Gain of 150-kb minimal deleted region on 1p36, and evaluation 67. Bartkova J, Horejsi Z, Koed K, et al. DNA damage chromosome arm 17q and adverse outcome in of CAMTA1 as a candidate tumor suppressor gene. Clin response as a candidate anti-cancer barrier in early patients with neuroblastoma. N Engl J Med 1999;340: Cancer Res 2005;11:1119–28. human tumorigenesis. Nature 2005;434:864–70. 1954–61. 53. White PS, Thompson PM, Gotoh T, et al. Definition 68. Blaheta RA, Hundemer M, Mayer G, et al. Expression 40. Chang CL, Zhu XX, Thoraval DH, et al. Nm23-1 and characterization of a region of 1p36.3 consistently level of neural cell adhesion molecule (NCAM) inversely mutation in neuroblastoma. Nature 1994;370:335–6. deleted in neuroblastoma. Oncogene 2005;24:2684–94. correlates with the ability of neuroblastoma cells to 41. Valentijn LJ, Koppen A, van Asperen R, Root HA, 54. Lahti JM, Valentine M, Xiang J, et al. Alterations in adhere to endothelium in vitro. Cell Commun Adhes Haneveld F, Versteeg R. Inhibition of a new differenti- the PITSLRE protein kinase gene complex on chromo- 2002;9:131–47. ation pathway in neuroblastoma by copy number some 1p36 in childhood neuroblastoma. Nat Genet 1994; 69. Collado M, Gil J, Efeyan A, et al. Tumour biology: defects of N-myc, Cdc42, and nm23 genes. Cancer Res 7:370–5. senescence in premalignant tumours. Nature 2005; 2005;65:3136–45. 55. He L, Yu JX, Liu L, et al. RIZ1, but not the alternative 436:642. 42. Saito-Ohara F, Imoto I, Inoue J, et al. PPM1D is a RIZ2 product of the same gene, is underexpressed in 70. De Preter K, Vandesompele J, Hoebeeck J, et al. No potential target for 17q gain in neuroblastoma. Cancer breast cancer, and forced RIZ1 expression causes G2-M evidence for involvement of SDHD in neuroblastoma Res 2003;63:1876–83. cell cycle arrest and/or apoptosis. Cancer Res 1998;58: pathogenesis. BMC Cancer 2004;4:55. 43. Maris JM, Weiss MJ, Guo C, et al. Loss of 4238–44. 71. Teitz T, Wei T, Valentine MB, et al. Caspase 8 is heterozygosity at 1p36 independently predicts for 56. Thompson PM, Gotoh T, Kok M, White PS, Brodeur deleted or silenced preferentially in childhood neuro- disease progression, but not decreased overall survival GM. CHD5, a new member of the chromodomain gene blastomas with amplification of MYCN. Nat Med 2000;6: probability, in neuroblastoma patients: a Children’s family, is preferentially expressed in the nervous system. 529–35. Cancer Group study. J Clin Oncol 2000;18:1888–99. Oncogene 2003;22:1002–11. 72. Yang Q, Zage P, Kagan D, et al. Association of 44. Matthay KK, Villablanca JG, Seeger RC, et al. 57. BilkeS,ChenQR,WestermanF,SchwabM, epigenetic inactivation of RASSF1A with poor outcome Treatment of high-risk neuroblastoma with intensive Catchpoole D, Khan J. Inferring a tumor progression in human neuroblastoma. Clin Cancer Res 2004;10: chemotherapy, radiotherapy, autologous bone marrow model for neuroblastoma from genomic data. J Clin 8493–500. transplantation, and 13-cis-retinoic acid. Children’s Oncol 2005;23:7322–31. 73. Yang QW, Liu S, Tian Y, et al. Methylation-associated Cancer Group. N Engl J Med 1999;341:1165–73. 58. Selzer RR, Richmond TA, Pofahl NJ, et al. Analysis of silencing of the thrombospondin-1 gene in human 45. Cooper R, Khakoo Y, Matthay KK, et al. Opsoclonus- chromosome breakpoints in neuroblastoma at sub- neuroblastoma. Cancer Res 2003;63:6299–310. myoclonus-ataxia syndrome in neuroblastoma: histo- kilobase resolution using fine-tiling oligonucleotide array 74. Misawa A, Inoue J, Sugino Y, et al. Methylation- pathologic features—a report from the Children’s CGH. Genes Chromosomes Cancer 2005;44:305–19. associated silencing of the nuclear receptor 1I2 gene Cancer Group. Med Pediatr Oncol 2001;36:623–9. 59. Mosse Y, Greshock J, King A, Khazi D, Weber BL, in advanced-type neuroblastomas, identified by 46. Gambini C, Conte M, Bernini G, et al. Neuroblastic Maris JM. Identification and high-resolution mapping of bacterial artificial chromosome array-based methyl- tumors associated with opsoclonus-myoclonus syn- a constitutional 11q deletion in an infant with ated CpG island amplification. Cancer Res 2005;65: drome: histological, immunohistochemical and molec- multifocal neuroblastoma. Lancet Oncol 2003;4:769–71. 10233–42. ular features of 15 Italian cases. Virchows Arch 2003;442: 60. Satge D, Moore SW, Stiller CA, et al. Abnormal 75. Banelli B, Gelvi I, Di Vinci A, et al. Distinct CpG 555–62. constitutional karyotypes in patients with neuroblasto- methylation profiles characterize different clinical 47. Shimada H. International neuroblastoma pathology ma: a report of four new cases and review of 47 others in groups of neuroblastic tumors. Oncogene 2005;24: classification for prognostic evaluation of patients with the literature. Cancer Genet Cytogenet 2003;147:89–98. 5619–28. peripheral neuroblastic tumors: a report from the 61. Guo C, White PS, Weiss MJ, et al. Allelic deletion at 76. Abe M, Ohira M, Kaneda A, et al. CpG island Children’s Cancer Group. Cancer 2001;92:2451–61. 11q23 is common in MYCN single copy neuroblastomas. methylator phenotype is a strong determinant of 48. Janoueix-Lerosey I, Novikov E, Monteiro M, et al. Oncogene 1999;18:4948–57. poor prognosis in neuroblastomas. Cancer Res 2005; Gene expression profiling of 1p35-36 genes in neuro- 62. De Preter K, Vandesompele J, Menten B, et al. 65:828–34. blastoma. Oncogene 2004;23:5912–22. Positional and functional mapping of a neuroblastoma 77. Mosse YP, Greshock J, Margolin A, et al. High- 49. Caren H, Ejeskar K, Fransson S, et al. A cluster of differentiation gene on chromosome 11. BMC Genomics resolution detection and mapping of genomic DNA genes located in 1p36 are down-regulated in neuroblas- 2005;6:97. alterations in neuroblastoma. Genes Chromosomes tomas with poor prognosis, but not due to CpG island 63. Kuramochi M, Fukuhara H, Nobukuni T, et al. TSLC1 Cancer 2005;43:390–403. methylation. Mol Cancer 2005;4:10. is a tumor-suppressor gene in human non-small-cell 78. Manning G, Whyte DB, Martinez R, Hunter T, 50. Teitz T, Wei T, Liu D, et al. Caspase-9 and Apaf-1 are lung cancer. Nat Genet 2001;27:427–30. Sudarsanam S. The protein kinase complement of the expressed and functionally active in human neuroblas- 64. Neumann HP, Bausch B, McWhinney SR, et al. Germ- human genome. Science 2002;298:1912–34. toma tumor cell lines with 1p36 LOH and amplified line mutations in nonsyndromic pheochromocytoma. N 79. Reiter JL, Brodeur GM. MYCN is the only highly MYCN. Oncogene 2002;21:1848–58. Engl J Med 2002;346:1459–66. expressed gene from the core amplified domain in 51. Abel F, Sjoberg RM, Nilsson S, Kogner P, Martinsson 65. Milne TA, Hughes CM, Lloyd R, et al. Menin and MLL human neuroblastomas. Genes Chromosomes Cancer T. Imbalance of the mitochondrial pro- and anti- cooperatively regulate expression of cyclin-dependent 1998;23:134–40.

Cancer Res 2006; 66: (12). June 15, 2006 6062 www.aacrjournals.org

Downloaded from cancerres.aacrjournals.org on September 27, 2021. © 2006 American Association for Cancer Research. Integrative Genomics Identifies Distinct Molecular Classes of Neuroblastoma and Shows That Multiple Genes Are Targeted by Regional Alterations in DNA Copy Number

Qun Wang, Sharon Diskin, Eric Rappaport, et al.

Cancer Res 2006;66:6050-6062.

Updated version Access the most recent version of this article at: http://cancerres.aacrjournals.org/content/66/12/6050

Supplementary Access the most recent supplemental material at: Material http://cancerres.aacrjournals.org/content/suppl/2006/06/20/66.12.6050.DC1

Cited articles This article cites 78 articles, 31 of which you can access for free at: http://cancerres.aacrjournals.org/content/66/12/6050.full#ref-list-1

Citing articles This article has been cited by 36 HighWire-hosted articles. Access the articles at: http://cancerres.aacrjournals.org/content/66/12/6050.full#related-urls

E-mail alerts Sign up to receive free email-alerts related to this article or journal.

Reprints and To order reprints of this article or to subscribe to the journal, contact the AACR Publications Subscriptions Department at [email protected].

Permissions To request permission to re-use all or part of this article, use this link http://cancerres.aacrjournals.org/content/66/12/6050. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.