ArrayCGH-based classification of neuroblastoma into genomic subgroups

Overall genomic pattern is a predictor of outcome in neuroblastoma

PURPOSE

For a comprehensive overview of the genetic alterations of neuroblastoma, their association and clinical significance, we conducted a whole-genome DNA copy number analysis.

PATIENTS AND METHODS

A series of 493 neuroblastoma (NB) samples was investigated by array-based comparative genomic hybridization in two consecutive steps (224, then 269 patients).

RESULTS

Genomic analysis identified several types of profiles. Tumors presenting exclusively whole-chromosome copy number variations were associated with excellent survival. No disease-related death was observed in this group. In contrast, tumors with any type of segmental chromosome alterations characterized patients with a high risk of relapse. Patients with both numerical and segmental abnormalities clearly shared the higher risk of relapse of segmental-only patients. In a multivariate analysis, taking into account the genomic profile, but also previously described individual genetic and clinical markers with prognostic significance, the presence of segmental alterations with (HR, 7.3; 95% CI, 3.7 to 14.5; P < .001) or without MYCN amplification (HR, 4.5; 95% CI, 2.4 to 8.4; P < .001) was the strongest predictor of relapse; the other significant variables were age older than 18 months (HR, 1.8; 95% CI, 1.2 to 2.8; P = .004) and stage 4 (HR, 1.8; 95% CI, 1.2 to 2.7; P = .005). Finally, within tumors showing segmental alterations, stage 4, age, MYCN amplification, 1p and 11q deletions, and 1q gain were independent predictors of decreased overall survival.

CONCLUSION

The analysis of the overall genomic pattern, which probably unravels particular genomic instability mechanisms rather than the analysis of individual markers, is essential to predict relapse in NB patients. It adds critical prognostic information to conventional markers and should be included in future treatment stratification.

Characterization of amplicons in neuroblastoma: high-resolution mapping using DNA microarrays, relationship with outcome, and identification of overexpressed genes

Somatically acquired chromosomal imbalances are a key feature of neuroblastoma, a heterogeneous pediatric solid tumor. Among these alterations, genomic amplification targeting the MYCN oncogene and observed in about 25-30% of the cases, strongly correlates with advanced stage and poor outcome. In this work, we have used BAC and SNP arrays as well as gene expression arrays to characterize amplifications in neuroblastoma. Eighty-eight distinct BACs defining high-level amplification events were identified in 65 samples, including 43 tumors and 22 cell lines. Although the highest recurrence was observed on chromosome 2, clones on chromosomes 8, 12, 16, and 17 also revealed genomic amplification in several samples. A detailed analysis of the 2p22-2p25 MYCN containing region indicated highly complex patterns in a number of cases. Coamplifications involving MYCN and other regions were explored by FISH in three cell lines. High-resolution arrays then allowed us to further refine the mapping of 25 amplicons in 19 samples, either reducing the size of a single continuous amplicon or increasing the complexity by highlighting multiple noncontiguous regions of amplification. Combined analysis of gene expression profiling and array-CGH data indicated that 12 to 25% of the genes that are targeted by genomic amplification are actually over-expressed in tumor cells, several of them having already been implicated in cancer. Finally, our results suggest that the presence of amplicons localized outside of chromosome 2, in addition to MYCN amplification, may be linked to a particularly severe outcome in neuroblastoma patients.

Aberrant methylation of candidate tumor suppressor genes in neuroblastoma

CpG island hypermethylation has been recognized as an alternative mechanism for tumor suppressor gene inactivation. In this study, we performed methylation-specific PCR (MSP) to investigate the methylation status of 10 selected tumor suppressor genes in neuroblastoma. Seven of the investigated genes (CD44, RASSF1A, CASP8, PTEN, ZMYND10, CDH1, PRDM2) showed high frequencies (> or =30%) of methylation in 33 neuroblastoma cell lines. In 42 primary neuroblastoma tumors, the frequencies of methylation were 69%, CD44; 71%, RASSF1A; 56%, CASP8; 25%, PTEN; 15%, ZMYND10; 8%, CDH1; and 0%, PRDM2. Furthermore, CASP8 and CDH1 hypermethylation was significantly associated with poor event-free survival. Meta-analysis of 115 neuroblastoma tumors demonstrated a significant correlation between CASP8 methylation and MYCN amplification. In addition, there was a correlation between ZMYND10 methylation and MYCN amplification. The MSP data, together with optimized mRNA re-expression experiments (in terms of concentration and time of treatment and use of proper reference genes) further strengthen the notion that epigenetic alterations could play a significant role in NB oncogenesis. This study thus warrants the need for a global profiling of gene promoter hypermethylation to identify genome-wide aberrantly methylated genes in order to further understand neuroblastoma pathogenesis and to identify prognostic methylation markers.

A functional screen identifies miR-34a as a candidate neuroblastoma tumor suppressor gene

MicroRNAs are small noncoding RNAs that have critical roles in regulating a number of cellular functions through transcriptional silencing. They have been implicated as oncogenes and tumor suppressor genes (oncomirs) in several human neoplasms. We used an integrated genomics and functional screening strategy to identify potential oncomirs in the pediatric neoplasm neuroblastoma. We first identified microRNAs that map within chromosomal regions that we and others have defined as frequently deleted (1p36, 3p22, and 11q23-24) or gained (17q23) in high-risk neuroblastoma. We then transiently transfected microRNA precursor mimics or inhibitors into a panel of six neuroblastoma cell lines that we characterized for these genomic aberrations. The majority of transfections showed no phenotypic effect, but the miR-34a (1p36) and miR-34c (11q23) mimics showed dramatic growth inhibition in cell lines with 1p36 hemizygous deletion. In contrast, there was no growth inhibition by these mimics in cell lines without 1p36 deletions. Quantitative reverse transcription-PCR showed a perfect correlation of absent miR-34a expression in cell lines with a 1p36 aberration and phenotypic effect after mimetic add-back. Expression of miR-34a was also decreased in primary tumors (n = 54) with 1p36 deletion (P = 0.009), but no mutations were discovered in resequencing of the miR-34a locus in 30 neuroblastoma cell lines. Flow cytometric time series analyses showed that the likely mechanism of miR-34a growth inhibition is through cell cycle arrest followed by apoptosis. BCL2 and MYCN were identified as miR-34a targets and likely mediators of the tumor suppressor phenotypic effect. These data support miR-34a as a tumor suppressor gene in human neuroblastoma.

CADM1 is a strong neuroblastoma candidate gene that maps within a 3.72 Mb critical region of loss on 11q23

Background

Recurrent loss of part of the long arm of chromosome 11 is a well established hallmark of a subtype of aggressive neuroblastomas. Despite intensive mapping efforts to localize the culprit 11q tumour suppressor gene, this search has been unsuccessful thus far as no sufficiently small critical region could be delineated for selection of candidate genes.

Methods

To refine the critical region of 11q loss, the chromosome 11 status of 100 primary neuroblastoma tumours and 29 cell lines was analyzed using a BAC array containing a chromosome 11 tiling path. For the genes mapping within our refined region of loss, meta-analysis on published neuroblastoma mRNA gene expression datasets was performed for candidate gene selection. The DNA methylation status of the resulting candidate gene was determined using re-expression experiments by treatment of neuroblastoma cells with the demethylating agent 5-aza-2'-deoxycytidine and bisulphite sequencing.

Results

Two small critical regions of loss within 11q23 at chromosomal band 11q23.1-q23.2 (1.79 Mb) and 11q23.2-q23.3 (3.72 Mb) were identified. In a first step towards further selection of candidate neuroblastoma tumour suppressor genes, we performed a meta-analysis on published expression profiles of 692 neuroblastoma tumours. Integration of the resulting candidate gene list with expression data of neuroblastoma progenitor cells pinpointed CADM1 as a compelling candidate gene. Meta-analysis indicated that CADM1 expression has prognostic significance and differential expression for the gene was noted in unfavourable neuroblastoma versus normal neuroblasts. Methylation analysis provided no evidence for a two-hit mechanism in 11q deleted cell lines.

Conclusion

Our study puts CADM1 forward as a strong candidate neuroblastoma suppressor gene. Further functional studies are warranted to elucidate the role of CADM1 in neuroblastoma development and to investigate the possibility of CADM1 haploinsufficiency in neuroblastoma.

Identification of 2 putative critical segments of 17q gain in neuroblastoma through integrative genomics

Partial gain of chromosome arm 17q is the most frequent genetic change in neuroblastoma (NB) and constitutes the strongest independent genetic factor for adverse prognosis. It is assumed that 1 or more genes on 17q contribute to NB pathogenesis by a gene dosage effect. In the present study, we applied chromosome 17 tiling path BAC arrays on a panel of 69 primary tumors and 28 NB cell lines in order to reduce the current smallest region of gain and facilitate identification of candidate dosage sensitive genes. In all tumors and cell lines with 17q gain, large distal segments were consistently present in extra copies and no interstitial gains were observed. In addition to these large regions of distal gain with breakpoints proximal to coordinate 44.3 Mb (17q21.32), smaller regions of gain (distal to coordinate 60 Mb at 17q24.1) were found superimposed on the larger region in a minority of cases. Positional gene enrichment analysis for 17q genes overexpressed in NB showed that dosage sensitive NB oncogenes are most likely located in the gained region immediately distal to the most distal breakpoint of the 2 breakpoint regions. Interestingly, comparison of gene expression profiles between primary tumors and normal fetal adrenal neuroblasts revealed 2 gene clusters on chromosome 17q that are overexpressed in NB, i.e. a region on 17q21.32 immediately distal to the most distal breakpoint (in cases with single regions of gain) and 17q24.1, a region coinciding with breakpoints leading to superimposed gain.