|
Advertisement |
|
|
||
 |
|
|||||
|
|||||
|
 |
||||||
|
||||||
|
||||||
|
Journal of Clinical Oncology, Vol 23, No 10 (April 1), 2005: pp. 2280-2299 © 2005 American Society of Clinical Oncology. DOI: 10.1200/JCO.2005.06.104 Unequivocal Delineation of Clinicogenetic Subgroups and Development of a New Model for Improved Outcome Prediction in Neuroblastoma astowskaFrom the Center for Medical Genetics, Department of Pathology, N. Goormaghtigh Institute for Pathology, Department of Pediatric Hematology and Oncology, Ghent University Hospital, Ghent; Department of Pediatric Hematology and Oncology, Cliniques St Luc, Université Catholique de Louvain, Brussels, Belgium; Department of Pathology, Stanford University Medical Center, Stanford, CA; Medizinische Klinik und Poliklinik V der Universität Heidelberg, Heidelberg; Gemeinschaftspraxis Pathologie, Am Fuchsengraben 3, Starnberg; Universitätskinderklinik, Marburg, Germany; Children's Cancer Research Institute, St Anna Kinderspital, Vienna, Austria; Institute of Human Genetics, University of Newcastle upon Tyne, Newcastle upon Tyne; Addenbrookes Hospital, Cambridge, United Kingdom; Centre d'Oncologie Génétique, Centre Léon Bérard, Lyon; Department of Pediatrics, University Hospital Center, Grenoble, France; Department of Oncology, Our Lady's Hospital for Sick Children, and National Centre for Medical Genetics, Our Lady's Hospital for Sick Children, Dublin, Ireland Address reprint requests to Jo Vandesompele, Center for Medical Genetics Ghent (CMGG), Ghent University Hospital, MRB, De Pintelaan 185, B-9000 Ghent, Belgium; e-mail: Joke.Vandesompele{at}UGent.be
PURPOSE: Neuroblastoma is a genetically heterogeneous pediatric tumor with a remarkably variable clinical behavior ranging from widely disseminated disease to spontaneous regression. In this study, we aimed for comprehensive genetic subgroup discovery and assessment of independent prognostic markers based on genome-wide aberrations detected by comparative genomic hybridization (CGH). MATERIALS AND METHODS: Published CGH data from 231 primary untreated neuroblastomas were converted to a digitized format suitable for global data mining, subgroup discovery, and multivariate survival analyses. RESULTS: In contrast to previous reports, which included only a few genetic parameters, we present here for the first time a strategy that allows unbiased evaluation of all genetic imbalances detected by CGH. The presented approach firmly established the existence of three different clinicogenetic subgroups and indicated that chromosome 17 status and tumor stage were the only independent significant predictors for patient outcome. Important new findings were: (1) a normal chromosome 17 status as a delineator of a subgroup of presumed favorable-stage tumors with highly increased risk; (2) the recognition of a survivor signature conferring 100% 5-year survival for stage 1, 2, and 4S tumors presenting with whole chromosome 17 gain; and (3) the identification of 3p deletion as a hallmark of older age at diagnosis. CONCLUSION: We propose a new regression model for improved patient outcome prediction, incorporating tumor stage, chromosome 17, and amplification/deletion status. These findings may prove highly valuable with respect to more reliable risk assessment, evaluation of clinical results, and optimization of current treatment protocols.
Neuroblastoma is the most frequent extracranial solid tumor in children and originates from neural crest–derived sympathetic nerve cells. In contrast to many other pediatric malignancies, progress in treatment (especially for advanced-stage tumors) has been relatively modest. Hence, at present, this tumor still poses a major challenge to the clinical oncologist. Although tumor stage and age at diagnosis are important predictors for patient outcome and are critical parameters for therapy stratification, a significant number of patients with apparently low-risk neuroblastoma at diagnosis relapse and require more aggressive treatment. Similarly, more than half of the patients with adverse prognostic features at present cannot be cured, despite intensive multimodal therapy. In view of the unpredictable tumor behavior, many studies have aimed at the identification of additional factors for prognostic stratification. Deletion of the short arm of chromosome 1, amplification of the MYCN proto-oncogene, and diploidy or tetraploidy were recognized quite early as genetic markers for aggressive neuroblastomas.1–3 However, the overall picture of the different genomic changes in the various tumor stages remained incomplete for a long time. The importance of loss of chromosome regions other than 1p (ie, 2q, 3p, 4p, 9p, 11q, 14q, and 18q) was demonstrated by loss-of-heterozygosity studies.4–6 Losses of 3p, 9p, and 11q may also have prognostic value.7–10 A major landmark was the recognition of the high incidence of unbalanced 17q rearrangements by molecular cytogenetics, and the subsequent demonstration that the resulting partial 17q gain was the most powerful genetic predictor for poor outcome.11,12 Many of the aforementioned studies focused on one or a few particular chromosomal regions. Hence, a comprehensive overview of the occurrence of the previously mentioned imbalances and their interrelationship could not be obtained. As successful karyotyping was not possible for a high proportion of neuroblastomas, it was not until the introduction of comparative genomic hybridization (CGH) that screening for DNA copy number gains and losses across the entire genome became feasible.13,14 This approach indeed proved to be particularly useful to address this issue, and consequently led to more general insights into the genetic heterogeneity of this enigmatic childhood tumor.15–21 Among others, evidence was obtained for a new genetic subgroup of MYCN single-copy neuroblastomas, characterized by loss of 11q, often in association with loss of 3p.15,20,21 Despite these significant new findings, published CGH data formats for representation of gains and losses (either ideogram or International System for Human Cytogenetic Nomenclature [ISCN] description) do not allow in-depth analysis of large series of tumors. Here we illustrate the use of an available online tool for conversion of conventional CGH data into a digitized format (a band-specific aberration matrix) amenable for a more comprehensive genetic subgroup discovery. Global data mining and multivariate survival analyses on 231 published CGH results from neuroblastoma tumors firmly established the existence of three major clinicogenetic subgroups in neuroblastoma and provided to be a comprehensive assessment of prognostic parameters with independent power, leading to identification of a survivor signature for neuroblastoma and a new mathematical model for improved outcome prediction.
Tumor Samples Previously published CGH data on primary neuroblastoma tumors were included in this study based on the availability of both individual band-specific CGH results for each tumor, as well as clinical patient information (stage, age at diagnosis, progression, and survival status). Samples without genetic aberration were excluded. This yielded a compilation of 231 unique cases from eight publications,15–22 of which 195 were also part of another multicenter study on 313 cases.11 In this previous study, the consistency of the CGH data among five European centers (from which the majority of the cases in this study originate) was assessed, and no significant differences between laboratories were detected. Each referenced study has properly validated their CGH methodology and results. Where possible, clinical information was updated with respect to the original publication. All CGH aberration data can be accessed through the public Progenetix repository,23 under ICD-O-3 (International Classification of Diseases for Oncology) code 9500/3. The sample identifiers from the Progenetix neuroblastoma case table are identical to those in the Supplemental Table (available online only in the full-text version; it is not included in the PDF [via Adobe® Acrobat Reader®] version), and are linked to the original publication (PubMed number) and CGH aberration pattern. Tumor stage was defined according to the International Neuroblastoma Staging System,24 and risk status, according to the International Neuroblastoma Pathology Classification.25 Event-free survival was defined as the time between initial diagnosis and relapse or death, or time between diagnosis and last follow-up if no event had occurred. The patients were treated according to the national protocols adapted to age and stage in use at that time in their country. Stratified survival analyses indicated that there were no significant differences in patient outcome in the various countries, considering either all patients together, or only the stage 4 tumors, or only infants (for these three groups, sufficient samples were available for reliable analysis; data not shown). The introduction of more homogeneous treatment regimens through the recently created International Society of Pediatric Oncology (SIOP) –Europe Neuroblastoma Group will allow confirmation of our results in a prospective study.
CGH Data Conversion The ISCN2matrix converter at Progenetix23 was used for both online conversion of the chromosome aberration list in ISCN format to a band-specific aberration matrix, as well as generation of a band-specific graphical representation of chromosomal gains and losses. During the conversion process, the CGH-based ISCN reverse in situ hybridization annotated karyotype (Supplemental Table) is resolved into 393 chromosome bands with categorical variables indicating the aberration status (–1: loss; 0: normal; +1: gain; +2: amplification), resulting in an average size of 7.9 megabases per annotation unit (band), well in the range of the spatial resolution achieved by chromosomal CGH.26
Exploratory Data Mining and Statistical Analysis
Definition of Genetic Aberrations in Statistical Analyses
Explorative and statistical analyses were applied to whole genome chromosome status aberration information obtained by CGH of 231 primary untreated neuroblastoma tumors (see Materials and Methods). The median age at time of diagnosis of these patients was 19.0 months. The median follow-up time for survivors was 42.8 months (interquartile range, 20.0 to 70.8). Of the 231 tumors, 88 were from infants younger than 1 year, and 143 were from children older than 1 year. According to the criteria of the International Neuroblastoma Staging System, 32 tumors were classified as stage 1, 35 as stage 2, 31 as stage 3, 100 as stage 4, and 33 as stage 4S.
Generation of a Band-Specific Aberration Matrix
Visualization of Independent Genetic Events by PCA To achieve this goal, we applied PCA to the 393-band aberration matrix for the 231 tumor samples. On projection of the 393 chromosome bands on the first two principal components, a clear picture of the chromosomal changes in neuroblastoma tumorigenesis emerges, with losses in the bottom left part, and gains in the upper right part of the projection (Fig 1). The PCA algorithm succeeds in grouping bands from the same chromosome or chromosome arm together, solely based on the aberration status of each band in the panel of tumors. It is also apparent from Figure 1 that some chromosomes are split into one or more band clusters, indicating that different regions in these chromosomes display different aberrations patterns in the tumor panel. In keeping with the PCA method, bands that cluster further away from the origin explain more variance and therefore represent significant alterations (eg, 2p23pter amplification, 17q gain, and loss of 1p and 11q). In contrast, bands closer to the center are not frequently altered in neuroblastoma.
Hierarchical Cluster Analysis From each band cluster in the PCA projection, one band was selected to represent this chromosomal region. If all bands from a single chromosome (except acrocentric chromosomes) were grouped into one cluster, two bands per chromosome were selected (one from each arm) to exclude the possibility that a smaller proportion of tumors could not be discriminated based on a possible differential aberration status for the short and long arms. Accordingly, the 393-band aberration matrix was reduced to a 45-band matrix, representing independent variables. Hierarchical clustering using different similarity or distance measures and clustering algorithms yielded a series of equivalent aberration patterns with three recurring distinct tumor clusters. A representative example, including visualization of the individual clusters on separate ideogram charts, is shown in Figures 2A and B.
Cluster 1 contains 94 tumors with predominantly numerical aberrations (typical gains of chromosomes 6, 7, and 17, and losses of chromosomes 3, 4, 11, and 14). The other clusters are both characterized by structural chromosome aberrations, such as partial 17q gain, 11q deletion, and to a lesser extent, 3p deletion in the 45 cases of cluster 2; and partial 17q gain, MYCN amplification, 1p deletion, and to a lesser extent, 1q gain in the 74 cases from cluster 3. More than 92% of all tumors fall into these three genetic subgroups. The remaining tumors are mainly characterized by absence of recurrent genetic aberrations.
Genetic Subgroups in Relation to Clinical Variables
While patients from cluster 1 are clearly diagnosed at a younger age compared with patients from cluster 2 or 3 (P < .0001), there is also a significant difference between clusters 2 and 3 (P = .0078; Table 2). The cases with 3p deletion seem to be partially responsible for the older age at diagnosis in cluster 2, as within this cluster, tumors with loss of 3p have a older age at diagnosis compared with 3p-intact cases (median age, 60.5 v 39.0 months; P = .027).
In keeping with the previously mentioned nonrandom distribution of several parameters within the three genetic subgroups, the Kaplan-Meier overall survival estimates are significantly different between clusters 1 and 2 or clusters 1 and 3 (P < .0001), while no difference was found between clusters 2 and 3 (P = .62; Fig 2C).
Univariate Survival Analysis
Besides confirmation of known significant parameters in univariate survival analysis, this study demonstrated for the first time that a normal chromosome 17 status is associated with a poor prognosis. Losses of 3p, as well as losses of other regions (4p, 9p, 14q, 16p, or 18q) have no predictive power (P > .1; data not shown).
Multivariate Survival Analysis When the Cox proportional hazards procedure was applied to model event-free survival probability, only a normal chromosome 17 status (hazard ratio, 4.80; 95% CI, 2.39 to 9.61; P < .0001) and partial chromosome 17q gain (hazard ratio, 5.66; 95% CI, 3.26 to 9.82; P < .0001) remained independent significant predictors for adverse prognosis. No Cox regression model could be tested taking both INPC risk status and ploidy into account, as only 18 cases for which both these parameters were determined were available. When tested separately (in conjunction with the other clinical and genetic parameters), ploidy had no independent predictive value, while the INPC risk status showed only marginal significance (P = .12 for retention and P = .046 for removal of this parameter in the model). Please note that these conclusions should be handled with care, as ploidy and INPC information was available for only 26% and 33% of the cases, respectively.
Stratified Survival Analyses for Chromosome 17 Status or Tumor Stage
Within the favorable-stage tumors (1, 2, or 4S), the following parameters were associated with lower overall survival probability (P values < 0.0001): MYCN amplification, 1p deletion, normal chromosome 17 status, and partial 17q gain. Again, it is important to note that both a normal chromosome 17 status as well as a partial 17q gain identify low-stage patients with increased risk. Patients with a low-stage tumor displaying whole chromosome 17 gain have a 100% survival probability (Fig 4C). For stage 3 or 4 tumors, no further significant parameters could be discriminated. Within the group of patients whose tumor showed whole chromosome 17 gain, the following parameters were associated with lower overall survival: 11q loss (P = .0019), 1q gain (P = .046), age older than 1 year at diagnosis (P < .0001), and tumor stage 3 (P < .0001) and stage 4 (P < .0001). In line with the stage-stratified survival estimates, patients with whole chromosome 17 gain diagnosed prior to age 1 year (62 of 97 patients) or with tumors belonging to stages 1, 2, or 4S (76 of 97 patients) displayed a 100% overall survival (Figs 4A and B). Within the group of patients whose tumors showed a normal chromosome 17 status (in a background of other genetic defects), borderline significant association with adverse outcome could be demonstrated for 1q gain or 1p deletion (P = .039 and P = .032, respectively). Within the subgroup of high-risk tumors with partial 17q gain, only stage 4 had additional predictive power (P = .022; Fig 4D). As the MYCN status is currently the only genetic parameter that is incorporated in therapy stratification, we tested whether the chromosome 17 status allowed further stratification in the group of MYCN-amplified and single-copy tumors. A normal chromosome 17 status or a partial 17q gain are both significantly associated with decreased survival probability (P < .001) of patients with a MYCN nonamplified neuroblastoma. In contrast, chromosome 17 status allows no further stratification in MYCN-amplified tumors because MYCN amplification is almost invariably associated with a normal chromosome 17 or partial chromosome 17q gain status (there is only one tumor with MYCN-amplification in combination with whole chromosome 17 gain).
Development of a New Regression Model to Predict Outcome in Neuroblastoma The findings given in this section indicate that a whole chromosome 17 gain status confers a good prognosis and is the most important genetic parameter for a high survival probability. While whole chromosome 17 gain is strongly associated with a numerical aberration pattern, a normal chromosome 17 and partial 17q gain status are almost always found in a background of other structural chromosome aberrations. To test whether the structural aberration pattern itself (partial chromosome deletions or amplifications) is indicative of poor survival, we determined whether patients whose tumor showed any structural chromosome aberration had poorer outcomes compared with patients whose tumor presented with numerical aberrations only. As expected, a structural aberration pattern was a significant predictor in univariate survival analyses (P < .0001). Multivariate outcome prediction, however, could not attribute independent prognostic power to a structural aberration pattern. Further analyses indicated that a surprisingly high fraction of tumors with whole chromosome 17 gain (45 of 97 tumors) also displayed one or more structural aberrations (while it was expected that virtually all tumors [131 of 134] with a normal chromosome 17 status or partial 17q gain would show a structural aberration pattern). Closer inspection, however, revealed that only a minority of these cases (11 of 97) presented with one of the three genetic defects typically found in higher-stage aggressive tumors (found in clusters 2 and 3; ie, MYCN amplification, 1p deletion, or 11q deletion), suggesting that many of the structural defects observed in conjunction with whole chromosome 17 gain are nonfrequently occurring genetic changes. This prompted us to determine whether the presence of either MYCN amplification or loss of 1p or 11q (hereafter referred to as a positive amplification/deletion status; see Materials and Methods) could serve as a new powerful and independent prognostic parameter. Indeed, Cox multivariate survival analysis indicated that the amplification/deletion status (hazard ratio, 2.04; 95% CI, 1.06 to 3.91; P = .032), the improved chromosome 17 status dichotomization (hazard ratio, 2.77; 95% CI, 1.14 to 6.72; P = .025), and tumor stage (3 and 4 v 1, 2, and 4S; hazard ratio, 3.49; 95% CI, 1.60 to 7.63; P = .0017) were independent predictors for adverse outcome. Table 5 gives an overview of the hazard ratios (relative to patients with a stage 1, 2, or 4S tumor showing whole chromosome 17 gain and a negative amplification/deletion status) for patients stratified according to previously mentioned independent outcome predictors. Subsequent confirmatory analyses indicated that the improved chromosome 17 status dichotomization indeed allowed further stratification in cases with a negative (P < .0001) or positive (P = .052) amplification/deletion status. Likewise, the amplification/deletion status parameter also allowed further stratification within the cases with whole chromosome 17 gain (P = .024), or either a normal chromosome 17 status or partial 17q gain (P = .016; Fig 5).
Deletion of the distal part of chromosome arm 1p and amplification of the MYCN oncogene were the first genetic markers of advanced-stage neuroblastoma recognized in the initial cytogenetic investigations of this tumor. Subsequently, an array of other recurrent structural chromosome aberrations was reported, and some were found to be associated with patient outcome. However, it was not until the advent of the whole genome scanning method CGH that a global view on the genetic heterogeneity and the interrelationship of the imbalances could be obtained. These studies not only provided insights in the genetic basis underlying the clinical variability associated with this childhood tumor, but most importantly, provided clues about the existence of two different genetic subgroups of advanced-stage unfavorable neuroblastomas, characterized by MYCN amplification and loss of 11q, respectively. Despite the wealth of information CGH brought about with respect to identification of chromosomal regions of consistent loss or gain, many publications suffer from a lack of uniform data presentation and detailed aberration information for the individual cases. As such, it has proven impracticable or impossible to reanalyze data or pool case results from different (often of relatively small sample size) studies to strengthen conclusions or gain further insight into the different aberration patterns or genetic subgroups that might exist. To overcome this problem, a public repository for published CGH data in a format suitable for data mining procedures has been initiated.23 Equipped with the ability to convert analogous CGH profiles to a digitized data mining–compatible format, we sought to apply common bioinformatics algorithms to CGH-based chromosome aberration information, and performed for the first time a comprehensive and in-depth analysis of genetic aberrations occurring in the various neuroblastoma subtypes. In the approach presented here, principal component analysis was an important step toward unbiased delineation of independent genetic events before explorative and statistical analyses. Subsequent hierarchical clustering firmly established the existence of three distinct genetic subgroups. Interestingly, these three subgroups not only differ with respect to their genomic aberration patterns, but are also characterized through significantly different age at time of diagnosis, tumor stage distribution, ploidy level, INPC risk status, and survival probability. Cluster 1 represents predominantly near-triploid low-stage tumors with numerical aberrations and favorable histology from infants (median age at diagnosis, 9 months) with excellent outcome. Cluster 2 contains predominantly near-diploid/tetraploid stage 4 tumors with unfavorable histology, partial 17q gain, and 11q loss, often in association with loss of 3p. These patients present with the oldest age at diagnosis (median age, 41 months) and have a poor overall 5-year survival of approximately 40%, which is similar to that of patients from cluster 3 (median age, 26 months), who present with near-diploid/tetraploid high stage tumors displaying unfavorable histology and partial 17q gain, loss of 1p, and MYCN amplification. The difference in age at diagnosis between high-stage clusters 2 and 3 may be explained by a difference in growth rate and/or the need for an additional genetic defect in a cluster 2 tumor to become clinically manifest (as suggested by a large proportion of 3p loss in cluster 2, associated with an older age at diagnosis). This indicates that clusters 2 and 3 are clinically distinct and might behave differently, which suggests discrimination between those tumor subtypes in future therapeutic strategies. While acknowledging the limitations of explorative cluster analysis (in that the answer is not absolute, and that different distance measures and cluster algorithms may result in differences between the generated patterns or clusters), we believe that the recurring patterns we observed, in addition to the significant nonrandom association of relevant neuroblastoma parameters to the observed clusters, render confidence to the presented clusters, delineating three major different clinicogenetic subgroups in neuroblastoma. Although the existence of clinicogenetic subgroups has been proposed in previous studies, these conclusions were not based on in-depth analysis owing to various limitations such as sample size, tumor stage distribution, or noninclusion of all relevant genetic aberrations.5,10,20,29 In contrast, this study provides an unbiased and comprehensive assessment of the subgroups based on a large series of tumor samples belonging to all stages, and based on all relevant genetic, histopathologic, and clinical information. Furthermore, our finding of 3p loss as a hallmark of older age at diagnosis provides additional insight into the series of genetic events leading to a subtype of aggressive neuroblastoma. To determine which of the individual parameters independently contribute to survival probability, stratified univariate and multivariate analyses were performed. Both analyses demonstrated that chromosome 17 status and tumor stage are the only significant independent predictors for patient outcome, corroborating previous studies11,29–31 and emphasizing the need to include at least tumor stage and chromosome 17 status in future multivariate evaluations of prognostic parameters. As many of the cases in this article were also compiled in an even larger multicenter study (reporting only on chromosome 1 and 17, MYCN status, and ploidy), our conclusion about the prognostic significance of chromosome 17 is not surprising. However, more importantly, we obtained new information regarding patients with a normal chromosome 17 status. Please bear in mind that these tumors always displayed other cytogenetic aberrations (Supplemental Table), as we excluded tumors with no detectable changes (see Materials and Methods). We were able to demonstrate that a normal chromosome 17 status (in conjunction with other defects) is an independent marker, and that patients with these tumors had an equally adverse outcome compared with those with partial chromosome 17q gain, a finding that is particularly true for favorable-stage tumors 1, 2, and 4S. In other words, a normal chromosome 17 status or partial chromosome 17q gain identify a new subgroup of increased-risk patients with presumed favorable-stage tumors. This finding may have important clinical consequences with respect to proper treatment and assessment of new therapeutic protocols. It is important to note that we could not confirm the recently reported adverse prognostic power of 3p or 11q loss in localized and stage 4S tumors.32 Taken together, these analyses indicate that the presence of an extra chromosome 17 (often in a background of other numerical aberrations) is the most important independent genetic indicator for excellent patient outcome, and suggest that these tumors represent a separate biologic entity of favorable neuroblastomas. A second important new observation, in line with the previous statement, was the identification of a specific survivor signature for children with neuroblastoma. Within the subgroup of tumors with whole chromosome 17 gain, tumor stage and age at diagnosis seem to have additional predictive power: the group of patients with either a favorable-stage tumor (stages 1, 2, and 4S) showing whole chromosome 17 gain, or a tumor with whole chromosome 17 gain diagnosed before age 1 year, show a 100% overall survival (Fig 4). Considering that whole chromosome 17 gain is strongly associated with near-triploidy, the above findings are in keeping with the report that ploidy is a strong prognostic factor for children younger than 1 year at diagnosis.3 The fact that the MYCN status as a standard marker for neuroblastoma therapy stratification drops out of the multivariate analysis is not unexpected, nor contrary to published data. As in many other reports based on univariate analysis, MYCN amplification is also a strong predictor for adverse outcome in this study. However, when other genetic, clinical, or histopathologic parameters are taken into account (multivariate analysis), MYCN status no longer has independent predictive power, corroborating published reports.11,29 This can be explained by the strong association of MYCN-amplification with other prognostic factors (eg, stage 4 and gain of 17q) that occur more frequently, and hence, can predict outcome for a larger proportion of patients. The fact that Cox proportional hazards modeling attributes independent prognostic value to those parameters that can predict an adverse outcome for as many patients as possible also explains why other genetic parameters that have univariate prognostic value (such as loss of 1p and 11q) drop out of the multivariate model when chromosome 17 status is included. This prompted us to investigate whether the combination of MYCN status and 1p or 11 deletion could serve as a new powerful independent predictor (besides chromosome 17 status) in multivariate survival analysis. Indeed, we could develop a new regression model that incorporates tumor stage (3 and 4 combined v 1, 2, and 4S); chromosome 17 status (whole chromosome gain v combined partial gain or normal chromosome status); and the combined status of MYCN, 1p, and 11q (amplification/deletion status). This new three-parametric model can predict patient outcome more accurately than any other tested model incorporating one or more of the available parameters (as evidenced by the relative risk ratio and overall score of the model). It will prove interesting to validate this model in other large cohorts of uniformly treated patients for whom all clinical, genetic, and histopathologic parameters are available. Again, the model demonstrates that in addition to the current incorporation of the MYCN status in therapy stratification, chromosome 17 status and preferably also chromosome 1 and 11 status should be routinely assessed and taken into consideration in the evaluation of clinical trials and possible future therapeutic stratifications. While many studies have confirmed the initial report on the clinical significance of unbalanced 17q gain in neuroblastoma,11 a recent fluorescence in-situ hybridization (FISH) –based study failed to attribute prognostic value to 17q gain.33 Our observation that a normal chromosome 17 status is also a strong predictor for adverse outcome, and that in the referenced FISH study, conventional dichotomization is used (hence grouping together favorable tumors with whole chromosome 17 gain and unfavorable tumors with a normal chromosome 17 status) might (partly) explain why in this specific study, 17q gain was not recognized as a prognostic marker, especially if a relatively high proportion of cases with a true normal chromosome 17 status was included. The discrepant results might also be explained by the particular method that is used to score chromosome 17 gain. Most studies attributing prognostic value to the chromosome 17 aberration status relied entirely or predominantly on CGH. The study that failed to confirm the chromosome 17 findings is based exclusively on FISH interpretation of 17q copy number gain. Preliminary data in our laboratory, based on FISH analyses, indicate large intratumoral heterogeneity of chromosome 17 copy numbers and aberration status (whole chromosome gain and partial gain present in the same tumor), complicating unequivocal interpretation. Multicolor interphase FISH and (microarray-based) CGH side-by-side comparisons are ongoing to assess concordance between these detection methods (data will be published elsewhere). The present study is also relevant in light of current efforts of tumor subgroup classification based on transcriptome profiling using oligonucleotide or cDNA microarrays. These methods are relatively expensive and require significant amounts of high-quality RNA. Assessment of genetic defects at the DNA level with strong diagnostic and prognostic relevance may provide a valuable alternative, as demonstrated by other studies.34–37 In this context, the recent report indicating the superiority of genomic aberration patterns with respect to tumor subtype identification compared with expression profiles, is of particular interest.38 Another important feature of our proposed scheme for data handling is its applicability to promising new whole genome profiling strategies such as Serial Analysis of Gene Expression–based digital karyotyping39 and microarray-based CGH in which metaphases have been replaced by arrayed genomic BAC clones,40,41 cDNA transcripts,42 or oligonucleotides.43,44 These new technologies not only offer a significant increase in resolution, but are also adaptable to automation, leading to higher throughput. Hence, going from a cytogenetic to a molecular analysis level, it is expected that new aberrations will be identified, hopefully speeding up the discovery of molecular targets for the development of enhanced therapeutic approaches.
The authors indicated no potential conflicts of interest.
We gratefully acknowledge Jan Kohler and all other CGH collaborators for providing individual CGH profiles and updated clinical patient information. We also highly value the comments we received from the anonymous reviewers.
Supported by BOF-grant 011F1200 and 011B4300, GOA-grant 12051203, FWO-grant G.0028.00 and VEO-grant 011V1302. Katleen De Preter is an aspirant with the Fund for Scientific Research Flanders (FWO-Vlaanderen). Nadine Van Roy is a post-doctoral researcher with the FWO. Jo Vandesompele is supported by a post-doctoral grant from the Institute for the Promotion of Innovation by Science and Technology in Flanders (IWT). Presented in part at the Advances in Neuroblastoma Research Meeting, Paris, France, June 17-19, 2002, and at the 8th European Workshop on Cytogenetics and Molecular Genetics of Human Solid Tumours, Barcelona, Spain, September 12-15, 2002. This text presents research results of the Belgian program of Interuniversity Poles of Attraction initiated by the Belgian State, Prime Minister's Office, Science Policy Programming. Authors' disclosures of potential conflicts of interest are found at the end of this article.
1. Seeger RC, Brodeur GM, Sather H, et al: Association of multiple copies of the N-myc oncogene with rapid progression of neuroblastomas. N Engl J Med 313:1111-1116, 1985[Abstract]
2. Fong CT, Dracopoli NC, White PS, et al: Loss of heterozygosity for the short arm of chromosome 1 in human neuroblastomas: Correlation with N-myc amplification. Proc Natl Acad Sci U S A 86:3753-3757, 1989 3. Look AT, Hayes FA, Shuster JJ, et al: Clinical relevance of tumor cell ploidy and N-myc gene amplification in childhood neuroblastoma: A Pediatric Oncology Group study. J Clin Oncol 9:581-591, 1991[Abstract]
4. Bown N: Neuroblastoma tumour genetics: Clinical and biological aspects. J Clin Pathol 54:897-910, 2001
5. Maris JM, Matthay KK: Molecular biology of neuroblastoma. J Clin Oncol 17:2264-2279, 1999 6. Westermann F, Schwab M: Genetic parameters of neuroblastomas. Cancer Lett 184:127-147, 2002[CrossRef][Medline] 7. Ejeskar K, Aburatani H, Abrahamsson J, et al: Loss of heterozygosity of 3p markers in neuroblastoma tumours implicate a tumour-suppressor locus distal to the FHIT gene. Br J Cancer 77:1787-1791, 1998[Medline] 8. Guo C, White PS, Weiss MJ, et al: Allelic deletion at 11q23 is common in MYCN single copy neuroblastomas. Oncogene 18:4948-4957, 1999[CrossRef][Medline] 9. Takita J, Hayashi Y, Kohno T, et al: Allelotype of neuroblastoma. Oncogene 11:1829-1834, 1995[Medline] 10. Luttikhuis ME, Powell JE, Rees SA, et al: Neuroblastomas with chromosome 11q loss and single copy MYCN comprise a biologically distinct group of tumours with adverse prognosis. Br J Cancer 85:531-537, 2001[CrossRef][Medline]
11. Bown N, Cotterill S, Lastowska M, et al: Gain of chromosome arm 17q and adverse outcome in patients with neuroblastoma. N Engl J Med 340:1954-1961, 1999 12. Speleman F, Bown N: 17q gain in neuroblastsoma, in Brodeur GM, Sawada T, Tsuchida Y, et al (eds): Neuroblastoma. Amsterdam, the Netherlands, Elsevier, 2000, pp 113-124
13. Kallioniemi A, Kallioniemi OP, Sudar D, et al: Comparative genomic hybridization for molecular cytogenetic analysis of solid tumors. Science 258:818-821, 1992 14. du Manoir S, Speicher MR, Joos S, et al: Detection of complete and partial chromosome gains and losses by comparative genomic in situ hybridization. Hum Genet 90:590-610, 1993[Medline] 15. Vandesompele J, Van Roy N, Van Gele M, et al: Genetic heterogeneity of neuroblastoma studied by comparative genomic hybridization. Genes Chromosomes Cancer 23:141-152, 1998[CrossRef][Medline] 16. Plantaz D, Mohapatra G, Matthay KK, et al: Gain of chromosome 17 is the most frequent abnormality detected in neuroblastoma by comparative genomic hybridization. Am J Pathol 150:81-89, 1997[Abstract] 17. Lastowska M, Nacheva E, McGuckin A, et al: Comparative genomic hybridization study of primary neuroblastoma tumors: United Kingdom Children's Cancer Study Group. Genes Chromosomes Cancer 18:162-169, 1997[CrossRef][Medline] 18. Brinkschmidt C, Christiansen H, Terpe HJ, et al: Comparative genomic hybridization (CGH) analysis of neuroblastomas: An important methodological approach in paediatric tumour pathology. J Pathol 181:394-400, 1997[CrossRef][Medline] 19. Vandesompele J, Speleman F, Van Roy N, et al: Multicentre analysis of patterns of DNA gains and losses in 204 neuroblastoma tumors: How many genetic subgroups are there? Med Pediatr Oncol 36:5-10, 2001[CrossRef][Medline] 20. Plantaz D, Vandesompele J, Van Roy N, et al: Comparative genomic hybridization (CGH) analysis of stage 4 neuroblastoma reveals high frequency of 11q deletion in tumors lacking MYCN amplification. Int J Cancer 91:680-686, 2001[CrossRef][Medline] 21. Breen CJ, O'Meara A, McDermott M, et al: Coordinate deletion of chromosome 3p and 11q in neuroblastoma detected by comparative genomic hybridization. Cancer Genet Cytogenet 120:44-49, 2000[CrossRef][Medline] 22. Brinkschmidt C, Poremba C, Christiansen H, et al: Comparative genomic hybridization and telomerase activity analysis identify two biologically different groups of 4s neuroblastomas. Br J Cancer 77:2223-2229, 1998[Medline]
23. Baudis M, Cleary ML: Progenetix.net: An online repository for molecular cytogenetic aberration data. Bioinformatics 17:1228-1229, 2001
24. Brodeur GM, Pritchard J, Berthold F, et al: Revisions of the international criteria for neuroblastoma diagnosis, staging, and response to treatment. J Clin Oncol 11:1466-1477, 1993 25. Shimada H, Ambros IM, Dehner LP, et al: The International Neuroblastoma Pathology Classification (the Shimada system). Cancer 86:364-372, 1999[CrossRef][Medline] 26. Bentz M, Plesch A, Stilgenbauer S, et al: Minimal sizes of deletions detected by comparative genomic hybridization. Genes Chromosomes Cancer 21:172-175, 1998[CrossRef][Medline]
27. Dysvik B, Jonassen I: J-express: Exploring gene expression data using Java. Bioinformatics 17:369-370, 2001 28. Holzman T, Kolker E: Statistical analysis of global gene expression data: Some practical considerations. Curr Opin Biotechnol 15:52-57, 2004[CrossRef][Medline]
29. Lastowska M, Cullinane C, Variend S, et al: Comprehensive genetic and histopathologic study reveals three types of neuroblastoma tumors. J Clin Oncol 19:3080-3090, 2001 30. Abel F, Ejeskar K, Kogner P, et al: Gain of chromosome arm 17q is associated with unfavourable prognosis in neuroblastoma, but does not involve mutations in the somatostatin receptor 2(SSTR2) gene at 17q24. Br J Cancer 81:1402-1409, 1999[CrossRef][Medline] 31. Cunsolo CL, Bicocchi MP, Petti AR, et al: Numerical and structural aberrations in advanced neuroblastoma tumours by CGH analysis: Survival correlates with chromosome 17 status. Br J Cancer 83:1295-1300, 2000[CrossRef][Medline]
32. Spitz R, Hero B, Ernestus K, et al: Deletions in chromosome arms 3p and 11q are new prognostic markers in localized and 4s neuroblastoma. Clin Cancer Res 9:52-58, 2003
33. Spitz R, Hero B, Ernestus K, et al: Gain of distal chromosome arm 17q is not associated with poor prognosis in neuroblastoma. Clin Cancer Res 9:4835-4840, 2003 34. Mattfeldt T, Wolter H, Kemmerling R, et al: Cluster analysis of comparative genomic hybridization (CGH) data using self-organizing maps: Application to prostate carcinomas. Anal Cell Pathol 23:29-37, 2001[Medline]
35. Joos S, Menz CK, Wrobel G, et al: Classical Hodgkin lymphoma is characterized by recurrent copy number gains of the short arm of chromosome 2. Blood 99:1381-1387, 2002
36. Dyer S, Prebble E, Davison V, et al: Genomic imbalances in pediatric intracranial ependymomas define clinically relevant groups. Am J Pathol 161:2133-2141, 2002
37. Burton EC, Lamborn KR, Feuerstein BG, et al: Genetic aberrations defined by comparative genomic hybridization distinguish long-term from typical survivors of glioblastoma. Cancer Res 62:6205-6210, 2002
38. Fritz B, Schubert F, Wrobel G, et al: Microarray-based copy number and expression profiling in dedifferentiated and pleomorphic liposarcoma. Cancer Res 62:2993-2998, 2002
39. Wang TL, Maierhofer C, Speicher MR, et al: Digital karyotyping. Proc Natl Acad Sci U S A 99:16156-16161, 2002 40. Pinkel D, Segraves R, Sudar D, et al: High resolution analysis of DNA copy number variation using comparative genomic hybridization to microarrays. Nat Genet 20:207-211, 1998[CrossRef][Medline] 41. Solinas-Toldo S, Lampel S, Stilgenbauer S, et al: Matrix-based comparative genomic hybridization: Biochips to screen for genomic imbalances. Genes Chromosomes Cancer 20:399-407, 1997[CrossRef][Medline] 42. Pollack JR, Perou CM, Alizadeh AA, et al: Genome-wide analysis of DNA copy-number changes using cDNA microarrays. Nat Genet 23:41-46, 1999[Medline]
43. Lucito R, Healy J, Alexander J, et al: Representational oligonucleotide microarray analysis: A high-resolution method to detect genome copy number variation. Genome Res 13:2291-2305, 2003
44. Bignell GR, Huang J, Greshock J, et al: High-resolution analysis of DNA copy number using oligonucleotide microarrays. Genome Res 14:287-295, 2004 Submitted June 21, 2004; accepted December 30, 2004.
This article has been cited by other articles:
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
|
|||||||||||
|
|||||||||||
|
|
Copyright © 2005 by the American Society of Clinical Oncology, Online ISSN: 1527-7755. Print ISSN: 0732-183X
|
TOP
ABSTRACT
INTRODUCTION
2 tests were used to address the distribution of the genetic and clinical parameters, and the International Neuroblastoma Pathology Classification (INPC) risk status in the three tumor clusters, or to demonstrate the significant correlation between two parameters. The Kaplan-Meier univariate survival analysis method and log-rank test were used to estimate overall survival and event-free survival. A stepwise Cox proportional hazards model (successive exclusion of nonsignificant covariates with P > .05) was applied to test the independent significant influence of different genetic and clinical parameters on patient outcome. Verification of the proportional hazards assumption (stating that a particular covariate should be proportionally related to the baseline hazard) was done by inspecting the log-minus-log plot of the survival function. Nonparametric Mann-Whitney testing was used to compare the age at diagnosis between the different genetic subgroups. 
