International consensus for neuroblastoma molecular diagnostics: report from the International Neuroblastoma Risk Group (INRG) Biology Committee

1Modern treatment strategies adjust the therapy of cancer patients according to the predicted biological behaviour of the individual tumour. This is especially important for neuroblastoma patients, as a subset of neuroblastic tumours will undergo spontaneous regression (in infants and in young children) or maturation (in children), whereas others will rapidly progress despite intensive multimodality therapy. This clinical heterogeneity has been known for decades. However, patient age and tumour stage alone cannot reliably predict tumour behaviour. Over the past two decades, tumour histology, status of the MYCN oncogene,, and tumour cell DNA content (ploidy;; ) have each been shown to be independently predictive of patient outcome in large retrospective and prospective studies. Besides these two genetic markers, recently 11q aberrations were also included in the INRG pretreatment risk classification. However, the optimal combination of these and additional recently described prognostic biomarkers to build a treatment stratification algorithm has not been determined yet. Here, we present the consensus of the INRG Biology Committee on which markers should be used currently, on the standardised operating procedures (SOPs; partly on the basis of SIOPEN activities) for analysing neuroblastoma tumour molecular diagnostics and genetic nomenclature in this rapidly evolving field.

2Favourable clinical behaviour of neuroblastic tumours is classically characterised by a propensity to undergo spontaneous regression or maturation without cytotoxic therapy (Figure 1). These tumours almost always show whole chromosome gains with few, if any, segmental chromosome aberrations and without gene amplifications. The DNA content is usually in the near-triploid (penta/hexaploid) range. This is also the case for the neuroblastic/ganglionic cell population in Schwann cell stroma-rich tumours, whereas the Schwann cell population itself has a diploid DNA content.

3Although tumour cell hyperdiploidy (usually near-triploid DNA content) is frequently associated with spontaneous maturation and possibly also with spontaneous regression, this ploidy level alone cannot guarantee benign tumour behaviour as near-triploid tumours may also have segmental chromosomal aberrations (i.e., gains or losses of only parts of chromosomes) and MYCN amplification, which can lead to clinically aggressive behaviour.

4Neuroblastomas with an unfavourable clinical behaviour (Figure 1) have a high propensity for locally invasive growth and widespread metastatic dissemination through the lymphatic and haematogenous systems. These tumours frequently show segmental aberrations and high-level amplification of the MYCN locus is detected in a substantial subset. MYCN amplification has been shown to be strongly associated with rapid tumour progression and poor prognosis in patients of all ages, with any stage of disease. As expected, MYCN amplification was also highly predictive of worse outcome in the INRG cohort of patients. Frequently, adjacent or more distantly located genes, such as DDX1, NAG, and ALK, are coamplified with MYCN, but amplification events in the absence of MYCN amplification are rare. Whether these coamplifications have a prognostic impact awaits clarification.

5Although MYCN status is central to the risk stratification systems in all cooperative clinical trial groups, it is important to emphasise that the majority of metastatic neuroblastomas do not show amplification of this oncogene (see also Figure 1). Other chromosomal aberrations (and a diploid DNA content) have been assumed to predict unfavourable tumour behaviour, including deletion at the chromosomal region 1p36.3 or 11q23, as well as unbalanced gain of the long arm of chromosome 17 (17q21 to 17qter;;; ). In addition, some studies have shown that deletions on chromosome 3p, 4p, 9p, and 12p also have prognostic significance. As 11q deletions are inversely associated with MYCN amplification, this aberration has emerged as a powerful biomarker of outcome in cases without MYCN amplification. Statistical analysis of the INRG database has confirmed this finding, and on the basis of these studies, 11q status has been included as a prognostic criterion in the INRG classification system.. Recent publications clearly show the potential of comprehensive genome-wide approaches to further refine the prognostic accuracy of somatically acquired chromosomal alterations,. These studies have shown that in tumours without MYCN amplification, segmental chromosome aberrations are associated with clinically aggressive disease. These findings have been extended in a series of 493 MYCN amplified and non-amplified neuroblastomas; in tumours with only whole-chromosome copy number variations, there were no disease-related deaths. In contrast, the presence of segmental alterations (see Table 3) with or without MYCN amplification was the strongest predictor of relapse.

6The majority of tumours have a hyperdiploid (near-triploid or penta/hexaploid) DNA content, whereas less than half of neuroblastic tumours are diploid. Locoregional tumours are commonly hyperdiploid; diploidy is more common in advanced-stage tumours. A number of studies have shown that in patients <18 months of age with metastatic disease, hyperdiploidy in combination with a non-amplified MYCN gene and the lack of specific segmental chromosome aberrations (such as 11q deletion) are predictive of a favourable outcome. However, in patients <18 months of age with metastatic disease with diploid, MYCN non-amplified tumours have a statistically significantly worse outcome.

7In the INRG analytic cohort (n=8800; see ) for details), 29% were diploid, 16% were MYCN amplified, 21% had an 11q aberration, 23% had a 1p aberration, and 48% had a 17q gain. Statistically significant associations of genetic factors were identified (Table 1). 11q aberration was associated with 17q gain (P<0.0001) and inversely associated with MYCN amplification (P=0.0006), but 11q aberration was not associated with diploidy or 1p aberration. Diploidy and 17q gain were not associated, but all other pair-wise comparisons were highly statistically significant (P<0.01).

8In the survival-tree regression approach of, patients were clustered into meaningful pretreatment groups that were homogeneous in terms of outcome and prognostic factors (clinical and biological). For the descriptive purposes of this manuscript, the INRG analytic cohort was further subdivided. Outcome by genetic factors vs INSS stage, age, and primary tumour site is shown in Tables 2a, and 2b, respectively. Within each genetic factor subgroup, the patterns of outcome differences that have been observed in the overall population prevailed: (a) EFS and overall survival (OS) decreased with increasing stage (Table 2a); (b) older patients had worse outcome than younger patients (Table 2b) and (c) patients with adrenal primary tumour site had worse outcome than those with non-adrenal tumour (Table 2b).

9However, as shown in Table 2a, only a limited number of data on the prognostic impact of some segmental aberrations are available, impeding reliable interpretation of some of the data (e.g., 17q gain in stage 1 and 4 s tumours). Thus, larger data sets on segmental aberrations are needed to allow final statements on their prognostic impact (Table 3).

10Tests of association were carried out using a χ2-test. P-values <0.05 were considered statistically significant. For event-free survival (EFS) analysis, time to event was defined as the time from diagnosis until the time of first occurrence of relapse, progression, secondary malignancy, or death, or until time of last contact if no event occurred. For OS, time to event was defined as time until death, or until last contact if the patient was alive. The methods of were used to calculate EFS and OS estimates, with s.e. according to.

11Because of the relative rarity of neuroblastomas, and the uniqueness of many of the molecular diagnostic factors, the INRG Biology Committee recommended that the genetic studies required for INRG classification be carried out in experienced laboratories, typically central reference laboratories for the cooperative groups (see Table 4), to guarantee high consistency and quality of results.

12As reported in the study by survival-tree regression analyses of the INRG database confirmed the prognostic significance of MYCN amplification, 11q aberrations, and ploidy, in addition to age and stage, in different subgroups of the INRG cohort. Table 3 includes these three genetic markers and also those, which should be analysed prospectively.

13An adequate amount of tumour material (i.e., 107 tumour cells) from at least two different regions of the tumour should be obtained. In collaboration with the institutional pathologist, the tumour cell content must be determined and recorded. The latter is an indispensable prerequisite to avoid false results and has to be carried out by the pathologist. A section (cryo- or paraffin) has to be kept to each tumour piece used for genetic or expression studies. In certain cases, interphase fluorescence in situ hybridisation (I-FISH) can be used to decide on the tumour cell number. A tumour cell content of more than 60% is required for most molecular studies and of more than 20% for ploidy measurement. Differentiating/maturing tumours can have fewer tumour cells, and therefore, the interpretation of these data need to de done with caution. For I-FISH, a low tumour cell number can also be sufficient, in case of numeric/structural chromosome aberrations.

14Molecular studies on maturing/mature tumours, such as ganglioneuroblastomas or ganglioneuromas, require meticulous identification of tumour and Schwannian stromal cells, and can thus only be undertaken when using an appropriate system (e.g., on ganglionic cells or by microdissection, ideally on paraffin sections). In case of needle-core biopsies, for obtaining tumour material from different areas, an exact determination of the tumour cell content is crucial again. Interphase fluorescence in situ hybridisation results of disseminated tumour cells in the bone marrow can be given only if MYCN amplification is present due to the higher error-proneness if segmental/numeric aberrations are evaluated in mixtures of normal and tumour cells. In case of immunological preselection and automatic relocation of DTCs, information on genetic aberrations other than MYCN amplification can be given as well.

15On account of the advent of new techniques enabling large retrospective studies, it is essential to store frozen tumour samples and/or extracted DNA, RNA, or protein (at −80°C or in liquid nitrogen) with exactly determined tumour cell content. Non-neoplastic reference cells from the same patient should be stored as well. The INRG Biology Subcommittee further emphasises the clear need for biobanking of high quality biological materials from neuroblastoma patients, and this must be central to the SOP of any cooperative group for the collection of diagnostic material. For further details and recommendations, see Figure 2 and the guidelines indicated by.

16The INRG Biology Committee agreed that MYCN status should be evaluated in every resected neuroblastic tumour, including the Schwann cell stroma-rich categories. Recommended and accepted techniques to detect structural and/or segmental aberrations are summarised in Table 3, and recommendations concerning the evaluation and terminology are given in Tables 5, 6, 7 and 8. Interphase fluorescence in situ hybridisation is preferred because it has a number of advantages compared with the other techniques, most important of which is direct quality control of the hybridization quality by the microscopist. Heterogeneous MYCN amplification requires meticulous analysis of the tumour specimen (Figure 3). At the present time, the prognostic significance of heterogeneously MYCN amplified neuroblastomas is not known, but amplification of a substantial number of tumour cells in a specimen is still considered an ominous sign.

17The term ‘structural’ aberrations (also known as structural chromosomal instability (s)-CIN) designates any kind of chromosomal alterations including gene amplification. To produce precise definitions, consensus was reached to introduce the term ‘segmental’ aberrations for gains and losses of chromosome parts. In neuroblastomas, most segmental chromosome aberrations are unbalanced, that is, they are associated with regional losses or gains of chromosome parts. Balanced aberrations, that is, reciprocal translocations without losses of genetic material, are thought to be relatively uncommon in neuroblastomas. Segmental aberrations (such as 11q23 and 1p36.3 deletions) can be identified by a number of techniques shown in Table 3. Currently, either I-FISH or PCR are carried out to detect segmental aberrations, but the INRG Biology Committee recommends array-based methods, multiplex ligation-dependent probe amplification (MLPA) or similar techniques in the future. The latter techniques are currently being validated (for nomenclature (I-FISH, PCR, and MLPA), see Tables 5, 6, 7 and 8; part of the nomenclature was developed by the SIOPEN Biology Group (for group members, see Table 4 and )).

18Ploidy was analysed in the INRG cohort using the definitions diploid (i.e., DI ⩽1.0) vs hyperdiploid (DI >1.0) as published in. More recent data suggest that prognostic classification may be further refined using specific ranges of DI The INRG Biology Committee recommends recording the exact numerical DI value for each tumour, so that clinically relevant cutoffs for DI can be determined (for techniques see Table 3). Any other method can be applied as long as it allows discrimination of the diploid DNA peak. However, an unknown number of normal cells contained in the tumour specimen under investigation, including tumour-derived, but non-neoplastic diploid Schwann cells are an important source of false data interpretation. Thus, the DNA content of maturing/mature tumours, which mostly develop from near-triploid neuroblastomas is easily misdiagnosed as diploid.

19There is clear consensus that the use of genetic data derived from diagnostic neuroblastoma tumours will remain central to patient treatment planning and gain even more power. The INRG Biology Committee addressed the issues of ongoing and future work directed at prognostic biomarker discovery and validation, as well as how these data sets can be leveraged to identify molecular targets for novel therapeutics. In Table 3, 12 genetic markers to be analysed prospectively are mentioned. Pan-genomic investigations may help to identify additional genomic areas of interest.

20A number of studies suggest that the pattern of DNA-based genomic changes is prognostic in neuroblastoma, and that whole genome analysis should be carried out rather than a series of individual locus-specific assays. Therefore, the INRG Biology Subcommittee suggests using pan- or multigenomic techniques enabling an analysis of all relevant genomic loci.

21Multiplex ligation-dependent probe amplification is a PCR-based technique that detects a large variety of segmental aberrations and gene amplifications in a robust manner (for technical details, see. Quantification of losses, gains, and amplifications in small amounts of DNA can be determined, and simultaneous investigation of a large number of loci, covering all currently known important aberrant regions found in neuroblastomas, can be carried out in a single assay. The robust nature of the results and the relatively low cost of the MLPA kits make this technique attractive for routine neuroblastoma analysis.

22Chip-based technologies have also been used to molecularly classify neuroblastoma tumours. Comparative genomic hybridisation has reached a high coverage of the target sequences and become more widely used. The advantage of single nucleotide polymorphism (SNP)-based platforms is the simultaneous detection of physical gains and losses and the detection of copy neutral LOH. Still, these techniques cannot be considered as routine techniques but as excellent tools to identify so far undetected genomic regions of prognostic impact. The INRG Biology Committee considers it a priority to work towards a single diagnostic tool that will reliably and accurately detect allelic deletions and gains, as well as MYCN amplification, and estimate overall DNA content (ploidy), on a single platform. This has been achieved using in-house arrays, but a commercially available platform will be needed to implement this or a similar technology (e.g., MLPA) uniformly in clinical laboratories across various groups and countries.

23Several groups have recently shown that genome-wide gene expression profiling can identify differentially expressed transcripts that provide reliable prognostic information. Although these studies are of outstanding quality, they each suffer from relatively small sample sizes, necessitating preliminary validation of proposed gene sets. In addition, there was very little overlap in the gene sets that were identified as prognostic by each group, so a consensus gene set, that is, predictive of outcome has yet to be identified. International collaboration will be required to unequivocally determine whether mRNA expression profiles, likely focused on a representative set of genes, are a sensitive and specific enough molecular assay to be used in the clinic, and/or if these data are synergistic with, or override the information derived from the assays of DNA alterations.

24Neuroblastoma treatment will continue to rely on risk grouping based on tumour genomic features. Increasing attention is now focused on utilising these data sets to discover therapeutic targets. It stands to reason that any prognostically relevant genomic aberration might also signal a molecular aberration that is critical to the maintenance of the malignant phenotype, and thus can be targeted for therapy. Regions of DNA copy number gain that result in the overexpression of a protein that is druggable in patients succumbing to the disease can theoretically be identified in carefully annotated data sets in which both high quality DNA- and RNA-based microarray data are available. Thus, a parallel focus on therapeutic target discovery and validation will further increase the significance of genomic efforts used in the discovery, validation, or even clinical application phases of implementation.

25Currently, neuroblastoma treatment planning is not possible without detailed knowledge of tumour cell genomics. International efforts were and are needed not only to identify the most significant genetic markers but also to identify and coordinate particularly suitable techniques for their evaluation in all patients. For future studies, new technologies, which provide a comprehensive picture of the tumour cell genome are recommended. Moreover, the INRG Biology Committee achieved consensus on the nomenclature of genetic aberrations and developed definitions of the terms to be used. The INRG Biology Committee is dedicated to providing the highest possible reproducibility and reliability of genetic markers enabling a uniform INRG classification and forming the basis for international clinical and translational studies. Finally, many of these recommendations apply not only to neuroblastoma tumours but also to any tumour entity for which genetic factors are essential for therapy decisions.