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

Neuroblastoma serves as a paradigm for utilising tumour genomic data for determining patient prognosis and treatment allocation. However, before the establishment of the International Neuroblastoma Risk Group (INRG) Task Force in 2004, international consensus on markers, methodology, and data interpretation did not exist, compromising the reliability of decisive genetic markers and inhibiting translational research efforts. The objectives of the INRG Biology Committee were to identify highly prognostic genetic aberrations to be included in the new INRG risk classification schema and to develop precise definitions, decisive biomarkers, and technique standardisation. The review of the INRG database (n=8800 patients) by the INRG Task Force finally enabled the identification of the most significant neuroblastoma biomarkers. In addition, the Biology Committee compared the standard operating procedures of different cooperative groups to arrive at international consensus for methodology, nomenclature, and future directions. Consensus was reached to include MYCN status, 11q23 allelic status, and ploidy in the INRG classification system on the basis of an evidence-based review of the INRG database. Standardised operating procedures for analysing these genetic factors were adopted, and criteria for proper nomenclature were developed. Neuroblastoma treatment planning is highly dependant on tumour cell genomic features, and it is likely that a comprehensive panel of DNA-based biomarkers will be used in future risk assignment algorithms applying genome-wide techniques. Consensus on methodology and interpretation is essential for uniform INRG classification and will greatly facilitate international and cooperative clinical and translational research studies.

Modern 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 (D'Angio et al, 1971;Evans et al, 1976). However, patient age and tumour stage alone cannot reliably predict tumour behaviour. Over the past two decades, tumour histology, (Shimada et al, 1999) status of the MYCN oncogene, (Brodeur et al, 1984;Seeger et al, 1985), and tumour cell DNA content (ploidy; Look et al, 1991;Ladenstein et al, 2001) 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 (Cohn et al, 2009). 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 (Ambros et al, 2003)) for analysing neuroblastoma tumour molecular diagnostics and genetic nomenclature in this rapidly evolving field.

Genetic features of neuroblastic tumours associated with favourable clinical behaviour
Favourable 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 (Ambros et al, 1995(Ambros et al, , 1996Brodeur, 2003;George et al, 2007;Mosse et al, 2007;Schleiermacher et al, 2007). 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 (Ambros et al, 1996;Bourdeaut et al, 2008).
Although tumour cell hyperdiploidy (usually near-triploid DNA content) is frequently associated with spontaneous maturation (Ambros et al, 1996) 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.

Genetic features of neuroblastic tumours associated with unfavourable clinical behaviour
Neuroblastomas 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 (Schwab et al, 1983;Brodeur et al, 1984;Seeger et al, 1985). 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 (Brodeur et al, 1984;Seeger et al, 1985;Ambros et al, 1995;Perez et al, 2000). As expected, MYCN amplification was also highly predictive of worse outcome in the INRG cohort of patients (Cohn et al, 2009). 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 (Manohar et al, 1995;Squire et al, 1995;George et al, 1997;De Preter et al, 2002;Weber et al, 2004).
Although 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, (Maris et al, 1995;Caron et al, 1996b;Attiyeh et al, 2005) as well as unbalanced gain of the long arm of chromosome 17 (17q21 to 17qter; Caron, 1995;Bown et al, 1999;Spitz et al, 2003). In addition, some studies have shown that deletions on chromosome 3p, 4p, 9p, and 12p also have prognostic significance (Caron et al, 1996a;Thompson et al, 2001;Schleiermacher et al, 2007). As 11q deletions are inversely associated with MYCN amplification, this aberration has emerged as a powerful biomarker of outcome in cases without MYCN amplification (Attiyeh et al, 2005). 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  Figure 1 Biological pathways and genetic features in neuroblastic tumours. Tumour cell ploidy (grey columns) can be used to subdivide neuroblastoma tumours into two broad groups (separated by the long punctuated line). Although the ploidy subgroups roughly correspond to the biologic subgroups (aggressive neuroblastomas marked by a red background -either with MYCN amplification in dark red and separated by a short punctuated line from neuroblastomas without MYCN amplification vs less aggressive behaving neuroblastomas indicated by a green background), they do not totally match. Although aggressive near-triploid neuroblastomas (in red below the long punctuated line) have been observed, it is less clear if 'benign' diploid neuroblastomas without any structural aberrations (in green above the long punctuated line) occur. 'Benign' clinical behaviour refers either to spontaneous regression/maturation without any therapy or with surgery only (no cytotoxic therapy).    (Cohn et al, 2009). Recent publications clearly show the potential of comprehensive genome-wide approaches to further refine the prognostic accuracy of somatically acquired chromosomal alterations, (Vandesompele et al, 2005;Michels et al, 2007;Mosse et al, 2007;Schleiermacher et al, 2007;Tomioka et al, 2008). 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 wholechromosome 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 (Janoueix-Lerosey et al, 2009).

Tumour cell ploidy
The majority of tumours have a hyperdiploid (near-triploid or penta/hexaploid) DNA content, whereas less than half of neuroblastic tumours are diploid (Look et al, 1984;Kaneko et al, 1987). Locoregional tumours are commonly hyperdiploid; diploidy is more common in advanced-stage tumours. A number of studies have shown that in patients o18 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 (George et al, 2005;Schleiermacher et al, 2007). However, in patients o18 months of age with metastatic disease with diploid, MYCN non-amplified tumours have a statistically significantly worse outcome.

Outcome by clinical and genetic subgroups
In the survival-tree regression approach of Cohn and Pearson et al (2009), 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 nonadrenal tumour (Table 2b).
However, 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).

Statistical considerations
Tests of association were carried out using a w 2 -test. P-values o0.05 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,  Kaplan and Meier (1958) were used to calculate EFS and OS estimates, with s.e. according to Peto and Peto (1972).

INTERNATIONAL CONSENSUS BY THE INRG BIOLOGY COMMITTEE
Because 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 (Ambros et al, 2003).
Consensus on the genetic markers to be currently used in the INRG risk classification system As reported in the study by Cohn and Pearson et al (2009) survival-tree regression analyses of the INRG database confirmed the prognostic significance of MYCN amplification, 11q aberra-  Table 3 includes these three genetic markers and also those, which should be analysed prospectively.
Tumour sampling/storing procedure and indication of the tumour cell content An adequate amount of tumour material (i.e., 10 7 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. Molecular studies on maturing/mature tumours, such as ganglioneuroblastomas or ganglioneuromas, (Ambros et al, 1996;Brodeur, 2003) 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 needlecore biopsies, for obtaining tumour material from different areas, an exact determination of the tumour cell content is crucial again (Frostad et al, 1999). 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.
On 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 À801C or in liquid nitrogen) with exactly determined tumour cell content. Nonneoplastic 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 (Qualman et al, 2005). For further details and recommendations, see Figure 2 and the guidelines indicated by .

Evaluation and the reporting of the MYCN gene copy status
The INRG Biology Committee agreed that MYCN status should be evaluated in every resected neuroblastic tumour, including the Schwann cell stroma-rich categories (Shimada et al, 1999). 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 5-8. 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.

Evaluation and reporting of segmental genetic aberrations: gains and losses of chromosome parts
The term 'structural' aberrations (also known as structural chromosomal instability (s)-CIN) designates any kind of chromosomal  Figure 2 Recommendations concerning the splitting of the tumour material for resected tumours or surgical biopsies. All specimens should be transported to the pathology laboratory as quickly as possible. From there, the snap frozen and/or OCT embedded material should then be transported immediately to the biology lab (can be used for any type of DNA, RNA or protein work). Normal reference cells (e.g., peripheral blood) should be sent to the reference laboratories. A balanced ratio between the MYCN-specific signals and the signal number of the reference probe on the chromosome 2q arm Neuroblastic tumours have a propensity to polyploidisation, especially but not exclusively after therapy with the occurrence of giant polyploid nuclei. The number of MYCN signals must not exceed the number of reference signals on 2q (caveat: the centromeric probe is unqualified as reference probe because of frequent centromeric associations in such tumours). The number of other chromosomes or the DNA index should be checked as well. 2p24 gain: The term 2p24 gain is suggested as a descriptive generic term for MYCN signal numbers exceeding up to 4-fold the number of reference signals on chromosomal arm 2q. This pattern could reflect either: The use of a 2p specific probe in addition to 2q (optionally a centromeric probe) is recommended to clarify the presence of a chromosome 2p gain vs a restricted MYCN gain. The discrimination could be important since a MYCN gain could indicate an 'incipient' MYCN amplification. Generic term for the coexistence of amplified as well as non-amplified tumour cells in the same tumour Single cells (at least two) or cell clusters or a multitude of cells with MNA besides proven tumour cells without MNA. The terms 'focal' and 'diffuse MNA' are specifications, which, however, can only be attributed after evaluation of the tissue sections. In case amplified tumour cells are found in addition to non-amplified cells, a very meticulous procedure is strongly recommended to exclude false-positive or false negative results (see Figure 3).

Terms Description Comments
Normal status Balanced ratio between the signal numbers of the chromosomal region of interest and the reference signals on the opposite arm of the chromosome In the case of a 2/2 ratio, an isodisomy with a complete LOH of all loci located on the investigated chromosome cannot be excluded. However, uniparental isodisomies are extremely rare in NB tumours. I-FISH imbalance (inconclusive with regard to LOH) Unbalanced ratio between the signal numbers of the chromosomal region of interest and the reference signals with more than one signals of the chromosomal region of interest Does not necessarily correspond to an LOH, and is therefore called inconclusive. PCR should be performed to clarify this result. In NB tumours, imbalances frequently, but not always, reflect LOH. International consensus for neuroblastoma molecular diagnostics PF Ambros et al 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 ligationdependent 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 -8; part of the nomenclature was developed by the SIOPEN Biology Group (for group members, see Table 4 and Ambros et al, 2003)).

Tumour cell DNA content (ploidy)
Ploidy was analysed in the INRG cohort using the definitions diploid (i.e., DI p1.0) vs hyperdiploid (DI 41.0) as published in Look et al (1991). 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.  Balanced ratio between the majority of signals (signal intensity is visualised in the graphic representation of the MLPA results as height of the bars) of both chromosomal arms In the case of a balanced ratio, a uniparental isodisomy (a uniparental isodisomy means that both chromosomes are derived from one parent. Thus, the two chromosomes are not homologous but identical. However, uniparental isodisomies are rare in NB tumours) with a complete LOH of all loci located on the investigated chromosome cannot be excluded Segmental chromosome aberration Loss Unbalanced ratio between the signals of the chromosomal region of interest (at least two adjacent probes) and the reference signals (at least two probes) signals This result could correspond to a FISH deletion that reflects an LOH, or a FISH imbalance, which does not necessarily indicate an LOH.

Gain
Unbalanced ratio (low signal excess) between the signals of the chromosomal region of interest (at least two adjacent probes) and the reference signals (at least two) of the chromosomal region of interest The threshold between gain and amplification needs to be determined for the experimental system with the help of other techniques,for example, I-FISH Amplification Unbalanced ratio (high signal excess) between the signals of a gene and all other probes located on the same chromosome The threshold between gain and amplification needs to be determined for the experimental system with the help of other techniques, for example, I-FISH No result Should be specified: unclear or not interpretable; inadequate tumour cell content; no tumour; not carried out FISH ¼ fluorescence in situ hybridisation; MLPA ¼ multiplex ligation-dependent probe amplification; I-FISH ¼ interphase fluorescence in situ hybridisation.
International consensus for neuroblastoma molecular diagnostics PF Ambros et al

FUTURE PROSPECTS: NEW TECHNIQUES
There 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.

DNA-based biomarkers
A 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. Multiplex ligation-dependent probe amplification is a PCRbased technique that detects a large variety of segmental aberrations and gene amplifications in a robust manner (for technical details, see (Schouten et al, 2002). 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 (Villamon et al, 2008). The robust nature of the results and the relatively low cost of the MLPA kits make this technique attractive for routine neuroblastoma analysis.
Chip-based technologies have also been used to molecularly classify neuroblastoma tumours (Bilke et al, 2005;George et al, 2005;Selzer et al, 2005;Spitz et al, 2006;Stallings et al, 2006;Lastowska et al, 2007;Michels et al, 2007;Mosse et al, 2007;Tomioka et al, 2008). 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 (George et al, 2005). 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  Figure 3 Heterogeneous MYCN amplification and recommended procedure for the clarification of the underlying genetic pattern.
International consensus for neuroblastoma molecular diagnostics PF Ambros et al 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 Janoueix-Lerosey et al, 2009), 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.

RNA-based biomarkers
Several groups have recently shown that genome-wide gene expression profiling can identify differentially expressed transcripts that provide reliable prognostic information (Wei et al, 2004;Ohira et al, 2005;Schramm et al, 2005;Asgharzadeh et al, 2006;Oberthuer et al, 2006;Lastowska et al, 2007;Tomioka et al, 2008). 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.

Molecular targets
Neuroblastoma 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.

CONCLUSION
Currently, 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.