Studies of the tumour genome by chromosome banding analysis suffer from several shortcomings, e.g. difficulties in obtaining a sufficient number of mitotic cells for analysis, overgrowth of normal cells in tissue culture, poor quality of chromosome preparation, and an inherently low resolution. Nevertheless, several large cytogenetic series of NB, some of which have been complemented by multicolour fluorescence in situ hybridisation analysis, have been reported with a high frequency of cases showing cytogenetic abnormalities and with detailed karyotype descriptions [6
]. In the present study, we have summarised the published cytogenetic data in NB and performed a comprehensive and unbiased statistical analyses. PCA of chromosomal imbalances revealed three distinct clusters of imbalances that tended to occur together in the same tumours, i.e. whole-chromosome gains, whole-chromosome losses, and structural changes. Of the structural changes, 1p- and dmin were by far the most common, occurring in 48% and 20% of the tumours, respectively. Indeed, when PCA of the tumour cases was performed, these two changes largely dominated the first two principal components. The third principal component was dominated by the spectrum of whole chromosome changes. In summary, the PCAs indicated that the accumulated cytogenetic data could warrant sub-classification of NBs into at least four groups, i.e. those characterised by (1) dmin and few other imbalances, (2) dmin, 1p-, and few other imbalances, (3) 1p-, other structural changes such as +17q, and whole chromosome losses, and finally (4) whole-chromosome gains and losses. Groups 1–3 were dominated by the structural abnormalities associated with high-risk tumours in the present study (dmin/hsr, 1p), while group 4 was dominated by numerical abnormalities associated with low-risk tumours (+9, +12, +17, and -4).
Several authors have suggested that NB can be sub-divided into at least two biological entities, reflecting the heterogeneous scenario observed in NB patients [26
]. The first group is characterised by a near-diploid karyotype and genetic imbalances based on structural cytogenetic changes. Among these, the NB with MYCN
amplification, also often showing 1p-, generally present with high-stage disease and are rapidly progressive, whereas the tumours without MYCN
amplification but with other structural changes such as loss of heterozygosity for 11q typically occur in patients with slowly progressive but often fatal disease. These two subgroups most closely resemble groups 1–2 and 3, respectively in the present study. The other main biological group is characterised by a near-triploid karyotype with few structural chromosome changes, occurring in infants with confined disease and a very good prognosis. This group corresponds to our group 4. Hence, despite its inherent methodological limitations, the present study identified clinico-genetic sub-groups that were largely similar to those identified by other recent studies using comparative genome hybridisation and oligonucleotide arrays [3
It has been suggested that diploid and triploid NB may arise from a common tetraploid precursor undergoing tripolar cell division [26
]. The first type of daughter cell (2n) may give rise to the near-diploid, typically aggressive, tumours characterised by structural chromosomal changes, particularly 1p-, 11q-, +17q, and dmin (our groups 1–3). The second type of daughter cell (3n) may evolve into the near-triploid, typically indolent, tumours harbouring whole-chromosome changes, such as +9, +12, and +17 (our group 4). Although a highly attractive hypothesis, little empirical data have so far been presented to substantiate that these two biological NB entities in fact evolve through different evolutionary pathways. If the near-diploid tumours were actually derived from a tripolar tetraploid mitosis, segregating into haploid sets as originally suggested by Kaneko and Knudson [27
], this could result in three possible sex chromosome complements in the resulting 2n cells, i.e. XY, XX, and YY. Of these, XY would be twice as probable as each of the two other two. It can be further argued that the YY cells, having nullisomy for the X chromosome, are not likely to survive. Hence, if the 2n tumours resulted from a tripolar cell division of a tetraploid cell, this would imply that 1/3 of the NB with 2n ploidy in male patients should have an XX chromosome complement. None of the 90 karyotypes from male patients with 2n tumours in the present series, showed such sex-reversal. However, this chain of reasoning assumes that somatic cell feminisation would not be lethal on the cellular level, which cannot be excluded at present. Furthermore, it does not rule out that tripolar division of a tetraploid cell lies behind the generation of near-triploid tumours, in which such somatic sex reversal would not occur if one assumed segregation into haploid sets.
In the present study we used the number of chromosome imbalances at which a certain chromosomal aberration typically occurred to model the evolution of genomic changes in NB. This approached has previously been used to delineate the temporal development of chromosome aberrations in several other tumours [9
]. Our analysis showed that the structural changes occurred in a step-wise fashion, with loss of 1p sequences and the emergence of double minute chromosomes as the earliest cytogenetic events, followed by gain of 1q and 17q, loss of 3p and 6q, acquisition of homogeneously staining regions, and finally loss of 11q. In contrast, the gains and losses of whole chromosomes occurred more or less simultaneously, in a fashion similar to that suggested for the acquisition of whole-chromosome gains in hyperdiploid childhood leukemias [30
]. This supports the notion that the numerical chromosome changes in 3n NB may arise through a single rare event, such as an asymmetrical mitotic cell division. In contrast, the 2n tumours appear to develop through the sequential acquisition of structural changes. Such a step-wise evolution of genome imbalances has been found in many common carcinomas, such as colorectal cancer, breast cancer, ovarian cancer, and bladder cancer [8
]. It is possible that this step-wise acquisition of clonal changes reflects genome plasticity, which might explain why 2n tumours are more resistant to treatment than 3n tumours.