Frequency of CN-LOH in NB
CN-LOH has previously been reported in many types of cancers  and to some extent in neuroblastoma . However the amount of LOH that is truly copy neutral found among primary NB tumors is rather low compared to many other tumors . In this paper we report a frequency of near-CN-LOH in primary tumors (35%) that differs from our previously reported frequency for NB (3.3%) . This difference is due to two important extensions; first we have changed the definition of CN-LOH to near-CN-LOH, thus accepting a wider range of abnormalities in order to include the near-triploid tumors. Second, we have extended the dataset by including 62 tumors not included in the first dataset.
The statistical analysis show that near-CN-LOH is more common among NB-cell lines than primary NB tumors (p = 0.002). This could implicate either that CN-LOH is more common among relapsing tumors, from which most cell lines are derived, or that CN-LOH is an effect of cell line immortalization. Near-CN-LOH was also found to be more frequent in the primary tumors than in the healthy controls (p = 0.004), although the healthy controls did present with near-CN-LOH as well. Due to the lack of matched control samples in the material, we were unable to discriminate between IBD and tumor derived CN-LOH in the tumors. However, the difference in frequency seen between these groups indicates that at least some of the near-CN-LOH in the tumors represents true tumor acquired events.
Possible clinical implications
As can be seen in Figure 2, the primary tumors sometimes display near-CN-LOH in regions of clinical and/or prognostic importance for NB tumors, such as 1 p, 11 q or 17 q. This highlights that even though CN-LOH is not very common in NB, it still adds an extra complexity to tumor genetics, and the total number of aberrations in these regions might be higher than previously thought. Thus by combining CN-LOH data with data about deletions and gains, new regions of interest can be identified and previously characterized regions can be further defined. However, the possible clinical implications of CN-LOH are at the moment difficult to evaluate, ranging from potentially harmless to resulting in tumor suppressor gene silencing, depending on mutational and imprinting status of the remaining allele. Moreover, in contrast to a hemizygous deletion, CN-LOH would not be expected to cause any haploinsufficiency effects, since the total copy number remains normal.
Loss of heterozygosity in combination with gain of the same region (Gain-LOH) is another very intriguing aberration that was found in the material, e.g. in the 17 q region often affected with gain in aggressive NB tumors (Figure 2). This abnormality is especially interesting due to its dual nature, which makes simultaneous promotion of both tumor suppressor genes and oncogenes possible. For some oncogenic gain-of-function mutations, such as the V617F mutation in the tyrosine kinase JAK2 in myeloproliferative disorders, loss of the wild type allele could actually give an advantage to the tumor cell. This mutation targets the auto inhibitory region of JAK2 and results in a constitutively active tyrosine kinase that confers growth factor independency as well as cytokine hypersensitivity. The region surrounding JAK2 (9 p) frequently contains CN-LOH or Gain-LOH in these diseases [22–25]. It has also been shown that adding wild type JAK2 to cells carrying the mutated JAK2 allele restores the growth factor dependency, indicating that in this case it is in fact favorable if not necessary for tumor progression to lose the wild type allele .
The subtypes of CN-LOH might represent different mechanisms
The subtypes of CN-LOH were found to differ in their relative frequency between the three sample groups. This probably reflects different modes of acquisition, more or less common in the different groups. The whole chromosome CN-LOH events could be explained by either anaphase lag, causing loss of one allele followed by reduplication, or nondisjunction causing two sister chromatids to end up in the same cell. This could occur either at mitosis as a tumor acquired event or at meiosis as a germ line event [17, 26]. Smaller segmental CN-LOH (interstitial and telomeric), on the other hand, probably arise in the tumors either through mitotic recombination between low copy repeats or due to a double-stranded-break repair error followed by reduplication .
It is also possible that some of the CN-LOH seen in the tumors might in fact represent IBD and not true tumor acquired events. This mechanism can in theory give rise to any of the different subgroups of CN-LOH. However due to meiotic recombination, the CN-LOH regions would be expected to become smaller with each generation, thus making small, interstitial events more likely than the larger telomeric or whole chromosome events. This could explain at least part of the differences in subtypes seen between the tumors and healthy controls. It is also important to note that although tumor derived CN-LOH is the primary interest here, IBD events might be important for tumor development and act as a first predisposing event towards cancer.
Genetic differences between cell lines
The array profiles, as well as the STR-genotyping, of the three related cell lines SK-N-SH, SH-EP and SH-SY5Y clearly confirmed that these cell lines are highly related to each other, but some distinct features separating these cell lines from each other was also found (Figure 4). It could be argued that these alterations were all present in a mosaic manner in the parental cell line and perhaps also in the original tumor, and that subcloning later has resulted in more purified cell lines. An alternative hypothesis would be that aberrations found to differ between the cell lines are in fact cell line derived events, and thus a product of cell line evolution. The finding of a gain at 2p16-pter, common to SK-N-SH and SH-SY5Y but not SH-EP give some support to the former theory. However, since each of the three cell lines also contain unique aberrations that differ from the others, it seems unlikely that this theory would account for all of the differences seen. Even though the SK-N-SH line might have lost some of its mosaicism over time, thus becoming a more homogeneous cell line, the fact that extra aberrations such as the deletion of 4q26-28.3 is visible only in SK-N-SH, still indicates that at least some of the aberrations are acquired secondary to the subcloning.
The data also show that both SH-SY5Y and SH-EP contain an extra copy of 1 q, although with different breakpoints and different copy number, thus presumably acquired as two separate events. The parental cell line SK-N-SH however, shows no sign of this aberration in the material, which indicates that this is a later event. According to early karyotypic data for these cell lines , SHEP was found to contain an isochromosome 1 q in 35% of the cells, resulting in 4 copies of the 1 q region, while SH-SY5Y presented with a complex duplication of a large 1 q segment, resulting in 3 copies of this region. SK-N-SH however was found to be karyotypically normal for chromosome 1. This corresponds well with the data presented here and explains the observed difference in copy number seen between the cell lines. However, other karyotypic information for these cell lines has also been found in the literature [27–29], resulting in a somewhat complex picture. Although some of these differences can be explained by limitations of the different methods used, for example the lack of resolution in CGH and M-FISH studies or the inability to find balanced translocations using SNP-arrays, there are still some discrepancies between the datasets, suggesting that different clones of these cell lines are used at different laboratories. The fact remains that even relatively stable cell lines change over time and that cell lines keep evolving, thus giving rise to aberrations specific not to the original tumor but rather to the cell culture environment itself. It is important to be aware of the effects that might result from culturing the cells for an extensive amount of passages and to check the cell lines for genetic changes on a regular basis.
Another interesting finding is that karyotypic information in itself sometimes can be misleading. The cell line SK-N-DZ used in this study was recently bought from European Collection of Cell Cultures (ECACC, HPA Culture Collections), where its karyotype is presented as lacking both copies of chromosome 2. However, when this cell line was analyzed with SNP-array it was found that not only does chromosome two exist but a high grade amplification of the MYCN gene was also present. The copy number plot for chromosome 2 (except the MYCN region) was mainly in the diploid range although somewhat fragmented, indicating that the chromosome might be broken into smaller pieces that are translocated to other regions of the genome and thus hard to detect using karyotyping. In fact the American Type Culture Collection (ATCC) describes the karyotype of this cell line as carrying 5 marker chromosomes of unknown origin, which supports this finding. This highlights that high density SNP-arrays can give information that is complementary to classical karyotype data, and that it is important to compare data from different sources in order to derive full information.