We demonstrate here that histologically similar tumors with a proposed different genetic etiology can be differentiated based upon their genomic profiles using array CGH as an effective method to generate informative profiles in a genome-wide manner. Previously, large-scale gene expression profiling using RNA extracted from tumor tissue has demonstrated its utility in classifying different tumor subtypes of leukemia and lymphoma [33, 34] and solid tumors [35, 36]. It has also been applied to study the molecular signatures of meningiomas of different grades and locations [37–41]. Notably, in a recent study, gene expression patterns were shown to be predictive of major cytogenetic patterns seen in meningiomas (chromosome 22 or 22q loss, loss of 1p36 alone, complex karyotypes with loss of 1p36 and/or 14q and diploid) based upon altered expression of a combination of genes within the regions of altered dosage and genes outside these regions . Each individual gene expression change represented an apparent correlate of the corresponding cytogenetic abnormality that could potentially play a role in meningioma initiation or progression. In particular, those altered genes not actually located in a region of cytogenetic abnormality may well represent components of pathways critical to meningioma formation. While it is difficult to assign such a role to each individual gene without additional studies, the overall pattern observed was clearly associated with patient outcome, though not with tumor histopathology.
Like gene expression profiling, whole genome profiling by array CGH using DNA extracted from tumor tissue has also shown its power to differentiate solid tumors of both mouse and human [27, 43–45]. Solid tumors often exhibit genomic imbalance events and some of these imbalance events are specific to certain types or subtypes of tumors. In view of the stable nature of DNA compared with RNA, array CGH is a potentially more widely applicable, easy and effective method for histo-molecular classification. In addition, array CGH provides information about genome stability of the tumor, which may be critical information for assessing its biological state and pathogenic potential.
Our array CGH analysis of sporadic solitary meningiomas detected a wide variety of genomic events, largely consistent with those which have been detected previously by a variety of cytogenetic and molecular techniques, and have often been associated with tumor progression [32, 46–52]. Notably, through unsupervised cluster analysis using our entire genome array CGH dataset, we observed a clean separation of meningiomas with or without chromosome 22 deletion for both solitary and multiple meningiomas. Chromosome 22 deletion is a frequent mechanism for somatic inactivation of the NF2 gene, thought to account for ~60% of meningiomas . The initiating event(s) in meningiomas that retain chromosome 22 has not yet been identified. The distinction between the two groups of meningiomas by array CGH cluster analysis apparently reflects the existence of at least two main types of genetic mechanism for meningioma tumorigenesis. Since no consistent imbalance event predominates among tumors without chromosome 22 deletion, there may be a greater genetic diversity in the initiation of these tumors. The lower number of ICS in tumors that retain chromosome 22 is consistent with the lack of genomic events detected previously using multi-allele marker techniques .
Our data also revealed segregation, based upon a variety of genomic events, of tumors that lost chromosome 22 into two major groups, one consisting predominantly of benign tumors and the other predominantly of tumors of advanced histological grade (Figure 3). The cluster analysis therefore illustrates some potential utility and power of array CGH profiling as an objective means for tumor subclassification and subgrouping. Due to the existence of genetic heterogeneity among meningiomas and the nature of histological examination, we did not expect all tumors to be assigned by cluster analysis precisely in concert with their pathological grade and indeed, that was the case. However, the tendency, with some exceptions, of Grade I tumors to cluster separately from higher grade tumors suggests that further refinement of the genomic profiling strategy is warranted, in combination with more extensive molecular analysis to determine whether differences are due to misclassification by array CGH or rather represent tumors whose molecular changes and predicted behavior are not efficiently recognized in the standard histological examination.
Whole genome array CGH analysis revealed a significant distinction between sporadic meningiomas with or without chromosome 22 deletion. Advanced grade tumors with chromosome 22 deletion had significantly more imbalance events in the rest of the genome than those without, suggesting not only a different genetic etiology, but also a different pathway for tumor progression. This distinction also raises the possibility that prior chromosome 22 loss itself may contribute directly to subsequent genomic instability. Therefore, the array CGH analysis points to the need to unravel potential alternative genetic pathways for a complete understanding of meningioma tumorigenesis. The cDNA array used here did not provide a sufficiently dense coverage to identify small regions of loss that could target specific suppressor genes and we detected no cases of homozygous deletion. There may be potential clues to alternative pathways revealed by the amplification events readily detected in some tumors, though these also typically covered relatively large segments and occurred primarily in malignant meningiomas. For example, meningioma 126M displayed amplification of an ~500 kb region (1.9–2.4 Mb) on chromosome 11 spanning, among other genes, the loci for insulin (INS) and insulin-like growth factor 2 (IGF2), which has been implicated previously in some meningiomas through expression studies . However, the same tumor also displays larger amplifications on chromosome 1 (78.6–87.5 Mb) and chromosome 17 (41.8–64.0 Mb), both of which contain numerous genes that could potentially play a role in meningioma growth. Similarly, tumor 128M displayed a focal amplification of the c-myc oncogene (MYC) on chromosome 8, whose expression may correlate with proliferative index and help predict recurrence of meningiomas [54, 55], but the same tumor also had a high level gain on chromosome 17 (49.2–64.0 Mb), overlapping with the large region amplified in 126M. When compared with all tumors that showed a gain on chromosome 17, the minimum shared segment is 58.9–64 Mb and spans more than 40 genes, including such known growth control-related genes as the growth hormone-chorionic somatomammotropin cluster and a protein kinase C family member (PKCA) among others. One benign meningioma (63B) had a focal amplification in a gene-rich segment on chromosome 12 (56.19–56.45 Mb), which includes CDK4, a known regulator of G1-S progression that phosphorylates the retinoblastoma gene product, Rb, along with several other potential candidates. These disparate events support the possibility that different mechanisms of progression may occur in different tumors but their large size, inconsistent occurrence and lack of a unified predictive pattern suggests that an integration of detailed genomic, genetic, gene expression and functional studies will be needed to delineate the array of alternative pathways to meningioma growth in vivo.
Most importantly in our study, array CGH analysis revealed strikingly different genomic profiles between familial multiple meningioma and sporadic multiple meningioma. Though this represents only a small number of tumors and even a single case of loss in one familial multiple meningioma of chromosome 22 or of one of the other chromosomal regions altered in the sporadic multiple meningiomas would reduce the apparent distinction, the absence of all genetic rearrangements in the familial multiple meningiomas is noteworthy. For example, the likelihood that all 4 familial multiple tumors retained chromosome 22 (based upon its overall rate of retention in meningiomas) is ~0.04. Combined with the absence of any other genomic rearrangement in all 4 such tumors, the retention of chromosome 22 provides further evidence that familial multiple meningioma is a genetically distinct tumor predisposition syndrome. Given that NF2 mutation was not seen and that chromosome 22 deletion is not a shared feature in these tumors, we believe that one or more other tumor suppressor genes or oncogenes is responsible for the tumor initiation of familial multiple meningioma. Although no genomic imbalance event was identified, even with the higher resolution oligonucleotide array, it is conceivable that loss of heterozygosity without copy number change may be uncovered using high density single nucleotide polymorphisms (SNP) arrays. The finding that sporadic multiple meningiomas harbor a similar level of genomic imbalance events to atypical solitary meningiomas should also draw attention to the potential for greater subarachnoid spreading or higher aggressiveness that would produce a worse prognosis than familial multiple meningiomas. Further, as sporadic multiple meningioma and familial multiple meningioma cannot be distinguished histopathologically, array CGH can provide an important means of differential diagnosis.