Anaplastic oligodendrogliomas with 1p19q codeletion have a proneural gene expression profile

The 1p19q codeletion and EGFR amplification are mutually exclusive and related to dramatically different outcomes in high grade gliomas. The 1p19q codeletion is strongly associated with an oligodendroglial phenotype and favorable prognosis [1]. It has recently been shown to be mediated by a specific t(1;19)(q10;p10) translocation [2]. To date the efforts performed to identify the genes specifically involved in the breakpoint have failed, mostly because both 1p and 19q centromeric regions contain highly repeated sequences. As a consequence the molecular mechanisms underlying the particular phenotype and the favorable outcome of this subset of gliomas remain completely unknown. Reliable detection of 1p19q codeletion requires an appropriate technique, such as CGH-array. Indeed, the most widely used LOH studies may not distinguish this signature from partial distal 1p and 19q deletion or gain, which have radically different prognostic implications [1]. On the other hand, EGFR amplification is tightly associated with chromosome 10 loss and gain of chromosome 7, representing another characteristic genomic signature [3]. EGFR amplification is more frequent in glioblastomas, but it is also found in a subset of anaplastic oligodendrogliomas and, in this setting, is predictive of extremely poor prognosis [4]. Recently, malignant gliomas have been separated into three expression profiles with distinct outcomes and histological correlations: 1) the proneural profile with a better prognosis, mostly corresponding to anaplastic gliomas (oligodendrogliomas and astrocytomas); 2) the proliferative and 3) mesenchymal profiles, corresponding mainly to glioblastomas [5]. However, correlation with 1p19q codeletion is still missing. Based on a set of gliomas analyzed by CGH-array [3], we selected tumors displaying one of these two characteristics and mutually exclusive patterns -1p19q codeletion or EGFR amplification- and compared their gene expression profiles.

EGFR amplification and whole 1p19q codeletion are mutually exclusive and predictive of completely different outcomes [3, 4]. To date, no studies have compared the gene expression profile of these two types of gliomas. Indeed, among microarray studies of gliomas [5, 28, 30, 31], only a few have compared genetically well-defined tumors [32‐35]. In addition, these studies were based on LOH or FISH [32‐34], and not on CGH-array. Yet, there is a need when interpreting a difference in gene expression to analyze it in relation to the genomic profile. Our data reveals clearly distinct gene expression profiles in these 2 groups of gliomas: those with EGFR amplification express the proliferative and mesenchymal gene set defined by Phillips et al., while 1p19q codeleted gliomas express the proneural group [5]. Moreover, gliomas with EGFR amplification clustered close to tumor stem cells. Indeed, the EGFR pathway is involved in the proliferation of normal neural stem cells and cancer stem cells [36]. This result is consistent with the fact that several studies isolated stem like tumor cells from glioblastoma but not from oligodendroglioma. Such studies did not include genetic profiles of the tumors, but data from our group suggest indeed that the capacity of cell renewal (as reflected by the formation of spheroids derived from the tumor) in vitro is tightly correlated with the presence of EGFR amplification (unpublished results). EGFR activation upregulates genes involved in neural stem cell proliferation: one of these genes is ASPM (abnormal spindle-like microcephaly associated) that promotes neuroblast proliferation and symmetric division and is strongly upregulated in glioblastomas. Inhibition of ASPM inhibits glioblastoma cell growth and neural stem cell proliferation [37].

A proneural/normal brain gene expression profile is a factor related to good prognosis and correlates with younger age and grade III histology, with most anaplastic oligodendrogliomas being classified as proneural [5, 28]. As shown here, this gene expression profile can be determined by a simple, highly discriminating RT-PCR test, and this may be useful for clinical practice. Until now, a proneural gene expression profile has not been reported to be associated with 1p19q codeletion. In Freije's study the number of gliomas with 1p19q codeletion was too small (4 out 74 patients) to address this question [28]. In Phillips' study the genomic/transcriptomic correlation was limited to patients with astrocytoma histology, and this may have limited the possibility of finding an association between 1p19q codeletion and the proneural gene expression profile [5]. However, the authors noticed a negative correlation between the proneural gene expression profile and EGFR amplification, similar to the negative correlation between 1p19q codeletion and EGFR amplification [3, 5]. Our study demonstrates that there is a strong correlation between 1p19q codeletion and the expression of proneural genes, suggesting that gliomas with a 1p19q codeletion represent a subgroup of proneural gliomas. In addition, the expression of neuronal genes in 1p19q codeleted tumors is consistent with a previous study showing selective expression of neuronal genes in oligodendrogliomas with 1p loss [33]. Whether there is a link between the good prognosis of proneural gliomas and the fact that gliomas with 1p19q codeletion display a proneural gene expression profile remains to be elucidated. We make the hypothesis that gliomas without 1p19q codeletion but with a gene expression profile similar to the 1p19q codeleted gliomas might also harbor a better prognosis.

The expression of "neuronal genes" in 1p19q codeleted gliomas can be interpreted in different ways. As advocated by some authors, this expression is probably due in part to the presence of infiltrated neurons in the tumor [30]. Indeed, 1p19q codeletion has been suggested to be more frequent in tumors with indistinct, irregular borders, which therefore, are more likely to be contaminated with normal brain tissue [38]. However, as shown here, this normal brain infiltration cannot completely explain the expression of neuronal genes by 1p19q codeleted gliomas. Indeed, these tumors only express a specific subset of neuronal genes (Figure 3). In addition, if the expression of neuronal genes was only due to infiltration of normal brain tissue, the expression pattern of the neuronal genes in these tumors would be similar to their expression in the normal brain samples, which was not the case. Furthermore, we have demonstrated that alpha-internexin (INA), a neuronal protein, was specifically expressed by 1p19q codeleted glioma tumor cells. Thus INA expression might be used as a simple surrogate marker of 1p19q codeletion. This hypothesis is currently being tested in a larger series of gliomas. Interestingly, recent ultrastructural analysis of oligodendrogliomas has shown neuronal structures such as synapses and neurosecretory granules [39]. Thus, another hypothesis for the expression of neuronal genes in 1p19q codeleted glioma tumor cells is that the cell of origin of these tumors could be a progenitor cell giving rise to both neurons and oligodendrocytes [40, 41]. This progenitor has less capacity of self renewal than the more multipotent neural stem cells. This is consistent with the fact that 1p19q codeleted oligodendroglioma fails in our hands to form spheroids in vitro (unpublished data). In this setting it is interesting to note that concomitant overexpression of both BMPs and BMP antagonists, such as the concomitant overexpression of BMP2 and NOG observed in 1p19q codeleted gliomas in our study, has been demonstrated in white matter progenitor cells, which can give rise to both oligodendrocytes and neurons [42]. Another non-exclusive explanation for the expression of "neuronal" genes in oligodendrogliomas could rely on the fact that some genes involved in neurogenesis and classified as "neuronal" may also play a role in oligodendroglial development, e.g. ASCL1/MASH1. This proneural gene specifies a population of telencephalic oligodendrocytes [43] and is also required for oligodendrocyte development in the spinal cord [44]. On the other hand, Olig2 -implicated in oligodendroglial specification- is also involved in neurogenesis: during development, Olig2+ progenitors give rise to both motoneurons and oligodendrocytes in the ventral spinal cord, [45]. Consistently with our results, these data, illustrating the tight connection that exists between neurons and oligodendrocytes fates, bring a new light on the pathogenesis of oligodendrogliomas with 1p19q codeletion. Finally it is important to remember that current WHO classification is only based on morphological similarity between normal cells and tumor cells, and the link between oligodendrocytes and oligodendrogliomas has never been demonstrated.