Array‐Based Genomics in Glioma Research

Abstract

Over the years, several relevant biomarkers with a potential clinical interest have been identified in gliomas using various techniques, such as karyotype, microsatellite analysis, fluorescent in situ hybridization and chromosome comparative genomic hybridization. Despite their pivotal contribution to our understanding of gliomas biology, clinical application of these approaches has been limited by technological and clinical complexities. In contrast, genomic arrays (array‐based comparative genomic hybridization and single nucleotide polymorphisms array) have emerged as promising technologies for clinical use in the setting of gliomas. Indeed, their feasibility and reliability have been rigorously assessed in gliomas and are discussed in this review. The well‐known genomic biomarkers in gliomas are in fact readily and reliably identified using genomic arrays. Moreover, it detects a multitude of new cryptic genomic markers, with potential biological and/or clinical significances. The main studies dedicated to genomic characterization of gliomas using genomic arrays are reviewed here. Interestingly, several recurrent genomic signatures have been reported by different teams, suggesting the validity of these genomic patterns. In light of this, genomic arrays are relatively simple and cost‐effective techniques whose implementation in molecular diagnostic laboratories should be encouraged as a valuable clinical tool for management of glioma patients.

INTRODUCTION

Glial tumors are the most common primary cerebral tumors in adults (13). The pathological classification of glial tumors edited by the World Health Organization (WHO) distinguishes astrocytomas, oligodendrogliomas, and oligoastrocytomas or mixed gliomas based on the tumor cell phenotype (32). Pilocytic (WHO grade I), diffuse (WHO grade II), anaplastic (WHO grade III) and glioblastoma (GBM) (WHO grade IV) are the main different subtypes of astrocytomas. Oligodendrogliomas and oligoastrocytomas or mixed gliomas are both classified into two subtypes (WHO grade II or low grade and WHO grade III, anaplastic or high grade). While this system serves well to predict the general behavior of most patient tumors, a significant number of tumors are extremely difficult to classify using a histopathologic approach alone, leading to a certain degree of intra‐ and interobserver variability 5, 10, 29, 32, 34, 41.

Over the last decades, nonrandom genomic alterations in gliomas have been identified as potentially interesting biomarkers associated with tumor class, tumor grade of malignancy, patients survival and tumor response to treatments. Several techniques have been used to identify and/or to test such aberrations in the research setting, including microsatellite analysis (LOH), fluorescent in situ hybridization (FISH), spectral karyotyping (SKY), chromosome comparative genomic hybridization (cCGH) and multiplex ligation‐dependent probe amplification (MLPA) 4, 17, 46, 48, 49.

As an alternative, genomic arrays, including array‐based comparative genomic hybridization (aCGH) and single nucleotide polymorphism (SNP) array, represent an emerging technology (50). In this review, we discuss the discoveries obtained in gliomas using genomic arrays to date, the potential of these findings to complement current pathologic diagnosis, and the potential for transfer of this technology from the bench to the bedside in the clinical care of glioma patients.

The genomic signatures reported in astrocytic and oligodendroglial tumors are reviewed below, as well as the studies dedicated to genomic arrays analysis of candidates' genomic regions.

ASTROCYTIC TUMORS

Diffuse astrocytoma (WHO grade II astrocytoma)

Using a customized 518 b, acterial artificial chromosome (BAC)/ P1‐derived artificial chromosome (PAC) aCGH, Roerig et al have investigated a series of gliomas with various histology, including 11 cases of diffuse astrocytomas (42). The aCGH included: (i) 308 BAC/PAC clones containing genomic regions classically imbalanced in cancers and gliomas; and (ii) 210 reference clones plotted across the genome every 15 Mb. Diffuse astrocytomas generally showed few genomic aberrations as compared with high‐grade astrocytomas. However, several recurrent genomic imbalances were consistently observed. The most highly recurrent genomic abnormality was chromosome 7q gain observed in 55% of cases. Chromosome 3q gain, 1p loss, 4q loss, 6q loss and 10q loss were reported in 45% of cases. Chromosome 4q gain, 8q gain, 11q gain, 2q loss, 12p loss, 12q loss, 14q loss, 17q loss and 19q loss were less frequent and reported in 36% of cases. Interestingly, none of the diffuse astrocytomas exhibited amplification of EGFR or any other locus tested by the array (42).

Similarly, in our personal series of 10 diffuse astrocytomas investigated using a 1 Mb resolution BAC‐aCGH, the most frequent genomic imbalance was chromosome 7q gain (50%). Chromosome 19p gain was observed in 40% of cases, and chromosomes 7p gain, 20p gain, 9p gain, 16p loss, 16q loss and 22q loss were observed in about 1/3 of cases. EGFR amplification was seen in only one case (21).

Taken together, these results are in agreement and show that the most frequent genomic aberration in diffuse astrocytomas is chromosome 7q gain. The others genomic imbalances are not consistently reported across each study and seem to be less recurrent in diffuse astrocytomas.

Anaplastic astrocytoma (WHO grade III astrocytoma)

In his series of gliomas, Roerig et al also studied 10 WHO grade III astrocytomas (AIII) (42). Surprisingly, chromosome 15q loss was found as the most frequent genomic imbalance (60%). Gain of chromosomes 3q, 7p, 10q, 11q, 17q and 20p and losses of chromosomes 1p, 3q, 9p and 10q were observed in 50% of cases. Chromosome 7q gain, 10p gain, 17p gain, 6q loss, 7p loss, 13q loss, 14q loss and 22q loss were observed in 40% of cases and EGFR amplification in 2 out of 10 tumors.

In our smaller personal series of five anaplastic astrocytomas, chromosome 19p gain, 19q gain, 20q gain, 9q loss and 10p loss were the most recurrent chromosome disequilibrium seen in 3/5 cases (21). EGFR amplification was observed in 1/5 case. In contrast to Roerig et al findings, chromosome 15q loss was observed in only 1/5 case (42).

Bredel et al, have screened a series of 54 gliomas of various pathological diagnosis using a 42 K cDNA based aCGH (3). In this series of gliomas, three tumors were diagnosed as anaplastic astrocytomas. Two out of the three tumors were classified into the group of gliomas harboring chromosome 7 gain, chromosome 10 loss and preferential EGFR amplification.

To summarize, a robust genomic signature of WHO grade III astrocytomas has not been identified yet by genomic arrays. This may be related to the small number of anaplastic astrocytomas (n = 18) analyzed in these three series. However, chromosome 10q loss seems to be a prominent finding in anaplastic astrocytomas. Further studies are warranted to better identify molecular portrait of these tumors.

Primary GBMs

GBM is the most frequent glioma in adults, and several studies have investigated this subtype using aCGH.

Misra et al used an aCGH containing ∼2250 BACs to screen copy number abnormalities (CNA) in a series of 50 primary WHO grade IV astrocytomas. Three major genomic patterns were revealed by Misra et al in this series of GBMs: (i) gain of chromosome 7 and loss of chromosome 10; (ii) loss of chromosome 10 without gain of chromosome 7; and (iii) intact chromosomes 7 and 10. The subgroups 1, 2 and 3 included 18, 9, and 23 tumors respectively. The authors did not observe any significant difference in terms of overall survival between the three genomic signatures (35).

Similar signatures were reported by Korshunov et al, despite a lower resolution aCGH. Indeed, the authors analyzed ∼300 P1/BAC targets statuses in 70 frozen newly diagnosed GBMs using the Genosensor CGH array (Vysis, Downers Grove, IL, USA). Based on an analysis of the entire set of clones, the authors identified two genomic signatures in the series of GBMs: (i) 25 tumors exhibiting mainly whole chromosome 7 gain and/or whole chromosome 10 loss; and (ii) 45 GBMs with chromosome arms 12q, 15q, 19p, 19q and 22q gains and chromosome arms 11q and 17q deletions. No clinicopathological difference was observed between these two genomic groups. Then, a set of 46 CNA was shown to be predictive of prognosis by Korshunov et al. According to this set of genes, two subgroups of GBMs were distinguished in a genetic point of view: (i) the subgroup 1 including 56 tumors; and (ii) the subgroup 2, with 14 tumors. There was a statistically significant difference in terms of median survival between both groups (13.1 months in the subgroup 1 vs. 34.5 months in the subgroup 2). The 46 clones with clinical value were mainly located in chromosomes 11, 3, 1, 9 and 17. The genetic signature associated with poor prognosis, including 26/46 clones, was shown to be also an independent prognostic factor (27). Despite this dramatic and promising statistical difference, this genomic signature has not been evaluated in an independent set of GBMs and requires confirmation in a new series using the same clones. In our hands, possibly due to the fact that the BACs were not exactly the same ones, analysis of a series of 32 patients do not confirm this signature (21).

Nigro et al focused their attention on GBMs with “favorable prognosis” (still alive > 24 months after the initial pathological diagnosis of GBM) (37). Thus, they studied, using a ∼1 Mb aCGH, a series of 34 GBMs including 10 samples from GBM patients with long survival. Two genomic groups of GBMs were identified: (i) the type 1 characterized by chromosome 7 gain and/or chromosome 10 loss; and (ii) the type 2 exhibited any major CNA on chromosome 7 and chromosome 10. The proportion of GBM patients with favorable prognosis was significantly higher in group 2 (9/18) vs. group 1 (1/16)(37). Thus, the type 2 genomic signature seems to predict a better outcome in GBM. The results obtained by Nigro et al have been included as a part of the molecular study conducted by Philips et al (40). Moreover, Philips et al analyzed additional samples using aCGH as described by Misra et al (35). Although this study was mainly focused on gene expression of high‐grade gliomas, the authors have reported recurrent chromosome 7 gain and chromosome 10 loss in grade IV tumors.

The results obtained by Roerig et al in a series of 19 primary GBMs were in accordance with the genomic pattern reported in the previous studies. Indeed, the most frequent genomic abnormalities were chromosome arm 7q gain (84%), 7q gain (84%), 10q loss (79%), 10p loss (68%), 1p loss (53%), 9p loss (53%), and 12q gain (53%). The most frequent high level amplifications targeted PDGFRA/KIT (21%), EGFR (32%) and CDK4/SAS (16%) (42). We reported similar results in our series of 32 primary GBMs; the most frequent CNA were chromosome arm 7p gain (81%), 10q loss (81%), 7q gain (75%), 10p loss (75%), 19p gain (69%), 19q gain (66%), 20p (50%), 20q gain (50%) and 9p loss (47%). EGFR amplification was observed in ∼2/3 of cases (21).

Maher et al reported a genomic pattern termed K3‐1 and observed in 18/20 primary GBMs. This pattern was characterized on the one hand by gains of whole chromosomes 7, 19 and 20, and partial gain of chromosome 12, and on the other hand, by whole chromosome 10 and partial chromosomes 9 and 11 deletions (33).

Kotliarov et al have investigated a large series of 82 GBMs using the Genechip Human Mapping 100 K SNP array from Affymetrix® (Affymetrix, Santa Clara, CA, USA). In agreement with previous works and in addition to other results, CDKN2A homozygous deletion, LOH on chromosome arm 9p, LOH on chromosome 10 and EGFR high‐level amplification were found the most frequently genomic alterations in GBMs (28). Addressing additional questions, Beroukhim et al have analyzed a series of 107 primary GBMs using a 100 K SNP array similar to the one used by Kotliarov et al 2, 28. In the line of previous studies, the authors showed that the most frequent genomic abnormalities in primary GBMs were chromosome 7 gain with EGFR high‐level amplification, chromosome arm 9p loss with homozygous deletion of CDKN2A and chromosome 10 loss (2).

Finally, Wong et al have screened, using an Affymetrix® 10 K SNP array, a series of 14 pediatric high‐grade gliomas including 13 GBMs. They identified chromosome arms 13q, 14q and 4q LOH in more than 50% of cases. In addition, EGFR high amplification was detected in 2/13 pediatric GBMs (54).

All these studies support a robust consensus genomic signature of primary GBM in adults. Indeed, chromosome 7 gain with or without EGFR amplification, chromosome arm 9p loss with or without CDKN2A homozygous deletion, and chromosome 10 loss are the most recurrent genomic imbalances observed in primary GBMs. GBMs without these genomic abnormalities, particularly intact 7 and 10, survive longer 21, 27, 33, 35, 37 (Figure 1A and B, respectively).

Figure 1.

Genomic profiling of tumors combining chromosome 7 gain and chromosome 10 loss with EGFR high level amplification (A) or without EGFR high level amplification (B). The x axis represents the genome from 1p telomere to Yq telomere. Each dot is a Bacterial Artificial Chromosome (BAC). For each BAC, the color code red, green and blue indicates gain, loss and normal status. Vertical bar indicates high‐level amplification.

Secondary GBMs

Secondary GBMs are classically defined as WHO grade IV astrocytoma developed through progression of a WHO grade II or grade III glioma.

Secondary GBMs have been investigated in details using aCGH by Maher et al (33). In this study, 17 secondary GBMs were analyzed using a highly resolutive oligonucleotide‐based aCGH (∼54.8 kb). Two distinct genomic subgroups were identified in the group of secondary GBMs, which appears heterogeneous: (i) the subgroup termed K3‐2; and (ii) the subgroup called K3‐3. The K3‐2 secondary GBMs presented chromosomes 4, 8 and 12 large gains and chromosomes 7 and 11 thin gains. In contrast, the K3‐3 secondary GBMs exhibited mainly chromosome losses, notably affecting chromosomes 6, 9, 10, 13, 18 and 19.

In his series, Roerig et al investigated also a series of 10 secondary GBMs (42). The most frequent losses involved 9p and 18q (60%). Losses of chromosomes 13q, 16p, 17p, 19q, 1p, 4q, 6q, 10q, 11q, 12q and 16q observed in more than 40% were also frequent. Chromosome gains were less frequent and involved chromosomes 4q, 7p, 7q, 8q, 10p and 17q (20%). In light of Maher's results, the series of Roerig seems to be enriched with type K3‐2 secondary GBMs. Our series including only four secondary GBMs revealed chromosomes 1q gain, 5p gain, 7p gain, 7q gain, 8q gain, 11p loss, 12q loss, 14q loss, 19p gain, 19q gain and 20q in 2/4 cases (21). Bredel et al analyzed three secondary GBMs and did not disclose recurrent genomic signature (3). A part of the study of Beroukhim et al was dedicated to a series of 15 secondary GBMs and showed, in agreement with previous studies, that chromosome 7 gain without EGFR high‐level amplification occur more frequently in secondary GBMs compared with primary GBMs (2).

Thus, two main genomic groups of secondary GBMs are emerging (33) : (i) those driven by chromosome gains; and (ii) those driven by chromosome losses. The clinical significance of these genomic patterns was investigated by Maher et al. The authors noted a shorter time to progression in the group driven by losses (42 months) in comparison with the group driven by gains (100 months).

Pilocytic astrocytoma (WHO grade I astrocytoma)

Few is known about genetic abnormalities‐driven oncogenesis of pilocytic astrocytomas (WHO grade I astrocytomas). Jones at al have screened a series of 44 pilocytic astrocytomas (including 12 tumors from patients aged above 15 years old) using a ∼1 Mb resolution aCGH (24). They did not find any CNA in ∼2/3 of cases, suggesting that the genome of pilocytic astrocytoma is poorly imbalanced in the majority of cases. Wong et al obtained similar results in a series of six pilocytic astrocytomas (part of a larger series of pediatric brain tumors) investigated using a 10 K SNP array from Affymetrix®(54). Interestingly, Jones et al showed that besides the group of “balanced” tumors, ∼1/3 of pilocytic astrocytomas exhibited gains involving particularly the entire chromosomes 5 and 7. Pilocytic astrocytomas from patients aged above 15 years old exhibited more frequently chromosome gains and chromosomes imbalanced in comparison with pilocytic astrocytoma observed in younger patients (24). Three pilocytic astrocytomas were also studied by Bredel et al and classified into the genomic group of gliomas harboring partial gain of chromosome 7 gain and additional genetic alterations (3).

Deshmukh et al have studied a series of 10 cerebellar pilocytic astrocytomas using both 385 K high‐resolution commercial oligonucleotide‐based aCGH from NimbleGen® (Roche NimbleGen, Madison, Wisconsin, USA) and 500 K SNP array from Affymetrix®. An amplification and an overexpression of HIPK2 on chromosome region 7q34 was observed in 60% of cases, suggesting the potential role of this gene in pilocytic astrocytomas oncogenesis (8). More recently, three independent studies conducted by Pfister et al, Bar et al and Jones et al have investigated more than 100 pilocytic astrocytomas and have showed that the critical gene of the 7q34 locus in pilocytic astrocytomas is BRAF 1, 25, 39. These three last studies showing duplication/gain of BRAF have used BAC or oligonucleotides‐based aCGH to find these genomic abnormality occurring in most of pilocytic astrocytomas. In addition, and, interestingly, Jones et al revealed that BRAF domain kinase is rearranged into a chimeric gene with KIAA1549 in the majority of pilocytic astrocytomas (25).

Pleomorphic xanthoastrocytoma (WHO grade II astrocytoma)

Pleomorphic xanthoastrocytomas is a rare entity whose genetics is poorly known. Interestingly, Weber et al detected a CDKN2A/p14ARF/CDKN2B homozygous deletion in 6/10 pleomorphic xanthoastrocytomas using a <0.5 Mb resolution aCGH precising the genetic background of this tumor group (53). However, involvement of this locus is also observed in other subtypes of gliomas and is not specific to this pathological entity. Therefore, additional studies are warranted to identify putative, more specific genetic abnormalities of this pathological subgroup.

OLIGODENDROGLIAL TUMORS

Low‐grade oligodendroglial tumors (WHO grade II oligodendroglioma and oligoastrocytoma)

Kitange et al conducted the first study dedicated to whole genome characterization of oligodendroglial tumors using BAC aCGH (26). The population included 21 WHO grade II oligodendroglial tumors (20 oligodendrogliomas and 1 mixed glioma). The most frequent chromosome aberrations identified were losses of chromosome arms 1p and 19q both in approximatively one‐third of cases. Chromosomes arms 1p/19q co‐deletion was observed in 29% of cases. Additional chromosome disequilibrium such as partial or complete losses of chromosome arms 4q, 9p and 13q, and gains of chromosome arms 7q, 7p, 8q and 11q were reported in at least 10% of tumors.

Rossi et al, have studied a series of 15 low‐grade (WHO grade II) oligodendrogliomas using a ∼0.4 Mb BAC aCGH. Deletion of chromosome arm 1p, chromosome arm 19q and both chromosome arms 1p and 19q were seen in 73%, 87% and 73% tumors, respectively. Less frequently, losses of chromosomes arms 11q, 13q, 4p and 11p were detected in low‐grade oligodendrogliomas. Finally, gain of chromosome 7 was observed in 40% tumors (43).

In our series of 31 WHO grade II oligodendroglial tumors (28 oligodendrogliomas and 3 oligoastrocytomas), chromosomes 1p/19q co‐deletion signature was observed in 21/31 tumors (68%) (21). This result is very similar to that reported by Rossi et al (43). In addition, chromosomes 4p loss, 11p loss, 7q gain and 7p gain were observed in 10 (32%), 8 (26%), 10 (32%) and 9 (29%) analyzed tumors (21). Roerig et al reported in a series of 10 WHO grade II oligodendroglial tumors, chromosomes 1p and 19q losses in eight cases (80%) and gain of chromosome 6q in three tumors (42). Similarly, Bredel et al observed chromosome 1p/19q co‐deletion in 4/5 WHO II oligodendroglial tumors (80%) (3).

Chromosome 1p/19q co‐deletion was thus confirmed as a strong marker associated with low‐grade oligodendroglial tumors. Interestingly, the whole arm 1p deletion combined with the whole arm 19q deletion have been shown recently to be the result of a translocation (Figure 2A) 11, 22. The aCGH technology provides a useful solution in that it allows detection of whole arm deletion of chromosome 1p and 19q and the centromeric breakpoints. This genomic combination suggests strongly the translocation.

Figure 2.

Genomic profiling of chromosome arms 1p and 19q co‐deleted tumors. A. aCGH indicates that the chromosomes 1p/19q co‐deletion pattern is “pure,” without any significant additional chromosome imbalance. B. In addition to the chromosomes 1p and 19q deletions, the aCGH identifies whole chromosome 4 loss. The legend is the same as the one on Figure 1.

Several questions remain to be solved in low‐grade oligodendroglial tumors. Indeed, the clinical significance of chromosome 7 gain, which is the second most frequent chromosome abnormality, is not elucidated. Moreover, few WHO grade II oligoastrocytomas have been investigated in depth, and their genomic pattern was not clearly compared with pure WHO grade II oligodendroglioma in order to identify putative new markers of mixed gliomas.

High‐grade oligodendroglial tumors (WHO grade III oligodendroglioma and oligoastrocytoma)

In a study dedicated to oligodendroglial tumors, Kitange et al analyzed 14 high‐grade oligodendroglial tumors (11 oligodendrogliomas and 3 oligoastrocytomas). The most frequent chromosome aberrations were loss of 1p (64% of tumors) and 19q (57% of tumors). Chromosome arms 1p/19q co‐deletion was seen in half of the tumors. Losses of chromosomes 4q, 9p and 13q were also reported in six, four and seven cases, respectively (26).

In our series of 30 high‐grade oligodendroglial tumors (21 oligodendrogliomas and 9 oligoastrocytomas), chromosomes 1p/19q co‐deletion signature was found with a similar frequency, 14/30 tumors (47%). Gain involving chromosome 7q, 7p and 19p occurred in 47% of cases. Chromosomes 9p and 10q losses were seen in 13 and 11 cases, respectively (21). More recently, we addressed specifically the question of the frequency of high‐level genetic amplifications, whatever the gene targeted and their prognostic value in a series of 52 WHO grade III oligodendrogliomas. Patients with gene‐amplified tumors (14/52) had worse prognosis than patients with non gene‐amplified tumors (38/52) (19). Roerig et al analyzed 10 WHO grade III oligodendrogliomas. These tumors exhibited mainly losses of chromosomes 1p (nine cases) and 19q (seven cases) and gain of chromosome 3p (five cases) (42). EGFR amplification was observed in one case. Chromosome 1p/19q co‐deletion was observed in 4/10 WHO grade III oligodendroglial tumors studied by Bredel et al (3).

While their study was not precisely dedicated to characterize CNA in oligodendroglial tumors, but rather to assess the usefulness of aCGH, Cowell et al analyzed first a series of 14 oligodendroglial tumors using a ∼0.5 Mb CGH. The pathological diagnosis was low‐grade oligodendroglioma in six cases, WHO grade III oligodendroglioma in five cases and oligoastrocytoma in three cases. The authors detected chromosomes 1p/19q co‐deletion in 57% of tumors (6).

Thus, in high‐grade oligodendroglial tumors, there is agreement that chromosome 1p/19q co‐deletion is also the most important marker observed in around 50% of cases. aCGH would contribute to differentiate WHO grade II and WHO grade III oligodendroglial tumors. Based on the literature data, the alterations most frequently associated with a grade III subtype seem to chromosome 19q gain, 22q loss, 10q loss and 9p loss.

The prognostic interest of these biomarkers has not been studied. While the chromosome arm 1p/19q co‐deletion has been found to be a favorable prognostic factor in oligodendroglial tumors, the clinical significance of additional alterations associated to this co‐deletion has not been yet clarified and need further investigations (Figure 2B). Of interest is the suggestion that chromosome 8q gain or the presence of an amplicon in the oligodendrogliomas population could be indicative of a poor outcome 19, 26.

Role of CGH‐a to investigate putative candidate genomic regions

aCGH clearly participates to precise genomic signatures of gliomas. The clinico‐biological significance of some genetic patterns remains unclear (Figure 3). Besides this whole genome screen, aCGH was also used to study putative candidate genomic regions.

Figure 3.

Other genomic profiling of gliomas different from those reported in 1, 2. The legend is the same as the one on Figure 1.

We have analyzed chromosome arm 1p status in a series of 108 diffuse gliomas (of various histologies) searching for the putative tumor suppressor gene located on chromosome arm 1p and associated with a favorable prognostic factor in glial tumors using a ∼1 Mb aCGH. On the one hand, whole chromosome arm 1p loss was associated highly with chromosome arm 19q loss, oligodendroglial phenotype and favorable prognosis. On the other hand, segmental chromosome 1p deletion predicting short survival was not associated with chromosome arm 19q loss and oligodendroglial phenotype (Figure 4, Idbaih A et al (2005); Reprinted with permission of John Wiley & Sons, Inc.). Thus, FISH and LOH assessment of chromosome 1p36 region status does not allow distinction between the “favorable” and “unfavorable” 1p loss in gliomas 18, 20. Ichimura et al have shown, in a series of 108 gliomas and using a chromosome 1 dedicated genomic array, that focal chromosome arm 1p loss encompassing 1p36 region was recurrent in malignant astrocytomas (close to 1/3 of cases) and should be distinguished from whole 1p deletion classically observed in oligodendrogliomas (16).

Figure 4.

Prognosis of gliomas patients according to the chromosome arm 1p status of their tumor (n = 83). For each tumor, the extent of the chromosome arm 1p deletion (if exists) is drawn as a vertical line on the left of the chromosome arm 1p. A. Total loss of chromosome arm 1p arm (dotted line in Figure 4D). B. Segmental chromosome arm 1p loss (broken line in Figure 4D) C. No chromosome arm 1p deletion (continuous line in Figure 4D) D. Kaplan–Meier overall survival curves of the three groups of patients (P < 0.0001). Idbaih A et al (2005); Reprinted with permission of John Wiley & Sons, Inc.

Chromosome 6 loss is a recurrent CNA in astrocytomas. Using a homemade chromosome 6 aCGH array covering almost all chromosome 6 known coding regions, Ichimura et al screened a series 104 astrocytomas (WHO grades II to IV) from adult patients (15). The authors precised the high frequency of chromosome arm 6q (and at a lesser extent chromosome arm 6p) loss in high‐grade astrocytomas in 16/30 WHO grade III astrocytomas and in 35/64 GBMs. Moreover, they discovered two minimal common regions of loss containing several candidates genes putatively involved in astrocytoma oncogenesis, such as PACRG, QK1 and ARID1B (15).

Similarly to chromosome 6 loss, chromosome 22 loss is also a recurrent chromosome imbalance in gliomas (47). Therefore, Diaz de Stahl et al analyzed 50 WHO grade IV astrocytomas using a highly resolutive chromosome 22 aCGH. The most frequent abnormality was the loss of one out the two chromosomes 22 detected in 20% of cases. The authors highlighted also two regions of chromosome loss and two regions of genomic gain, including putative tumor suppressor genes and putative oncogenes in GBM biology (9).

Rossi et al used a ∼0.5 Mb BAC aCGH in order to dissect chromosome arm 7p high‐level amplifications in GBMs. The authors identified in a series of 37 GBMs: (i) new amplified genomic regions on chromosome arm 7p, which do not include EGFR loci; and (ii) new amplified/ highly expressed genes on chromosome arm 7p, such as LANCL2 and SEC61G, which might participate to gliomagenesis as EGFR amplification/high expression (44).

Similarly, Ruano et al studied a series of 20 primary GBMs using a ∼30 000 probes cDNA array and pinpointed new amplified/ highly expressed candidate oncogenes in glioma biology, such as STIM2, TNFSF13B and COL4A2, notably located on chromosome arm 4p, 13q and 13q, respectively (45).

Thus, in addition to identification of genomic signatures of gliomas, aCGH should facilitate the identification of new putative genes involved in gliomagenesis.

DISCUSSION

Currently, the diagnostic criteria used for classification of gliomas are based solely on morphological features of tumors (32). However, in some subgroups of gliomas, this classification is challenging and results in low reproducibility for a significant percentage of tumors 12, 34, 37, 39, 46. It is now well established that even consensus histological diagnostic categories may consist of patients with different profiles of molecular alterations associated with radically different prognostic value. It is now clear that the analysis of genomic biomarkers is rapidly becoming part of the routine diagnostic work‐up of gliomas and will become an essential complement to histopathologic diagnosis. However, the optimal techniques to implement in departments of pathology are debated.

The first study dedicated to genomic characterization of gliomas using aCGH was reported by Hui et al (14). In this study, 14 GBMs were investigated using BAC/PAC/P1 aCGH containing 58 oncogenes. The authors reported high‐level amplifications or gains of several genes, in particular, CDK4, GLI, MYCN, MYC, MDM2, PDGFRA, PIK3CA, EGFR, CSE1L and NRAS in different cases. In addition to the biological results, this first study demonstrated the feasibility of aCGH for detection of several genomic imbalances in one‐time experiment, thereby obviating the need for multiple experimental procedures.

The feasibility and utility of aCGH were soon further confirmed by Cowell et al, who rigorously demonstrated the robustness and reliability of aCGH for CNA screening in gliomas (7). Indeed, they compared CNA detected on the one hand by SKY, and on the other hand, by a 6000 BACs aCGH, in four recently established glioma cell lines. The authors demonstrated that aCGH is able to detect genomic abnormalities seen by SKY, but also thinner chromosome imbalances undetectable by SKY. This methodological study concluded that aCGH is reliable compared with classical cytogenetic technique and more resolutive than cCGH (7). In addition, aCGH is certainly less time consuming and labor intensive than SKY, cCGH and metaphasic FISH mainly because it does not require tumor and/or normal cells short‐term culture and chromosome metaphases.

In the two previous studies, aCGH was conducted using high‐quality DNA extracted from fresh/frozen tissue. However, analysis of DNA extracted from routine clinical pathology samples, which are formalin fixed and paraffin embedded (FFPE) using aCGH, is more challenging. Given the great advantages of this technique, interesting studies have investigated the feasibility of screening DNA from FFPE tissue using aCGH. Johnson et al have characterized a series of 15 FFPE astrocytomas and their 15 paired frozen tumors using a homemade customized aCGH testing ∼3000 clones in order to identify DNA criteria predictive of a reliable aCGH experiment from FFPE samples (23). According to the authors, the total DNA extracted from FFPE tumor should meet the following key criteria to ensure the reliability of the aCGH experiment: (i) it should contain more than 70% of tumor DNA; (ii) it should not be too fragmented with genomic regions greater than 300 bp; (iii) it should not be extracted from too necrotic tissue areas; and (iv) its extraction should provide at least 10 to 20 ng of DNA or should be done from at least 2000 microdissected cells (in these lower limits of material, DNA amplification is required, highlight the authors). Interestingly, the period of fixation (below 4 days) does not influence badly the aCGH data (23).

In the same line, Mohapatra et al investigated a series of 28 FFPE oligodendrogliomas and oligoastrocytomas (stored less than 5 years) using: (i) a home‐built 200‐BACs aCGH testing chromosomes 1, 7, 19 and X and (ii) FISH and LOH techniques evaluating chromosomes arms 1p/19q statuses. The conclusions of the three techniques used by the authors were almost perfectly identical, supporting the use of aCGH for FFPE gliomas samples (36).

Further progress has been made as platforms for aCGH have matured. Of particular value to FFPE pathologic specimens is the advent of oligonucleotide‐based aCGH. The shorter length of DNA fragments employed in most oligonucleotides aCGH platforms improved analysis of highly fragmented DNA samples, such as FFPE sections. Recent studies have demonstrated the power of such approach to archival samples with good success, reproducibility and at high resolution relative to conventional methods (33). However, validation and further studies of such methods will be needed to determine how readily such methods can be applied to a wide range of samples with variable preservation and quality.

Based on the results of the literature, aCGH appears as a very reliable and promising technique to implement in neuropathological laboratories to search for CNA abnormalities in gliomas samples (Table 1). Indeed, this technique has the ability to identify the key genetic alterations in gliomas in one single experiment in comparison with FISH or LOH analyses, which present some limitations when considered for clinical application. Indeed, FISH and LOH, provide segmental, one‐time results whose interpretation is not highly comprehensive in some cases (20). Moreover, LOH requires matched patient blood, which is not always available. In comparison with SKY and cCGH, aCGH present some significant advantages. Indeed, SKY and cCGH are challenging to implement in practice because they require normal or tumor short‐term cultures and metaphases, which are difficult to obtain, particularly in low‐grade and 1p/19q co‐deleted gliomas. In addition, these lower‐resolution techniques need significant cytogenetic expertise for interpretation and are more labor and time consuming. More recently, new technologies have been developed to assess CNA, notably, those in gliomas. MLPA is based on a semi‐quantitative polymerase chain reaction (PCR) and investigates several genomic loci in one‐time experiment. Therefore, it is clearly an interesting alternative to aCGH and cCGH in order to assess reliably, and, quite easily, statuses of critical genomic regions in gliomas. However, in comparison with aCGH, the number of genomic loci tested is limited, 45 for instance in MLPA vs. several thousands in aCGH (17). In the future, identification and validation of the really relevant genomic loci with clinical impact in gliomas will permit the precise determination of the number of candidate genomic regions to test in glial tumors through MLPA assay and aCGH. Besides the well‐known techniques described above, new technologies are emerging. For instance, molecular inversion probes (MIP), which are able to screen several hundred SNPs in one time assay, seems to be an interesting technique to test in gliomas (12). More recently, deep sequencing‐based genomic study appears to be a promising technology. Indeed, this innovative technique will permit in the future the simultaneous detection of gene sequence and CNA in one single experiment. However, to our knowledge, these last two techniques have not been published yet in gliomas. Deep sequencing‐based expression studies have been recently reported in gliomas, showing the power of this approach to identify new critical genes involved in gliomas oncogenesis, such as isocitrate deshydrogenase 1 (IDH1) (38).

Table 1.

Relevant and reliable signatures observed in gliomas. Abbreviations: WHO = World Health Organization; HD = homozygous deletion; AMP = amplification.

References	Pathologic tumor class	Genomic signature	Frequency
3, 6, 21, 26, 42, 43	Oligodendroglial tumor (WHO Grades II and III)	Chr 1p/19q co‐deletion	37%–90%
3, 21, 33, 35, 37, 40	Glioblastoma	Chr 7 gain and chr 10 loss with/without CDKN2A HD with/without EGFR AMP	36%–75%
1, 8, 25, 39	Pilocytic astrocytoma	7q34 gain with BRAF rearrangement	53%–68%
53	Pleomorphic xanthoastrocytoma	CDKN2A HD	60%

Despite the potential advantages, genomic array presents several limitations in application to patients. Indeed, aCGH fails in estimation of ploïdy and does not currently reliably detect balanced or heterogeneous genomic abnormalities, such as translocations or internal deletions (EGFRvIII), although higher‐resolution oligonuleotides aCGH and SNP array platforms may in the future have some ability to detect such alterations (31). In addition, establishing quantitative thresholds for differentiating between high‐level amplification and simple genomic gain at a locus can be difficult and is platform dependent and tissue dependent. This can be more challenging when attempting to discriminate between homozygous vs. heterozygous deletion at a locus, as they differ by only a single copy. However, these limitations in practice are rare and easily overcome for ambiguous cases by using additional techniques to validate such results.

Taken together, these data suggest that genomic (CGH or SNP) arrays are powerful tools for research and are almost ready for a clinical use. However, the gold standard for CNA detection in clinical laboratories is FISH (and LOH at some degree). The interests of genomic arrays in comparison with FISH were discussed above. Now, prospective additional studies, notably clinical trials, are warranted to validate the robustness of genomic arrays faced compared with FISH and to allow their transfer from the bench to the bedside. Despite the fact that BAC array‐based CGH is as reliable as SNP array for CNA detection, it is difficult to implement in clinical laboratories because some FFPE gliomas samples do not meet the quality criteria for BAC array‐CGH 23, 36. Oligonucleotide‐based CGH and SNP array screening the whole genome of frozen or FFPE glioma samples in one‐time experiment and at a higher resolution seems to be, up to now, the best genomic array technologies to implement in clinical laboratories for CNA detection. The automation of DNA extraction and hybridization of the labeled DNAs upon the array will encourage the transfer genomic array to clinics. The quality of the DNA extracted could be checked using gel agarose or PCR. The most critical and heavy equipment to acquire is the scanner. Emerging techniques such as MLPA, MIP and genomic deep sequencing need further studies to assess their potential transfer in clinical practice.

Up to now, the most critical genomic biomarker in gliomas and particularly in anaplastic oligodendrogliomas is the chromosome arms 1p/19q co‐deletion distinguishing favorable vs. unfavorable anaplastic oligodendrogliomas. Genomic arrays are particularly interesting in this group of gliomas. Indeed, through its ability to detect whole 1p and 19q arms deletion compared with FISH or microsatellite analysis, genomic arrays can provide adjunct information to improve diagnostic evaluation and reduce interobserver variation (21). In the entire set of gliomas, genomic arrays seem also to be of interest. For example, Bredel et al identified a set of 170 genes associated with of the tumor phenotype in gliomas. These genes were mainly located on chromosomes 19q, 1p, 10 and 15q (3).

Perhaps more importantly, aCGH could also be of help to detect more effective predictors of survival within well‐categorized phenotypes. Thus, Korshunov et al and Maher et al found genomic imbalances predictive of prognosis in GBM and of time to progression from low‐grade tumor to secondary GBM (see above) 27, 33. Recent multivariate analysis also indicates that genomics could be an independent prognostic factor in a retrospective series of gliomas (21).

In summary, genomic arrays have clearly emerged as the most promising tool for complementing neuropathology analysis of human gliomas, as well as in discovery of novel classes and prognostic indicators in these tumors. While aspects of implementation will need additional investigations for optimal clinical delivery, there is little doubt that tools such as genomic arrays will soon be making the transition from bench to bedside at a rapid pace to improve our clinical treatment of these challenging diseases. Besides genomic arrays, new promising genetic techniques whose evaluation is ongoing are emerging and will be compared soon with genomic arrays. Comprehensive molecular characterization of gliomas, including epigenomics, expression profiling and/or sequencing data as conducted by the TCGA project but also by others teams will improve the “pure” genomic portrait of gliomas and will identify the key genes in glial tumors biology 30, 38, 51, 52