Gliomas are the most common primary brain tumors and constitute a heterogeneous group of neoplasms with respect to morphological appearance, biological behavior, genetic alterations, as well as response to therapy and clinical outcome. Astrocytic tumors are the most common gliomas, with glioblastoma multiforme being the most malignant and, at the same time, most frequent astrocytic tumor. In contrast, oligodendroglial tumors, comprising oligodendrogliomas and mixed oligoastrocytomas, are estimated to account for only 5–18% of all gliomas.1 Molecular genetic studies have revealed that the development of astrocytic and oligodendroglial gliomas is caused by distinct genetic alterations, which may be important for the refinement of glioma classification based on genomic profiling.2 The genetic hallmark of oligodendroglial tumors is the combined loss of heterozygosity (LOH) on chromosomal arms 1p and 19q (LOH 1p/19q), which is rare in astrocytic tumors but found in up to 80% of oligodendrogliomas and ∼50% of oligoastrocytomas.3 Importantly, LOH 1p/19q has emerged as an independent predictive marker of better response to radio- and chemotherapy as well as longer survival in patients with anaplastic oligodendroglial tumors.4, 5, 6 The frequent LOH on 1p and 19q in oligodendroglial tumors indicates that these chromosomal arms carry yet unknown tumor suppressor genes whose inactivation is of paramount importance for oligodendroglioma development. Using microsatellite-based deletion mapping, we delineated 3 distinct candidate tumor suppressor gene regions on the short arm of chromosome 1, which were mapped to 1p36.31-pter (distal to the anonymous marker D1S2633), 1p36.22-p36.31 (between D1S489 and D1S2642) and 1p34.2-p36.1 (between D1S2743 and D1S482), respectively.7 On the long arm of chromosome 19, a single common region of deletion was mapped to 19q13.3 between the markers D19S219 and D19S246.8
To identify novel genes that might be involved in the pathogenesis of oligodendroglial tumors with LOH 1p/19q, we performed a cDNA microarray-based expression profiling of 35 gliomas and screened specifically for differential gene expression between tumors with and without LOH 1p/19q. Since we were particularly interested in the identification of novel candidate genes from 1p and 19q, we used customized cDNA microarrays that were not only enriched for cancer-relevant genes9, 10 but additionally covered all genes located within the chromosomal segments 1p36.13-p36.31 and 19q13.2-q13.33, i.e., 2 of our previously identified regions of common deletion in gliomas.7, 8 Expression profiles of gliomas with and without LOH 1p/19q were compared by using significant analysis for microarrays (SAM)11 to determine the genes showing significant differential expression, followed by prediction analysis for microarrays (PAM)12 to define classifiers based on these preselected genes. Microarray findings were validated by quantitative real-time reverse transcription-PCR analyses of selected candidate genes. Thereby, we identified a distinct set of candidate genes that were significantly downregulated in gliomas with LOH 1p/19q when compared to gliomas without 1p/19q losses.
Material and methods
The investigated human tumor samples were collected at the Department of Neuropathology at Heinrich-Heine-University Düsseldorf and the Department of Neuropathology, Charité, Universitätsmedizin Berlin, and analyzed in an anonymized manner as approved by the local institutional review boards. In total, we studied tumors from 35 glioma patients (19 male, 16 female, mean age at diagnosis: 42 years, range: 11–68 years). All cases were histologically classified according to the WHO classification of tumors of the nervous system.1, 13 The tumor series comprised 8 diffuse astrocytomas of WHO grade II (AII), 7 well-differentiated oligodendrogliomas of WHO grade II (OII), 14 anaplastic oligodendrogliomas of WHO grade III (AOIII) and 6 anaplastic oligoastrocytomas of WHO grade III (AOAIII). Samples of each tumor were snap-frozen immediately after operation and stored at −80°C. A tumor cell content of at least 80% was histologically determined for each specimen used for nucleic acid extraction.
Nucleic acid extraction and LOH analysis
Extraction of high-molecular-weight DNA and RNA from frozen tumor tissue was carried out by ultracentrifugation as described elsewhere.14 Genomic DNA from peripheral blood leukocytes was extracted according to a standard protocol.15 At least 5 microsatellite loci located on chromosome arm 1p (D1S200, D1S211, D1S507, D1S489, D1S468) and 5 microsatellite loci located on chromosome arm 19q (D19S572, D19S219, D19S1182, D19S596, D19S210) were analyzed for LOH in each tumor using nondenaturing polyacrylamide gel electrophoresis and silver staining, as reported earlier.7, 16 In 17 cases, up to 30 microsatellites on 1p and 7 microsatellites on 19q had been analyzed for LOH.7
cDNA microarray generation and analysis
For microarray-based RNA expression profiling, we used custom-made cDNA microarrays that represented ∼7,000 distinct human gene-specific fragments, with particular enrichment for cancer-relevant genes.9, 10 In addition, each microarray carried gene-specific cDNA fragments representing 215 genes located within the chromosomal regions 1p36.13-p36.31 and 19q13.2-q13.33, which span our previously identified glioma-associated regions of common deletion.7, 8, 16 The respective genes were retrieved from the University of California Santa Cruz (UCSC) human genome browser database (http://genome.ucsc.edu/, assembly April 2002). From each gene, we amplified a specific DNA-fragment of ∼250–400 base pairs located in the 5′-region of the respective cDNA sequence. PCR primers were designed using sequence information from the NCBI Entrez Nucleotide data base (http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?CMD=search&DB=nucleotide). A complete list of the genes represented on the customized microarrays is available online at http://www.dkfz.de/genetics/supplement/Tews/IntJCancer. All cDNAs were spotted in triplicates as reported.9
For microarray hybridization, tumor RNA and commercially available universal human reference RNA (Stratagene, La Jolla, CA) were amplified according to a published protocol17 and hybridized to the microarrays in a GeneTac chamber (Genomic Solutions, Ann Arbor, MI) using an automated hybridization and washing protocol. After hybridization and stringent washing, fluorescence intensity images were acquired using a GenePix™ 4000 A (Axon Instruments, Foster City, CA) dual laser scanner and analyzed with the GenePix™ Pro 3.0 imaging software. For each case, we performed color switch experiments, in which tumor and reference DNA were labeled vice versa with Cy3-dUTP and Cy5-dUTP, respectively. Results of the 2 color-switched experiments were averaged. Data sets for spots not recognized by the GenePix™ software were excluded from the analysis. All remaining data sets were ranked according to spot homogeneity (as assayed by the ratio of median and mean fluorescence intensities), spot intensity and the standard deviation of logarithmic ratios for replicate spots. Data points that ranked among the lower 20%, based on the criteria just described, were removed from the data set. For each hybridization, fluorescence ratios (Cy5/Cy3) were normalized by variance stabilization.18 Results obtained for 2,956 genes were used for further analysis and will be deposited in the gene expression omnibus database.
We used the expression analysis systematic explorer-algorithm19 (EASE) to get an overview of the transcriptional profiles in gliomas with combined LOH 1p/19q vs. gliomas without this alteration and to assign differentially expressed genes across the whole genome to gene ontology (GO) categories like Biological Process or Molecular Function (Fig. 1). With this application, it is possible to test for significant co-regulation of identified genes within each GO category in the different tumor groups. We then identified those genes from the candidate regions on 1p36.13-p36.31 and 19q13.2-q13.33 that showed the strongest up- or downregulation (ln x > 1 or ln x < −1) relative to the universal reference cDNA pool (Fig. 2). To elucidate which genes significantly discriminate between the tumors with and without LOH 1p/19q, we performed a bootstrapping algorithm based on a permutative student's t test called significance analysis of microarrays (SAM).11 This algorithm was used to preselect significant genes, which were subsequently analyzed with PAM.12 These analyses were performed for the following comparisons (Fig. 3): (i) 14 tumors with LOH 1p/19q vs. 18 tumors with retention of heterozygosity on 1p and 19q; (ii) 5 WHO grade II oligodendrogliomas with LOH 1p/19q vs. 9 WHO grade II gliomas with retention of heterozygosity on 1p and 19q; (iii) 9 WHO grade III tumors with LOH 1p/19q vs. 9 WHO grade III tumors with retention of heterozygosity on 1p and 19q; (iv) 7 WHO grade II oligodendrogliomas vs. 14 WHO grade III anaplastic oligodendrogliomas and (v) 5 WHO grade II oligodendrogliomas with LOH 1p/19q vs. 8 WHO grade III anaplastic oligodendroglioma with LOH 1p/19q. All comparisons were performed either with the whole set of genes or with the limited set of genes located within the candidate regions on 1p and 19q. To determine whether expression ratios of genes from the candidate regions on 1p and 19q are related to gene dosage effects, we plotted normalized expression ratios between 1p/19q-intact tumors and 1p/19q-deleted tumors for each gene from these regions in a continuous manner according to the order of the individual genes along the chromosome (Fig. 4). The genes showing normalized mRNA expression ratios of 1p/19q-intact vs. 1p/19q-deleted tumors of more than 3 were considered as possible candidates for tumor suppressor genes that are downregulated in 1p/19q-deleted gliomas (Fig. 4).
The mRNA expression levels of selected candidate genes identified by microarray analysis (ICMT, MGC4399, RPL18 and PLA2G4C) were additionally determined by quantitative real-time reverse transcription-PCR. The transcript levels of the following housekeeping genes served as references: lamin B1 (LMNB1), phosphoglycerate kinase 1 (PGK1) and cyclophilin A (PPIA). The respective PCR primers are summarized in Table I. For each gene, forward and reverse primers were designed to bind to different exons. To further minimize the risk of amplification from genomic DNA, all tumor RNA samples were digested with DNase I prior to reverse transcription. Each cDNA sample was analyzed in triplicate using the ABI PRISM 7700 Sequence Detection System (Applied Biosystems, Weiterstadt, Germany). The protocol was carried out as reported,10 efficiency was calculated as described elsewhere20 and the expression level of each gene relative to the housekeeping genes was calculated according to a published algorithm.21
Table I. Primer Sequences Used for Quantitative Real-Time Reverse-Transcription PCR
Allelic status on 1p and 19q
Microsatellite analysis revealed combined allelic losses on 1p and 19q (LOH 1p/19q) in 15 of the 35 investigated gliomas (40%), including 6 of 7 oligodendrogliomas, 8 of 14 anaplastic oligodendrogliomas and 1 of 6 anaplastic oligoastrocytomas. Retention of heterozygosity at all informative markers on 1p and 19q was found in 18 tumors (51%), including 8 of 8 diffuse astrocytomas, 1 of 7 oligodendrogliomas, 4 of 14 anaplastic oligodendrogliomas and 5 of 6 anaplastic oligoastrocytomas. Two anaplastic oligodendrogliomas demonstrated LOH restricted to either 1p or 19q. Microarray data of these 2 cases and 1 case of WHO grade II oligodendroglioma were only included in the comparison of the expression profiles between WHO grade II and WHO grade III oligodendrogliomas irrespective of the 1p/19q status.
cDNA microarray-based expression profiling
The raw data of the microarray experiments are available online at http://www.dkfz.de/genetics/supplement/Tews/IntJCancer. After quality control, results obtained for 2,956 genes were used for further analyses. To identify crucial biological subjects in gliomas with and without LOH 1p/19q, the set of genes that was differentially expressed between these 2 groups was analyzed for significant enrichment with respect to different GO categories by using the EASE package. This algorithm ranks functional clusters by the statistical frequency of individual genes in a given category relative to all genes of this category present on the microarray. Figure 1a summarizes groups of mRNAs enriched among the transcripts found to be upregulated relative to the reference cDNA pool in tumors with or without LOH 1p/19q, while Figure 1c shows groups of mRNAs enriched among the transcripts that were downregulated relative to the reference cDNA pool in tumors with or without LOH 1p/19q, respectively. Figures 1b and 1d depict groups of transcripts that were significantly up- or downregulated relative to the reference cDNA pool in the investigated gliomas irrespective of the 1p/19q allelic status. Table II (http://www.dkfz.de/genetics/supplement/Tews/Intjcancer) lists all transcripts significantly enriched (EASE-score < 0.05) in the different gene categories listed in Figure 1.
Since the primary goal of our study was the identification of novel candidate genes within the chromosomal regions 1p36.13-p36.31 and 19q13.2-q13.33, we mainly concentrated the further biostatistical analysis of the microarray data on genes located within these candidate regions. First, we determined those transcripts from these regions that demonstrated the highest levels of up- or downregulation relative to the universal human reference RNA pool (Fig. 2). Next, all gene expression ratios were tested for significance using the SAM algorithm. Genes determined to be differentially expressed at significant levels were used for PAM analysis, which allowed for the identification of specific sets of genes discriminating between 2 given entities. All analyses were performed using the whole set of genes as well as the genes located within the candidate tumor suppressor gene regions on 1p36 and 19q13.3. For the latter, only those genes are depicted in the centroids that were identified as significantly discriminating genes in both analyses.
PAM analysis of 14 tumors with LOH 1p/19q vs. 18 tumors with retention of heterozygosity on both chromosomal arms revealed a total of 37 genes whose expression discriminated between both tumor groups (Fig. 3a). Among these were 5 genes from 1p36.13-p36.31 or 19q13.2-q13.33, respectively, which turned out to be significant. These genes code for the mitochondrial carrier protein MGC4399 (MGC4399, 1p36.22), the ribosomal protein L18 (RPL18, 19q13.2-q13.4, OMIM 604179), the light polypeptide of ferritin (FTL, 19q13.33, OMIM 134790), the spermidine synthase protein (SRM, 1p36.23-p36.13, OMIM 182891) and zinedin (ZIN, synonymous: STRN4, 19q13.32). Besides these genes, the tea domain family member 2 (TEAD2; 19q13.3) appeared as an important gene, but was not significant in the closing analysis using genes located on 1p36 and 19q13.3 (Fig. 3a).
The comparison between anaplastic oligodendroglial tumors (WHO grade III) with and without LOH 1p/19q revealed a total set of 20 discriminating genes, among which RPL18, ZIN, the D-site of albumin promoter-binding protein gene DBP (19q13.33, OMIM 124097) and FLJ10781 (19q13.33) mapped to the 19q13.3 candidate region (Fig. 3b). PAM analysis of WHO grade II gliomas with and without LOH 1p/19q revealed a total of 14 discriminating genes, only 2 of which, namely the genes for isoprenylcysteine carboxyl methyltransferase (ICMT, 1p36.31, OMIM 605851) and FTL, were from our 1p or 19q candidate regions (Fig. 3c). In all the 3 comparisons (Figs. 3a–3c), the discriminating genes from 1p and 19q were always expressed at higher levels in tumors without allelic losses on 1p and 19q when compared to tumors with LOH 1p/19q.
PAM analysis of the expression profiles in WHO grade II oligodendrogliomas (n = 7) vs. those in anaplastic oligodendrogliomas (WHO grade III) (n = 14) independent from the allelic status on 1p and 19q revealed 35 discriminating genes, which included 5 genes located within 1p36.13-p36.31 or 19q13.2-q13.33, namely the gene for cytoplasmic phospholipase A2 gamma (PLA2G4C, 19q13.3, OMIM 603602), the v-rel avian reticuloendotheliosis viral oncogene homolog B gene (RELB, 19q13.2-q13.3, OMIM 604758), the transcription factor ZNF114 gene (ZNF114, 19q13.2), the gene for neuronal PAS domain protein 1 (NPAS1, 19q13.32) and the clone NT_028054.21 (1p36) (Fig. 3d). All 5 genes were expressed at higher levels in WHO grade II oligodendrogliomas when compared to the anaplastic tumors, suggesting that their downregulation may be important for the malignant progression of oligodendrogliomas. Among the genes from other chromosomes that were downregulated in anaplastic oligodendrogliomas was the putative tumor suppressor gene EP300 at 22q13.2, which constituted the top discriminating gene in the respective centroid (Fig. 3d). PAM analysis of WHO grade II oligodendrogliomas with LOH 1p/19q (n = 5) vs. WHO grade III anaplastic oligodendrogliomas with LOH 1p/19q (n = 8) revealed only 4 discriminating genes, none of which mapped to 1p or 19q (Fig. 3e).
Mapping of candidate tumor suppressor genes on 1p and 19q
To elucidate whether certain regions on 1p and 19q show reduced expression of several neighboring genes, e.g., due to local genomic imprinting, we plotted the gene expression ratios in the order of genes along each chromosome-axis to construct a transcription map of the relevant regions on 1p36.13-p36.31 and 19q13.2-q13.33, respectively (Fig. 4). Corresponding to the model of gene dosage effects, a linear averaged gene expression ratio of tumors with intact 1p and 19q vs. tumors with LOH 1p/19q would correspond to the factor 2 (i.e., 2 functional alleles vs. 1 functional allele). To exclude errors caused by gene expression variances within the samples, we chose an expression ratio between 1p/19q-intact vs. 1p/19q-deleted tumors of more than 3 as threshold value suggestive of downregulated expression in tumors with LOH 1p/19q.
This approach revealed several candidate genes showing reduced expression in 1p/19q-deleted tumors when compared to 1p/19q-intact tumors. Among these genes were FTL, MGC4399, RPL18, SRM and ZIN, which also were identified as significant discriminators by PAM analysis (see earlier). Furthermore, among the genes showing a marked difference in expression between gliomas with and without LOH 1p/19q was EMP3 (19q13) (Fig. 4), which recently was reported as a putative tumor suppressor gene frequently downregulated by promoter hypermethylation in gliomas and neuroblastomas.22 Figure 4 also shows that genes with lower expression in tumors with LOH 1p/19q are distributed along the candidate regions without any evident clustering at a circumscribed position.
Validation of discriminating genes by quantitative real-time reverse transcription-PCR
The differential expression of 4 genes from the 1p/19q-region (RPL18, MGC4399, ICMT, PLA2G4C) with the highest discrimination potential in the PAM analyses was validated by using quantitative real-time reverse transcription-PCR analysis (RQ-PCR). Figure 5 shows the RQ-PCR results. RPL18 turned out to be expressed at 4 times higher levels in gliomas without LOH 1p/19q when compared to gliomas with LOH 1p/19q. The same relationship was confirmed when only anaplastic oligodendroglial tumors with and without LOH 1p/19q were compared with each other. MGC4399 transcripts were expressed at 3 times higher levels in gliomas without 1p and 19q losses when compared to tumors with combined allelic losses. The microarray results could also be confirmed for ICMT, which was found to be slightly downregulated in WHO grade II tumors with LOH 1p/19q when compared to WHO grade II tumors without 1p/19q losses. Finally, in line with the microarray data, RQ-PCR revealed a 2-fold higher expression level of PLA2G4C transcripts in well-differentiated oligodendrogliomas (WHO grade II) when compared to anaplastic oligodendrogliomas (WHO grade III) (Fig. 5).
LOH on chromosomal arms 1p and 19q is the most common genetic alteration in oligodendroglial tumors.3 In spite of intensive efforts, however, the relevant tumor suppressor genes located on 1p and 19q that are consistently inactivated in oligodendrogliomas with LOH 1p/19q are still elusive. Previous studies suggested several candidate genes, such as the cyclin-dependent kinase inhibitor 2C (CDKN2C) gene at 1p32, that was found to be mutated or homozygously deleted in a small subset of anaplastic oligodendrogliomas.23, 24 A more recent study pointed to the calmodulin-binding transcription activator 1 gene (CAMTA1) at 1p36, which showed reduced expression but no mutations in oligodendrogliomas with 1p deletion.25 On 19q13.3, the product of the p190RhoGAP gene, which functions as a regulator of Rho kinases, was reported to inhibit PDGF-induced murine oligodendrogliomas.26 However, mutations of p190RhoGAP in human oligodendrogliomas have not been identified (data not shown). A more recently reported candidate gene on 19q13.3 is the myelin-related epithelial membrane protein gene 3 (EMP3), which was shown to demonstrate promoter hypermethylation and transcriptional downregulation in neuroblastomas as well as subsets of astrocytic and oligodendroglial gliomas.22 In addition, the EMP3 gene product displayed tumor suppressive properties both in vitro and in vivo.22 In the present study, we found a reduced expression of EMP3 transcripts in gliomas with LOH 1p/19q as opposed to gliomas without this alteration. Thus, our data support a role of EMP3 inactivation in gliomas, although more detailed analyses assessing these tumors for EMP3 mutations and investigating the precise relationship between EMP3 promoter hypermethylation, loss of EMP3 expression and LOH 19q in gliomas remain to be carried out.
Our study follows up on previous work from our groups that delineated distinct tumor suppressor gene candidate regions on 1p and 19q in human gliomas.7, 8 To identify genes that are located within 2 of these candidate regions, i.e., 1p36.22-p36.31 and 19q13.3, and are downregulated in oligodendroglial tumors with LOH 1p/19q, we investigated a series of 35 primary gliomas with known 1p/19q status by using cDNA microarray-based expression profiling. Thereby, we identified a distinct set of novel candidate genes located on 1p or 19q. Among these, a total of 8 genes located within 1p36.13-p36.31 (MGC4399, SRM, ICMT) or 19q13.2-q13.33 (RPL18, FTL, ZIN, FLJ10781, DBP) demonstrated significantly lower expression in gliomas with LOH 1p/19q when compared to gliomas with retention of heterozygosity on both chromosomal arms. These findings were independently confirmed for selected genes (MGC4399, RPL18 and ICMT) by using quantitative real-time reverse transcription-PCR analyses.
MGC4399 (synonym: HuBMSC-MCP; NCBI Entrez Nucleotide database accession no. NM_032315) encodes a mitochondrial carrier protein originally reported to enhance dendritic cell endocytosis.27 The inner membrane of mitochondria harbors a number of specific carrier proteins that regulate the transport of various metabolites, nucleotides and cofactors in and out of the matrix space, and are also involved in the induction of apoptosis.28, 29 These proteins have similar structures and functions and, therefore, are assigned to the mitochondrial carrier superfamily, which includes MGC4399. Concerning the putative role of MGC4399 in oligodendroglial tumor cells, we can only speculate to date, since a distinct function of MCG4399 in glial cells has not been identified yet. Using confocal imaging, MCG4399 was not only localized in mitochondria but also detected in pseudopodial protrusions of MCF-7 breast cancer cells.27 Therefore, one might hypothesize that MCG4399 transports substrates related to mitochondrial oxidative phosphorylation and provide energy for tumor cell migration.27 However, further studies are needed to clarify the functional significance of MGC4399 in gliomas and the mechanisms underlying its downregulation in both low-grade and anaplastic tumors with LOH 1p/19q.
The gene product of RPL18 (19q13.2-13.4) is a protein of 22 kDa, which is part of the 60S ribosomal subunit. It was shown that RPL18 can bind to the double-stranded RNA activated protein kinase (PKR), thereby preventing PKR activation. In vitro, this leads to inhibition of PKR-dependent phosphorylation of the eukaryotic translation initiation factor eIF-2α and stimulation of protein biosynthesis.30 The EASE-analysis displayed in Figure 1 indicates that a whole set of ribosome-associated genes (Table II, online supplementary information), among them RPL18, is downregulated in gliomas with LOH 1p/19q when compared to gliomas without this alteration. A similar finding was observed for ZIN, which is expressed predominantly in brain tissue and functions as a scaffolding or signaling protein binding calmodulin in the presence of Ca2+.31 Another gene found to be downregulated in 1p/19q-deleted gliomas was the FTL gene, which encodes the light polypeptide of ferritin, an intracellular molecule that stores iron in a soluble, nontoxic, readily available form. Recently, FTL was reported to be downregulated in gestational trophoblastic neoplasia.32 However, so far neither FTL nor RPL18 nor ZIN have been implicated in glioma pathogenesis. The same applies for the spermidine synthase gene SRM, whose transcript levels were also lower in 1p/19q-deleted tumors.
The D-site of albumin promoter-binding protein (DBP) is a member of the proline and acidic amino acid-rich basic leucine zipper transcription factor family and seems to play pivotal roles in cell cycle control, carcinogenesis, circadian gene regulation, liver regeneration, apoptosis and liver-specific gene regulation.33 This protein is also expressed in many brain regions with highest amplitudes of circadian expression changes detected in the suprachiasmatic nucleus.34 Our findings indicate a lower expression of DBP transcripts in 1p/19q-deleted as opposed to 1p/19q-intact oligodendroglial tumors of WHO grade III. A similar result was obtained for the anonymous gene FLJ10737, which was discussed as a putative tumor suppressor in neuroblastoma earlier.35
Both microarray experiments and RQ-PCR analysis revealed lower expression of ICMT in WHO grade II gliomas with LOH 1p/19q when compared to WHO grade II gliomas with intact 1p/19q. Isoprenylcysteine carboxyl methyltransferase may stimulate cell proliferation by methylation of the carboxyl-terminal isoprenylcysteine of CAAX proteins, such as Ras and Rho.36 Carboxyl-terminal methylation of Ras by ICMT is important for the targeting of Ras proteins to the plasma membrane. Consequently, ICMT inactivation was found to inhibit the transformation of fibroblasts by oncogenic K-Ras and Braf proteins36 and has been suggested as a promising novel approach to cancer treatment.37 The functional role of ICMT in glioma pathogenesis and its significance as a potential target for glioma therapy remain to be elucidated.
In addition to the identification of genes showing differential expression between gliomas with and without 1p/19q deletions, we were interested to find oligodendroglioma progression-associated candidate genes whose expression changed from WHO grade II to WHO grade III. PAM analysis suggested a set of 35 genes discriminating between WHO grade II oligodendrogliomas and anaplastic oligodendrogliomas (WHO grade III). The top discriminating gene was EP300, a histone acetyltransferase and tumor suppressor gene being involved in the regulation of several tumorigenic pathways, such as the transforming growth factor beta (TGF-β), retinoblastoma protein and p53 pathways.38EP300 maps within an astrocytoma-associated candidate tumor suppressor gene region in 22q13.2, but lacked detectable mutations in astrocytic gliomas.39 Nevertheless, it remains to be investigated whether epigenetic changes, such as promoter hypermethylation, may contribute to the reduced expression of EP300 transcripts in anaplastic when compared to well-differentiated oligodendrogliomas. Five of the 35 genes that discriminated between WHO grade II and WHO grade III oligodendrogliomas mapped to 19q13.3. These 5 genes were distinct from the set of 19q13.3 genes that were found to be downregulated in gliomas with LOH 1p/19q when compared to gliomas without LOH 1p/19q. Interestingly, previous microsatellite studies have shown that LOH 19q is not only an early aberration in oligodendroglial tumors but also a progression-associated change in astrocytic gliomas.40 Therefore, our finding may support the view that distinct genes in 19q13.3 contribute to oligodendroglioma initiation and glioma progression, respectively. The gene encoding the gamma isoform of the cytosolic phospholipase A2 (PLA2G4C) had the highest discrimination potential among the 5 genes from 19q13.3. Cytosolic phospholipase A2 gamma seems to be predominantly expressed in the heart and skeletal muscle.41 Other authors demonstrated that this enzyme is constitutively expressed in the endoplasmic reticulum and plays important roles in remodeling and maintaining membrane phospholipids under various conditions, including oxidative stress.42 Biochemically, phospholipase A2 catalyzes the hydrolysis of the sn-2 acyl bond of glycerolipids to produce lysophospholipids and release arachidonic acids. High amounts of arachidonic acids, which are converted into biologically active eicosanoids, are present in glioma when compared to normal white or grey matter.43 Several eicosanoids have been shown to promote cell survival, stimulate proliferation and modulate cell adhesion as well as angiogenesis.44, 45 However, the precise role of cytosolic phospholipase A2 gamma in gliomas and the significance of its lower expression in WHO grade III when compared to WHO grade II oligodendrogliomas remain to be determined. The same applies to the other 3 genes from 19q13.3 (RELB, ZNF114 and NPAS1) as well as the single anonymous gene from 1p36 (NT_028054.21) found to be downregulated in anaplastic oligodendrogliomas, i.e., so far none of these genes has been implicated in glioma pathogenesis.
In summary, we identified a distinct set of genes that are located within previously defined candidate tumor suppressor gene regions on 1p36.22-p36.31 and 19q13.3 and showed significantly lower transcript levels in 1p/19q-deleted gliomas when compared to 1p/19q-intact gliomas. We additionally found novel oligodendroglioma progression-associated candidate genes whose expression differed significantly between WHO grade II and WHO grade III oligodendrogliomas. Further studies are needed to elucidate the molecular mechanisms underlying the differential expression of these newly identified candidate genes and to clarify their functional roles in oligodendroglial tumors.
We thank Felix Kokocinski and Nicolas Delhomme for database management and support, Frauke Devens and Andreas Danner for excellent technical support as well as Robert Tibshirani and Axel Benner for statistical advice. This work was supported by grants from the Deutsche Forschungsgemeinschaft to G. R. (SFB503-B7), from the Deutsche Krebshilfe to G. R. (70-3088-Sa1) as well as to G. R. and A. v. D. (70-3163-Wi3), and from the Bundesministerium für Bildung und Wissenschaft within the national genome research network 2 (NGFN-2) to P. L. and M. H. (01GS0460), to G. R. (01GS0462), and to A. v. D (01GS0463). B.T. is a scholar of the Studienstiftung des Deutschen Volkes.