Introduction

Gliomas are the most common primary brain tumors in adults. The World Health Organization (WHO) classification divides gliomas into three main subgroups: astrocytomas, oligodendrogliomas, and oligoastrocytomas (mixed gliomas). It further distinguishes between four malignancy grades (WHO grades I–IV). Gliomas exhibiting oligodendroglial features include oligodendrogliomas (WHO grade II) and anaplastic oligodendrogliomas (WHO grade III) as well as oligoastrocytomas (WHO grade II), anaplastic oligoastrocytomas (WHO grade III) and glioblastomas with an oligodendroglial component (GBMO, WHO grade IV) [1]. Oligodendroglial tumors account for 15-20% of all gliomas [2,3].

The identification of the genes targeted by complete 1p/19q co-deletion, a characteristic of oligodendrogliomas, has been a long-standing quest. Combined loss of whole chromosome arms 1p and 19q is the most frequently detected genetic imbalance in oligodendroglial tumors, occurring in 60-90% of oligodendrogliomas and 30-50% of oligoastrocytomas while they are rarely found in GBMO [4-6]. The 1p/19q co-deletion is due to an unbalanced translocation, der(1;19)(q10;p10) [7,8] and has been highly associated with chemosensitivity and a less aggressive course of progression [3,9-11]. Thus, the co-deletion has become an important prognostic and predictive marker.

Recurrent mutations in the capicua transcriptional repressor gene (CIC), on 19q and to a much lesser extent in the far upstream element (FUSE) binding protein 1 gene (FUBP1) on 1p31.1 have only recently been identified in oligodendrogliomas by using DNA next generation sequencing [12]. Based on the assumption that the 1p/19q co-deletion might help to unmask mutations that result in tumorigenesis, numerous efforts were made in the past to identify putative tumor-associated genes, but with limited success. Therefore, the detection of mutations in CIC and FUBP1 marks an important step in deciphering the process of oligodendroglial tumor development.

Genomic sequencing has also led to the identification of mutations of the isocitrate dehydrogenase genes (IDH1/2) in human gliomas. These mutations preferentially occur in lower grade gliomas, including 80-92% of oligodendrogliomas [5,13-15]. Most 1p/19q co-deleted tumors analyzed so far display IDH1/2 mutations [14,16]. However, IDH1/2 mutations are not exclusively found in oligodendroglial and oligoastrocytic gliomas, but also in the majority of grade II and III astrocytic tumors, indicating the existence of a common initiating event among these histologically and clinically diverse glioma subtypes [6].

In an effort to further characterize oligodendroglial tumors, we analyzed a set of 17 oligodendrogliomas and oligoastrocytomas by applying a comprehensive approach of genome-wide profiling by array comparative genomic hybridization (array CGH), expression analyses by transcriptome next generation sequencing (RNA-seq) and DNA Sanger sequencing for mutations in CIC, FUBP1 and IDH1/2.

Materials and Methods

Results

We analyzed a cohort of 17 oligodendroglial tumor samples, consisting of nine cases diagnosed as oligodendroglioma (O) and eight as oligoastrocytoma (OA) for copy number variations (CNV); CIC, FUBP1 and IDH1/IDH2 mutations; and gene expression changes. Results are summarized in Table 1.

By array CGH, we observed the three distinct genetic subtypes that we have previously described for oligodendroglial tumors [4]. The majority of the O (8/9) and half of the OA (4/8) carried the 1p/19q co-deletion (combined loss of the entire chromosome arms 1p and 19q) typically found in O. These cases were therefore genetically classified as the “oligodendroglial” subtype. Three OA, but none of the O displayed aberrations characteristic of astrocytomas (A), such as gain of chromosome 7 or a combined gain of chromosome 7 and loss of chromosome 10. These cases were classified genetically as the “astrocytic” subtype. A single case each of O and OA had none of the aberrations typically found in glial tumors and therefore, these cases were genetically classified as the “other” subtype.

The IDH1 mutation c.395G>A (R132H) was present in 14 of the 17 oligodendroglial tumors, including all twelve tumors with 1p/19q co-deletion and two tumors of the “astrocytic” genetic subtype. None of the “other” oligodendroglial tumors had mutations in IDH1. IDH2 mutations were not observed in any case.

Identification of novel point mutations in CIC

We detected unique somatic single nucleotide variants (SNVs) in CIC in 7 of the 12 (58.3%) oligodendroglial tumors with 1p/19q co-deletion. The frequency was 75% (6/8) in cases histologically classified as O, whereas it was only 25% (1/4) in those classified as OA. Four of the seven CIC SNVs have not been described so far (Figure 1 and Table 1). The seven SNVs included one frameshift mutation that led to a premature stop codon, one splice-site mutation, one nonsense mutation, and four missense mutations. To evaluate the impact of the four missense mutations on protein structure, we used the PolyPhen-2, Mutation Taster, and Mutation Assessor software [18-20]. All four missense mutations were predicted to be “probably damaging” with a PolyPhen-2 score of ≥ 0.96 on a scale of 0 to 1 (Mutation Taster: “disease causing” with a probability >0.99; Mutation Assessor: low to medium). Moreover, the missense mutations clustered either in exon 5, the HMG box region, or exon 19, the globular domain (GlobDom) that includes the recently described protein-binding domain of CIC [22]. Both gene regions are predicted to be essential to unimpeded protein function (Figure 1). Therefore, all CIC mutations found in the 1p/19q-codeleted tumors have the potential to abolish or diminish the function of the remaining allele, either by creating mRNA that is subjected to nonsense mediated decay or by creating a non- or sub-functional CIC protein. None of the SNVs was detected in the DNA from the patients’ blood.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 1. Distribution of CIC mutations in this study.

The dark blue boxes represent exons. HMG-box denotes the highly conserved DNA-interacting high-mobility group domain. Exon 19 and exon 20 harbor a globular domain (GlobDom). An annotated protein-protein interacting domain is located within exon 20 [22]. Mutations marked by a frame indicate novel mutations not yet described in the literature. While somatic missense mutations were only found in the HMG-box and GlobDom of CIC (BT14,5,2,16), stop-, frameshift- and splice site mutations were found across the CIC protein (BT4,15,7). The two missense mutations that were not located in exons 5 or 19 (BT17,11) were already present in the normal DNAs and had a very low PolyPhen-2 Score, indicating that these SNVs are probably polymorphisms. The red bar represents the exon-spanning deletion in BT1 which extends up to the adjacent PAFAH1B3 and PR19 gene (see also Figure 2).

https://doi.org/10.1371/journal.pone.0076623.g001

In addition we identified novel CIC SNVs in two of the five oligodendroglial tumors lacking the 1p/19q co-deletion (Figure 1). Both SNVs were missense mutations with a very low PolyPhen-2 score (0.037 and 0.000) [Mutation Taster: Disease causing with a probability 0.958 for the former and polymorphism with a probability of 0.999 for the latter; Mutation Assessor predicted both SNVs to be neutral], indicating that these SNVs are unlikely to damage protein function, and they were also present in the DNA from the patients’ blood. Moreover, they were located in exons 10 and 15, outside the HMG box and protein binding regions. Missense mutations in these exons have not been described in oligodendroglial tumors so far. Therefore, we assume that these variants are non-functional germ-line variations rather than causally related with tumor origin or growth.

A novel partial deletion of the remaining CIC allele in a 1p/19q co-deleted case

RNA-seq identified two tumors, BT1 and BT7, with notably decreased CIC mRNA levels (Figure 2a, Table S3). Sanger sequencing revealed a frameshift mutation in BT7 (see above), but no mutation in BT1. Nevertheless, array CGH of BT1 showed one oligonucleotide-probe situated within CIC and a second probe within the PAFAH1B3 gene just downstream of CIC displayed a log₂-ratio of -2 (Figure 2b), in contrast to the entire chromosome arm 19q, which displayed a log₂-ratio of -0.9 to -1. This result indicated a deletion of at least 8.6 kb and a maximum size of 17.3 kb on the remaining 19q-arm. In order to further evaluate and define the breakpoints of the deletion, we performed qPCR on the DNA of BT1 compared to DNA from BT2 with a single copy of 19q and BT11 with two copies. The resulting Ct values of BT1 clearly distinguished between homozygously deleted regions and regions with one remaining copy, where Ct values corresponded to the Ct values of the other 1p/19q-codeleted tumors. The 5’ position of the breakpoint of the deletion was refined to [hg19] chr19:42,790,756-42,790,839 and the 3’ breakpoint to [hg19] chr19:42,805,880-42,808,040 (Figure 3a). We concluded that exons 2-20 of CIC and all of PAFAH1B3 were deleted. These results also correlated with the expression data that showed a significant down-regulation of both CIC and PAFAH1B3 whereas PRR19, a gene downstream of PAFAH1B3 and presumably outside of the deletion, was upregulated (Figure 3b, Table S3).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 2. CIC Expression analysis and homozygous deletion of CIC in array CGH in tumor sample BT1.

2a Column plot of expression of the CIC gene in tumor samples (BT1-15) and normal brain (nb1-3) analyzed using RNA-seq. Tumor samples harboring point mutations are marked as green columns, samples without point mutations as gray columns. Compared to other tumor samples, BT1 (gray arrow) and BT7 (green arrow) show a strong down regulation of CIC. This can be explained by a homozygous deletion of CIC in BT1 (compare Figure 2b) and nonsense-mediated mRNA decay due to a frameshift mutation in BT7. RPKM = Reads Per Kilobase of exon model per Million mapped reads (gene counts/total counts of each sample). 2b: Array CGH results of BT1 harboring a partial homozygous deletion of CIC (red circle). One oligonucleotide-probe situated within CIC and another in PAFAH1B3 (A_16_P21013642: [hg19] 42,795,949-42,796,008 and A_14_P124269: [hg19] 42804518-42804577) showed a log₂-ratio of -2, while the whole long arm of chromosome 19 displayed a log₂-ratio of -0.9 (corresponding to one allele in the tumor and two alleles in the control probe in array CGH). The log₂-ratio of -2 indicates a homozygous deletion in about 75% of the cell population, corresponding to a background due to normal cells in the tumor tissue. Array CGH narrows the exon-spanning deletion of the second allele of CIC to a minimum size of 8.6 kb and a maximum size of 17.3 kb.

https://doi.org/10.1371/journal.pone.0076623.g002

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 3. Exon-spanning deletion in BT1 leads to change in expression of CIC, PAFAH1B3 and PR19.

3a Genomic region chr19:42,788,249-42,815,094 that includes the CIC, PAFAH1B3 and PR19 gene. The red bar represents the deletion in BT1. The minimum extent (thick) and the maximum extend (thin) of the deletion according to array CGH and qPCR results are shown. 3b: Expression data (RNA-seq) for tumor samples (BT1-15) with both alleles of 1p and 19q (middle gray), 1p/19q co-deletion (light gray) and normal brain (nb1-3, dark gray). BT1 harboring the exon-spanning deletion shows a strong down regulation of expression in CIC and PAFAH1B3 and an upregulation in PR19 (red arrows). RPKM = Reads Per Kilobase of exon model per Million mapped reads (gene counts/total counts of each sample).

https://doi.org/10.1371/journal.pone.0076623.g003

Discussion

Recently, CIC and FUBP1 have been identified as candidate tumor suppressor genes in oligodendrogliomas with the typical 1p/19q co-deletion [12]. In an attempt to uncover the extent and diversity of these mutations, and their impact on gene expression, we carried out a targeted analysis of 17 oligodendroglial tumors. The assessment of mutations in CIC and FUBP1 using a comprehensive approach combining high-resolution array CGH, sequencing and expression data was the main objective of this study.

Through array CGH, three distinct genetic subtypes that we had previously reported could be confirmed in the 17 oligodendroglial tumors [4]. (i) Tumors carrying the 1p/19q co-deletion typically found in O were assigned to the “oligodendroglial” subtype. In our tumor set, 8 out of 9 O and 4 out of 8 OA carried this co-deletion compared to 60-90% of O and 30-50% of OA reported in the literature [5,6]. (ii) Three OA, but no O showed aberrations characteristic of astrocytomas (for example gain of chromosome 7 or a combined gain of chromosome 7 and loss of chromosome 10) and were genetically classified as the “astrocytic” subtype (iii). One O and one OA showed none of these aberrations and were therefore classified as the “other” subtype. We identified the IDH1 c.395G>A mutation in 14 of the 17 oligodendroglial tumors (82.4% in total, 8/9 O, 6/8 OA), including all 12 tumors with the 1p/19q co-deletion (100%) and two of the three tumors with the “astrocytic” genetic subtype (66.7%), while none of the tumors of the “other” subtype harbored this mutation. This is in accordance with recent findings, where IDH mutations were identified in about 75-94% of oligodendrogliomas and 71-98% of oligoastrocytomas [5,13-15]. A strong correlation between the 1p/19q co-deletion and mutation of IDH1 was previously reported [24], and is consistent with our findings (Fisher’s exact test, p=0.015).

We detected protein altering somatic mutations in the remaining CIC allele in eight of the 1p/19q-codeleted tumors, including five novel mutations, and the first description of an exon-spanning deletion. The frequency of mutations in CIC in our study corroborates previously published data [12,13,17,22]. In contrast to Yip et al. [22], we did not observe an association between somatic CIC and IDH1 mutation (p=0.206), but were able to confirm that somatic CIC mutations were associated with the 1p/19q co-deletion (p=0.0294). Since there is a strong association between 1p/19q co-deletion and IDH1 mutation, the association of CIC mutations with IDH1 mutation reported by Yip et al. is more likely due to the association with the 1p/19q co-deletion. In agreement with previous studies, we also found a higher occurrence of CIC mutations in O (78%) as compared to OA (12.5%) [12,13,17,22]. However, this finding was not significant if only 1p/19q co-deleted tumors were considered (Fisher’s exact test, p = 0.067), and may only reflect the higher occurrence of 1p/19q co-deletion in O compared to OA.

According to the two-hit hypothesis, tumor initiation generally requires both copies of a tumor suppressor gene to be inactivated [25]. In oncogenesis, the first hit could be acquired through the deletion of one allele via a gross chromosomal event, such as loss of an entire chromosome or chromosome arm [26,27]. It can be assumed that the translocation between chromosomes 1 and 19 leading to a subsequent loss of 1p and 19q corresponds to this process in oligodendrogliomas. Inactivation of the second allele may arise from mutations resulting in truncation, missense mutations at residues essential for protein function, from deletions or insertions, or from epigenetic silencing [27]. This scenario is exactly the case for the CIC mutations detected in our tumor set. All of the somatic CIC alterations we found in the remaining allele of 1p/19q-codeleted tumors either led to the truncation of the CIC protein (4/8) or were missense mutations predicted to be damaging by PolyPhen-2 (4/8). We have found a partial deletion of the CIC gene - an event not yet reported in the scientific literature - that extends from exon 2 of CIC to the adjacent PAFAH1B3 gene. We evaluated previously published data and confirmed that all reported tumors with 1p/19q co-deletion and CIC mutation showed a protein damaging mutation of the remaining allele: all were either truncating mutations (nonsense, frameshift, and splice acceptor mutations), missense mutations that are predicted to be protein damaging (PolyPhen-2 score of at least 99%, disease causing according to Mutation Taster with probability > 0.68; functional impact according to Mutation Assessor ranging from low to high), or in-frame-deletions in exons 5 and 20 (see Table S3) [12,13,17,22]. It can therefore be stated that CIC mutations are not only frequent in oligodendroglial tumors, but also bear a high potential of inactivating their encoded protein. Our findings together with published data therefore support the hypothesis that CIC functions as a tumor suppressor in 1p/19-codeleted oligodendrogliomas and oligoastrocytomas and is commonly altered in these tumors.

All four somatic CIC missense mutations in our 1p/19q-codeleted tumors were located in exons 5 and 19. Mutations have been previously shown to cluster in exons 5 and 20 [22]. Exon 5 and parts of exon 6 encode the HMG box domain, which seems to include several mutational hotspots, such as the c.6014C>T (p.Arg201Trp) [12,13,17,22]. Base exchange is very likely to occur at the cytosine at this position, because it is part of a CpG dinucleotide and almost certainly methylated in the mammalian genome. One third of all human point mutations appear in such sites through deamination of the 5-methylcytosine to uracil leading to a transition from CpG to TpG [28]. Yip et al. also described missense mutations in exon 20, where they predicted the location of a protein binding domain [22]. In our cohort, we found missense mutations in exon 19 but not in exon 20. Bettegowda et al. and Jiao et al. found also four missense mutations in exon 19. We therefore assume an association with an annotated GlobDom that includes both exon 19 and 20 (compare Figure 1). Data mining of the existing literature and the TCGA dataset revealed that out of a total of 77 missense mutations with high Polyphen-2 score, 60 were located within the HMG-box (77.9%), 15 within the GlobDom (19.5%) and only two outside of these structural units (compare Table S2) [12,13,17,22].

It is important to point out that we and others have found protein truncating CIC mutations (nonsense, frameshift, and splice site mutations) that were located outside exons 5, 19, and 20 (see Table S4). Since a loss of function mutation (which one would expect for tumor suppressor gene to promote cancer) can be acquired through different types of mutations and along almost the whole gene, one would not expect to find mutations (except for missense mutations) only in certain domains of a gene. Our results emphasize that sequencing only “hotspot” regions of the CIC gene, such as exons 5 and 20, as indicated by others, may not be sufficient to detect the whole spectrum of CIC mutations, some of which are deleterious to the CIC protein function [22]. The results of the low grade brain tumor characterization by TCGA seem to underscore our findings; the mutations in CIC in the 213 analyzed samples with complete information are not restricted to exons 5 and 20. The five novel mutations reported here were not found in the TCGA samples including any microdeletions in CIC. The lack of microdeletions could be due to the fact that small (intragenic) deletions in CIC could be below the resolution limit of the platform used to detect CNVs in the TCGA samples (Affymetrix’s Genome-Wide Human SNP Array 6.0). We have, however, identified two cases within the TCGA dataset with 1p/19q co-deletion, no CIC mutations, and very low expression level of CIC, which is very similar to our findings in one case with partial deletion of CIC. This might indicate a second hit in the CIC gene in these tumors – such as a microdeletion below the detection limit - that was missed by the applied methods.

Our tumor set also included two tumor samples that harbor novel SNVs of CIC but did not carry the 1p/19q co-deletion. Both SNVs were transmitted via or occurred in the germline as they are present in the DNA from the patients’ blood. They were located in exons 10 and 15, and were missense mutations not likely to be protein damaging according to in silico prediction. We therefore assume that these SNVs are rare variants and not mutations in the sense of a pathogenic effect. To the best of our knowledge, only two tumors with a CIC mutation that retained both copies of 19q have been described in the literature so far. One reported by Sahm et al. had a missense mutation located in exon 3 [17]. We evaluated this mutation using in silico prediction, and found it to not have a deleterious effect on protein function (PolyPhen-2 score 0.000; ‘polymorphism’ according to Mutation Taster [probability 0.894]; see Table S4). The other case, reported by Yip et al., was diagnosed as a WHO IV astrocytoma/gliomatosis cerebri [22]. Using PCR, the authors did not find a 1p/19q co-deletion in this tumor. However, since this method is less sensitive if the relative proportion of tumor to stromal cells is too small, as often the case in gliomatosis cerebri, the 1p/19q co-deletion could simply not have been detected. In addition, we reevaluated the functional effect of all other missense mutations in 1p/19q-codeleted tumors reported so far using in silico prediction. Only one additional tumor harbored a missense mutation with Polphen-2 score < 0.01 (Mutation Taster: ‘polymorphism’ with probability 0.996; Mutation Assessor: ‘neutral’) [22]. This mutation was also located in exon 3 and the tumor had an additional frameshift mutation in the same allele of the CIC gene (see Table S4). Therefore, we assume that the two missense mutations with a low PolyPhen-2 score reported by Yip et al. and Sahm et al. are also rather rare non-functional variants [17,22]. Unfortunately, the authors did not analyze normal tissue from the patients; therefore it is not known, whether these variants were also germline. We propose that the aforementioned CIC SNVs in our cases and the reported tumors with low PolyPhen-2 scores do not constitute a ‘hit’ in the Knudson model. Nevertheless, we cannot exclude that these rare variants contribute to tumor development.

Expression analysis showed no significant differences in the transcript level of CIC in tumors with and without 1p/19q co-deletion. CIC expression did not differ significantly between tumors with and without CIC mutation, concordant with recently published data [22]. However, there were two exceptions with notably decreased transcripts: BT1 where the remaining allele of CIC was partially deleted and BT7 with a frameshift mutation causing a premature translation-termination codon (PTC) in the remaining allele of CIC. In eukaryotes, mRNA harboring PTCs is detected and eliminated by nonsense-mediated mRNA decay [29]; this could explain loss of CIC expression in BT7. Our results indicate a tight regulation of the CIC promoter activity in oligodendroglial tumors, leading generally to normal transcript level of CIC even in tumors with 1p/19q co-deletion.

Mutations of FUBP1 were recently reported in a subset of oligodendrogliomas [12,17]. In our study, somatic mutations of FUBP1 on chromosome arm 1p were only found in 3 of 17 tumors (17.6%) and only in tumors of the “oligodendroglial” genetic subtype (3/12, 25%). The frequency of FUBP1 mutations in our cohort corresponds well to the frequency of FUBP1 mutations reported in the literature, ranging between 11.1% and 22.2% for O, 6.3% and 12.5% for OA [12,17], and 6.3% and 28.6% for 1p/19q-codeleted tumors [17]. Frameshift and nonsense mutations were the predominant types of alterations found in previous studies [17], but only one of the three mutations in FUBP1 detected in our tumor set is predicted to generate a truncated protein (BT7). The relevance of the other two mutations is unclear: one was a silent mutation (BT4) and the other occurred in an intronic region (BT5).

FUBP1 mutations occurred only in tumors that also had both CIC and IDH1 mutations combined with 1p/19q co-deletion, consistent with observations of Sahm et al [17]. Interestingly, the one case that presented deleterious alterations in both CIC and FUBP1 (BT7) was the only recurrent tumor in our cohort, and the patient died only 3 months after surgery. We have previously reported on a different anaplastic oligodendroglioma with the same mutational status that was also a recurrent tumor and had engrafted in mice [30]. Based on the relatively low incidence of FUBP1 mutations, their concurrence with CIC mutations, and their occurrence in recurrent tumors, we hypothesize that FUBP1 mutations might constitute a later event in oligodendroglioma tumorigenesis.

To date, not much is known about the function of the protein capicua homolog that is encoded by the CIC gene in humans. It contains an annotated high mobility group (HMG) domain and was classified as a member of a new Sox-related HMG subfamily [31]. This highly conserved DNA-interacting domain allows the CIC protein to bind nucleosomes and thus to regulate chromosome architecture and gene transcription [32]. Since CIC is predominantly expressed in immature granule cells in the CNS, it is assumed that CIC has a role in CNS development [31]. In addition, CIC function has been explored in Drosophila where it regulates terminal and dorso-ventral patterning of the embryonic body by repressing genes downstream of multiple receptor tyrosine kinase (RTK) pathways, such as Torso and EGFR, by binding to octameric sequences in the promoter regions of the affected genes. CIC itself is under negative post-transcriptional control by RTK signaling via MAPK-mediated degradation of the CIC protein by phosphorylation [33-35]. FUBP1 codes for the far upstream binding protein 1 (FUBP1 also named FBP1), which has an important role in cell proliferation and is implicated in multiple types of cancers [36]. FUBP1 stimulates the transcription of the c-Myc proto-oncogene by binding to the single strand DNA of the far upstream element (FUSE) in the c-Myc promotor region [37,38]. By complexing with the FUBP interacting repressor (FIR) it negatively regulates c-Myc expression [38]. A number of recent reports have indicated that independent of the c-myc pathway, FUBP1 acts as an RNA-binding protein to cellular mRNA or viral RNA [39,40] and is involved in the development of the neural system [36,41,42].

We were able to show here that CIC mutations are common in oligodendroglial tumors with 1p/19q co-deletion whereas FUBP1 mutations seem to occur more rarely. We also broadened the spectrum of CIC mutations and provided the first report of a partial homozygous deletion of CIC that was not detected by standard sequencing or array CGH alone. Our results highlight the importance of a comprehensive approach for the detection of CIC mutations. Based on our results, it is likely that previous sequencing studies have underestimated the frequency of CIC mutations. Exome sequencing complemented by detection of smaller rearrangements and deletions, for example, by using paired-end sequencing, may illuminate the full mutational spectrum of CIC in human oligodendrogliomas. The high frequency of CIC mutations in this and previous studies, their co-occurrence with 1p/19q co-deletion, and their inactivating character support the theory of CIC being a tumor suppressor gene in the development of human oligodendrogliomas. However, not all tumors with the 1p/19q co-deletion carried an alteration in the second allele of CIC (or FUBP1), indicating that other mechanisms or genes are involved in oligodendroglioma development. It will be important to delineate the pathway through which CIC acts and to analyze CIC mutations in a larger cohort of tumors so that conclusions about their impact on prognosis and treatment can be made. Our results in conjunction with other studies indicate that CIC mutations play a critical role in oligodendroglioma development and form a molecular feature distinctive of this glioma subtype.

Supporting Information

Acknowledgments

We thank Andreas Polten (Agilent Technologies) and Götz Frommer (Agilent Technologies) for their technical support of this study.