|Home | About | Journals | Submit | Contact Us | Français|
Oligodendrogliomas are the second most common malignant brain tumor in adults and exhibit characteristic losses of chromosomes 1p and 19q. To identify the molecular genetic basis for this alteration, we performed exomic sequencing of seven tumors. Among other changes, we found that the CIC gene (homolog of the Drosophila gene capicua) on chromosome 19q was somatically mutated in six cases and that the FUBP1 gene [encoding far upstream element (FUSE) binding protein] on chromosome 1p was somatically mutated in two tumors. Examination of 27 additional oligodendrogliomas revealed 12 and 3 more tumors with mutations of CIC and FUBP1, respectively, 58% of which were predicted to result in truncations of the encoded proteins. These results suggest a critical role for these genes in the biology and pathology of oligodendrocytes.
Oligodendrogliomas (ODs) account for 20% of brain tumors in adults and, as their name suggests, they consist primarily of cells resembling oligodendroglia (1, 2). These tumors generally arise in the white matter of the cerebral hemispheres, commonly in the frontal lobes. Well-differentiated ODs can evolve into high-grade “anaplastic” ODs, although it is often difficult to clearly distinguish these two types from each other or from other brain tumors (1, 2). Because this distinction is important for the management of patients, molecular biomarkers for ODs are of great interest.
To date, the best biomarker for ODs is loss of heterozygosity (LOH) of chromosomes 1p and 19q (2–4). Assessment for LOH events is now commonly performed in patients with ODs because of their important implications for therapeutic responses (2–4). The chromosome losses occur in 50% to 70% of tumors and are often associated with a pericentromeric translocation of chromosomes 1 and 19, producing marker chromosome der(1;19) (q10;p10) (2–6). This translocation is unbalanced, leaving the cells with one copy of the short arm of chromosome 1 and one copy of the long arm of chromosome 19. The functional basis for most cancer translocations involves one of the genes residing near the breakpoints, producing fusions that alter the gene’s product. In contrast, the der(1;19) (q10;p10) breakpoints are in gene-poor centromeric regions and are always associated with LOH (4, 5). This suggests that the basis for the t(1;19) translocation is the unmasking of a tumor suppressor gene(s) on either chromosome 1p or 19q (2–4, 7, 8). This is supported by the fact that some tumors lose only chromosome 1p sequences, while others lose only chromosome 19q sequences (2–4, 7, 8).
To identify this putative tumor suppressor gene(s), as well as to increase understanding of OD pathogenesis, we sequenced the coding exons of 20,687 genes in DNA from seven anaplastic ODs using the Illumina HiSeq platform (9). The clinical characteristics of the patients and their tumors are listed in table S1. The average distinct coverage of each base in the targeted regions was high (135-fold) and 94% of the bases were represented by at least ten distinct reads (table S2). Loss of heterozygosity (LOH) of chromosomes 1p and 19q was confirmed using common single nucleotide polymorphisms (SNPs) identified as heterozygous in DNA from corresponding normal cells (Fig. 1 and fig. S1).
We have previously described methods for the accurate identification of somatic mutations in next-generation sequencing data from Illumina instruments (10). Using these stringent criteria to avoid false positive calls, we identified a total of 225 non-synonymous somatic mutations, affecting 200 genes among the seven tumors (table S3). There were an average of 32.1 ± 10.7 non-synonymous somatic mutations per tumor (table S2), similar to the number found in the most common type of adult brain tumor [glioblastoma, 35.6 non-synonymous somatic mutations per tumor (11)].
There were a number of notable mutations identified in these seven tumors. We identified three tumors with mutations in PIK3CA, encoding the catalytic subunit of the PI3Kα enzyme, and one tumor with a mutation in PIK3R1, encoding the regulatory subunit (table S3). The NOTCH1 gene was mutated in two tumors and at least one of these was inactivating (a 1 bp deletion), consistent with the recently described tumor suppressor role for this gene (12). Finally, the IDH1 (isocitrate dehydrogenase 1) gene was mutated in all seven tumors at the same residue, resulting in an amino acid substitution of His for Arg at codon 132, as expected for this tumor type (13, 14).
One of the major goals of this study was the investigation of the target gene(s) on chromosome 1 or 19. By analogy with other tumor suppressor genes (15, 16) we expected that the residual copy of the target gene(s) would contain mutations in most tumors with LOH of the relevant region. On chromosome 1p, there were eight somatically mutated genes, but only two with mutations in more than one tumor: FUBP1 (Far Upstream Element [FUSE] Binding Protein 1) and NOTCH2 (table S3). On chromosome 19q, there were three genetically altered genes identified, two of which were mutated in a single tumor each. The third, CIC (homolog of the Drosophila capicua gene), was mutated in six of the seven tumors. In each of these six cases, the fraction of mutant alleles was high (80.5 ± 10.7%), consistent with loss of the non-mutated allele. The mutations were confirmed to be homozygous by Sanger sequencing (Fig. 2A).
To validate these results and determine the spectrum of FUBP1, NOTCH2, and CIC mutations in ODs, we examined tumor DNA from an additional 27 tumors and matched normal cells. No additional mutations of NOTCH2 were found, but FUBP1 and CIC mutations were identified in 3 and 12 of the additional cases respectively and generally (14 of 16 mutations) appeared to be homozygous (Fig. 2B and table S4). The probability that these mutations were passengers rather than drivers was <10−8 for both genes [binomial test (17)]. All FUBP1 mutations and more than 25% of the CIC mutations were predicted to inactivate their encoded proteins, as they altered splice sites, produced stop codons, or generated out-of-frame insertions or deletions (Fig. 2B and table S4). This type of mutational pattern is routinely observed in tumor suppressor genes such as TP53 or FBXW7 (18) but is never observed in bona fide oncogenes.
The capicua gene was discovered in a screen for mutations affecting the anteroposterior pattern of Drosophila embryos (19). In Drosophila, the protein encoded by CIC has been shown to be a downstream component of receptor tyrosine kinase (RTK) pathways that includes EGFR, Torso, Ras, Raf, and mitogen-associated protein kinases (MAPKs) (20). In the absence of RTK signaling, cic, in combination with other transcription factors such as Groucho (Gro), blocks transcription by binding to canonical octameric elements in regulatory regions (21). RTK signaling blocks the function of cic via MAPK-mediated phosphorylation or docking, resulting in degradation of cic and the consequent activation of the genes it normally represses (22). The most highly conserved functional domain of the cic protein is the HMG (high mobility group) box responsible for its binding to DNA. Importantly, 8 of the 11 missense mutations we observed in ODs were located in this domain (Fig. 2B).
In addition to the high conservation of CIC sequences among metazoans, the human cic protein contains nine consensus phosphorylation sites for MAPK (23). This suggests that human cic functions similarly to its Drosophila counterpart. This hypothesis is supported by mass spectroscopic studies that have shown human cic protein to be phosphorylated within 10 min of EGF treatment of HeLa cells (24).
The protein encoded by FUBP1 binds to single stranded DNA, in particular the far-upstream element (FUSE) of MYC, a well-studied oncogene (25). Although overexpression of FUBP1 can stimulate MYC expression (25), it has also been shown that FUBP1 protein participates in a complex with PUF60 that negatively regulates MYC expression (26). Our data, showing that FUBP1 is inactivated by mutations, are consistent with the idea that FUBP1 mutations lead to MYC activation in these tumors by relieving the negative effects of the FUBP1-PUF60-FUSE complex.
There are only a small and statistically insignificant number of point mutations of FUBP1 or CIC recorded in the COSMIC database (18). However, CIC has been shown to be translocated in two cases of Ewing’s sarcoma-like tumors that harbored t(4;19)(q35;q13) translocations. Unlike the mutations observed in ODs, the translocations in these two cases seemed to activate the cic protein by fusing it to the C-terminus of DUX4, conferring oncogenic properties to the new protein (27).
Overall, 23 mutations of CIC or FUBP1 were identified in the 34 tumors analyzed in this study. Notably, of the 27 cases with 19q loss, 18 cases (67%) contained intragenic mutations of CIC, while none of the 8 ODs without 19q loss contained CIC mutations (table S1). As our mutational screens would not detect some types of inactivating mutations (e.g., large deletions or promoter mutations) or epigenetic alterations, the fraction of tumors with detectable CIC and FUBP1 mutations is likely an underestimate of their actual contribution. To evaluate the prevalence of CIC and FUBP1 mutations, we sequenced 92 tumors of the nervous system and 206 non-nervous system tumors and found only 3 missense mutations in CIC (breast, prostate, and medulloblastoma) and no truncating alterations.
The identification of inactivating mutations of CIC or FUBP1 in a substantial fraction of ODs is expected to provide important insights into the pathogenesis of these tumors as well as help refine methods currently used for their diagnosis, prognosis, and treatment.
We thank M. Whalen, N. Silliman, J. Ptak, L. Dobbyn, J. Schaeffer for expert technical assistance. Supported by The Virginia and D. K. Ludwig Fund for Cancer Research, The Pediatric Brain Tumor Foundation, the Duke Comprehensive Cancer Center Core, the Burroughs Wellcome Fund, James S. McDonnell Foundation, Fundacao de Amparo a Pesquisa do Estado de Sao Paulo (FAPESP) grants 04/12433-6, 01/12898-4, the NCI Division of Cancer Prevention contract N01-CN-43302 and National Institutes of Health grants CA43460, CA57345, CA11898, CA121113, CA62924, RC2DE020957 and NS20023. C.B. is a recipient of T32 CA009574 NIH/NCI Inst. National Research Service Award. Johns Hopkins University has filed a patent relating to the application of the mutations described in this work to the diagnosis and treatment of cancer. Under agreements between the Johns Hopkins University, Genzyme, Exact Sciences, Inostics, Qiagen, Invitrogen and Personal Genome Diagnostics, N.P., B.V., K.W.K. and V.E.V are entitled to a share of the royalties received by the University on sales of products related to genes and technologies described in this manuscript. N.P., B.V., K.W.K. and V.E.V are co-founders of Inostics and Personal Genome Diagnostics are members of their Scientific Advisory Boards, and own Inostics and Personal Genome Diagnostics stock, which is subject to certain restrictions under Johns Hopkins University policy. The sequence data reported in this manuscript are deposited in http://www.sanger.ac.uk/perl/genetics/CGP/cosmic.
Supporting Online Material
Materials and Methods
Figs. S1 and S2
Tables S1 to S4
References (28, 29)