Malignant astrocytomas are a deadly solid tumor in children. Limited understanding of their underlying genetic basis has contributed to modest progress in developing more effective therapies. In an effort to identify such alterations, we performed a genome-wide search for DNA copy number aberrations (CNA) in a panel of 33 tumors encompassing grade 1 through grade 4 tumors. Genomic amplifications of 10-fold or greater were restricted to grade 3 and 4 astrocytomas and included the MDM4 (1q32), PDGFRA (4q12), MET (7q21), CMYC (8q24), PVT1 (8q24), WNT5B (12p13), and IGF1R (15q26) genes. Homozygous deletions of CDKN2A (9p21), PTEN (10q26), and TP53 (17p3.1) were evident among grade 2 to 4 tumors. BRAF gene rearrangements that were indicated in three tumors prompted the discovery of KIAA1549-BRAF fusion transcripts expressed in 10 of 10 grade 1 astrocytomas and in none of the grade 2 to 4 tumors. In contrast, an oncogenic missense BRAF mutation (BRAFV600E) was detected in 7 of 31 grade 2 to 4 tumors but in none of the grade 1 tumors. BRAFV600E mutation seems to define a subset of malignant astrocytomas in children, in which there is frequent concomitant homozygous deletion of CDKN2A (five of seven cases). Taken together, these findings highlight BRAF as a frequent mutation target in pediatric astrocytomas, with distinct types of BRAF alteration occurring in grade 1 versus grade 2 to 4 tumors. Cancer Res; 70(2); 512–9
- pediatric astrocytoma
- genomic profiling
- gene alterations
Targeted molecular therapeutics holds great promise for the development of less toxic and more effective personalized treatment strategies for cancer (1). This approach, however, requires a detailed understanding of the molecular alterations that drive tumor formation and malignant progression. The identification of seminal changes in cancer can be used to inform the development of therapeutic agents that specifically inhibit the gene products and pathways that are deregulated in association with these changes. This strategy has been used successfully in adult high-grade astrocytomas, leading to clinical trials that target the epidermal growth factor receptor (EGFR), for example. However, astrocytomas in children are distinct clinical entities from those seen in adults and do not harbor many of the critical genetic alterations found in their adult counterparts (2–5). It is therefore important that the molecular alterations unique to pediatric astrocytomas are identified and characterized.
Recently, it was shown that the majority of pilocytic astrocytomas (PA; WHO grade 1) harbor 7q34 duplications (6), which result in gene fusions between KIAA1549 and BRAF (7), and concomitant expression of KIAA1549:BRAF fusion transcripts (7–9). Although this discovery suggests that targeted inhibition of BRAF or relevant downstream signaling mediators may be especially effective for the treatment of PA, identification of therapeutically informative molecular alterations in pediatric malignant astrocytomas (WHO grades 2, 3, and 4) has been an elusive goal.
To provide a more detailed characterization of the spectrum of genetic alterations in all malignancy grades of pediatric astrocytoma, we report a genome-wide analysis of DNA copy number alterations for a series of these tumors, presented in combination with the sequence analysis of five relevant genes, including three frequently mutated in adult astrocytomas (TP53, PTEN, and IDH1) and two involved in mitogen-activated protein kinase (MAPK) signaling (BRAF and KRAS). Our results confirm that genes targeted for sequence alteration in adult astrocytomas are less frequent mutation targets in pediatric astrocytomas. In addition, we show that BRAF fusion is a signature event in PA, but not in the malignant astrocytomas, whereas BRAF activating mutation is common in grade 2 to 4 tumors. Moreover, we found that the majority of grade 2 to 4 tumors with BRAF activating mutation have concomitant CDKN2A homozygous deletion (HD). Collectively, these findings suggest that the combination of BRAF activation and CDKN2A inactivation may be unique drivers of malignancy in a subset of pediatric astrocytomas, and whose alteration suggest targets for future therapeutic drug design.
Materials and Methods
A total of 41 astrocytomas were included in this study. Nonneoplastic brain samples derived from epileptic surgeries from pediatric (n = 2) and adult (n = 6) patients were also included. All samples were subjected to detailed histopathologic review (S.R.V.) and approved for use according to Committee on Human Research guidelines. Sufficient RNA was obtained from all cases for determination of presence or absence of KIAA1549-BRAF fusion transcripts, and sufficient DNA was obtained from 40 cases for gene sequence analysis. DNA from 33 of the cases was examined for copy number alterations.
Identification of copy number aberrations
To identify copy number gains and losses, we used the custom-designed molecular inversion probe (MIP) cancer panel consisting of 24,037 single-nucleotide polymorphisms (SNP; Affymetrix). The MIP assay and analysis of resultant data were performed as described (10, 11) using 37 ng of genomic DNA as starting material. The fraction of the genome altered (FGA) for each tumor was calculated as the total length of gained and deleted loci divided by the total genome length (2,866 Mb; National Center for Biotechnology Information Build 35).
Assessment of KIAA1549-BRAF fusion transcripts
Total RNA was extracted using the miR-Vana RNA isolation kit (Applied Biosystems), and first-strand cDNA was generated using the High Capacity cDNA Reverse Transcription kit (Applied Biosystems). PCR was conducted using 60 ng cDNA and final concentrations of 2.5 mmol/L MgCl2, 25 nmol/L deoxynucleotide triphosphates (dNTP), 400 nmol/L primers, and 0.05 unit/μL AmpliTaq Gold (Applied Biosystems). Cycling conditions were as follows: 95°C for 10 min, 35 cycles of 95°C for 15 s, 55°C for 30 s, 72°C for 1 min, followed by 72°C for 10 min and 4°C hold. The forward primer, 5′-CGGAAACACCAGGTCAACGG-3′ (KIAA1549 exon 15), and reverse primer, 5′-GTTCCAAATGATCCAGATCCAATTC-3′ (BRAF exon 11), used in this study detect three fusion transcripts: KIAA1549-exon-15/BRAF-exon-9, KIAA1549-exon-16/BRAF-exon-11, and KIAA1549-exon-16/BRAF-exon-9 (7). Wild-type BRAF primers were as follows: forward, 5′-TTGTGACTTTTGTCGAAAGCTGC-3′, and reverse, 5′-AAGGGGATGATCCAGATGTTAGG-3′.
Gene sequence analysis
For BRAFV600E mutation analysis, genomic DNA was PCR amplified using BRAF exon 15 primers with M13 tails: forward, 5′-TGTAAAACGACGGCCAGTCATAATGCTTGCTCTGATAGGA-3′; reverse, 5′-AGCGGATAACAATTTCACACAGGCCAAAAATTTAATCAGTGGA-3′. PCR final concentrations (25 μL volume) were 2.0 mmol/L MgCl2, 250 nmol/L dNTPs, 0.02 unit/μL AmpliTaq Gold, 0.2 mg/mL bovine serum albumin, 500 nmol/L primers, and 10 ng of genomic DNA. Cycling conditions were as follows: 95°C for 10 min, 35 cycles at 95°C for 15 s, 64°C for 45 s, 72°C for 1 min, followed by 72°C for 10 min and 4°C hold. PCR products were treated with ExoSAP (USB) per the manufacturer's protocol and sequenced (ELIM Biopharmaceuticals).
For analysis of IDH1 sequence alterations, we used the protocol described by Balss and colleagues (12). Briefly, a fragment of 129-bp length spanning the catalytic domain of IDH1, including codon 132, was amplified using the sense primer IDH1f (5′-CGGTCTTCAGAGAAGCCATT-3′) and the antisense primer IDH1r (5′-GCAAAATCACATTATTGCCAAC-3′). PCR conditions, using 20 ng DNA, were for 35 cycles with denaturing at 95°C for 30 s, annealing at 56°C for 40 s, and extension at 72°C for 50 s. Sequencing was with sense primer IDH1f (5′-CGGTCTTCAGAGAAGCCATT-3′).
KRAS mutations were identified as described by Lièvre and colleagues (13). Specific probes for mutated and nonmutated alleles labeled with fluorescence reporter dyes were included in 5 μL of reaction mixtures containing 10 ng DNA, 1 μg of specific primers and probes, and 1 μL of Taqman Universal PCR Master Mix (Applied Biosystems). Reaction mixtures were subjected to the following cycle conditions: 95°C for 15 min; 40 cycles, 95°C for 15 s; and 60°C for 1 min. The single mutation detected by allelic discrimination was confirmed by direct sequencing of exon 2 of the KRAS gene.
All tissue was routinely fixed in either phosphate-buffered 4% formalin, dehydrated by graded ethanols, and embedded in wax (Paraplast Plus, McCormick Scientific) using routine techniques. Immunohistochemistry was performed using the Benchmark XT (Ventana Medical Systems) and the Ultraview (multimer) detection system. All sections were cut at 5 μm and mounted on SuperFrost/Plus slides (Fisher Scientific). Epitope retrieval was performed for 30 min in Tris buffer (pH 8) at 90°C before application of the primary antibodies. Rabbit polyclonal antibodies to phospho-MAPK (pMAPK; extracellular signal-regulated kinase 1/2; Invitrogen) were used at a dilution of 1:200 with an incubation time of 2 h at 37°C.
To identify regions of genomic copy number gain and loss, we analyzed DNA from 33 astrocytomas (Table 1) using a panel of ∼24,000 MIPs enriched for sequences that directly interrogate allele-specific copy number of ∼1,000 known cancer genes. Included in this analysis were 2 grade 1 tumors, 11 grade 2 tumors, 9 grade 3 tumors, and 11 grade 4 tumors. Calculation of the mean FGA for each grade of tumor revealed increasing FGA with increasing malignancy grade (grade 1 < 0.000; grade 2 = 0.016; grade 3 = 1.375; grade 4 = 2.964), with the mean FGA for combined grade 3 + grade 4 tumors significantly higher than for grade 1 + grade 2 tumors (P < 0.003, two-sided Wilcoxon rank-sum test; Table 1; Fig. 1).
Several locations of copy number alteration were observed among grade 2 to 4 tumors. Among these, high-level amplifications (>10 copies) were observed exclusively in grade 3 and 4 tumors and included known oncogenes MDM4 (1q32), PDGFRA (4q12), MET (7q21), CMYC (8q24), PVT1 (8q24), WNT5B (12p13), and IGF1R (15q26; Fig. 2A). In contrast to adult grade 3 and 4 astrocytomas, no high-level or focal copy number gains for EGFR were observed. Regions of HD (less than one copy) were observed in grade 2 to 4 astrocytomas, involving CDKN2A (seven tumors; Fig. 2B), PTEN (one tumor), and TP53 (one tumor; Table 2; Supplementary Table S1).
As revealed by the copy number aberration results (Fig. 1; Table 1), genomic alterations in grade 1 and 2 tumors are exceedingly rare. Nonetheless, a single recurrent alteration was identified in two grade 1 tumors and a single grade 2 tumor: a 2-Mb duplication at 7q34, with breakpoints encompassing the KIAA1549 and BRAF genes (Fig. 3A). Duplications of this region have been associated with KIAA1549-BRAF rearrangements, resulting in the production of fusion mRNAs detectable by reverse transcription-PCR (RT-PCR; refs. 7–9). To determine whether the 7q34 duplications represented KIAA1549-BRAF fusion transcripts, mRNA was isolated from each of the initial cohort of 33 tumors and examined for these products. Both of the grade 1 tumors expressed BRAF fusion transcripts, but not the grade 2 tumor with a similar duplication of 7q34 (Fig. 3B). Eight additional grade 1 astrocytomas subsequently examined by RT-PCR all revealed expression of KIAA1549-BRAF fusion transcripts. In total, 10 of 10 PAs expressed one of the three common KIAA1549-BRAF fusion transcripts previously described (7), whereas none of 31 grade 2, 3, and 4 tumors expressed these fusion transcripts (Fig. 3B).
In addition to the gene rearrangements of BRAF resulting in the synthesis of KIAA1549-BRAF fusion transcripts, a missense activating BRAF mutation (BRAFV600E) has been previously reported in grade 1 and 2 astrocytomas at low incidence (6, 7). Here, DNA sequence analysis revealed BRAFV600E mutation in 0 of 10 grade 1, 2 of 11 (18%) grade 2, 3 of 9 (33%) grade 3, and 2 of 11 (18%) grade 4 astrocytomas (Fig. 4A; Tables 1 and 2). MIP analysis indicated that both glioblastoma multiforme (GBM) tumors with BRAFV600E had 7q34 copy number gains (Fig. 4B), consistent with the relative peak areas of the BRAF DNA sequence tracings seen in these tumors (Fig. 4A). BRAF copy number gains were not evident in the three grade 3 tumors or two grade 2 tumors with BRAFV600E mutations.
Genes that are frequent mutagenic targets in the development of adult astrocytoma include TP53, PTEN, and IDH1 (4, 5, 12). To address the incidence of these alterations in pediatric astrocytomas, we sequenced relevant coding regions of each of these genes in the current set of tumors. Sequence alterations of TP53 were identified in five cases (two grade 3 and three grade 4 tumors), PTEN in one case (one grade 4 tumor), and IDH1 in none of the tumors (Table 1). Activating mutation of KRAS, an infrequent occurrence in both adult and pediatric astrocytoma, was found in one grade 4 astrocytoma lacking the BRAFV600E alteration (Table 1). Together, these data show that BRAF is the most common mutagenic target in pediatric astrocytomas (Table 2). Specifically, 17 of 41 tumors had BRAF alterations followed by CDKN2A HD (7 instances) and TP53 sequence alteration or HD (6 instances). Inactivating mutation and/or deletion of PTEN and amplification of PDGFRA were the only other specific gene alterations observed in multiple tumors (two tumors each).
We next examined gene alterations that occur concomitant with BRAF mutation and found that CDKN2A HD was observed in five of the seven tumors with the BRAFV600E alteration (Table 1). Among grade 2 to 4 tumors, the association between CDKN2A HD and BRAFV600E mutation was significant: Whereas 71% of grade 2 to 4 tumors with BRAFV600E have CDKN2A HD (5 of 7), only 8% (2 of 24) of the remaining grade 2 to 4 astrocytomas lacking the BRAFV600E mutation harbored CDKN2A HD (P = 0.0016, two-sided Fisher's exact test). In fact, no additional specific gene alterations, including those resulting from high-level genomic amplifications, were evident in the two remaining grade 3 tumors with BRAFV600E mutation (Table 1).
Because increased MAPK signaling accompanies BRAF activation, we next examined pMAPK expression in astrocytomas (14, 15). Interestingly, all pediatric astrocytomas examined (40 cases), irrespective of malignancy grade or BRAF mutation status, showed strong immunoreactivity for pMAPK (Supplementary Fig. S1).
Grade 2 to 4 pediatric astrocytomas share histopathologic similarities with their corresponding adult counterparts, and, in general, pathologic examination does not distinguish pediatric and adult tumors of the same malignancy grade. Numerous studies have identified signature genetic mutations in high-grade adult astrocytomas, including the recent TCGA analysis of GBM tumors (4). The TCGA study showed that EGFR, PTEN, and TP53 changes are among the most frequently observed alterations in adult GBM, whereas recent studies using parallel genomic approaches showed that mutations in the IDH1 gene predominate in adult gliomas (5, 16). With the possible exception of TP53 mutation, the frequent genetic alterations seen in adult astrocytomas have been identified at lower frequencies in pediatric astrocytomas (2, 3, 17). This lack of signature genetic mutations in pediatric astrocytomas is unfortunate, as these serve as important reference points with which to test therapeutic hypotheses.
The results of the current study further emphasize the distinct genetic etiologies of pediatric and adult astrocytomas. In our series, the only common adult astrocytoma gene alterations observed at appreciable frequencies were CDKN2A HD (7 of 31 cases, 23%) and alterations affecting p53 function (6 of 31 cases, 19%). The incidence of these alterations, however, is much lower than that reported by the TCGA project for adult GBM (4). Although a negative observation, the lack of IDH1 mutation in the current series of astrocytomas is further evidence of the distinct genetic nature of the pediatric tumors. In this regard, a recent study that included 42 pediatric gliomas examined for IDH1 mutation also concluded that these alterations are rare in childhood astrocytomas (18).
The approach taken here to identify common alterations in pediatric astrocytoma involved microarray analysis for high-resolution whole-genome examination. The SNP platform used has particularly high-density coverage of 1,000 genes associated with human cancer, yet has sufficient coverage for identifying focal gains and deletions across the majority of the genome (median distance between probes, across the entire genome, is 43 kb), and certainly has sufficient density to identify copy number alterations affecting all large chromosomal regions. Whereas recurrent copy number alterations indicating low-level gains or hemizygous deletions are of importance, and are presented in Supplementary Table S2, it is the regions of high-level amplification and HD that are of particular interest in revealing specific oncogene and tumor suppressor gene targets, respectively (Tables 1 and 2; Supplementary Table S1). These array data highlight PDGFRA and CDKN2A as important oncogene and tumor suppressor gene targets, respectively, in the development of pediatric malignant astrocytoma.
The identification of focal 7q34 copy number gains with rearrangement in three tumors (Fig. 3A) prompted an analysis of BRAF alterations, revealing KIAA1549-BRAF fusion transcript expression in 10 grade 1 tumors but not in a single grade 2 tumor with this genomic alteration. Although it is possible that there are additional KIAA1549-BRAF alterations not detected with the primer pairs we used, it seems that BRAF fusion transcripts are highly diagnostic of pediatric grade 1 astrocytoma. This is particularly important in clinical practice, as biopsy material from these tumors arising in the brainstem or optic nerve may not be representative of the entire tumor, and PA may be misdiagnosed as grade 2 or 4 astrocytoma. The development of reagents for immunohistochemical detection of the junction sequences in paraffin-embedded tissues would therefore be of great value.
In light of prior reports indicating a few instances of BRAFV600E activating mutations in grade 1 and 2 pediatric astrocytomas (6, 7), we determined the presence or absence of this alteration in the current series of tumors to address the overall incidence of BRAF alterations. Unexpectedly, seven tumors with BRAFV600E mutation were identified among the 31 grade 2 to 4 tumors (23%), suggesting that this alteration is one of the most common gene alterations in pediatric malignant astrocytomas (Table 2). Observation of KRAS activating mutation (Table 1), high-level PDGFRA or MET amplification (Table 1), and low-level copy number increases of PTK2 (Supplementary Table S2) in seven additional grade 3 and 4 tumors that lack BRAFV600E indicates the occurrence of BRAF or surrogate genetic alterations for achieving heightened BRAF activity in most, and perhaps all (as implied by Supplementary Fig. S1 results), pediatric astrocytomas.
Even more surprising than the number of instances of BRAFV600E in pediatric grade 2 to 4 astrocytomas was the occurrence of CDKN2A HD in five of the seven tumors with BRAFV600E: a frequency of co-occurrence that is highly significant. This association is underscored by the lack of other specific gene alterations occurring in combination with BRAFV600E. BRAFV600E in combination with CDKN2A inactivation has been described in other cancers, especially melanoma (19), and in a recent report, this combination was identified in 2 of 18 grade 2 pediatric astrocytomas (8). Our results reinforce the importance of coincident BRAFV600E with CDKN2A HD in pediatric malignant astrocytomas and suggest that this combination of gene alterations occurs at similar frequencies across grade 2 to 4 tumors.
Finally, there are several BRAF signaling pathway inhibitors currently in clinical trial (20), including inhibitors of MAPK, a key downstream effector of BRAF. Interestingly, our immunohistochemical results indicate that MAPK activity is not differentially elevated in tumors with BRAF alterations (Supplementary Fig. S1), and suggest that additional activators of MAPK may be deregulated in pediatric astrocytoma. Studies are under way to determine how BRAF controls astrocyte cell growth as well as to assess the effect of MAPK inhibitors on deregulated BRAF-mediated astrocytoma growth. Moreover, the new pediatric astrocytoma–associated genetic mutations identified in this report should inform the development of more relevant animal models of pediatric astrocytoma.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
We thank Cynthia Cowdrey for her outstanding assistance in tissue acquisition and processing.
Grant Support: Pediatric Low Grade Astrocytoma Foundation (J.G. Hodgson, D.H. Rowitch, and C.D. James), National Brain Tumor Society (D.H. Gutmann), Pediatric Brain Tumor Foundation (J.G. Hodgson, S.R. VandenBerg, D.H. Rowitch, M.S. Berger, and C.D. James), Harriet H. Samuelsson Foundation (J.D. Schiffman), Center for Children's Brain Tumors at Lucile Packard Children's Hospital (J.D. Schiffman, P.G. Fisher, J.M. Ford, and H. Ji), and NIH grants CA097257 (M.S. Berger and C.D. James), CA101777 (J.G. Hodgson), CA121940, and HG000205 (P. Flaherty). D.H. Rowitch is a Howard Hughes Medical Institute Investigator.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/).
- Received May 22, 2009.
- Revision received October 17, 2009.
- Accepted November 5, 2009.
- ©2010 American Association for Cancer Research.