Advertisement
Human Cancer Biology

IDH Mutation and Neuroglial Developmental Features Define Clinically Distinct Subclasses of Lower Grade Diffuse Astrocytic Glioma

Daniel Gorovets, Kasthuri Kannan, Ronglai Shen, Edward R. Kastenhuber, Nasrin Islamdoust, Carl Campos, Elena Pentsova, Adriana Heguy, Suresh C. Jhanwar, Ingo K. Mellinghoff, Timothy A. Chan and Jason T. Huse
DOI: 10.1158/1078-0432.CCR-11-2977 Published May 2012

Abstract

Purpose: Diffuse gliomas represent the most prevalent class of primary brain tumor. Despite significant recent advances in the understanding of glioblastoma [World Health Organization (WHO) IV], its most malignant subtype, lower grade (WHO II and III) glioma variants remain comparatively understudied, especially in light of their notable clinical heterogeneity. Accordingly, we sought to identify and characterize clinically relevant molecular subclasses of lower grade diffuse astrocytic gliomas.

Experimental Design: We conducted multidimensional molecular profiling, including global transcriptional analysis, on 101 lower grade diffuse astrocytic gliomas collected at our own institution and validated our findings using publically available gene expression and copy number data from large independent patient cohorts.

Results: We found that IDH mutational status delineated molecularly and clinically distinct glioma subsets, with IDH mutant (IDH mt) tumors exhibiting TP53 mutations, platelet—derived growth factor receptor (PDGFR)A overexpression, and prolonged survival, and IDH wild-type (IDH wt) tumors exhibiting EGFR amplification, PTEN loss, and unfavorable disease outcome. Furthermore, global expression profiling revealed three robust molecular subclasses within lower grade diffuse astrocytic gliomas, two of which were predominantly IDH mt and one almost entirely IDH wt. IDH mt subclasses were distinguished from each other on the basis of TP53 mutations, DNA copy number abnormalities, and links to distinct stages of neurogenesis in the subventricular zone. This latter finding implicates discrete pools of neuroglial progenitors as cells of origin for the different subclasses of IDH mt tumors.

Conclusion: We have elucidated molecularly distinct subclasses of lower grade diffuse astrocytic glioma that dictate clinical behavior and show fundamental associations with both IDH mutational status and neuroglial developmental stage. Clin Cancer Res; 18(9); 2490–501. ©2012 AACR.

Translational Relevance

The clinical and molecular heterogeneity exhibited by lower grade gliomas indicates the likely underlying presence of distinct disease subclasses, the understanding of which is required for developing more effective treatment modalities. Our study shows the importance of IDH mutation, TP53 abnormalities, and receptor tyrosine kinase pathway activity in the establishment of clinically relevant stratification schemes for diffuse glioma variants. Furthermore, we identify, through expression profiling, robust subclasses of lower grade diffuse astrocytic glioma, whose distinct molecular profiles support more diversified clinical management algorithms. The survival analyses presented in this report clearly show the relative importance of molecular stratification over conventional World Health Organization (WHO) grading in the prognostic assessment of lower grade diffuse astrocytic gliomas. Finally, the associations we establish between IDH mutant tumors and specific neuroglial precursor pools further implicate the mammalian subventricular zone as the main point of origin for a large subset of diffuse gliomas and an anatomic/cellular target for therapeutic intervention.

Introduction

Diffuse gliomas represent the most common primary brain tumors in adults with a combined annual incidence of approximately 15,000 cases (CBTRUS 2008; http://www.cbtrus.org/). Their intrinsically invasive nature precludes definitive surgical resection, mandating the development of more effective medical therapies (1). The World Health Organization (WHO) groups diffuse gliomas by histopathology into astrocytomas, oligodendrogliomas, and oligoastrocytomas while also designating clinical grades that predict biologic behavior. As such, WHO II (low-grade) and even WHO III (anaplastic) gliomas can follow protracted clinical courses—years to decades—whereas the WHO IV glioblastoma exhibits median survival on the order of 15 months (1, 2).

All lower grade gliomas eventually undergo some degree of malignant transformation, with the plurality of diffuse astrocytic gliomas—astrocytomas and a significant subset of oligoastrocytomas—evolving into so-called secondary glioblastomas. These tumors contrast with the majority of glioblastomas, which actually arise de novo in a fully malignant state (primary glioblastoma) driven by pathogenic mechanisms differing from those used in stepwise evolution (3). The recent discovery of isocitrate dehydrogenase enzyme (IDH1 and IDH2) mutations in substantial majorities of WHO II and III gliomas and secondary glioblastomas, but not primary glioblastomas, has provided a strong genetic foundation for this distinction (4–6). Glioma-associated IDH mutations lead to overproduction of the oncometabolite R(−)-2-hydroxyglutarate (7, 8), and although precise downstream tumorigenic processes remain unclear, sweeping disruptions of the epigenome are likely involved (9, 10). Importantly, IDH mutation in diffuse gliomas confers favorable prognosis in a manner that transcends histopathology (5, 11).

WHO II and III diffuse astrocytic gliomas exhibit considerable variability in disease outcome, far greater than that seen for glioblastoma, supporting the existence of clinically and therapeutically relevant molecular heterogeneity. IDH mutational status likely plays an important role in the delineation of prognostically distinct subgroups. Yet even IDH mutant tumors show notable variability in outcome measures such as overall survival, indicating the likely involvement of additional influential molecular parameters. Recently, large-scale integrated genomics has successfully identified clinically relevant disease stratification in a variety of tumor types (12–15). More specifically, global expression profiling in primary glioblastoma has revealed 4 molecular subclasses, termed proneural, neural, classical, and mesenchymal, each associated with specific patterns of genomic alterations (15). Applying this sort of multidimensional analysis to diffuse astrocytic gliomas would greatly facilitate the discovery of prognostic and/or predictive biomarker sets. Unlike their oligodendroglial counterparts (16, 17), diffuse astrocytic gliomas infrequently harbor loss of chromosomes 1p and 19q (1p/19q deletion; ref. 18), and instead exhibit mutations in TP53 and/or PTEN, promoter methylation at the PTEN, CDKN2A and/or MGMT loci, and overactivity in receptor tyrosine kinase signaling networks, particularly those driven by EGF receptor (EGFR), platelet—derived growth factor receptor (PDGFRA), and the downstream PI3K/AKT pathway (19–25). The extent to which these abnormalities correlate with each other in distinct tumor subclasses within lower grade diffuse astrocytic gliomas is unclear.

To address the molecular foundations of clinical heterogeneity in diffuse astrocytic glioma, we conducted detailed molecular profiling on a large (N = 101) set of tumors incorporating global expression analysis along with several disease-relevant genomic, epigenomic, and immunophenotypic parameters. We then validated our major findings in a large, independent data set derived from the Repository for Molecular Brain Neoplasia Data (REMBRANDT). Our results delineate genomically and clinically distinct subclasses of diffuse astrocytic glioma that also show intriguing associations with specific neuroglial developmental compartments.

Materials and Methods

Glioma sample cohort

Formalin-fixed, paraffin-embedded (FFPE) tissue blocks and clinical data were obtained following study approval from the Memorial Sloan-Kettering Cancer Center (MSKCC; New York, NY) Institutional Review Board under auspices of a blanket biospecimen utilization protocol. The following specific search terms were used to identify study cases: “low-grade astrocytoma,” “astrocytoma, WHO grade II,” “anaplastic astrocytoma,” and “astrocytoma, WHO grade III.” Seven oligoastrocytomas known to be 1p/19q intact were also included. Following histopathologic quality control, genomic DNA and total RNA were extracted from 10-μm tissue sections using commercially available reagents (Qiagen).

IDH genotyping

Mutational analysis for IDH1 and IDH2 was conducted using iPLEX mass spectrometry–based genotyping (Sequenom). Briefly, iPLEX is based on a single-base primer extension assay. A multiplexed PCR reaction using amplification primers bracketing a specific mutation is followed by a one-base extension at the nucleotide of interest. The difference in mass between extended products allows the distinction between wild-type and mutant alleles. For additional methodologic details including specific primer sequences, see Supplementary Materials.

TP53 sequencing

Exons 5 to 8 of TP53 were subjected to bidirectional Sanger sequencing. Primers for 350-bp amplicons were designed using Primer3 (http://frodo.wi.mit.edu/primer3/) to cover exonic regions and around 50 bp of flanking intronic sequence. PCR reactions were carried out in 384-well plates, in a Duncan DT-24 water bath thermal cycler, with 10 ng of template DNA (Repli-G Midi, Qiagen) using a touchdown PCR protocol with Kapa2G Fast HotStart Taq (Kapa Biosystems) consisting of 1 cycle of 95°C for 5 minutes; 3 cycles of 95°C for 30 seconds, 64°C for 15 seconds, 72°C for 30 seconds; 3 cycles of 95°C for 30 seconds, 62°C for 15 seconds, 72°C for 30 seconds; 3 cycles of 95°C for 30 seconds, 60°C for 15 seconds, 72°C for 30 seconds; 37 cycles of 95°C for 30 seconds, 58°C for 15 seconds, 72°C for 30 seconds; and 1 cycle of 70°C for 5 minutes. Templates were then purified using AMPure (Beckman Coulter Genomics) and sequenced bidirectionally with M13 forward and reverse primers at Beckman Coulter Genomics. Mutations were detected using an automated detection pipeline at the MSKCC Bioinformatics Core (see Supplementary Materials for more detail). All positive results were confirmed by repeat Sanger sequencing at 4× coverage.

EGFR copy number analysis

EGFR copy number was assessed by quantitative PCR (Q-PCR). Specific primer sets were designed for EGFR (for-cttcaaaaactgcacctcca; rev-caagcaactgaacctgtgact) and glyceraldehyde-3-phosphate dehydrogenase (GAPDH) (for-cagcaagagcacaagaggaa; rev-caactgtgaggaggggagat). Q-PCR reactions were based on SYBR green chemistry (Applied Biosystems) and included 20 ng of genomic DNA and 500 nmol/L of each primer. Ten-microliter reactions in quadruplicate were run on a 7900 HT Fast Real-Time PCR machine (Applied Biosystems). Final relative quantity (RQ) values were scaled relative to euploid DNA after normalization to GAPDH. Examining results for all tumors revealed a natural division between cases with high-level amplification (>18 copies) and cases with either euploid status or low-level copy number gain/polysomy (<8 copies).

PTEN copy number

FISH for PTEN was conducted as previously described (26), using probes for both the PTEN gene and the centromere of chromosome 10 (Abbott Molecular). Tumors exhibiting one signal for PTEN and/or chromosome 10 in more than 20% of examined cells (200) were considered to harbor PTEN loss. This conservative cutoff point accounts for incomplete nuclear profiles and is routinely used by the MSKCC Clinical Cytogenetics Laboratory.

Promoter methylation analysis

Five hundred nanograms of gDNA was bisulfite-modified using the EZ DNA Methylation Kit (Zymo). End product was eluted in a 10-μL volume. CpGenome Universal Methylated DNA (Millipore) and locally generated whole-genome–amplified DNA were used as positive and negative controls, respectively. Complete methodology for CDKN2A methylation-specific PCR (MSP) has been previously reported (24). For PTEN MSP, previously reported primer sequences were used (20). PCR reaction mixtures included 1 μL of bisulfite-treated DNA, 1 μL of 10 mmol/L dNTPs, 0.5 μL of dimethyl sulfoxide, 1.5 μL of 50 mmol/L MgSO4, 0.2 μL Platinum Taq DNA polymerase (Invitrogen), 2.5 μL 10× buffer, and 150 nmol/L of each primer. Reaction conditions were as follows: 95°C for 2 minutes, 95°C for 30 seconds, 56°C for 30 seconds, 72°C for 1 minute, 43 cycles, 72°C for 3 minutes. Half of each reaction mixture was run on a 2% agarose gel.

Immunohistochemistry

Immunostaining was conducted on a Discovery Ultra automated device (Ventana) with the following antibodies and concentrations: p53 (1:40), phospho-PRAS40 (p-PRAS40; 1:40), IDH R132H (Dianova; 1:30), PDGFRA (1:100), Hu (Invitrogen; 1:500), GLAST (1:50), and EGFR (1:100). All antibodies were procured from Cell Signaling Technology, unless otherwise indicated. See figures for representative micrographs and scoring metrics.

Expression profiling

Total RNA quality was assessed by RPL13a reverse transcription-PCR (RT-PCR) for all samples (27), and an exclusion cutoff was set at ΔCt (tumor) − ΔCt (normal brain) < 6.75. Qualified samples were analyzed on HT-12 Expression BeadChips following cDNA-mediated annealing, selection, extension, and ligation (Illumina). Quantile-normalized expression data from the 61 highest quality tumor samples, along with the one normal brain, were filtered using coefficient of variation (top 10%), median expression (top 25%), and average deviation to generate lists of 200, 250, 300, and 350 genes. Consensus k-means clustering (for k = 2, 3, 4, and 5) was applied using GenePattern software, followed by positive silhouette width (R package silhouette) to identify “core samples” within each cluster. Significance analysis of microarray (SAM; ref. 28) was then conducted on core samples from each cluster relative to samples from the remaining 2 clusters followed by receiver operating characteristic (ROC) analysis (R package colAUC) to rank subclass genes with significant upregulation (>2.0-fold). Subsequent classification of the entire MSKCC data set (N = 81) proceeded by uncorrelated shrunken centroid (USC) methodology (29) with statistical parameters r = 1.0 and d = 0. Centroids were initially established using a training set consisting of the best 50% of samples for each subclass as determined by positive silhouette width for the refined signature gene set (N = 24).

Affymetrix HGU133 expression data for an additional 148 WHO II and III astrocytomas were downloaded directly from the REMBRANDT website (https://caintegrator.nci.nih.gov/rembrandt/). Expression levels for genes with more than one probe were averaged following quantile normalization. SAM analysis was repeated on the MSKCC data set clusters based on USC classification at a fold change value of zero and significant genes also present in the REMBRANDT data set were ranked by ROC. The top 60 genes from each cluster list (180 in all) were then selected and USC classification was trained on MSKCC samples before its application to REMBRANDT samples. To assemble a classifier optimized for MSKCC and REMBRANDT data, SAM analysis was reapplied to clusters identified by independent USC classification in both sample sets. A common list of significant genes for each cluster in both data sets was identified. ROC analysis was then conducted to rank cluster genes for both data sets and the sum of ROC values was then taken as a measure to order genes. The top 100 upregulated genes for each subclass were assembled in a final 300-gene signature. Consensus k-means clustering (k = 3) in MeV was then applied to generate final cluster assignments.

Global copy number

Affymetrix 100K SNP array data were downloaded from the REMBRANDT website and analyzed using GenePattern software (Broad Institute, Cambridge, MA). Raw copy number data for tumor/normal sample pairs were converted into single-nucleotide polymorphism (SNP) files using the SNPFileCreater module applying invariant set normalization with median reference. SNP files were then converted to copy number files using the CopyNumberDivideByNormals module followed by GLAD segmentation analysis. Significant copy number alterations were determined using the GISTIC module with a q value (false discovery rate) of less than 0.05.

Developmental signature analysis

Correlations between REMBRANDT tumor expression profiles and selected gene signatures were made using single-sample gene set enrichment analysis (GSEA) as described previously (15). Gene signatures for differentiated neurons, oligodendrocytes, or astrocytes were obtained from a published murine database (30). Gene Ontology (GO) lists (GO: 0048666; neuron_development and GO:0009888; tissue_development) were retrieved from the Molecular Signatures Database (Broad Institute).

Statistics

Survival curves were analyzed by the Gehan–Breslow–Wilcoxon test along with multivariate Cox regression (R package). Age distributions were evaluated by 2-tailed t test. All other univariate associations were evaluated using Fisher's exact test, whereas multinomial or binomial regression models (R package) were used to assess multivariate associations. When applicable, significance cutoff values account for Bonferroni correction.

Results

IDH mutation in diffuse astrocytic gliomas correlates with molecular features and biologic behavior

We collected FFPE tissue blocks and, when available, clinical and radiographic data from 101 cases of diffuse astrocytic glioma (WHO II and III) diagnosed at MSKCC from 2000 to 2010. All material was reviewed by a neuropathologist (J.T. Huse) before inclusion. Our final cohort consisted of 35 WHO II and 66 WHO III tumors, 76 primary and 25 recurrent. Of the recurrent tumors, 9 had been previously treated with surgery alone with the remaining cases receiving additional therapy in the form of radiation (2 cases), chemotherapy (1 case), or both (13 cases). Kaplan–Meier analysis following WHO grade stratification revealed no significant differences in overall survival between cases having received either surgery alone or surgery combined with additional therapy (Supplementary Fig. S1). A number of standard demographic and clinical parameters such as gender, age at diagnosis, extent of surgical resection, and Karnofsky performance status did not vary significantly between WHO grades (Table 1).

View this table:
Table 1.

Demographic and clinical features of diffuse astrocytic gliomas (MSKCC sample set) stratified by WHO grade and IDH mutational status

To screen for IDH mutations in our sample set, we used mass spectrometry–based genotyping for IDH1 and IDH2 along with immunohistochemical (IHC) staining for IDH1 R132H (31), by far the most common glioma-associated IDH mutation. We identified IDH mutations in 69 of 101 cases (69%; 67 IDH1 and 2 IDH2), with 95% concordance between methods. We then conducted a multidimensional molecular profiling analysis of our sample set, incorporating an array of genomic, epigenomic, and IHC parameters previously implicated in glioma biology, and integrated these results with IDH mutational status (Fig. 1A; Supplementary Fig S2). We found that IDH mutant (IDH mt) tumors were significantly enriched for p53 abnormalities—point mutations in TP53 (P = 0.036) and/or extensive (>30% of tumor cell nuclei) p53 immunopositivity (P = 0.001)—as well as detectable PDGFRA IHC expression (P < 0.0001). IDH wild-type (IDH wt) tumors instead exhibited enhanced PI3K/AKT signaling, as indicated by both strong IHC staining for the AKT effector phospho-PRAS40 (p-PRAS40; P = 0.0002) and genomic loss of the phosphoinositide 3-kinase (PI3K) regulator PTEN as measured by FISH (P = 0.0001). In addition, PTEN promoter methylation, as assessed by MSP, was associated with IDH mutation (P = 0.0003), whereas focused Q-PCR revealed enrichment in IDH wt tumors for high-level EGFR amplification (P < 0.0001). Finally, CDKN2A promoter methylation showed no preferential association with either IDH mt or IDH wt tumors (P = 0.532) and Q-PCR for PDGFRA revealed no instances of high-level amplification (data not shown).

Figure 1.
Figure 1.

IDH mutational status delineates molecularly and clinically distinct subsets of diffuse astrocytic gliomas (MSKCC sample set). A, schematic showing associations between IDH mutational status (wt/mt), PDGFRA expression (IHC), p53 immunopositivity (IHC), TP53 mutational status (mt/wt), CDKN2A promoter methylation (Meth), EGFR amplification (CN), PTEN copy number loss (FISH), PTEN promoter methylation (Meth), and p-PRAS40 expression (IHC). B, amalgamated radiographic findings (MRI) showing the frequency of lobar involvement across IDH mt and IDH wt tumor subsets (blue, frontal; red, parietal; green, temporal; orange, occipital). Sample sizes are in parentheses. C, Kaplan–Meier curves showing overall survival for subsets of the diffuse astrocytic glioma sample set stratified by either IDH mutation, WHO grade, or both. Sample sizes are in parentheses. α, P = 0.038; β, P = 0.002; other P values shown.

IDH mutation also delineated distinct clinical profiles. IDH mt tumors occurred in significantly younger patients (Table 1) with longer overall survival (Fig. 1C) than their IDH wt counterparts. This latter trend was stronger than that seen in the same patient cohort following conventional WHO grade stratification and remained even when WHO II and WHO III tumors were analyzed separately (Fig. 1C). Furthermore, IDH mt tumors exhibited a predilection for frontal lobe localization by MRI scan that was not apparent for their IDH wt counterparts (Fig. 1B). These findings echo similar data from various WHO II–IV glioma populations (5, 11, 32, 33) and underscore generalizable trends for IDH mutation across glioma biology.

Transcriptional profiling reveals three molecular subclasses of diffuse astrocytic glioma

To further investigate biologic heterogeneity within diffuse astrocytic gliomas, we conducted global transcriptional profiling on 80 MSKCC samples for which sufficient RNA of adequate quality was available (see Materials and Methods; GEO accession number GSE35158). Material from one normal brain was also analyzed. Consensus k-means clustering conducted on 62 high-quality samples (61 tumors and 1 normal brain; see Materials and Methods for quality control metrics) based on 300 genes with strong, yet variable, expression across tumors revealed 3 defined subgroups, with clustering stability peaking at k = 3 (Supplementary Fig. S3C). These findings were stable when similarly filtered gene lists of different sizes—200, 250, or 350 genes—were analyzed (Supplementary Fig. S3A, S3B, and S3D). Positive silhouette width was then applied as a metric to determine core members within each subclass (Supplementary Fig S4). Given that cytotoxic therapy may alter gene expression profiles, we also conducted an analogous workflow excluding the 6 pretreated recurrent tumors that were present in the primary clustering set of 62 samples. Reassuringly, 95% of the remaining core cluster members (41 of 43) retained their prior classification in this setting (data not shown). Expression data from the core sample set were collectively used in SAM and ROC analyses to identify and rank subclass-specific genes. A balanced signature set of 177 significantly upregulated genes (59 per subclass; Supplementary Table S1) was generated and applied in an USC classifying algorithm to all 80 samples (Fig. 2A). On the basis of molecular and clinical characteristics (see below), the 3 subclasses were named neuroblastic, early progenitor-like (EPL), and preglioblastoma.

Figure 2.
Figure 2.

A and B, standardized heatmaps showing expression subclasses for diffuse astrocytic gliomas. A, a signature of 177 genes (59 per subclass), optimized from Illumina HT-12 data, delineates subclasses in 81 FFPE samples from the MSKCC set (MSKCC). B, a derivative of this initial signature (180 genes, 60 per subclass), where genes absent on the Affymetrix HGU133 platform have been replaced by others delineating subclass on the HT-12 platform, recapitulates subclasses in 148 REMBRANDT fresh-frozen samples (REMB). C, SAM analysis based on subclass constituents in both MSKCC and REMBRANDT data sets identifies shared signature genes for all subclasses (highlighted in yellow along with other tabulated overlaps). Numbers of genes obtained from each SAM analysis are shown in parentheses. D, a final shared signature set of 300 genes (100 per subclass) effectively delineates subclasses in both data sets (MSKCC/REMB). NB, neuroblastic; PG, preglioblastoma.

The robustness of these subclass distinctions was validated using expression data from an independent set of 148 WHO II and III astrocytomas, available from the REMBRANDT database. Only 132 of the 177 genes in our initial classifier were present on the Affymetrix HGU133 platform used by REMBRANDT. Consequently, we supplemented our primary signature with additional subclass-specific genes also present on the HGU133 array to reassemble a balanced classifier (60 genes per subclass; Supplementary Table S2). Remarkably, despite significant differences in tissue substrate between MSKCC and REMBRANDT cohorts (FFPE blocks vs. fresh-frozen specimens), this modified signature distinguished neuroblastic, EPL, and preglioblastoma subgroups within the REMBRANDT set, whose proportional composition approximated that of the MSKCC samples (Fig. 2B).

To formulate a classifying gene signature operative from FFPE as well as fresh-frozen tissue, we repeated SAM analysis for neuroblastic, EPL, and preglioblastoma tumors in both MSKCC and REMBRANDT data sets and identified intersections between analogous subgroups. Reassuringly, common genes were almost entirely confined to overlaps between identical subclasses (Fig. 2C). We then generated a final signature of 100 upregulated genes per subclass (Supplementary Table S3), which by consensus k-means clustering (k = 3) recapitulated prior classifier assignments for 83% (190 of 229) of tumors across both cohorts (Fig. 2D). Ingenuity pathway analysis showed distinct patterns of molecular and functional annotation associated with each tumor subclass (Supplementary Tables S4–S9). Neuroblastic signature genes mapped extensively to molecular networks concerned with mature neuronal biology, such as serotonin receptor signaling and γ—aminobutyric acid (GABA) receptor signaling. In contrast, EPL signature genes were enriched in developmental pathways such as bone morphogenetic protein (BMP) signaling and WNT signaling. Finally, preglioblastoma signature genes repeatedly showed Ingenuity-based functional associations with brain cancer, glioma, and glioblastoma.

Transcriptional subclasses of diffuse astrocytic gliomas demonstrate distinct molecular and clinical features

Integrating subclass assignments with additional molecular profiling data for MSKCC samples yielded several significant correlations (Table 2). Perhaps most importantly, IDH wt tumors were largely restricted to preglioblastoma subclass (21 of 24), which, parenthetically, also contained the one normal brain specimen, whereas IDH mt tumors were almost exclusively found in neuroblastic and EPL subclasses (55 of 56). Additional molecular and clinical associations followed as would be predicted by IDH mutational status. For instance, the IDH wt preglioblastoma subclass was enriched in EGFR amplification, PTEN loss, and high p-PRAS40 levels but relatively deficient in TP53 abnormalities and PDGFRA expression. EPL and neuroblastic tumors, both largely IDH mt, notably differed in their frequency of TP53 mutation and/or p53 immunopositivity, each significantly associated with EPL subclass. Moreover, PDGFRA expression correlated more robustly with EPL tumors, whereas a near absence of strong p-PRAS40 immunostaining characterized neuroblastic subclass. Neuroblastic tumors also showed the highest percentage of WHO II constituents whereas preglioblastoma tumors were predominantly WHO III. Notably, several of these associations remained significant following multivariate regression modeling. When our molecular profiling data were similarly stratified by WHO grade, correlations were generally weaker, with only 2 parameters—p53 immunohistochemistry and p-PRAS40 immunohistochemistry, both enriched in WHO III tumors—showing statistical significance following multivariate analysis (Table 2). Moreover, no compelling molecular or demographic distinctions were identified between primary and recurrent tumors (Supplementary Table S10). Finally, examining an array of semiquantitatively scored histopathologic characteristics revealed minimal differences between primary and recurrent tumors, particularly when compared with those distinguishing WHO grade and molecular subclass in our sample set (Supplementary Table S11).

View this table:
Table 2.

Demographic and molecular features of diffuse astrocytic gliomas (MSKCC sample set) stratified by expression subclass and WHO grade

While many of the molecular profiling parameters detailed in the preceding paragraph, most notably IDH mutational status, were not available for REMBRANDT samples, global copy number data were obtained for 50 cases. Raw data segmentation using the GLAD module (GenePattern software) was followed by GISTIC analysis to determine statistically significant regions of genomic gain and/or loss (Supplementary Tables S12–S14). This process revealed distinct copy number profiles associated with the different subclasses of diffuse astrocytic tumor (Fig. 3A). Preglioblastoma tumors exhibited copy number abnormalities commonly seen in primary glioblastomas (34), including EGFR amplification, low-level 7p gain, CDKN2A/B deletion, and 10q loss including but not limited to PTEN. In contrast, EPL tumors featured broad gains of 8q, a region that includes MYC, along with losses of 13q, 19q, and 9p23. Finally, neuroblastic tumor copy number profiles were relatively silent, with only 8q gain reaching a comparable level of statistical significance (q < 0.05).

Figure 3.
Figure 3.

Molecular subclasses of diffuse astrocytic glioma show distinct genomic and clinical characteristics. A, copy number (CN) profiles for diffuse astrocytic gliomas derived from SNP array data arranged by molecular subclass. Statistically significant CN abnormalities (q < 0.05) as determined by the GISTIC algorithm are also listed. B, charts showing age distributions for molecular subclasses. C–E, Kaplan–Meier curves showing overall survival for molecular subclasses of diffuse astrocytic gliomas. Statistically significant survival differences are indicated along with their associated P values. Sample sizes are in parentheses. REMBRANDT data are depicted with (E) and without (D) WHO grade stratification. NB, neuroblastic; PG, preglioblastoma.

Molecular subclass showed strong clinical correlations across both MSKCC and REMBRANDT sample sets. As would be predicted by their IDH wt status, preglioblastoma tumors occurred in older patients with shorter overall survival, whereas EPL and neuroblastic tumors arose in younger patients with better outcome (Fig. 3B–D). Interestingly, when each transcriptional subclass was viewed independently in the REMBRANDT data set—which contained sufficient samples to power analysis—survival differences between WHO II and WHO III tumors were not observed (Fig. 3E), although statistical significance was approached for neuroblastic tumors (P = 0.138). This despite the fact that WHO grade stratification across the entire REMBRANDT cohort did delineate distinct survival groups (Supplementary Fig S5). Furthermore, multivariate Cox regression analysis revealed that molecular subclass correlated better with overall survival than WHO grade. Specifically, preglioblastoma subclass designation was associated with an HR of 4.05 relative to EPL subclass [95% confidence interval (CI), 3.11–5.26; P < 0.0001], whereas WHO grade III carried an HR of 1.15 relative to WHO grade II (95% CI, 0.91–1.45; P = 0.54).

Molecular subclassification of diffuse astrocytic glioma reflects tissue lineage and neuroglial developmental stage

To further investigate biologic distinctions between neuroblastic, EPL, and preglioblastoma subclasses, we explored potential links to neuroglial lineage and development. We obtained gene sets enriched in differentiated neurons, oligodendrocytes, and astrocytes from a published mouse brain transcriptome (30), along with GO lists associated with developmental processes in either neurons or a variety of nonneuronal, mostly mesenchymal tissues. We then projected these signatures onto REMBRANDT expression data using single-sample GSEA and derived correlation scores with respect to each tumor's transcriptional profile (Fig. 4A). We found that both neuroblastic and EPL subclasses strongly correlated with neuronal development genes. However, only neuroblastic tumors exhibited consistent association with a more differentiated neuronal signature, whereas EPL tumor expression profiles were more strongly enriched in oligodendrocytic genes. Preglioblastoma tumors robustly correlated with signatures of nonneuronal tissue development and differentiated astrocytes, although subsets showed clear associations with mature neuronal and oligodendrocytic gene sets. This latter finding indicates fundamental heterogeneity in the preglioblastoma subclass with regard to differentiation phenotype and/or lineage derivation and recapitulates known data for primary glioblastoma (15).

Figure 4.
Figure 4.

Molecular subclasses of diffuse astrocytic gliomas resemble distinct tissue lineages and neuroglial precursor cell pools. A, expression signatures designating differentiated murine astrocytes (astro), neurons (neuro), and oligodendrocytes (oligo; ref. 30), along with GO gene sets associated with neuronal (neuron_dev) and nonneuronal (tissue_dev) tissue development were projected by single-sample GSEA on the transcriptional data from the REMBRANDT sample set. Correlation scores are shown in a standardized heatmap. B, schematic showing neurogenesis in the SVZ. The association of each developmental stage with EGFR, GLAST, and Hu expression is also shown. C, photomicrographs (400×) of representative NB and EPL tumors immunostained for Hu, GLAST, and EGFR. Hu staining is nuclear, whereas GLAST and EGFR staining are cytoplasmic. Immunopositivity for Hu was defined as nuclear staining in more than 50% of tumor cells. For EGFR, patchy, low-level staining was differentiated from stronger, more uniform expression. D, tabulated IHC results for Hu, GLAST, and EGFR across molecular subclasses. P values are indicated. NB, neuroblastic; PG, preglioblastoma.

Signature-based analysis linking neuronal development with neuroblastic and EPL subclass is consistent with recent work implicating neuroglial precursors in the forebrain subventricular zone (SVZ) as likely cells of origin for IDH mt glioma subtypes (32, 33). In the adult mammal, SVZ neurogenesis primarily serves to resupply olfactory bulb interneurons, although oligodendrocytic precursors may also derive from this anatomic niche (35). Recent work has identified both EGFR and the glutamate aspartate transporter, GLAST, as robust markers of early-stage SVZ progenitors (36, 37). In contrast, nuclear expression of the RNA-binding protein Hu (ELAVL2) has been associated with more differentiated precursors, such as migrating neuroblasts, along with mature neurons (Fig. 4B; refs. 36, 37). Immunohistochemistry in MSKCC samples for these 3 markers revealed that neuroblastic subclass significantly correlated with strong Hu immunopositivity, whereas EPL tumors were notably enriched in both detectable GLAST expression and strong, uniform staining for EGFR (Fig. 4C and D). These findings show additional associations between IDH mt diffuse astrocytic glioma subclasses and distinct stages of SVZ neurogenesis, with EPL and neuroblastic tumors resembling early-stage and late-stage neuroglial progenitors, respectively.

Discussion

WHO II and III diffuse astrocytic gliomas represent a contiguous spectrum of closely related disease entities, united by similar histopathology, frequent IDH mutation, and lack of 1p/19q deletion. Nevertheless, their variable clinical behavior almost certainly reflects underlying biologic heterogeneity. To probe these foundations, we undertook a multidimensional molecular profiling study aimed at integrating disease-relevant biomarkers with global expression profiling. Findings made in our MSKCC sample cohort, composed entirely of FFPE tissue, were recapitulated in an independent data set from REMBRANDT derived from fresh-frozen specimens. Such cross-validation underscores the utility of archival FFPE material in large-scale molecular profiling studies, especially in the setting of relatively uncommon disease entities.

Not surprisingly, we found IDH mutational status to be the single most important factor delineating biologically and prognostically distinct disease subgroups. More specifically, IDH mt tumors were associated with genomic abnormalities in TP53, upregulated PDGFRA expression, and PTEN promoter methylation, whereas IDH wt tumors exhibited a molecular profile more akin to that of primary glioblastomas, with enrichment for EGFR amplification, PTEN loss, and PI3K/AKT pathway activity. As such, our data recapitulate earlier studies on diffuse glioma (5, 6, 38), while also formally linking dysregulated PDGF signaling and PTEN promoter methylation with IDH mutation. In addition, IDH mutation in diffuse astrocytic gliomas correlated with a marked survival benefit that remained even after WHO grade stratification. While this finding is consistent with multiple published reports on WHO II, III, and IV gliomas (5, 6, 11), other large studies have failed to identify any increase in overall survival associated with IDH mutation specifically in WHO II astrocytomas (39, 40). While the precise reasons underlying this discrepancy within the literature are unclear, differences in sample cohorts with regard to histopathologic inclusion/exclusion criteria and patient management may contribute. Further prospective studies addressing this specific question would be of considerable value.

Global expression profiling has been effectively used to formulate clinically relevant stratification within primary glioblastoma (14, 15, 41). Recent computational work applied classifying signatures initially developed for primary glioblastoma—proneural, neural, classical, and mesenchymal—directly to a range of diffuse gliomas including WHO II and III astrocytomas (42). We chose a contrasting approach for the present study, focusing expression profiling on diffuse astrocytic gliomas alone, in the absence of other variants, with the aim of identifying endogenous disease subgroupings within this more restricted tumor spectrum. Our analysis revealed 3 robust subclasses, whose designations reflected not only the compositions and functional associations of their respective gene signatures but also the biologic behavior and developmental characteristics of their tumors.

Integrating molecular and clinical data with expression subclass showed compelling correlations, whose number and statistical significance exceed those associated with either WHO grade or primary/recurrent tumor status. And while many simply followed from the IDH mutational composition of the subclass in question, definitive distinctions between the 2 IDH mt subclasses, neuroblastic and EPL, were also revealed. Most strikingly, TP53 point mutations and/or strong nuclear p53 staining, the latter an established marker of p53 dysfunction, occurred with significantly higher frequency in EPL tumors. While the mechanisms underlying this disparity are unclear at this time, differential mutational susceptibilities in distinct precursor cell pools (see below) may play a role.

Genomic differences between molecular subclasses were further established by global copy number analysis. Both neuroblastic and EPL subclasses frequently showed 8q gain. However, neuroblastic tumors otherwise exhibited relatively silent copy number profiles compared with their EPL counterparts, raising an intriguing correlation between aneuploidy and TP53 mutational status. While p53 deficiency does not, in itself, appear to directly cause aneuploidy (43), recent work has shown that TP53 mutations may conspire with aneuploidy to promote oncogenesis (44, 45). EPL tumor copy number profiles also differed dramatically from those of preglioblastoma tumors, which prominently featured 7p gain coupled with 9p and 10q loss, a pattern highly reminiscent of primary glioblastoma. These findings echo an earlier report that established a clinically relevant stratification scheme for astrocytic tumors based on the presence of either 7p or 8q gain (46).

As would be predicted from its largely IDH wt composition, preglioblastoma subclass was associated with both higher age and lower overall survival relative to either neuroblastic or EPL subclass in 2 independent patient cohorts. Coupled with the molecular data discussed above, these clinical correlations indicate that preglioblastoma tumors likely represent variants of primary glioblastoma whose histopathology has yet to fully evolve into a bona fide WHO IV pattern. Indeed, a close pathogenic relationship between primary glioblastomas and lower grade IDH wt astrocytomas has been suggested in at least one earlier report (11). WHO grade–based stratification revealed no significant survival differences between WHO II and III tumors within each individual molecular subclass. Furthermore, multivariate Cox regression definitively showed subclass designations to be more predictive of overall survival than WHO grades. These findings suggest that formal WHO II versus III assessment in diffuse astrocytic gliomas may not be clinically relevant once molecular subclass has been established. Furthermore, they imply that survival differences between unstratified WHO II and III tumors, which we find in both patient cohorts examined in this study, may largely reflect fundamental differences in their respective molecular subclass compositions.

Recent data have implicated SVZ neuroglial progenitors as potential glioma cells of origin, particularly for IDH mutant variants (32, 33). In support of this conjecture, we find that IDH mt diffuse astrocytic gliomas exhibit both a geographic predilection for the frontal lobes and a strong correlation with neuronal development in gene expression space. Within IDH mt subclasses, the association of EPL tumors with GLAST and EGFR immunopositivity along with expression signatures of both mature oligodendrocytes and developing neurons implies origin from a more primitive, multipotent progenitor, whereas the association of neuroblastic tumors with Hu immunopositivity and a mature neuronal gene set suggests fundamental derivation from a late-stage precursor, perhaps closer to neuroblastic differentiation. That being said, we do not interpret these latter findings to suggest that neuroblastic tumors directly derive from postmitotic neuroblasts.

In conclusion, we have shown that molecularly and clinically distinct subclasses exist within lower grade diffuse astrocytic gliomas and are delineated by gene expression signatures and IDH mutation. The links between IDH mt tumor subclasses and SVZ progenitor compartments are particularly intriguing and invite thorough exploration in relevant in vivo systems to drive development of novel therapeutic strategies.

Disclosure of Potential Conflicts of Interest

No potential conflicts of interests were disclosed.

Authors' Contributions

Conception and design: D. Gorovets, T.A. Chan, J.T. Huse

Development of methodology: D. Gorovets, K. Kannan, E.R. Kastenhuber, N. Islamdoust, S.C. Jhanwar, I.K. Mellinghoff, J.T. Huse

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): D. Gorovets, E. Pentsova, S.C. Jhanwar, A. Heguy, T.A. Chan, J.T. Huse

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): D. Gorovets, K. Kannan, R. Shen, E.R. Kastenhuber, S.C. Jhanwar, A. Heguy, T.A. Chan, J.T. Huse

Writing, review, and/or revision of the manuscript: D. Gorovets, E.R. Kastenhuber, S.C. Jhanwar, T.A. Chan, J.T. Huse

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): D. Gorovets, E.R. Kastenhuber, I.K. Mellinghoff, T.A. Chan, J.T. Huse

Study supervision: S.C. Jhanwar, J.T. Huse

Generating data: C. Campos

Provided clinical information: E. Pentsova

Grant Support

J.T. Huse is a Leon Levy Foundation Young Investigator and a recipient of the AACR/Landon Foundation Innovator Award for Research in Personalized Cancer Medicine. This work was supported by the MSKCC Brain Tumor Center.

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.

Acknowledgments

The authors thank Eric Holland, Cameron Brennan, and Marc Rosenblum for their thoughtful comments during manuscript preparation; Daoqi Yu, Igor Dolgalev, and Olga Aminova for their technical assistance; and Lakshmi Nayak for her advice about the collection of clinical data.

Footnotes

  • Note: Supplementary data for this article are available at Clinical Cancer Research Online (http://clincancerres.aacrjournals.org/).

  • Received November 21, 2011.
  • Revision received February 24, 2012.
  • Accepted February 26, 2012.

References

  1. 1.↵
    1. Wen PY,
    2. Kesari S
    . Malignant gliomas in adults. N Engl J Med 2008;359:492–507.
    OpenUrlCrossRefPubMed
  2. 2.↵
    1. Louis DN,
    2. Ohgaki H,
    3. Wiestler OD,
    4. Cavenee WK,
    5. Burger PC,
    6. Jouvet A,
    7. et al.
    The 2007 WHO classification of tumours of the central nervous system. Acta Neuropathologica 2007;114:97–109.
    OpenUrlCrossRefPubMed
  3. 3.↵
    1. Huse JT,
    2. Holland EC
    . Targeting brain cancer: advances in the molecular pathology of malignant glioma and medulloblastoma. Nat Rev 2010;10;319–31.
    OpenUrlCrossRef
  4. 4.↵
    1. Parsons DW,
    2. Jones S,
    3. Zhang X,
    4. Lin JC-H,
    5. Leary RJ,
    6. Angenendt P,
    7. et al.
    An integrated genomic analysis of human glioblastoma multiforme. Science 2008;321:1807–12.
    OpenUrlAbstract/FREE Full Text
  5. 5.↵
    1. Sanson M,
    2. Marie Y,
    3. Paris S,
    4. Idbaih A,
    5. Laffaire J,
    6. Ducray F,
    7. et al.
    Isocitrate dehydrogenase 1 codon 132 mutation is an important prognostic biomarker in gliomas. J Clin Oncol 2009;27:4150–4.
    OpenUrlAbstract/FREE Full Text
  6. 6.↵
    1. Yan H,
    2. Parsons DW,
    3. Jin G,
    4. McLendon R,
    5. Rasheed BA,
    6. Yuan W,
    7. et al.
    IDH1 and IDH2 mutations in gliomas. N Engl J Med 2009;360:765–73.
    OpenUrlCrossRefPubMed
  7. 7.
    1. Dang L,
    2. White DW,
    3. Gross S,
    4. Bennett BD,
    5. Bittinger MA,
    6. Driggers EM,
    7. et al.
    Cancer-associated IDH1 mutations produce 2-hydroxyglutarate. Nature 2009;462:739–44.
    OpenUrlCrossRefPubMed
  8. 8.↵
    1. Ward PS,
    2. Patel J,
    3. Wise DR,
    4. Abdel-Wahab O,
    5. Bennett BD,
    6. Coller HA,
    7. et al.
    The common feature of leukemia-associated IDH1 and IDH2 mutations is a neomorphic enzyme activity converting alpha-ketoglutarate to 2-hydroxyglutarate. Cancer Cell 2010;17;225–34.
    OpenUrlCrossRefPubMed
  9. 9.↵
    1. Figueroa ME,
    2. Abdel-Wahab O,
    3. Lu C,
    4. Ward PS,
    5. Patel J,
    6. Shih A,
    7. et al.
    Leukemic IDH1 and IDH2 mutations result in a hypermethylation phenotype, disrupt TET2 function, and impair hematopoietic differentiation. Cancer Cell 2010;18;553–67.
    OpenUrlCrossRefPubMed
  10. 10.↵
    1. Xu W,
    2. Yang H,
    3. Liu Y,
    4. Yang Y,
    5. Wang P,
    6. Kim SH,
    7. et al.
    Oncometabolite 2-hydroxyglutarate is a competitive inhibitor of alpha-ketoglutarate-dependent dioxygenases. Cancer Cell 2011;19;17–30.
    OpenUrlCrossRefPubMed
  11. 11.↵
    1. Hartmann C,
    2. Hentschel B,
    3. Wick W,
    4. Capper D,
    5. Felsberg J,
    6. Simon M,
    7. et al.
    Patients with IDH1 wild type anaplastic astrocytomas exhibit worse prognosis than IDH1-mutated glioblastomas, and IDH1 mutation status accounts for the unfavorable prognostic effect of higher age: implications for classification of gliomas. Acta Neuropathologica 2010;120;707–18.
    OpenUrlCrossRefPubMed
  12. 12.↵
    1. Alizadeh AA,
    2. Eisen MB,
    3. Davis RE,
    4. Ma C,
    5. Lossos IS,
    6. Rosenwald A,
    7. et al.
    Distinct types of diffuse large B-cell lymphoma identified by gene expression profiling. Nature 2000;403:503–11.
    OpenUrlCrossRefPubMed
  13. 13.↵
    1. Perou CM,
    2. Sorlie T,
    3. Eisen MB,
    4. van de Rijn M,
    5. Jeffrey SS,
    6. Rees CA,
    7. et al.
    Molecular portraits of human breast tumours. Nature 2000;406:747–52.
    OpenUrlCrossRefPubMed
  14. 14.↵
    1. Phillips HS,
    2. Kharbanda S,
    3. Chen R,
    4. Forrest WF,
    5. Soriano RH,
    6. Wu TD,
    7. et al.
    Molecular subclasses of high-grade glioma predict prognosis, delineate a pattern of disease progression, and resemble stages in neurogenesis. Cancer Cell 2006;9:157–73.
    OpenUrlCrossRefPubMed
  15. 15.↵
    1. Verhaak RGW,
    2. Hoadley KA,
    3. Purdom E,
    4. Wang V,
    5. Qi Y,
    6. Wilkerson MD,
    7. et al.
    An integrated genomic analysis identifies clinically relevant subtypes of glioblastoma characterized by abnormalities in PDGFRA, IDH1, EGFR, and NF1. Cancer Cell 2009;17;98–110.
    OpenUrl
  16. 16.↵
    1. Cairncross JG,
    2. Ueki K,
    3. Zlatescu MC,
    4. Lisle DK,
    5. Finkelstein DM,
    6. Hammond RR,
    7. et al.
    Specific genetic predictors of chemotherapeutic response and survival in patients with anaplastic oligodendrogliomas. J Natl Cancer Inst 1998;90:1473–9.
    OpenUrlAbstract/FREE Full Text
  17. 17.↵
    1. Reifenberger G,
    2. Reifenberger J,
    3. Liu L,
    4. James CD,
    5. Wechsler W,
    6. Collins VP
    . Molecular genetic analysis of oligodendroglial tumors shows preferential allelic deletions on 19q and 1p. Am J Pathol 1994;145:1175–90.
    OpenUrlPubMed
  18. 18.↵
    1. Ducray F,
    2. El Hallani S,
    3. Idbaih A
    . Diagnostic and prognostic markers in gliomas. Curr Opin Oncol 2009;21:537–42.
    OpenUrlCrossRefPubMed
  19. 19.↵
    1. Funari FB,
    2. Fenton T,
    3. Bachoo RM,
    4. Mukasa A,
    5. Stommel JM,
    6. Stegh A,
    7. et al.
    Malignant astrocytic glioma: genetics, biology, and paths to treatment. Genes Dev 2007;21:2683–710.
    OpenUrlAbstract/FREE Full Text
  20. 20.↵
    1. McBride SM,
    2. Perez DA,
    3. Polley MY,
    4. Vandenberg SR,
    5. Smith JS,
    6. Zheng S,
    7. et al.
    Activation of PI3K/mTOR pathway occurs in most adult low-grade gliomas and predicts patient survival. J Neuro Oncol 2010;97:33–40.
    OpenUrlCrossRefPubMed
  21. 21.↵
    1. Paz MF,
    2. Yaya-Tur R,
    3. Rojas-Marcos I,
    4. Reynes G,
    5. Pollan M,
    6. Aguirre-Cruz L,
    7. et al.
    CpG island hypermethylation of the DNA repair enzyme methyltransferase predicts response to temozolomide in primary gliomas. Clin Cancer Res 2004;10:4933–8.
    OpenUrlAbstract/FREE Full Text
  22. 22.↵
    1. Stander M,
    2. Peraud A,
    3. Leroch B,
    4. Kreth FW
    . Prognostic impact of TP53 mutation status for adult patients with supratentorial World Health Organization Grade II astrocytoma or oligoastrocytoma: a long-term analysis. Cancer 2004;101:1028–35.
    OpenUrlCrossRefPubMed
  23. 23.↵
    1. Varela M,
    2. Ranuncolo SM,
    3. Morand A,
    4. Lastiri J,
    5. De Kier Joffe EB,
    6. Puricelli LI,
    7. et al.
    EGF-R and PDGF-R, but not bcl-2, overexpression predict overall survival in patients with low-grade astrocytomas. J Surg Oncol 2004;86:34–40.
    OpenUrlCrossRefPubMed
  24. 24.↵
    1. Watanabe T,
    2. Katayama Y,
    3. Yoshino A,
    4. Komine C,
    5. Yokoyama T
    . Deregulation of the TP53/p14ARF tumor suppressor pathway in low-grade diffuse astrocytomas and its influence on clinical course. Clin Cancer Res 2003;9:4884–90.
    OpenUrlAbstract/FREE Full Text
  25. 25.↵
    1. Wiencke JK,
    2. Zheng S,
    3. Jelluma N,
    4. Tihan T,
    5. Vandenberg S,
    6. Tamguney T,
    7. et al.
    Methylation of the PTEN promoter defines low-grade gliomas and secondary glioblastoma. Neuro Oncol 2007;9:271–9.
    OpenUrlAbstract/FREE Full Text
  26. 26.↵
    1. Qian X,
    2. Jin L,
    3. Shearer BM,
    4. Ketterling RP,
    5. Jalal SM,
    6. Lloyd RV
    . Molecular diagnosis of Ewing's sarcoma/primitive neuroectodermal tumor in formalin-fixed paraffin-embedded tissues by RT-PCR and fluorescence in situ hybridization. Diagn Mol Pathol 2005;14:23–8.
    OpenUrlCrossRefPubMed
  27. 27.↵
    1. Abramovitz M,
    2. Ordanic-Kodani M,
    3. Wang Y,
    4. Li Z,
    5. Catzavelos C,
    6. Bouzyk M,
    7. et al.
    Optimization of RNA extraction from FFPE tissues for expression profiling in the DASL assay. Biotechniques 2008;44:417–23.
    OpenUrlCrossRefPubMed
  28. 28.↵
    1. Tusher VG,
    2. Tibshirani R,
    3. Chu G
    . Significance analysis of microarrays applied to the ionizing radiation response. Proc Natl Acad Sci U S A 2001;98:5116–21.
    OpenUrlAbstract/FREE Full Text
  29. 29.↵
    1. Yeung KY,
    2. Bumgarner RE
    . Multiclass classification of microarray data with repeated measurements: application to cancer. Genome Biol 2003;4:R83.
    OpenUrlCrossRefPubMed
  30. 30.↵
    1. Cahoy JD,
    2. Emery B,
    3. Kaushal A,
    4. Foo LC,
    5. Zamanian JL,
    6. Christopherson KS,
    7. et al.
    A transcriptome database for astrocytes, neurons, and oligodendrocytes: a new resource for understanding brain development and function. J Neurosci 2008;28:264–78.
    OpenUrlAbstract/FREE Full Text
  31. 31.↵
    1. Capper D,
    2. Zentgraf H,
    3. Balss J,
    4. Hartmann C,
    5. von Deimling A
    . Monoclonal antibody specific for IDH1 R132H mutation. Acta Neuropathologica 2009;118:599–601.
    OpenUrlCrossRefPubMed
  32. 32.↵
    1. Lai A,
    2. Kharbanda S,
    3. Pope WB,
    4. Tran A,
    5. Solis OE,
    6. Peale F,
    7. et al.
    Evidence for sequenced molecular evolution of IDH1 mutant glioblastoma from a distinct cell of origin. J Clin Oncol 2011;29:4482–90.
    OpenUrlAbstract/FREE Full Text
  33. 33.↵
    1. Stockhammer F,
    2. Misch M,
    3. Helms HJ,
    4. Lengler U,
    5. Prall F,
    6. von Deimling A,
    7. et al.
    IDH1/2 mutations in WHO grade II astrocytomas associated with localization and seizure as the initial symptom. Seizure 2012;21:194–7.
    OpenUrlCrossRefPubMed
  34. 34.↵
    1. The Cancer Genome Atlas Research Network
    . Comprehensive genomic characterization defines human glioblastoma genes and core pathways. Nature 2008;455:1061–8.
    OpenUrlCrossRefPubMed
  35. 35.↵
    1. Kriegstein A,
    2. Alvarez-Buylla A
    . The glial nature of embryonic and adult neural stem cells. Annu Rev Neurosci 2009;32:149–84.
    OpenUrlCrossRefPubMed
  36. 36.↵
    1. Abrous DN,
    2. Koehl M,
    3. Le Moal M
    . Adult neurogenesis: from precursors to network and physiology. Physiol Rev 2005;85:523–69.
    OpenUrlAbstract/FREE Full Text
  37. 37.↵
    1. Platel JC,
    2. Gordon V,
    3. Heintz T,
    4. Bordey A
    . GFAP-GFP neural progenitors are antigenically homogeneous and anchored in their enclosed mosaic niche. Glia 2009;57:66–78.
    OpenUrlCrossRefPubMed
  38. 38.↵
    1. Hartmann C,
    2. Meyer J,
    3. Balss J,
    4. Capper D,
    5. Mueller W,
    6. Christians A,
    7. et al.
    Type and frequency of IDH1 and IDH2 mutations are related to astrocytic and oligodendroglial differentiation and age: a study of 1,010 diffuse gliomas. Acta Neuropathologica 2009;118:469–74.
    OpenUrlCrossRefPubMed
  39. 39.↵
    1. Thon N,
    2. Eigenbrod S,
    3. Kreth S,
    4. Lutz J,
    5. Tonn JC,
    6. Kretzschmar H,
    7. et al.
    IDH1 mutations in grade II astrocytomas are associated with unfavorable progression-free survival and prolonged postrecurrence survival. Cancer 2011;118:452–60.
    OpenUrlPubMed
  40. 40.↵
    1. Kim YH,
    2. Nobusawa S,
    3. Mittelbronn M,
    4. Paulus W,
    5. Brokinkel B,
    6. Keyvani K,
    7. et al.
    Molecular classification of low-grade diffuse gliomas. Am J Pathol 2010;177:2708–14.
    OpenUrlCrossRefPubMed
  41. 41.↵
    1. Huse JT,
    2. Phillips HS,
    3. Brennan CW
    . Molecular subclassification of diffuse gliomas: seeing order in the chaos. Glia 2011;59:1190–9.
    OpenUrlCrossRefPubMed
  42. 42.↵
    1. Cooper LA,
    2. Gutman DA,
    3. Long Q,
    4. Johnson BA,
    5. Cholleti SR,
    6. Kurc T,
    7. et al.
    The proneural molecular signature is enriched in oligodendrogliomas and predicts improved survival among diffuse gliomas. PLoS One 2010;5:e12548.
    OpenUrlCrossRefPubMed
  43. 43.↵
    1. Bunz F,
    2. Fauth C,
    3. Speicher MR,
    4. Dutriaux A,
    5. Sedivy JM,
    6. Kinzler KW,
    7. et al.
    Targeted inactivation of p53 in human cells does not result in aneuploidy. Cancer Res 2002;62:1129–33.
    OpenUrlAbstract/FREE Full Text
  44. 44.↵
    1. Li M,
    2. Fang X,
    3. Baker DJ,
    4. Guo L,
    5. Gao X,
    6. Wei Z,
    7. et al.
    The ATM-p53 pathway suppresses aneuploidy-induced tumorigenesis. Proc Natl Acad Sci U S A 2010;107:14188–93.
    OpenUrlAbstract/FREE Full Text
  45. 45.↵
    1. Rausch T,
    2. Jones DT,
    3. Zapatka M,
    4. Stutz AM,
    5. Zichner T,
    6. Weischenfeldt J,
    7. et al.
    Genome sequencing of pediatric medulloblastoma links catastrophic DNA rearrangements with TP53 mutations. Cell 2012;148:59–71.
    OpenUrlCrossRefPubMed
  46. 46.↵
    1. Nishizaki T,
    2. Kubota H,
    3. Harada K,
    4. Harada K,
    5. Ito H,
    6. Suzuki M,
    7. et al.
    Clinical evidence of distinct subgroups of astrocytic tumors defined by comparative genomic hybridization. Human Pathol 2000;31:608–14.
    OpenUrlCrossRefPubMed
View Abstract
Back to top
View this article with LENS

Open full page PDF
Article Alerts
Email Article
Citation Tools
IDH Mutation and Neuroglial Developmental Features Define Clinically Distinct Subclasses of Lower Grade Diffuse Astrocytic Glioma
Daniel Gorovets, Kasthuri Kannan, Ronglai Shen, Edward R. Kastenhuber, Nasrin Islamdoust, Carl Campos, Elena Pentsova, Adriana Heguy, Suresh C. Jhanwar, Ingo K. Mellinghoff, Timothy A. Chan and Jason T. Huse
Clin Cancer Res May 1 2012 (18) (9) 2490-2501; DOI: 10.1158/1078-0432.CCR-11-2977
del.icio.us logo Digg logo Reddit logo Twitter logo CiteULike logo Facebook logo Google logo Mendeley logo

Jump to section

Advertisement