IDH1 and IDH2 Mutations Are Prognostic but not Predictive for Outcome in Anaplastic Oligodendroglial Tumors: A Report of the European Organization for Research and Treatment of Cancer Brain Tumor Group

Recent sequencing of the genome of glioblastoma identified novel mutations in the isocitrate dehydrogenase1 gene (IDH1; refs. 1–5). These mutations occur in the highly conserved residue R132 in the active site of IDH1 and mostly concern Arg→His amino acid substitutions (codon CGT→CAT change). Less frequently, CGT→TGT (substitution of Arg→Cys), CGT→GGT (Arg→Gly), and CGT→AGT (Arg→Ser) mutations are present. IDH1 R132 mutations are present in 55% to 80% of grade II and III oligodendroglioma and astrocytoma but rarely in primary glioblastoma nor in a variety of other primary brain tumors including pilocytic astrocytoma. Although TP53 mutations and 1p/19q loss are mutually exclusive in glioma, IDH1 mutations are present in TP53 mutated and in 1p/19q codeleted tumors. This, in combination with the findings in patients with multiple biopsies in which there were no cases that acquired IDH1 mutations after the acquisition of either TP53 mutations or combined 1p19q loss suggests that the occurrence of IDH1 mutations is an early event in the tumorigenesis of diffuse glioma (6). Although the initial observations did not identify this mutation in other tumors, sequencing of the acute myeloid leukemia genome recently identified IDH1 mutations in 15 of 187 acute myelogenous leukemia patients, predominantly in patients with normal cytogenetic status (7–9). Less frequently, glial tumors show mutations in the corresponding codon 172 of the IDH2 gene, which codes for a mitochondrial enzyme with a similar function (3). Although initial studies focused on the decreased enzymatic conversion of isocitrate to α-ketoglutarate in the presence of an IDH1 mutation, a recent study showed that the mutated IDH1 enzyme increased the NADPH-dependent conversion of α-ketoglutatrate to R(−)-2-hydroxyglutarate (10). Of note, the presence of the wild-type IDH may actually provide the substrates required for this conversion. Inborn errors of R(−)-2-hydroxyglutarate metabolism are related to the development of brain tumors (11, 12). These data suggest that by an altered substrate affinity with a gain of function, IDH1 mutations may indeed act as an oncogene, and presents a potentially drugable target.

Retrospective studies have found a major favorable prognostic effect of the presence of IDH1 and IDH2 mutations on survival of grade II and grade III glial tumors. However, because patients in these studies patients were managed heterogeneously, it is still unknown if IDH1 and IDH2 mutations are predictive markers for outcome to treatment or prognostic markers. Moreover, these studies are on retrospective data sets. The prospective randomized phase III European Organization for Research and Treatment of Cancer (EORTC) study 26951 investigated the benefit of six cycles of adjuvant procarbazine, 1-(2-chloroethyl)-3-cyclohexyl-l-nitrosourea, and vincristine (PCV) chemotherapy in anaplastic oligodendroglial tumors (13). The study showed that adjuvant PCV chemotherapy improves progression-free survival (PFS) but not overall survival (OS), most likely due to the effects of crossover at the time of progression. Further details including molecular analysis of the study have been published elsewhere (14). We used this study to investigate if IDH mutations predict outcome to chemotherapy and to correlate IDH mutation status to other known markers related to outcome.

Patients were eligible for EORTC study 26951 if they had been diagnosed by the local pathologist with anaplastic oligodendroglioma or anaplastic oligoastrocytoma (AOA) with at least 25% oligodendroglial elements according to the 1993 edition of the WHO classification of brain tumors (15), had at least three of five anaplastic characteristics (high cellularity, mitoses, nuclear abnormalities, endothelial proliferation, or necrosis), were between ages 16 and 70 y, had an Eastern Cooperative Oncology Group performance status of 0 to 2, and had not undergone prior chemotherapy or radiotherapy to the skull. Patients were randomized between treatment with radiotherapy alone versus radiotherapy followed by six cycles of adjuvant PCV chemotherapy. The clinical and molecular details of this study have been published elsewhere; for eligibility, the study required written informed consent (13, 14). Patients were included on the local diagnosis, and central pathology review was part of the study. For this study, the WHO criteria 2007 were used (14). Those criteria do no longer consider necrosis consistent with the diagnosis anaplastic oligoastrocytoma; for this analysis, these tumors are considered together with glioblastoma multiforme (GBM).

All molecular studies were done using selected areas enriched for a high tumor cell percentage. DNA was extracted from formalin-fixed, paraffin-embedded tissues as previously described (16). Fluorescent in situ hybridization was used to assess copy number aberrations of chromosome 1p, 19q, 7, 10, and 10q, and the EGFR gene as described elsewhere (13, 14). MGMT promoter methylation was assessed with methylation-specific multiplex ligation-dependent probe amplification, as previously described (17).

For statistical analysis, results on IDH1 and IDH2 mutations were taken together and correlated to clinical characteristics [age, performance status, involved lobe (frontal versus other), molecular features (polysomy chromosome 7, EGFR amplification, loss of chromosome 1p/19q, loss of chromosome 10, loss of chromosome 10q, and MGMT promoter methylation), histologic features, diagnosis (pure versus mixed), and presence or absence of necrosis and endothelial proliferation] and to PFS and OS in treatment groups. PFS and OS were measured from the day of randomization. Patients provided written informed consent according to national and local regulations.

IDH1 and IDH2 alterations of the mutational hotspot codons R132 and R172, respectively, were assessed by bidirectional cycle sequencing of PCR-amplified fragments. Primers used were IDH1-forward 5′-CTCCTGATGAGAAGAGGGTTG-3′ and IDH1-reverse 5′-TGGAAATTTCTGGGCCATG-3′, and IDH2-forward 5′-TGGAACTATCCGGAACATCC-3′ and IDH2-reverse 5′-AGTCTGTGGCCTTGTACTGC-3′, respectively.

The Kaplan-Meier technique was used to estimate PFS and OS. The prognostic significance of the IDH1 and IDH2 mutations for PFS and OS were first univariately analyzed. For multivariate analysis, the major prognostic clinical variables used were as follows: type of surgery (resection or biopsy), WHO performance status (0, 1, 2), age (<50, ≥50), location (frontal versus nonfrontal), the central histology review diagnosis (AOD or AOA), endothelial abnormalities, necrosis, and the molecular factors combined (1p/19q loss, EGFR^amp, CHR7^poly, CHR10^loss, CHR10q^loss, and MGMT) promoter methylation. For the MGMT MLPA assay, the cutoff of 0.25 was used to distinguish between MGMT promoter–methylated and MGMT promoter–unmethylated tumors (17, 18). Association between factors except for the performance status was assessed by the Spearman correlation coefficient; the Fisher's exact test was used for inference. Correlations superior or equal to 0.2 (fair) were reported. For performance status (scored 0,1, and 2), the Wilcoxon rank-sum test was used. Survival analyses were done with the log-rank test and the Cox regression analysis with and without forward stepwise selection (5% significance). Peto's technique was used for interaction tests. Internal validation was done by bootstrap resampling technique (5% significance) to assess the generalizability of the models. Factors with a probability of inclusion in regression models of >60% based on 1,000 bootstrap samples were considered confirmed as independent prognostic factor. Factors with a probability of inclusion between 51% and 60% were retained in the final model but were not confirmed. Logistic regression with stepwise selection was used for multivariate analysis of the relationship between IDH1 and other factors (5% significance). These analyses were purely exploratory and no adjustment for multiplicity was done.

A total of 368 patients had been entered into the study. From 159 patients, enough material was present to assess the mutational status of IDH1 and IDH2. In 151 and 118 of these patients, respectively, the 1p/19q status and the MGMT promoter methylation status were known. The data from 29% of these patients have been presented previously as part of a single-center study (5). PFS and OS were similar for the patients with and without IDH1 and IDH2 mutation assessment (P = 0.195 and 0.565, respectively).

IDH1 mutations were identified in 73 cases (45.9%), and only one IDH2 mutation was identified. In all further analysis, this patient was considered among the IDH1 mutated.

Table 1 summarizes the baseline characteristics of the patients with and without IDH1 and IDH2 mutations. In the univariate analysis, the presence of IDH1 mutations was positively correlated with a previous resection for a low-grade tumor (r = 0.25), absence of necrosis (r = 0.33), absence of epidermal growth factor receptor (EGFR) amplification (r = 0.40), absence of polysomy of chromosome 7 (r = 0.31), absence of loss of chromosome 10 (r = 0.33), presence of 1p/19q loss (r = 0.43), and presence of MGMT promoter methylation (r = 0.42; all P < 0.01). Weaker positive correlations (r < 0.2; P between 0.05 and 0.01) were found with younger age (r = 0.16), the absence of mitosis (r = 0.17), frontal involvement (r = 0.18), and the central review histologic diagnosis according to the WHO 2007 classification (r = 0.18; ref. 19). In logistic regression, the presence of IDH1 and IDH2 mutations was predicted by younger age (P = 0.0021), the absence of necrosis (P = 0.0005), the absence of EGFR amplification P = 0.0007), the presence of 1p/19q loss (P = 0.001) and the presence of MGMT promoter methylation (P < 0.0001).

Both PFS and OS were strongly correlated with IDH1 mutational status. Median and percentage of 2-year PFS for wild-type IDH1 patients was 7.8 months and 19% versus 50 months and 65% for patients with IDH1-mutated tumors [hazard ratio (HR), 0.27; 95% confidence interval (95% CI), 0.18-0.40]. In the Cox multivariate analysis with stepwise selection, the presence of IDH mutations, necrosis, frontal localization, and 1p/19q loss were selected as independent prognostic factors. With bootstrap validation, both the presence of IDH1 mutations and of combined 1p/19q loss were confirmed.

Median and percentage of 2-year OS was 16 months and 37% for patients with IDH1 wild-type tumors and not reached and 83% for patients with IDH1-mutated tumors (HR, 024; 95% CI, 0.15-0.38). In the Cox multivariate analysis with stepwise selection, the presence of IDH1 mutations, of necrosis, and of 1p/19q loss were independent prognostic factors. Bootstrapping confirmed IDH1 mutations and 1p/19q loss, and necrosis was of borderline significance (included in 55% of samples).

In the subgroup of 91 that were confirmed anaplastic oligodendroglial tumors according to the WHO 2007 (anaplastic oligodendroglioma with or without necrosis and AOA without necrosis) at central review, 44 had an IDH1 mutation. This IDH1-mutated subgroup had a significantly better survival (HR, 0.19; 95% CI, 0.10-0.35; Fig. 1). Two-year survival for the IDH1 intact tumors was 38.3% (95% CI, 24.6-51.8) versus 88.4% (95% CI, 74.3-95.0) for the IDH1 mutated tumors.

For neither PFS nor OS, the presence of IDH1 mutations was related to outcome to adjuvant PCV chemotherapy (test for interaction, respectively, 0.70 and 0.94; Fig. 2). Table 2 summarizes median and 2-year PFS in relationship to treatment (radiotherapy or radiotherapy/PCV). For PFS, the HR (95% CI) reduction in the presence of an IDH1 mutation after radiotherapy only was 0.29 (0.17-0.50) and 0.24 (0.13-0.43) after treatment with radiotherapy/PCV.

As expected, IDH1 mutations were less frequent in tumors with glioblastoma features, but occurred in about half of the patients with a confirmed anaplastic oligodendroglial tumors at central review and in 86% of patients with combined 1p/19q loss. In the current data set, IDH2 mutations were rare (1:159). IDH1 mutations were more frequent in younger patients; patients having previously undergone a resection for a low-grade tumor; patients without the diagnosis of a glioblastoma or AOA with necrosis (synonym: glioblastoma with oligodendroglial differentiation) at central review; patients without necrosis in the tumor; patients with frontal involvement; and from the molecular perspective, patients without EGFR amplification, Trisomy 7, and loss of chromosome 10. These findings all fit with a low incidence of IDH1 mutations in primary glioblastoma, but with a high incidence of these mutations in grade II and grade III tumors. In addition, a strong correlation (P < 0.0001) was found with MGMT promoter methylation. IDH1 mutations were observed in 62% of the MGMT promoter–methylated tumors as opposed to only 10% of the MGMT-unmethylated tumors. Others also found a strong correlation with MGMT promoter methylation (58% IDH1 mutations in methylated versus 26% IDH1 mutations in unmethylated tumors; ref. 5). This contrasts with a German study, in which both IDH1 mutations and MGMT promoter methylation were independent prognostic variables, and in which analysis the presence of the 1p/19q codeletion lost its prognostic significance (20). Of note, combined loss of 1p/19q is also highly correlated with MGMT promoter methylation (5, 21, 22). Moreover, our current data and a previous report from our group confirm the presence of IDH1 mutations both in patients with the 1p/19q codeletion and with TP53 mutations, although in all the series, a substantial percentage of 1p/19q-codeleted or TP53-mutated tumors do not carry and IDH1 or IDH2 mutation (5, 6, 23).

The advantage of studying prognostic and predictive factors within a prospective randomized study is that the treatment heterogeneity that is usually disturbing retrospective studies is limited. This allows the assessment of the effect of the molecular marker on the outcome to the randomized treatment, thus making a distinction between markers of prognostic and of predictive significance. Clearly, the presence of IDH1 mutations are of major prognostic significance for outcome in this group of anaplastic oligodendroglial tumors. However, the present study gives no indication that the presence of IDH1 mutations predicts the outcome to adjuvant PCV chemotherapy. In a previous report on temozolomide chemotherapy in progressive low-grade astrocytoma, we observed no relationship between outcome and IDH mutations. This suggests that at present, the improved survival in IDH1-mutated tumors is primarily due to a less aggressive biological behavior, and not because of an improved outcome to chemotherapy treatment (24).

Previous analysis of EORTC study 26951 has shown that adjuvant PCV after radiotherapy increased PFS but not OS, presumably because of crossover at the time of recurrence. Combined loss of 1p/19q was found to be of prognostic but not of predictive significance for outcome to adjuvant PCV chemotherapy. Subsequent molecular analysis confirmed that because of the unclear histologic boundaries between oligodendroglial tumors and astrocytic tumors, a considerable percentage of the tumors included in this EORTC study carried a genotype that was resembles more the genetic alterations usually observed in primary glioblastoma (EGFR amplification, loss of 10, polysomy of 7; ref. 14). This also explains the relatively low frequency of IDH mutations in this series (46%) that contrast to most series on grade III gliomes (with IDH mutations in the range of 60-80%). Surprisingly, in the EORTC 26951 study, MGMT promoter methylation was found to be of prognostic significance, for both PFS and OS, and not predictive for outcome to PCV chemotherapy (17). To these data, the IDH1 and IDH2 data are now added. In the current analysis, which includes IDH status, the independent prognostic significance of MGMT promoter methylation is lost. This might be due to a lack of power in these exploratory analyses.

To conclude, the presence of IDH1 mutations do not predict outcome to adjuvant PCV chemotherapy. For grade III oligodendroglial tumors, the assessment of both the 1p/19q codeletion and IDH1 mutations offer additional prognostic information. The recent report of increased conversion of α-ketoglutaraat to R(−)-2-hydroxyglutarate by the mutated IDH1 enzyme suggests this mutation indeed acts as an oncogene and actually may present drugable targets (10). This opens an entire new horizon of potential therapeutic strategies for grade II and grade III glioma.

No potential conflicts of interest were disclosed.

Grant Support: Grants number 2U10CA11488-25 through 2U10CA11488-35 from the National Cancer Institute (Bethesda, Maryland, USA) and by a donation from the Dutch Cancer Society through the EORTC Charitable Trust. We thank the EORTC Translational Research Fund grant TRF 01/02, the AstraZeneca EORTC Translational Research grant AZ/01/02, as well as the Dutch Cancer Society grant DDHK 2005-3416 and EMC 2007-3932 for providing financial support for this work.

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.