Activation of the NRF2 pathway and its impact on the prognosis of anaplastic glioma patients

Reverse Transcription and Quantitative Real-time PCR

Quantitative PCR analysis was performed on an ABI7300 system. Primers and probes used for amplification of cDNAs are described in Supplementary Table S1.

Sequencing of IDH1, IDH2, NRF2, and KEAP1 Genes

Genomic DNAs from paired peripheral blood and primary tumor tissues were extracted from snap-frozen samples. Sequencing primers are shown in Supplementary Table S1.

Immunohistochemical Analysis

Tumor samples were fixed in 10% buffered formalin and embedded in paraffin. A guinea pig polyclonal antibody against p62 (Progen GP62, 1 : 3000) was used.

Microarray Analysis

Total RNA was extracted from 12 randomly selected anaplastic glioma tissues. The data were deposited to the Gene Expression Omnibus (accession no. GSE52942).

Quantitative High-resolution DNA Methylation Analysis

DNA methylation was examined using the MassARRAY technique (Sequenom). Primers used in this study are listed in Supplementary Table S1.

Establishment of T98 Cells Expressing IDH1 or IDH1 R132H

An expression vector for FLAG-tagged IDH1 or IDH1 R132H²¹ was introduced into T98 cells. Stable transformants were selected with 40 μg/mL blasticidin S.

Immunoblot Analysis

Whole cell extracts were prepared from frozen tumor samples. Nuclear extracts and whole-cell extracts were prepared from T98-derived cell lines. Band intensities were quantified using ImageJ64 (ImageJ 1.45s).

Chromatin Immunoprecipitation Assay

Anti-NRF2 antibody (Santa Cruz Biotechnology, sc-13032) was used for ChIP assay. Primers used in the ChIP assay are described in Supplementary Table S1.

NRF2 Knockdown Experiment

NRF2 small interfering (si)RNAs were purchased from Invitrogen and introduced into T98 cells using MP-100 MicroPorator (Digital Bio Technology).

Cell Viability Assay

Cells were challenged with BCNU (LKT Laboratories), and their viability was examined after 48 h using the Cell Counting Kit 8 (Dojindo Molecular Technologies).

Metabolite Measurements and Metabolomic Profiling of Cultured Cells

To measure the levels of intracellular metabolites, extracts were prepared from 2–6 × 10⁶ cells per sample and analyzed on a capillary electrophoresis–connected electrospray ionization–time-of-flight mass spectrometry system as previously described.²²

Examination of Possible Causes for NRF2 Activation in Anaplastic Gliomas and Glioblastomas

To explore underlying mechanisms for the elevated NRF2 activity, we first investigated whether anaplastic gliomas and glioblastomas harbored mutations in the NRF2 and KEAP1 genes. Among the 117 anaplastic gliomas, somatic missense and heterozygous mutations of KEAP1 were identified in 2 cases (1.7%). The clinical features of these cases are shown in Supplementary Table S3. One case was anaplastic astrocytoma (AA) (1 out of 58 cases; 1.7%) diagnosed as Turcot syndrome, given the familial history and coexistence of colon cancer, and the other was diagnosed as sporadic anaplastic oligodendroglioma (AO) (1 out of 31 cases; 3.2%). NRF2 mutations were not found in the 117 anaplastic gliomas. Mutations in the NRF2 and KEAP1 genes were not found in the 79 randomly selected glioblastomas or 5 glioblastoma cell lines (A172, U87, U251, U373, and T98).

We next examined other possible causes for the NRF2 activation. One is the reduced expression of KEAP1, and the other is the accumulation of p62, both of which inhibit NRF2 degradation and increase the expression of NQO1/GCLM. However, there were no correlations between KEAP1 and NQO1/GCLM expression in the tumor tissues from 95 anaplastic glioma or 52 glioblastoma cases (NQO1 vs KEAP1: R = 0.154, P = .136 and R = 0.070, P = .622 in anaplastic glioma and glioblastoma, respectively; GCLM vs KEAP1: R = 0.070, P = .498 and R = 0.148, P = .294 in anaplastic glioma and glioblastoma, respectively) (Fig. 1D and E and data not shown). Accumulation of p62 was not observed in any of the 95 anaplastic glioma or 10 glioblastoma tissues (data not shown). These results indicated that high expression of NRF2 target genes was not caused by the downregulation of KEAP1 or p62 accumulation. Thus, currently well-known causes did not account for the elevation of NRF2 target gene expression that was observed in the present cohort of anaplastic glioma and glioblastoma cases.

Correlations Between NRF2 Target Gene Expression and Prognosis in Anaplastic Glioma and Glioblastoma Cases

We examined the correlations between NRF2 target gene expression and the progression-free survival (PFS) and overall survival (OS) rates in anaplastic glioma and glioblastoma cases (Fig. 1F–I). In a univariate analysis, high expression of NRF2 target genes was associated with a worse PFS and OS rate in anaplastic gliomas (Fig. 1F and H). In contrast, the expression level was not associated with prognosis in glioblastomas (Fig. 1G and I). These results indicated that elevated expression of NRF2 target genes correlates with poor prognosis in anaplastic gliomas but not in glioblastomas.

Association of IDH1/2 Mutation Status With NRF2 Target Gene Expression

Based on the clear association between NRF2 target gene expression and prognosis, we focused on anaplastic gliomas for further study. Somatic mutations in the IDH1 or IDH2 gene have been found in a substantial portion of anaplastic glioma cases, for which the prognosis is significantly better than for those without the mutations.²⁰ We hypothesized that IDH1/2 mutation status correlates with expression levels of NRF2 target genes. To examine the NRF2 target gene expression in anaplastic gliomas with mutated IDH1/2, we conducted a microarray analysis to measure the expression levels of representative NRF2 target genes, including NQO1, HMOX1, GCLM, TXNRD1, and PRDX1, in 12 anaplastic gliomas with or without mutated IDH1/2 (Fig. 2A) as a pilot study. We found that the average relative expression levels of all 5 genes were lower in anaplastic gliomas with mutated IDH1/2 than in those with wild-type IDH1/2, and this difference was statistically significant for GCLM. To verify this observation that NRF2 target gene expression appeared to be reduced in anaplastic gliomas with mutated IDH1/2, we compared the expression levels of NQO1 and GCLM among all anaplastic glioma cases with and without mutated IDH1/2. The NQO1 and GCLM mRNA levels in IDH1/2-mutated tumors were significantly lower than those in IDH1/2 wild-type tumors (Fig. 2B), which was consistent with the protein abundance (Fig. 2C). When anaplastic gliomas were divided into 2 subgroups—AA and AO/anaplastic oligoastrocytoma (AOA)—an almost similar tendency was observed (Supplementary Fig. S1).

Open in a separate window

Fig. 2.

Expression of NRF2 target genes in anaplastic gliomas with and without mutated IDH1/2. (A) Heat map showing the expression of the 5 NRF2 target genes in anaplastic gliomas with (cases 1–6) and without (cases 7–12) mutated IDH1/2. A sample with a median value was defined as the control. The gray scale indicates the level of fold change in gene expression relative to the control value. The darkest and the brightest colors indicate a 4-fold increase and decrease, respectively. (B) Expression of NQO1 and GCLM in anaplastic gliomas with and without IDH1/2 mutations and glioblastomas relative to the normal brain. Data represent means ± SD. Each sample was analyzed in triplicate. *P < .05. (C) Protein abundance of NQO1 and GCLM in anaplastic gliomas with and without IDH1/2 mutations. Band intensities of NQO1 and GCLM were quantified and normalized with those of alpha-tubulin. Values of case 1 were set as one. (D–F) Profiles of CpG methylation ratio in the promoter regions of (D) MGMT, (E) NQO1, and (F) GCLM in anaplastic gliomas with mutated (n = 64) and wild-type IDH1/2 (n = 24). Data are expressed as the averages and SDs of the methylation ratios of each CpG unit. *P < .05. (G) Scattergrams demonstrating the correlations between NQO1 and NRF2 expression and between GCLM and NRF2 expression in anaplastic gliomas. (H) Expression of NRF2 in anaplastic gliomas with mutated and wild-type IDH1/2 genes relative to the normal brain. Data represent means ± SD. Each sample was analyzed in triplicate. *P < .05.

In contrast, there was no difference in NQO1 and GCLM expression between anaplastic gliomas with wild-type IDH1/2 and glioblastomas (Fig. 2B). The prevalence of high expressers of NRF2 target genes was significantly lower in anaplastic gliomas with mutated IDH1/2 than in those with wild-type IDH1/2 (5.8% vs 33%, P = .0012) and in glioblastomas (5.8% vs 33%, P = .0018). These results implied that NRF2 activity was lower in anaplastic gliomas with mutated IDH1/2 than in other malignant gliomas.

Examination of Possible Causes for the Reduced Expression of NRF2 Target Genes in Anaplastic Gliomas With Mutated IDH1/2

Given that hypermethylation phenotypes have been reported for malignant gliomas with IDH1 mutations,²³ we suspected that promoter methylation might reduce NRF2 target gene expression. To test this hypothesis, we exploited MassARRAY technology and accessed the methylation statuses of the cytosine–phosphate–guanine (CpG) islands residing in the promoter regions of NQO1 and GCLM. We also examined the O⁶-DNA methylguanine-methyltransferase gene (MGMT) as a positive control, for which the promoter CpGs are heavily methylated in IDH1-mutant gliomas.²⁴ Methylation ratios that spanned 16–18 CpG sites in the promoter region of each gene were quantified. We examined anaplastic gliomas with (64 cases) and without (24 cases) IDH1/2 mutations. The MGMT promoter was highly methylated in both groups, and, in particular, anaplastic gliomas with mutated IDH1/2 exhibited significantly higher methylation ratios in 7 of the 16 CpG sites, compared with those with wild-type IDH1/2 (Fig. 2D). In contrast, the methylation ratios in the promoter regions of NQO1 and GCLM were relatively low and mostly comparable, irrespective of the IDH1/2 status, except for a single CpG in the GCLM promoter (Fig. 2E and F). Thus, promoter methylation does not seem to account for the reduced expression of NRF2 target genes observed in the presence of IDH1/2 mutations.

We next examined the expression levels of KEAP1 and NRF2. There were no differences in KEAP1 expression between anaplastic gliomas with mutated IDH1/2 and those with wild-type IDH1/2 (data not shown). In contrast, the expression level of NRF2 positively correlated with those of NQO1 (R = 0.391, P < .001) and GCLM (R = 0.266, P = .02) in anaplastic gliomas (Fig. 2G). NRF2 expression was significantly lower in anaplastic gliomas with mutated IDH1/2 than in those with wild-type IDH1/2 (Fig. 2H). These results suggest that the reduced expression of NRF2 at the transcription level is one of the causes for the downregulation of NRF2 target genes in anaplastic gliomas with mutated IDH1/2. In subtypes of anaplastic gliomas, correlation between the decreased NRF2 expression and IDH1/2 mutation was clear in AO/AOA but not in AA (Supplementary Fig. S1). Of note, there was no significant association between the NRF2 mRNA level and the prognosis of anaplastic gliomas, probably due to the multilayered regulation of the NRF2 activity at transcription level and posttranscription level (Supplementary Fig. S2A and B).

An Exogenous IDH1 Mutant Molecule Suppresses NRF2 Activity in T98 Glioblastoma Cells

The clinical analysis revealed a correlation between IDH1/2 mutation and downregulation of NRF2 target genes. To examine whether the IDH1/2 mutation suppressed the NRF2 pathway, either exogenous FLAG-tagged IDH1 or IDH1 R132H, which is the most frequently encountered IDH1 mutant, was introduced into T98 glioblastoma cells, and stable transformants were obtained. Introduction of IDH1 R132H into T98 cells significantly slowed down the proliferation rate (data not shown). We first verified the similar expression levels of exogenous IDH1 and IDH1 R132H in an immunoblot analysis, using anti-FLAG antibody (Fig. 3A). Next, the expression levels of NRF2 target genes were quantified. Three major NRF2 target genes, NQO1, GCLC, and GCLM, were downregulated in IDH1 R132H–expressing cells (Fig. 3B), suggesting that the IDH1 mutation limits the NRF2 activity.

Open in a separate window

Fig. 3.

Effect of IDH1 R132H expression on NRF2 activity in T98 glioblastoma cells. (A) Expression levels of exogenously introduced IDH1 and IDH1 R132H. Exogenous IDH1 and IDH1 R132H were detected with anti-FLAG antibody. Lamin B was used as a loading control. (B and C) Relative mRNA levels of (B) NQO1, GCLC, and GCLM and (C) NRF2. Data represent the means ± SD of 5 independent experiments, and each sample was analyzed in triplicate. The expression levels in IDH1-expressing cells were set as one. **P < .005, ***P < .001. (D) NRF2 protein abundance in nuclear extracts. Lamin B was detected as a loading control. Band intensities of NRF2 were quantified and normalized with those of Lamin B. A value of IDH1-expressing cells was set as one. (E) Location and sequence of AREs in NQO1, GCLC, and GCLM gene loci. (F) ChIP assay with anti-NRF2 antibody in IDH1- and IDH1 R132H–expressing cells. In the control samples, rabbit IgG was added instead of anti-NRF2 antibody. NQO1 promoter, GCLC enhancer, and GCLM promoter, which contain AREs, and NQO1 exon 2, which does not contain AREs, were examined. Data represent the means ± SD of triplicate samples. A representative result of 3 independent experiments is shown. (G) Metabolomic profiling. Data represent the means ± SD of 3 independent samples. The expression levels in IDH1-expressing cells were set as one. P-values were calculated with Student's t-test. *P < .05, **P < .005, ***P < .001. Promoted and inhibited processes in IDH1 R132H–expressing cells are indicated with red and blue arrows, respectively. Abbreviations of the metabolites are described in the Supplementary information.

We next examined NRF2 expression at mRNA and protein levels. NRF2 mRNA level was decreased in the IDH1 R132H–expressing cells (Fig. 3C), which is consistent with the results obtained in the clinical study. Almost parallel to the mRNA reduction, the nuclear accumulation of NRF2 protein was also decreased in the presence of IDH1 R132H (Fig. 3D and Supplementary Fig. S3). To evaluate the DNA binding activity of NRF2, we conducted a ChIP assay with anti-NRF2 antibody. We examined the recruitment of NRF2 to AREs located in the NQO1 promoter, GCLC enhancer, and GCLM promoter (Fig. 3E). The second exon of NQO1 was used as a negative control locus without AREs. NRF2 was clearly recruited to the NQO1 promoter,²⁵ GCLC enhancer,²⁶ and GCLM promoter²⁷ in control cells expressing wild-type IDH1. In the presence of IDH1 R132H, however, the NRF2 recruitment to the NQO1 promoter was reduced, and no significant recruitment was observed to the GCLC enhancer or GCLM promoter (Fig. 3F). Thus, IDH1 mutation reduces the protein abundance and DNA binding of NRF2 in T98 cells. As the latter is more dramatic than the former, we surmise that the mutant IDH1 might affect, for unknown reasons, the DNA binding ability of NRF2.

We further conducted metabolomic profiling of T98 cells expressing IDH1 or IDH1 R132H to examine whether any NRF2-deficient signatures were detected in the latter cells (Fig. 3G). The elevation of 2-hydroxyglutarate (2-HG) (Fig. 3G, red box) verified the expected enzymatic activity of IDH1 R132H.²⁸ The clearest difference was observed in the glutathione synthesis pathway. In the presence of IDH1 R132H, glutathione and γ-glutamylcysteine were decreased (Fig. 3G, blue boxes), whereas glutamine and glutamate were increased. Because γ-glutamylcysteine is the first-step product in the glutathione synthesis catalyzed by γ-glutamylcysteine ligase (GCL), these results indicate that GCL activity was inhibited in IDH1 R132H–expressing cells and are consistent with the reduced expression of GCLC and GCLM, which encode the 2 subunits of GCL. Thus, the metabolomic profile of the IDH1 R132H–expressing cells exhibited the signature of NRF2 deficiency and supported our hypothesis that the IDH1 mutation suppressed glutathione synthesis of the NRF2 activities.

We would like to thank Drs Tomoyoshi Soga and Masahiro Sugimoto for providing their special skills of metabolome analysis; Nao Ota for technical support regarding metabolome analysis; and the Biomedical Research Core of the Tohoku University Graduate School of Medicine for their technical support.

Conflict of interest statement. None declared.