IDH -Mutant Gliomas

Isocitrate dehydrogenase ( IDH ) mutation is one of the most critical genomic alterations in lower grade and secondary glioblastoma patient. More than 90% of IDH mutation is located at codon R132 of IDH1 gene. IDH mutation produces oncometabolite “2-hydroxyglutarate” and induces epigenetic alteration, such as DNA global methylation and histone methylation. As a result, IDH mutation promotes early gliomagenesis. Since IDH mutation is the earliest genomic event and almost always retained during tumor progression, IDH mutation is expected as novel therapeutic target. Herein, we review the clinical characteristics of IDH -mutant gliomas, biological role of IDH mutation for gliomagenesis, and current and future therapeutic approach for IDH mutant tumors.


Introduction
The WHO 2016 classification integrates molecular and histological features in the diagnosis of gliomas. Among numerous genomic alterations, the isocitrate dehydrogenase (IDH) mutation is one of the most important genetic alterations found in this kind of tumor. As IDH mutation is a ubiquitous mutation in lower grade gliomas, the development of molecular target therapies against IDH mutations is expected. Here, we review IDH-mutant gliomas, focusing on their role in tumorigenesis and as novel therapeutic targets.

Discovery of IDH mutations in cancers
The presence of an isocitrate dehydrogenase (IDH) mutation was first discovered in colorectal cancers [1]. Parsons et al. [2] found mutations of the IDH1 (2q.33) in 12% of the glioblastomas (GBMs). Other large scale studies validated that IDH1 and IDH2 (IDH) mutations were found in the majority of secondary GBM and lower grade (WHO grade II and III) gliomas, whereas these were rarely found in adult primary and pediatric GBMs [2][3][4]. Almost all of the IDH1 mutations occur at codon 132, >90% of them exhibit a c.395G>A (R132H) substitution, followed by R132C [3, 5, 6]. Although the frequency was low, IDH2 mutations were also identified at codon 172 in gliomas [4,7].

Age distribution of IDH-mutant gliomas
According to some statistical analyses, the IDH-mutant GBM or anaplastic astrocytoma patients were more than 20 years younger than those with IDH-wildtype GBM [4]. In contrast, IDH-mutant GBM patients were only 4 years older than those with IDH1-mutant grade II and III astrocytoma [41]. This indicates that IDH-mutant glioma arises earlier than IDH-wildtype glioma (mostly GBM).

Prognosis of IDH-mutant gliomas
Parsons et al. [2] initially demonstrated that IDH1-mutant GBM patients survived about threefold longer than those with IDH1-wildtype GBM. Other groups verified that IDH1 mutation is a favorable prognostic biomarker in gliomas [4,42,43]. In addition to GBM, large amounts of clinical studies indicated that the IDH mutation was an independent prognostic factor in grade II and III gliomas [ 4,28,[43][44][45][46][47]. Notably, the prognosis of IDH1-mutant GBM is better than of IDH1wildtype AA [48]. Also, a prospective randomized study (NOA-04) revealed that IDH1 mutation, hypermethylation of the O 6 -methylguanine DNA-methyltransferase (MGMT) promoter, age, extent of resection, and oligodendroglial histology are independent prognostic factors in anaplastic gliomas [44]. Among them, the impact of IDH1 mutation conferred a stronger favorable prognosis than 1p/19q co-deletion, MGMT promoter methylation, and histology [44]. Collectively, IDH1 mutation is a convincing prognostic factor in gliomas, irrespective of tumor grade and histology.

Prognostic classification for gliomas
Suzuki et al. [28] distinguished lower grade gliomas on the basis of the presence of IDH1 mutation, TP53 mutation, and 1p/19q co-deletion. Accordingly, tumors were classified into three groups: type I (IDH1-mutant with 1p/19q co-deletion; favorable prognostic group), type II (IDH1-mutant with TP53 mutation; intermediate prognostic group), and type III (IDH1-wildtype; poor prognostic group). Eckel-Passow et al. [47] classified gliomas into five groups based on the mutation status of IDH1 and TERT promoter and on 1p/19q co-deletion. This group also demonstrated that TERT promoter mutations and ATRX alterations provided additional information for a tailored prognostic classification [49]. Besides, Arita et al. [50] proposed a classification of grade II-IV gliomas based on the mutations in IDH and the hotspot in TERT promoter.
Among IDH-mutant astrocytic tumors, CDKN2A/B homozygous deletion was demonstrated to be an unfavorable prognostic molecular marker [51]. Similarly, another group demonstrated that PIK3R1 mutation and altered retinoblastoma pathway genes, including RB1 and CDKN2A, were independent predictors of poor survival in astrocytic tumors. In oligodendrogliomas, NOTCH pathway inactivation and PI3K pathway activation were associated with poor prognosis [52,53]. Collectively, these molecular markers could predict prognosis in glioma patients.

IDH mutation drives production of oncometabolite D-2-hydroxyglutarate
In humans, IDH is composed of three types of isozymes (IDH1, IDH2, and IDH3). IDH1 is located in the cytoplasm and peroxisomes, whereas IDH2 and IDH3 are localized in the mitochondria and are involved in the TCA cycle. IDH1 and IDH2 are NADP+ dependent, whereas IDH3 is NAD+ dependent. IDH converts isocitrate into α-ketoglutarate (α-KG). No mutation in IDH3 has been detected in human cancers. If IDH is mutated, it blocks normal enzymatic activity and instead produces D-2-hydroxyglutarate (2-HG) from α-KG in an NADPH dependent manner, demonstrated that a G-CIMP-like phenotype and G-CIMP positive proneural glioblastomas were formed after the introduction of an IDH1 mutation into normal human astrocytes (NHA). These data indicate that mutant IDH induced TET2 suppression, followed by G-CIMP, in cancer cells. Consistent with IDH-mutant glioma patients, glioma patients with G-CIMP are younger at diagnosis and survive longer than those without G-CIMP [62]. Intriguingly, about 10% of G-CIMP tumors were relapsed as G-CIMP low tumors with poor clinical outcome [65].
The Cancer Genome Atlas (TCGA) performed comprehensive transcriptome analysis. Accordingly, GBM was classified into four groups (classic, mesenchymal, proneural, and neural groups). Aberrations and gene expression of EGFR and NF1 define the classical and mesenchymal subtypes, whereas tumors with an IDH1 mutation were classified within the proneural group. The proneural group is also accompanied by a PDGFRA gene abnormality and the G-CIMP feature [66]. DNA methylation induced by the IDH1 mutation caused hypermethylation at cohesion and CCCTC-binding factor (CTCF) binding sites and compromised the binding of the insulator protein. As a result, loss of CTCF at a domain permits a constitutive enhancer to interact aberrantly with the receptor tyrosine kinase gene PDGFRA [67].

IDH mutation promotes global histone methylation
IDH mutation is also known to increase histone methylation. Lysine methylation of histone tails modifies chromatin structure and regulates gene expression. By competition with α-KG, 2-HG inhibits histone demethylases including members of the Jumonji transcription factor family (JMJD2A, JMJD2C/KDM4C, and JHDM1A/FBXL11), resulting in histone hypermethylation [68]. Indeed, hypermethylation in H3K4me1, H3K4me3, H3K9me2, H3K27me2, H3K79me2, H3K27me3, H3K9me3, and H3K36me3 was observed in cells with exogenous 2-HG or mutant IDH1 induction [60, 63, 64, 69]. Sasaki et al. [63] also demonstrated that IDH1 R132H knock in mice showed significantly increased early hematopoietic progenitors, histone hypermethylation, and DNA methylation. Interestingly, the elevation of H3K9me3 levels was observed earlier than the DNA methylation change in NHA upon IDH1 R132H induction [69], suggesting that histone methylation may be an early event in IDH1-mutant cancers. The hypermethylation of histones blocks cell differentiation in cancer cells [60, 63, 64, 69]. Using a histone demethylating agent or a specific mutant IDH1 inhibitor, suppressed cell differentiation can be restored [70,71]. Besides, 2-HG impairs collagen maturation, which leads to basement membrane aberrations that play a part in glioma progression [72]. Taken together, these data show that DNA hypermethylation and histone methylation promote tumorigenesis through a wide range of gene function changes (Figure 2).

IDH mutation inducible metabolic alterations
In addition to the epigenetic changes, IDH1 mutation is known to alter hypoxia inducible factor 1α (HIF-1α) activity. Under oxidative conditions, α-KG-dependent prolyl hydroxylases (PHDs), which form the Egl nine homolog (EglN) families, induce HIF-1α hydroxylation. Hydroxylated protein is then bound by the von Hippel-Lindau tumor suppressor protein (VHL), ubiquitylated, and degraded via proteasome. In contrast, under hypoxia, the hydroxylation reaction is inhibited and HIF-1α is upregulated. HIF-1α then activates the transcription of several genes to mediate a switch from oxidative to glycolytic metabolism and induces angiogenesis by regulating the expression of vascular endothelial growth factor (VEGF) [73,74]. Koivunen et al. [33] demonstrated that IDH1 mutation attenuates HIF-1α through the activation of HIF prolyl 4-hydroxylase (EGLN), enhancing the proliferation and soft agar growth of NHA.
While several studies demonstrated that the IDH1 mutation induced aerobic glycolysis via HIF-1α activity [59,75], other group reported that HIF-1α responsive genes, including lactate dehydrogenase (LDHA) were downregulated; silenced LDHA was associated with increased methylation of the LDHA promoter [76]. Another group showed that IDH1 mutation reduces pyruvate flux to lactate and suppresses monocarboxylate transporters MCT1 and MCT4, which mediate lactate transmembrane transport [77]. IDH mutation also alters pyruvate metabolism, including pyruvate dehydrogenase and pyruvate carboxylase enzymes, resulting in anaplerosis of the TCA cycle [78,79].
Cancer cells are known to depend on reductive carboxylation (RC) of glutamine-derived α-KG for de novo lipogenesis under hypoxia [80]. It is worth noticing that the RC pathway is inhibited by IDH mutation [55]. Under hypoxia, IDH1 mutation upregulated the contribution of glutamine to lipogenesis [81,56].
Importantly, branched-chain amino acid transaminase (BCAT), which catalyzes the α-KG to glutamate conversion, was expressed at lower levels in IDH1-mutant gliomas than in IDH1-wildtype [86,87]. As a result, the glutamate level was decreased, and cell proliferation and invasiveness were suppressed in IDH-mutant gliomas [87].

Role of extensive resection in IDH1-mutant gliomas
There is a huge amount of evidence showing that surgical resection has a pivotal role in survival benefit of glioma patients. Extensive resection is known to prolong survival in low grade glioma and also in GBM (IDH1-wildtype) [88][89][90][91]. In IDH1mutant gliomas, an MRI study demonstrated that IDH1-mutant tumors were rarely located in high risk areas of the brain and show unilateral patterns of growth, sharp tumor margins, and less contrast enhancement [92,93]. Indeed, radiographic atlas revealed IDH1-mutant gliomas were frequently located at frontal lobe [94]. A diffusion-tensor imaging study demonstrated that IDH-mutant GBM has a less invasive phenotype than IDH-wildtype GBM [95]. Intriguingly, patients with IDH1wildtype gliomas had a reduced neurocognitive function and lower performance score than those with IDH1-mutant gliomas [96]. In addition, lesion volume was not associated with neurocognitive function for patients with IDH1-mutant tumors, but associated for those with IDH1-wildtype tumors [96]. Consequently, IDH1-mutant gliomas may be relatively less invasive to the surrounding eloquent area than IDHwildtype GBM.
In addition, Beiko et al. [97] reported that extensive resection, including nonenhancing area, prolonged survival in IDH1-mutant anaplastic astrocytoma and glioblastoma. They also mentioned, since IDH1-mutant gliomas were predominantly located at frontal lobe, that maximal resection was relatively amenable. Another group independently demonstrated that gross total resection extended survival in grade III IDH1-mutant gliomas without 1p/19q co-deletion [98]. In contrast, survival advantage was controversial in grade II astrocytoma [99,100]. These results suggest that for IDH1-mutant gliomas, especially grade III astrocytoma, maximal resection should be considered.

Prediction of IDH status
To establish IDH status-based treatment strategies, including surgery, advanced preoperative or intraoperative molecular analysis is important. Magnetic resonance spectroscopy (MRS) can be used to detect 2-HG and glutamate changes [101][102][103][104][105][106][107]. A recent MRS study demonstrated that 2-HG peaks rapidly decrease in accordance with tumor regression, whereas they increase with tumor progression in IDH-mutant gliomas [108], suggesting that 2-HG concentration, measured by MRS, may be a reliable approach to evaluate disease states in IDH-mutant gliomas.
In addition, several MR techniques, including diffusion tensor imaging and MR methods for determining relative cerebral blood volume, have been proposed to detect mutant IDH1 noninvasively [109][110][111]. Moreover, T2-FLAIR mismatch sign was found as a highly specific imaging marker for IDH-mutant astrocytoma [112][113][114]. Intraoperative technologies to assess IDH1 mutation have also been established [115][116][117]. These advanced technologies may allow the development of tailored surgical strategies for IDH-mutant gliomas. Other group demonstrated that urinary 2-HG is increased in patients with IDH1-mutant gliomas [118]. These findings indicate the possibility of application of indirectly assessed 2-HG as a clinical biomarker.

Treatment vulnerability in IDH-mutant gliomas 9.1 Radiotherapy for IDH-mutant gliomas
It has been shown that there is a higher relative sensitivity to radiotherapy and concurrent temozolomide (TMZ) in IDH1-mutant GBM patients than in those with IDH1-wildtype GBM [119], although there is no prospective clinical evidence of radiation therapy to extend survival in glioma patients with IDH1 mutation. As described above, IDH mutation inhibits NADPH and glutamate production, resulting in reduced glutathione levels and increased reactive oxygen species (ROS) [120][121][122][123]. Conversely, radiosensitivity in IDH1-mutant tumors was diminished by IDH1 inhibitor [124]. These findings support selective vulnerability to radiation therapy in IDH-mutant gliomas.

Temozolomide
Current standard management of GBM consists of surgical tumor resection, following local radiotherapy with temozolomide treatment [125]. Additionally, adjuvant TMZ prolonged survival in anaplastic astrocytoma [126]. Several studies demonstrated IDH1-mutation as a predictive biomarker for TMZ sensitivity in low grade gliomas and secondary GBM [127,128].
Cytotoxicity of TMZ is provoked by the formation of O 6 -methylguanine (O 6 G)-DNA adducts. O 6 G-DNA adducts induce DNA strand break and apoptosis through the O 6 G-thymine-mediated mismatch repair pathway [129,130]. It has also been established that the activation of DNA repairing pathways, including methylguanine methyltransferase (MGMT) repair enzyme, together with mismatch repair (MMR) system proteins deficiency, such as mutation-induced MSH2 and MSH6, result in drug resistance [131][132][133]. MGMT promoter methylation is highly methylated in IDH1-mutant gliomas, particularly oligodendrogliomas, compared with IDH-wildtype [43].
In contrast, TMZ-induced hypermethylation is a critical problem. Long-term TMZ exposure induces MMR inactivation, followed by DNA hypermutation phenotype. Among numerous mutations, gene alterations in RB and AKT-mTOR pathways promoted malignant progression in IDH1-mutant gliomas [27].

Other chemotherapeutic agents
Sulkowski et al. [144] demonstrated that 2-HG inhibits KDM4A and KDM4B, histone demethylases that play a critical role in double strand repair. As a result, IDH1 mutation suppresses HR and induces PARP inhibitor sensitivity. Additionally, IDH1mutant downregulates the DNA double strand break sensor ATM by altering histone methylation, resulting in impaired DNA repair. As a result, IDH1 mutation causes DNA damage susceptibility to radiation and daunorubicin and reduces self-renewal of hematopoietic stem cells in acute myeloid leukemia [145].

Specific IDH inhibitor
In 2013, specific inhibitors for IDH1 and IDH2 mutations were discovered [70,146]. In IDH2-mutant AML cells, an IDH2 R140Q inhibitor induced both histone and DNA demethylation [147]. These effects reversed blocked cell differentiation and resulted in cytotoxicity in vitro [146,147]. It is interesting to note that histone hypermethylation is more rapidly reversed than DNA hypermethylation [147]. In IDH1-mutant AML cells, differentiation and DNA demethylation were also induced by a next generation IDH1 inhibitor [148]. Since the IDH2 mutation is crucial for proliferation and maintenance of leukemia cells [149], an IDH inhibitor may be used as a novel and efficient chemotherapeutic agent against IDH-mutant AML cells. Indeed, clinical trials demonstrated durable response for IDH1/2-mutant refractory AML patients [150,151].
In IDH1-mutant glioma cells, Rohle et al. [70] reported that a specific IDH1 inhibitor, AGI-5198, blocked 2-HG production, histone demethylation, cell differentiation, and inhibited cell growth in endogenous IDH1-mutant glioma cells. Other group demonstrated that BAY 1436032, a pan inhibitor of IDH1 mutation, promoted mild cytotoxic effects in vivo [152]. In contrast, we established that, even with a long-term IDH1 inhibitor treatment, 2-HG depletion does not induce demethylation of global-DNA and histones, cell differentiation, nor cytotoxicity [141]. Studies using another IDH1 inhibitor also revealed minimal cytotoxicity despite a rapid decrease in 2-HG levels in glioma cells [153,154]. Similarly, treatment with an IDH1 inhibitor did not contribute to cytotoxicity, and the CpG island methylation status as well as histone trimethylation levels were largely retained in malignant glioma and chondrosarcoma [155,156]. Intriguingly, in immortalized human astrocytes with an inducible IDH1 R132H expression system, a specific IDH1 inhibitor induced demethylation and inhibited tumorigenesis when forced expression was prior or concomitant to inhibitor treatment, but these effects were not observed if the treatment was delayed [157]. These results indicate that 2-HG depletion or blocked mutant IDH1 might be insufficient to control tumor growth and reprogramming of epigenomic alterations in progressed IDH1-mutant gliomas. Indeed, preliminary results indicate that the 6-month progression-free survival of IDH1-mutant glioma, chondrosarcoma, and cholangiocarcinoma is 25, 56, and 43%, respectively, suggesting that the potential of the IDH1 inhibitor may be weaker in IDH1-mutant gliomas than in other cancers [158].

DNA demethylating agents
In addition to IDH1 inhibitor treatments, other strategies to control IDH1mutant tumor cells have been proposed. Because the IDH1 mutation promotes proliferation by blocking DNA demethylation, treatment with DNA demethylating agents reverses DNA methylation and inhibits proliferation in IDH1-mutant cells [71,159]. Intriguingly, treatment with both the DNA demethylating agent 5-azacytidine (5-Aza) and TMZ demonstrated extensively prolonged survival in an IDH1-mutant orthotopic xenograft model [160].

Bcl-2 family inhibitors
Since 2-HG suppresses the activity of cytochrome c oxidase in mitochondrial complex IV, the mitochondrial threshold for apoptosis was decreased after BCL-2 inhibition in IDH1 and IDH2-mutant AML [161]. Similarly, another Bcl-2 family member, the Bcl-xL inhibitor, induced apoptosis in IDH-mutant cells, including endogenous IDH1-mutant glioma cells [162]. Together, inhibition of Bcl-2 family members may be targetable to control growth in IDH-mutant cells.

DNA damaging agents
Because PLK1 activation provokes a rapid bypass through the G2 checkpoint after TMZ treatment in IDH1-mutant tumors, combination treatments with TMZ and a PLK1 inhibitor significantly suppressed tumor growth in an IDH1-mutant in vivo model [138]. In tumors with ATRX mutation-associated alternative lengthening telomeres (ALT), ATR inhibitor is highly sensitive [163], implying that such inhibition may be useful for treatments of IDH1-mutant astrocytic tumors with positive ALT. IDH1 mutation blocked HR, so-called "BRCA ness" phenotype provided specific sensitivity for PARP inhibitor both in vitro and in vivo [144].

Vaccination therapy
Schumacher et al. [165] reported an immunological approach to control IDH1mutant cells. They showed that an epitope derived from the IDH1-mutant amino acid sequence is presented in HLA class II molecules of antigen-presenting cells, which elicit a strong immune response via CD4 + T cells. In addition, they showed that constitutive stimulation with synthetic peptides having the IDH1-mutation sequence developed an immune response that eradicated IDH1 mutated tumors in a mouse model with human HLA molecules. Thus, vaccine therapy targeting for IDH1-mutation is expected to develop for future clinical trial [165,166]. Moreover, IDH1-mutation caused downregulation of leukocyte chemotaxis, resulting in repression of the tumor-associated immune system including immune cells, such as macrophages [167]. Additionally, tumor infiltrating lymphocytes (TILs) and programmed death ligand 1 (PD-L1) were expressed at low levels in IDH1-mutant gliomas [168]. In contrast, Kohanbash et al. [153] demonstrated reduced expression of cytotoxic T lymphocyte-associated genes and IFN-gamma inducible chemokines in IDH1-mutant cells; these results were reversed by specific IDH1 inhibitor. Therefore, combination treatments with vaccine immunotherapy and IDH1 inhibitor result in enhanced toxicity in IDH-mutant tumors.

Target for altered metabolism
IDH1 mutation induced altered metabolism is also expected as a novel therapeutic target. Based on the fact that the main carbon source for α-KG and 2-HG synthesis in IDH1-mutant cells is glutamine from glutaminolysis, a suitable target therapy would be the use of glutaminase (GLS) inhibitor or anti-diabetic drug metformin via the inhibition of mitochondrial complex I in the electron transport system [83,[169][170][171]. Since reduced glutamate blocks glutathione synthesis, inhibition of glutaminase specifically sensitizes IDH-mutant glioma cells to oxidative stress and radiation [86].
Mutant IDH1 alters steady state levels of NAD+ through inhibiting NAPRT1, one rate limiting enzyme for NAD+ biosynthesis. Therefore, inhibition of nicotinamide phosphoribosyltransferase (NAMPT), another rate limiting enzyme, induced high cytotoxicity in IDH1-mutant patient-derived glioma cells [141]. Since TMZ rapidly consumes NAD+ through PARP activation, combination treatments with TMZ and NAMPT inhibitor further enhanced NAD+ depletion-mediated cytotoxicity in IDH1-mutant cancers [142]. Similarly, Lu et al. [143] reported that the PARPassociated DNA repair pathway was extensively compromised in IDH1-mutant cells due to decreased NAD+ availability, thus sensitive to TMZ.
Because of the relationships between IDH1 mutation and MYC activation [38,40,172], target therapy to regulate MYC, by using bromodomain and extra-terminal (BET) inhibitors, CDK7 or MYC-induced glycolysis may be used for IDH-mutant gliomas [40, [173][174][175]. Given the results of these studies, IDH1 mutation-specific biological alterations and metabolic feature may be expected as novel therapeutic targets.

Conclusions
In summary, investigations on IDH mutations enabled distinctive tumor classification and may allow the development of specific therapeutic strategies. Further preclinical and clinical studies are warranted to overcome the outcomes of cancer development in IDH-mutant glioma patients.