Isocitrate dehydrogenase (IDH) enzymes, of which there are three isoforms, are essential enzymes that participate in several major metabolic processes, such as the Krebs cycle, glutamine metabolism, lipogenesis and redox regulation.^1,2,3

IDH1 is located in the cytoplasm and peroxisomes, whereas IDH2 and IDH3 are located in the mitochondrial matrix.⁴
The catalytic sites of IDH1 and IDH2 exhibit affinity for the substrate, isocitrate, together with nicotinamide adenine dinucleotide phosphate (NADP⁺) and a divalent metal cation, usually magnesium or manganese,⁵
resulting in the formation of α-ketoglutarate (α-KG).
IDH3, which also catalyses the transformation from isocitrate into α-KG, employs nicotinamide adenine dinucleotide (NAD⁺) as its cofactor. The catalytic activity of IDH requires homodimerisation along with an alteration in the enzyme conformation; isocitrate binding changes the structure of the enzyme from an open to a closed conformation.⁶

Substrate recognition depends on the amino acid residues in the active site, whereas the frequent mutated active site residue in cancer is arginine 132 (R132).⁵
Mutations in IDH are prevalent in human malignancies. In glioma, IDH mutations are recognised in >80% of World Health Organisation (WHO) grade II/III cases.⁷ In WHO grade IV glioblastoma (GBM), IDH mutations are also found frequent in secondary GBM, which account for 73% of clinical cases, whereas they are less seen in primary GBM (3.7%).⁸
A follow-up investigation showed that the presence of IDH mutations predict a favourable disease outcome with prolonged median survival in GBM (IDH wild type: 15 months; IDH mutant: 31 months) and anaplastic astrocytoma (IDH wild type: 20 months; IDH mutant: 65 months).⁷ Although IDH-mutated glioma generally exhibits a better disease outcome, the high incidence of IDH mutations in secondary GBM suggests that lower-grade glioma with IDH mutation often recur with having undergone malignant transformation to a higher grade. In addition, IDH-mutated glioma is more likely to develop a hypermutation phenotype, which is associated with worsened prognosis.⁹ In non-central nervous system (non-CNS) malignancies, IDH mutations are identified in acute myeloid leukaemia (AML; 16% among all clinical cases),¹⁰ intrahepatic cholangiocarcinoma (23% among all clinical cases)¹¹ and central/periosteal chondrosarcoma (56% among all clinical cases).¹² The investigation of these non-CNS tumours with similar IDH mutation provides valuable information for glioma research, whereas in the present review we tend to be focussed on IDH-mutated glioma.
IDH mutations that are associated with cancer tend to localise to the arginine residue that is crucial for the recognition of isocitrate (R132 for IDH1, R140 or R172 for IDH2).⁷ Missense mutations in the IDH1 gene result in the replacement of a strong, positively charged arginine residue at position 132 with lower-polarity amino acids such as histidine (H), lysine (K) or cysteine (C), which impedes the formation of hydrogen bonds with the α-carboxyl and β-carboxyl sites of isocitrate.^13,14

The mutant IDH enzyme therefore exhibits decreased affinity for isocitrate, along with an elevated preference for NADPH. However, only one copy of the IDH gene is mutated in tumours and, in tumour cells harbouring heterozygous IDH mutations, the main forms of IDH dimers are presumed to be heterodimers that contain a version of wild-type IDH1 and a version with the R132H mutation. As a result, in IDH-mutant cells, the IDH1 wild-type component of the dimer converts isocitrate into α-KG to produce NADPH, whereas the mutant part of the dimer exhibits neomorphic activity, converting α-KG into D-2-hydroxyglutarate (D-2-HG) in an NADPH-dependent manner

Metabolic reprogramming
IDH-mutant enzymes cause the accumulation of D-2-HG at concentrations as high as 5–30 mM¹⁵ in the cytoplasm, thereby draining carbohydrates from the Krebs cycle.¹⁸ The Krebs cycle is adjusted to compensate for fluctuations in the metabolic pathways.¹⁹ A ¹³C metabolic flux analysis suggested that IDH1-mutated cells exhibit increased oxidative metabolism in the Krebs cycle, whereas reductive glutamine metabolism is suppressed.²⁰ 
With the depletion of cellular metabolism, several non-Krebs-cycle sources of carbohydrates are recruited to compensate for the loss of α-KG.^21,22
Waitkus et al.²³ demonstrated that glutamate dehydrogenase 2, an enzyme that catalyses the conversion of glutamate into α-KG and that is expressed at high levels in the brain, is important for relieving the metabolic liabilities in the context of IDH mutants. This finding is confirmed by the observation that IDH-mutated glioma cells are more sensitive to the inhibition of glutaminase,²⁴ suggesting that glutaminolysis serves as a key compensatory pathway to maintain metabolic homoeostasis.

McBrayer et al.²⁵ further highlighted the dependency of IDH1-mutated cells on glutaminolysis, as D-2-HG functions as an inhibitor of the branched-chain amino acid transaminase (BCAT1/2), thereby decreasing the levels of glutamate. Furthermore, the consumption of NADPH by IDH mutants compromises de novo lipogenesis, resulting in an increased dependence on exogenous lipid sources for cellular growth.² This is accompanied by the stimulation, by D-2-HG, of glutamine-derived lipogenesis under hypoxic condition to meet the needs for lipid productivity.²⁶
Lactate dehydrogenase A (LDHA) catalyses the transformation of pyruvate formed by glycolysis into L-lactate,²⁷ and the expression of LDHA is thus considered to be a hallmark of Warburg phenotype, allowing rapid glycolytic flux to meet the demands for cellular proliferation.²⁸ Although LDHA is highly expressed in a variety of cancer cells, it is silenced in glioma tissue specimens and patient-derived glioma cells with IDH mutants.^29,30 
Silencing of LDHA (and of several other glycolysis genes including CA9 and VEGFA) has been found to be associated with hypermethylation in the promoter region of these genes in response to D-2-HG. The overall epigenetic silencing of the glycolytic pathway might explain the slow-growing nature of IDH-mutated glioma as compared with their IDH wild-type counterparts.^30,31 In support of this hypothesis, in a recent study, the acquisition of the Warburg phenotype was associated with more aggressive gliomas and was found to occur at the CpG island methylator phenotype (G-CIMP) in gliomas described below, which is specific for astrocytoma.³²

In addition, IDH mutations lead to the neomorphic enzyme activity, which redirects the Krebs cycle for D-2-HG production. The resultant decrease in α-KG levels might affect the level of hypoxia-inducible factor subunit HIF-1α,³³ as α-KG is normally needed for prolyl hydroxylases (PHD) to hydroxylate and promote the degradation of HIF. However, the detailed molecular mechanism on how HIF is regulated in the context of IDH mutation is currently unclear. Other lines of evidence showed that D-2-HG, but not L-2-HG, stimulates the activity of the prolyl hydroxylase PHD2, which results in the reduced expression of HIF-1/2α.³⁴ More effort is encouraged to elucidate the relationship between D-2-HG and the hypoxia-sensing pathway in glioma and other IDH1-mutated malignancies.
Overall, the acquisition of mutant IDH results in substantial reprogramming of cellular metabolism. Glutamine and/or glutamate serve as key substrates to compensate for the metabolic impact by strengthening synthetic pathways for lipids and glutathione. Interestingly, IDH-mutated glioma shows a distinctive metabolic pattern compared with other solid tumours—most notably, the remarkably reduced glycolysis, the metabolic hallmark of fast proliferating malignancies. The unique metabolic pathways in IDH-mutated glioma not only explain the slow-growing nature of this disease but also suggest that developing targeted strategies for IDH-mutant-specific metabolic patterns could be a valuable approach for future glioma therapeutics.

Gilbert et al.⁵³ showed that IDH1-mutated glioma cells exhibit strong oxidative stress, as evidenced by an enhanced expression of manganese superoxide dismutase and protein carbonylation. This increased stress was confirmed by subsequent investigations showing that IDH-mutated cancers are more prone to oxidative damage.^52,54 Also was confirmed elevated oxidative stress that is closely related to the acquisition of IDH mutants, leading to oxidative damage in biomolecules such as DNA and lipids.⁵¹ Owing to the substantially increased oxidative burden, inhibiting antioxidant pathways, such as the synthesis of glutathione, which is mediated by the transcription factor nuclear factor erythroid 2-related factor 2 (Nrf2), could be a valuable strategy for targeting IDH1-mutated solid tumours.⁵⁵ In addition, proline synthesis has been reported to maintain redox homoeostasis in mitochondria in IDH1-mutated cells. Enhanced activity of pyrroline 5-carboxylate reductase 1-mediated glutamate-to-proline transformation in IDH-mutated cells alongside the oxidation of NADH partially uncouples the electron transport chain from Krebs cycle activity, thus maintaining anabolism in cancer cells

Direct targeting of mutant IDH
Given that the neomorphic activity of IDH mutants correlates with malignant transformation, direct targeting of the mutant enzyme has been a heavily pursued strategy. Rohle et al.⁵⁷ reported the first synthetic inhibitor of the IDH mutant, AGI-5198, which blocks the production of D-2-HG and impairs IDH1-mutated xenograft growth in vivo.
Second generation of IDH-mutant inhibitors, ivosidenib (AG-120) and vorasidenib (AG-881), are currently approved by the Food and Drug Administration as a therapeutic option for IDH-mutated AML.⁵⁸ These IDH-mutant inhibitors exhibit an improved brain-to-plasma ratio, suggesting that they might be effective for IDH1-mutated glioma.⁶⁰ Several other IDH-mutant inhibitors, such as BAY1436032, have shown tumour-suppressing effects as experimental therapeutics for the treatment of AML and astrocytoma in animal models.^61,62 

Two clinical studies (NCT03127735 and NCT02746081) are currently ongoing to confirm these findings in patients with IDH1-mutated AML or advanced solid tumours, respectively.
Despite the promising success of the IDH-mutant inhibitors, several studies have indicated the potential limitations of their application. For example, Johannessen et al.⁶³ discovered that, although the IDH-mutant inhibitor AGI-5198 successfully reduces neomorphic activity, it relieves hypermethylation phenotype but to a much less extent, as evidenced by elevated histone-3 methylation. In addition, Sulkowski et al.⁶⁴ reported that
AGI-5198 relieves the burden of DNA damage in cancer cells, which might increase their resistance to genotoxic therapies, such as radiation and chemo agents. This phenomenon has been confirmed by another study showing that AGI-5198 confers radioprotective effects on IDH1-mutated cancer cells.⁶⁵ Overall, targeting IDH-mutant neomorphic activity is a straightforward strategy and has shown efficacy against haematopoietic malignancies in humans and several experimental models for solid cancers. In addition to suppressing D-2-HG production, a combined approach with other agents, such as inhibitors of critical enzymes in metabolic or DNA repair pathways, might be helpful to improve the disease outcome (see the discussion below on synthetic lethality).

RTK pathway.
RTK (receptor tyrosine kinase) signaling is the most frequently altered signaling pathway in GBM, especially in IDH-wildtype GBM tumors. RTK is a cell-surface receptor that binds growth factors, the family of which includes EGFR, PDGFR, TGFR, FGFR, MET, and VEGFR, and is an essential component of signal transduction pathways that mediate cell-to-cell communication.

In GBM, the activation of RTK signaling through the PI3K/AKT/mTOR pathway induces cell proliferation, migration, differentiation, and survival.
The most common targets of the RTK pathway are EGFR and PTEN, the former acting in an oncogenic role while the latter acting as a tumor suppressor.
In GBM cells, the activation of EGFR and the PI3K/AKT/mTOR signaling could be achieved either through amplification of the EGFR (resulting in overexpression of EGFR) and/or EGFR mutation. The negative regulator of the pathway, PTEN, could be inactivated through mutation or deletion, and thus facilitates the pathway activation and induces cell migration, invasion, and survival.

Another commonly altered RTK pathway in GBM is the Ras pathway (Ras/BRAF/MEK) [67,68]. Active Ras (Ras-GTP) promotes cell cycle progression, cell survival, and migration through a cascade of downstream effectors. RTK has been suggested as a druggable target in GBM and is extensively investigated in clinical trials.

The retinoblastoma protein (RB) pathway is also found to be frequently altered in GBM and plays a crucial role in regulating tumorigenesis in GBM. The phosphorylation of RB protein, which is accomplished by the CDK4/Cyclin D1 complex, can inhibit the cell cycle progress from the G1 to S phase by binding with the E2F transcription factor. RB pathway could be joined with the TP53 pathway through CDKN2A, which encodes Ink4a and Arf proteins and plays an important role in activating RB and TP53, respectively.

The growth inhibition function of the RB pathway is often disrupted in GBM, most commonly due to inactivation of CDKN2A/CDKN2B and RB1 and amplification of CDK4 and CDK6. Methylation of the RB1 promoter, which is frequent in secondary GBM (IDH-mutant ones), can also result in decreased RB1 expression and cell-cycle checkpoint function and finally leads to dysregulated cell cycle and uncontrolled cell proliferation. CDK4 and CDK6 inhibitors have shown promising antitumor efficacy in GBM and are being studied in clinical trials