Molecular Pathways: Isocitrate Dehydrogenase Mutations in Cancer

Background

The first identification of a cancer-associated isocitrate dehydrogenase (IDH) mutation was in a patient with colorectal cancer, and emerged from one of the earliest comprehensive analyses of mutations in protein-coding genes (1). Two years later, the same group reported a substantially higher frequency of IDH mutations (12%) when they applied whole genome sequencing to a small number of glioblastomas, the most common malignant brain tumor in adults (2). Interestingly, the majority of IDH-mutant glioblastomas (5/6) were from patients whose tumors had developed overtime from lower grade (WHO grade II and WHO grade III) tumors. This seminal finding was confirmed in a follow-up study with a much larger number of tumors, which reported IDH mutations found in the vast majority (>70%) of WHO grade II and WHO grade III gliomas (3).

Since these initial studies, many human cancers were examined for the presence of mutations in IDH1 and IDH2. IDH mutations were observed in a number of hematopoietic neoplasms, most commonly acute myeloid leukemia (AML; ∼10%–15%; refs. 4–6) and angio-immunoblastic T-cell lymphoma (∼20%; ref. 7). IDH mutations were also found in chondrosarcoma (∼50%; ref. 8), intrahepatic cholangiocarcinoma (∼15%–20%; ref. 9), and at lower frequency (< 5%) in other solid tumors (e.g., glioblastomas, colorectal cancer, esophageal cancer, bladder cancer, melanoma, prostate carcinoma, breast adenocarcinoma; ref. 10). Somatic heterozygous mutations in IDH1 or IDH2 were also found in the majority of enchondromas and spindle cell hemangiomas in patients with the Ollier disease and Maffuci syndrome, nonhereditary skeletal disorders (11).

More recent DNA resequencing projects have provided additional information regarding the timing of IDH mutations during tumor development. Analyzing over 300 gliomas, Watanabe and colleagues found that in 51 cases with multiple biopsies, neither acquisition of a mutation in TP53 nor loss of 1p/19q occurred prior to a mutation in IDH1 (12). Further analysis of matched biopsy pairs, collected from glioma patients at the initial diagnosis and at the time of tumor recurrence showed that IDH1^R132H was the only mutation that was consistently present in both the initial and recurrent biopsy specimen (13). In leukemia patients, IDH2 mutations were observed in the absence of NPM1c mutations in both mature and progenitor cell populations, suggesting that IDH2 mutation might be an early and perhaps preleukemic event (14, 15).

The vast majority of cancer-associated mutations in IDH1 and IDH2 map to an arginine within the catalytic pocket of the enzyme. Mutations in IDH1 mostly occur at arginine 132, with substitutions including R132H, R132C, R132L, R132S, and R132G. Mutations in IDH2 typically occur at arginine 172 or arginine 140 (which is analogous to R132 in IDH1). The clustering of cancer-associated IDH mutations in the functional domain of the enzyme suggested that these mutations might endow the mutant protein with a novel and presumably oncogenic enzymatic activity. This question has been explored through untargeted metabolomic profiling of cells engineered to express the mutant enzyme. Compared with parental cells, cells expressing the IDH1^R132H-mutant enzyme were found to produce the R(-) enantiomer of the metabolite 2-hydroyglutarate (R-2-HG), which accumulates in IDH-mutant human gliomas (16) and leukemias (5, 17). Production of R-2-HG involves direct conversion from α-KG and relies on the presence of a wild-type allele (18), likely explaining the rareness of loss of heterozygosity.

The identification of an “oncometabolite” in IDH-mutant tumors strengthened the hypothesis that IDH mutations are oncogenic, and led many investigators to examine the ability of mutant IDH to transform nonmalignant cells. Expression of mutant Idh in mouse myeloid progenitor 32D cells and primary mouse bone marrow cells impaired hematopoietic differentiation and increased stem/progenitor cell marker expression, suggesting a proleukemogenic effect (19). A more recent study reported that retrovirally mediated expression of mutant Idh2^R140Q in murine primary hematopoeitic bone marrow stem and progenitor cell populations induced myeloproliferative-like neoplasms, T-cell lymphoma or B-cell lymphoma when transplanted into irradiated mice (20). However, these hematologic malignancies occurred at low penetrance and with long latency, suggesting that they did not arise solely due to mutant Idh2 expression. Expression of mutant Idh2 in a nontransformed mesenchymal multipotent mouse cell line (C3H, 10T) impaired their differentiation into adipocytic and chondrocytic lineages and resulted in loss of contact inhibition and tumor formation in vivo (21). In immortalized human astrocytes, expression of mutant IDH, but not wild-type IDH or a catalytically inactive IDH mutant promoted their anchorage-independent growth (22).

Further insights into the role of mutant IDH in tumor initiation have emerged from experiments with genetically engineered mice. Tamoxifen-induced global expression of Idh2^R140Q or ^R172K, driven from the chicken β-actin promoter, resulted in cardiomyopathy, white matter abnormalities throughout the central nervous system, and muscular dystrophy; mice engineered to express mutant Idh2 in specific tissues reportedly developed carcinomas with very long latencies (23). In another model, mice that expressed a doxycycline-inducible Idh2^R140Q allele from the Collagen A1 locus did not develop leukemia, even after one year of continuous doxycycline treatment (24). The most common cancer-associated IDH mutation, IDH1^R132H, has also been inserted into the endogenous murine Idh1 locus and expression of the mutant enzyme subsequently targeted to specific cell populations. Expression of Idh1^R132H in hematopoietic cells (using the LysM and Vav promoters to express Cre-recombinase) resulted in increased numbers of hematopoietic progenitors, but no overt leukemia (25). Expression of Idh1^R132H in nestin (Nes)-expressing neural stem cells resulted in perinatal lethality due to cerebral hemorrhage (26). Expression of Idh1^R132H in GFAP-expressing astrocytes resulted in impaired collagen maturation and basement membrane function, but again no tumors. Tamoxifen-inducible expression of mutant Idh1 in chondrocytes (using the Col2A1 promoter to express Cre recombinase) resulted in enchondromas, benign cartilage tumors that are precursors to malignant chondrosarcomas (27). Doxycycline-induced expression of mutant Idh2 (Idh2^R140Q or Idh2^R172K) increased hepatocyte proliferation in a liver injury model, but was not sufficient to induce tumors (28).

More recent studies have examined whether IDH mutations might collaborate with other genetic events to induce cancer in mice. Using a mouse transplantation assay, Chaturvedi and colleagues found that mutant IDH1 was not sufficient to transform hematopoietic cells but accelerated the onset of leukemia in cooperation with HoxA9 (29). Chen and colleagues applied a mosaic mouse modeling approach in which hematopoietic stem and progenitor cells (HPSC) from Flt3-ITD or Nras^G12D mice were transduced with a retroviral vector expressing mutant Idh2 and then assessed for tumorigenic potential following transplantation into syngeneic recipient mice. Idh2 mutants were found to cooperate with Flt3 or Nras alleles to drive leukemia formation (30). Cooperativity between mutant Idh2^R140Q and other leukemia-relevant pathway alterations (e.g., Flt3-ITD or homeobox proteins HoxA9 and Meis1a) was confirmed in the aforementioned doxycycline-inducible mouse model (24) and, more recently, another mosaic mouse modeling approach (31). Saha and colleagues showed that Idh2^R172K cooperated with mutant KRAS (Kras^G12D) to induce intrahepatic cholangiocarcinomas (28). In aggregate, these studies demonstrate that mutant IDH cooperates with other oncogenic events to initiate cancer, consistent with the finding that IDH-mutant human cancers typically harbor alterations in multiple other cancer genes. In glioma, for example, IDH mutations are associated with missense mutations in ATRX, TP53, and TERT (diffuse astrocytomas) or codeletion of chromosome arms 1p and 19q (oligodendrogliomas; refs. 32,33). In AML patients with a normal cytogenetic profile, IDH mutations are associated with mutations in NPM1, FLT3-ITD, and DNMT3A (refs. 34, 35; Fig. 1).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1.

Co-occurring genetic lesions in IDH-mutant glioma and AML. Oncoprint figure (generated using Memorial Sloan Kettering cBio Portal; ref. 10) showing frequency of commonly co-occurring genetic lesions in glioma (top) and AML (bottom). Note that TET2 is shown to illustrate near complete mutual exclusivity with mutations in IDH1/2. Results are based upon data generated by the TCGA Research Network: Glioma data set (ref. 32); AML dataset (73). For simplicity changes in copy number were not included. TET2, Tet Methylcytosine Dioxygenase 2; DNMT3A, DNA (Cytosine-5-)-Methyltransferase 3 Alpha; FLT3, Fms-Related Tyrosine Kinase 3; NPM1, Nucleophosmin (Nucleolar Phosphoprotein B23, Numatrin); TP53, Tumor Protein P53; ATRX, Alpha Thalassemia/Mental Retardation Syndrome X-Linked.

Recent studies have begun to address the question of whether the activity of the mutant IDH enzyme remains important for the growth of IDH-mutant cancers once they are fully established. Studies in experimental cancer models suggest that this is indeed the case, with the strongest evidence coming from leukemia models. In TF-1 human erythroleukemia cells, ectopic expression of mutant IDH promotes growth factor independence, a phenotype that can be reversed with a small-molecule mutant IDH inhibitor (36, 37). Ex vivo treatment of freshly isolated IDH-mutant leukemic blasts with a mutant selective IDH inhibitor induces a cellular differentiation program (36). Pharmacologic inhibition of the mutant IDH enzyme blocks colony formation of human AML cells but not of normal CD34⁺ bone marrow cells (29). In a genetically engineered leukemia model, pharmacologic or genetic inhibition of mutant Idh2 triggered the differentiation and death of AML cells (30), and doxycycline-induced silencing of mutant Idh2 similarly eliminated Idh2^R140Q/Hoxa9 or Idh2^R140Q/Meis1a-driven leukemia cells (24).

Clinical–Translational Advances

The above mentioned results in experimental leukemia models are supported by the preliminary findings of an ongoing phase 1 clinical trial, which showed that the mutant IDH2 inhibitor AG-221 produces clinical responses, including complete and durable responses, in about 40% of patients with AML and MDS (38).

The contribution of the mutant enzyme to the maintenance of IDH-mutant solid tumors remains currently unknown, and further insights are likely to emerge from an ongoing single-arm dose escalation study (ClinicalTrials.gov NCT02073994) with the mutant-selective IDH1 inhibitor AG-120 (39). Data from experimental models suggest that IDH1-mutant solid tumors remain, at least in part, dependent on the activity of the mutant enzyme. In HT1080 human fibrosarcoma cells, RNAi-mediated suppression of endogenous mutant IDH1 significantly inhibited anchorage-independent growth (40). Knocking out the endogenous mutant IDH1 gene using TALEN technology similarly impaired anchorage-independent growth and in vivo growth of IDH-mutant human sarcoma cells (41). In IDH1-mutant glioma cells, pharmacologic blockade of the mutant enzyme retarded their growth in soft agar and in mice (42).

Targeted therapy of IDH-mutant cancers is likely to grow beyond inhibition of the mutant enzyme itself and may include strategies directed against the epigenetic and metabolic changes that are associated with IDH mutations. Current data suggest that the R-2-HG “oncometabolite” is responsible for many, if not all, biologic effects of cancer-associated IDH mutations. Cell-permeable esters of R-2-HG phenocopy the effects of mutant IDH in experimental models (43), and ectopic expression of the dehydrogenase (44) that converts R-2-HG into α-KG, and thus counteracts the activity of the mutant IDH enzyme, is sufficient to reverse the cellular effects of cancer-associated IDH mutants (45). R-2-HG competitively inhibits a large family of α-KG–dependent enzymes, a protein family with over 60 members. These include the ten-eleven translocation (TET) family of 5-methyl cytosine hydroxylases, the jumonji domain containing (JmjC) family of histone lysine demethylases (46, 47), enzymes involved in nucleic acid metabolism (AlkB, FTO), and many enzymes with still unknown functions (ref. 48; Fig. 2).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2.

Molecular mechanisms of IDH-associated tumorigenesis. Wild-type IDH1/2 (wtIDH1/2) converts isocitrate (generated through the citric acid (TCA) cycle) into α-KG, producing NADPH in the process_. Mutant IDH1/2 (mutIDH1/2) converts α-KG to the oncometabolite R-2-HG, consuming NADPH. R-2-HG inhibits members of the protein family of α-KG–dependent dioxygenases. Epigenetic modifications result from inhibition of histone lysine demethylases (KDM) and the 5-methyl cytosine hydroxylase TET2. Inhibition of prolyl-hydroxylation impairs collagen maturation. R-2-HG has been reported to activate the enzyme prolyl-hydroxylase 2 (PHD2) that inhibits hypoxia-inducible factor 1-alpha (HIF-1α; ref. 22), although other studies suggest that it may be inhibited (46, 47). Mutant IDH may also change the cellular redox environment by altering the ratio of NADPH to NADP⁺. IDH3 has been omitted from this figure for simplicity.

It is currently unknown which α-KG–dependent enzymes function as context-dependent tumor suppressors and are inhibited at the relevant R-2-HG concentrations in human tumors. Inhibition of TET2 is likely to mediate the effects of mutant IDH in AML, given the well documented role of TET2 in hematopoietic differentiation, the almost completely mutual exclusivity of TET2 and IDH mutations in AML (Fig. 1), the decreased 5-hydroxymethylcytosine (5-hmc) levels in IDH-mutant AMLs, and the identification of TET2 as a candidate effector of mutant IDH in a short hairpin RNA library screen of α-KG–dependent dioxygenases (19, 37, 49–51). Inhibition of the histone demethylases JMJD2A and JMJD2C likely contributes to the effects of mutant IDH on cellular differentiation in many cell lineages. These enzymes are particularly sensitive to inhibition by R-2-HG (46, 47), and knockdown of Jmjd2C (also known as Kdm4c) was sufficient to phenocopy the effects of mutant Idh on adipocyte differentiation in 3T3-L1 cells (43). In fact, it seems plausible than many hallmarks of mutant IDH-associated human malignancies, such as restricted cellular differentiation and DNA hypermethylation, result from coordinate effects of R-2-HG on DNA and histone methylation (43,52). These findings raise the intriguing question of whether IDH-mutant cancers cells might show an increased sensitivity to the DNA methyltransferase inhibitors decitabine or 5-azacytidine (53–55) or drugs targeting histone modifications.

It is less clear whether epigenetic alterations are also responsible for the oncogenic effects of mutant IDH in solid tumors. We observed that growth inhibition of IDH1-mutant glioma xenografts by a mutant-selective IDH1 inhibitor was not clearly linked to epigenetic changes in tumor tissue (42). Similarly, ectopic expression of the R-2-HG-dehydrogenase in HT1080 sarcoma cells effectively inhibited their growth in vivo without clear changes in histone methylation marks or TET2-induced 5-hmC production (41). It is plausible that the oncogenic effects of mutant IDH in solid tumors are linked to metabolic alterations associated with IDH mutations, including reductions in α-KG levels, changes in the NADP⁺/NADPH ratio and mitochondrial bioenergetics, and an impaired ability of the IDH enzyme to catalyze the reverse carboxylation of α-KG to form isocitrate. These metabolic changes can have a profound effect on the ability of cells to proliferate, differentiate, and escape cell death, and may introduce therapeutic vulnerabilities (56–64). Recent studies reported an enhanced susceptibility of IDH1-mutant cancers to inhibitors of BCL-2 (63), due to the effect of mutant IDH on mitochondrial bioenergetics, or to depletion of the coenzyme NAD+ (65).

The success of therapeutic strategies to curb the growth of IDH-mutant cancers will hinge on a deeper understanding of the molecular mechanisms of IDH-associated tumorigenesis, and R-2-HG effector pathways, in each cancer type. In addition to the epigenetic and metabolic effects of mutant IDH outlined above, R-2-HG has been reported to stimulate the activity of the EGLN prolyl 4-hydroxylases, leading to diminished levels of hypoxia-inducible transcription factor (HIF) and enhanced soft agar growth of human astrocytes (66). Consistent with this result, the expression of HIF1α-responsive genes, including many essential for glycolysis, appears to be lower in IDH-mutant gliomas (67), perhaps explaining the often poor retention of the radiotracer 2-[(18)F] Fluoro-2-deoxy-D-glucose (FDG) by these tumors. The group of α-KG–dependent dioxygenases also includes enzymes that hydroxylate proline and lysine residues in collagen and promote its stability. It is intriguing to speculate that the observed brain hemorrhages in Nes-Cre Idh1^R132H mice might have resulted from impaired collagen maturation and basement membrane architecture (68).

A remarkable body of knowledge has been generated in the few years since the first description of cancer-associated IDH mutations, and the future seems bright for the development of rationale therapeutic approaches to IDH-mutant cancers. The ability to monitor the activity of the mutant IDH enzyme through quantification of R-2-HG levels in blood (69) or magnetic resonance spectroscopy in tumors (70–72) will be instrumental in developing such strategies.

Disclosure of Potential Conflicts of Interest

K. Yen has ownership interest (including patents) in Agios Pharmaceuticals. No potential conflicts of interest were disclosed by the other authors.

Disclaimer

The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

Authors' Contributions

Conception and design: O. Clark, I.K. Mellinghoff

Development of methodology: O. Clark, I.K. Mellinghoff

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): O. Clark, I.K. Mellinghoff

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): O. Clark, I.K. Mellinghoff

Writing, review, and/or revision of the manuscript: O. Clark, K. Yen, I.K. Mellinghoff

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): O. Clark, I.K. Mellinghoff

Study supervision: O. Clark, I.K. Mellinghoff

Grant Support

Research reported in this publication was supported by the NIH under award numbers R01NS080944 (to I.K. Mellinghoff), T32CA160001 (to O. Clark), and P30CA008748 (to O. Clark and I.K. Mellinghoff, through their institution). I.K. Mellinghoff was supported by the Ben and Catherine Ivy Foundation and the Defeat GBM Research Collaborative of the National Brain Tumor Society.