Oncogenic Isocitrate Dehydrogenase Mutations: Mechanisms, Models, and Clinical Opportunities

In 2008 and 2009, two independent cancer genome sequencing projects identified mutations in isocitrate dehydrogenase 1 (IDH1) in both glioblastoma multiforme (GBM; ref. 1) and acute myeloid leukemia (AML; ref. 2). Further investigation revealed that mutations of the homologous enzyme isocitrate dehydrogenase 2 (IDH2) were present in other cases of these diseases (3). In IDH1, the disease-associated alterations were all missense mutations confined to a single arginine residue, R132, in the enzyme active site. An R132H substitution constituted the vast majority of mutational events. In IDH2, the disease-associated mutations were present at the homologous R172 residue (with the alterations being predominantly R172K), and at another active site arginine R140 (with the alterations being predominantly R140Q; ref. 4). In all cases, the mutations in IDH1 and IDH2 were mutually exclusive and present in the hemizygous state. More than 75% of grade 2/3 gliomas and secondary glioblastomas, and more than 20% of cytogenetically normal AML cases, were found to harbor an IDH1 or IDH2 mutation. At the time, this discovery represented a major success for cancer genome sequencing projects and sparked a flurry of investigations aimed at understanding the role that IDH mutations play in tumorigenesis.

IDH1 is localized in the cytoplasm and peroxisomes, whereas IDH2 is found in the mitochondrial matrix. These homodimeric enzymes both catalyze a redox reaction that converts isocitrate to α-ketoglutarate [α-KG; also known as 2-oxoglutarate (2-OG)] while reducing NADP to NADPH and liberating CO₂ (Fig. 1). IDH1 and IDH2 are highly homologous to each other but are structurally, functionally, and evolutionarily distinct from the NAD-dependent, heterotrimeric IDH3 enzyme that functions in the tricarboxylic acid (TCA) cycle to produce the NADH required for oxidative phosphorylation. The physiologic function of the NADP-dependent IDH1/2 enzymes has not been well characterized, but they are thought to play roles in the metabolism of glucose, fatty acids, and glutamine, and to contribute to the maintenance of normal cellular redox status.

The reaction catalyzed by IDH1 and IDH2 is thought to be one of only three major mechanisms of NADPH production in mammalian cells (5). NADPH is the major carrier of reducing equivalents in cells. It supplies reducing power to key reactions in a number of macromoecular biosynthetic pathways, and to systems that defend against the oxidative stress imposed by reactive oxygen species (ROS). For example, both the glutathione and thioredoxin antioxidant systems rely on the reducing power of NADPH to regenerate their capacity to detoxify oxidative damage. Therefore, mutations that disrupt the normal functions of IDH1 and IDH2 may have significant consequences for cellular redox balance. In fact, the initial examinations of this question produced data suggesting that the tumor-associated mutations in IDH1 acted via dominant negative inhibition of the wild-type enzyme's activity (6). It was proposed that this inhibition could lead to a drop in 2-OG levels and disrupt intracellular signaling processes sensitive to the concentration of this metabolite (6). However, this loss-of-function explanation did not fit well with the very restricted spectrum of observed mutations in the IDH enzymes or their heterozygosity.

The key breakthrough in understanding the effects of IDH1 and IDH2 mutations was a discovery made by scientists at Agios Pharmaceuticals in late 2009. Using a metabolite profiling strategy, they discovered that the mutant IDH enzymes acquire a neomorphic activity in which the normal product α-KG is converted to 2-hydroxyglutarate (2-HG) in a reaction that consumes, rather than produces, NADPH (ref. 7; Fig. 2). Structural data revealed that these mutations not only alter the affected arginine but also substantially reorganize several other key residues within the active site. As a result, the affinity of the mutant enzymes for isocitrate is reduced and their affinity for α-KG and NADPH is increased, allowing the neomorphic reaction to proceed.

The product of the novel reaction, 2-HG, is a poorly understood metabolite that does not participate in any known productive metabolic pathway. This chiral molecule is very similar in structure to the achiral 2-OG and is normally present at low levels in both its D- (or R-) and L- (or S-) enantiomeric forms. (This article will use the D/L rather than the R/S nomenclature to describe these molecules.) Several wild-type metabolic enzymes produce both 2-HG enantiomers as low efficiency by-products, and both are maintained at low concentrations in normal cells and tissues by 2-HG dehydrogenase housekeeping enzymes that recycle these compounds back to α-KG (Fig. 3). In vitro and in vivo experiments have shown that the mutant IDH enzymes exclusively produce the D- form of 2-HG (D2HG). In patients whose tumors bear somatic IDH1 or IDH2 mutations, D2HG production outstrips its elimination such that it builds up to concentrations that are orders of magnitude higher than normal, in some cases to millimolar levels (4, 7, 8). This outcome prevails no matter what amino acid substitution occurs at any of the 3 mutated arginine residues in IDH1/2. Although the detailed biochemical mechanism of catalysis by mutant IDH enzymes is not yet fully understood, it seems that the presence of the wild-type enzyme derived from the normal IDH allele increases the capacity of the mutant IDH enzyme to produce 2-HG (9). These findings, combined with the pattern of mutation in these genes, support the concept that the production of D2HG by the mutant enzyme is responsible for driving tumor progression. Hence, D2HG has been described as an “oncometabolite” in glioma and AML.

There is another clinical situation that results in high levels of 2-HG. A group of rare, inherited, neurometabolic disorders called 2-hydroxyglutaric acidurias are characterized by dramatically elevated D2HG or L2HG levels in the central nervous system (CNS), serum, and urine of affected patients (10, 11). All L2HG acidurias, and half of D2HG acidurias, are caused by homozygous loss-of-function mutations in the 2-HG dehydrogenase enzymes responsible for converting L2HG and D2HG back into α-KG. Subsequent to the discovery of the D2HG-producing IDH mutations, the approximately 50% of D2HG aciduria patients for whom the underlying genetic defect was unknown were found to harbor germline IDH2 mutations (12).

The phenotypes of 2-HG acidurias are diverse, with patients exhibiting a range of developmental CNS symptoms, neurologic deficits, and a reduced lifespan. Some of these patients are at higher risk of developing brain tumors, consistent with the proposed effects of the 2-HG produced by mutant IDH enzymes in gliomas. However, malignant brain tumors have been described only in patients with L2HG acidurias. Furthermore, the majority of these tumors were medulloblastomas or pediatric neuroectodermal tumors, which is inconsistent with the histologic spectrum of IDH-related brain tumors (13). Early mechanistic studies in rats aimed at understanding the pathology of these diseases suggested that 2-HG can increase ROS levels in the CNS, but few additional details have emerged (14). Patients with a 2-HG aciduria therefore provide a unique physiologic model of the effects of long-term exposure to 2-HG, and suggest that 2-HG alone is not sufficient to drive tumorigenesis. This lack of tumorigenicity seems to be especially true for the hematopoietic system, as no hematologic malignancies have been observed in 2-HG aciduria patients. Therefore, although IDH mutations appear to be major contributors to tumorigenesis in glioma and AML, additional cooperating mutations must be required for complete transformation and disease progression.

As an interesting aside, the discovery of the unique gain-of-function resulting from these specific arginine mutations in the IDH enzymes has had implications beyond the field of medicine. The industrial production of certain classes of synthetic polymers, including nylon and some pharmaceuticals, relies on adipic acid, which is currently produced from fossil fuels. Biochemical engineering approaches have identified an alternative method of adipic acid production from simple sugars that involves a series of enzymatic steps. However, the efficient production of a key enzyme activity, 2-hydroxyadipate dehydrogenase, required for this process has been lacking. By translating the cancer-associated gain-of-function mutations in the IDH enzymes to the related Saccharomyces cerevisiae homoisocitrate dehydrogenase enzyme, investigators were able to reengineer this protein to catalyze the required reaction and produce chirally pure (R)-2-hydroxyadipic acid (15). It remains to be seen whether this approach will enable the cost-effective and scalable fossil fuel–independent production of this key chemical.

Focused genotyping efforts across all major human cancer types in 2009 and 2010 revealed that IDH1 and IDH2 mutations were not present to any significant extent in most common tumors (16, 17). These data have been corroborated more recently by results from various cancer genome projects, in that only a few cases of sporadic IDH mutations have been observed. However, examinations of rarer tumor types have revealed a significant number of IDH mutations in chondrosarcoma (18), including enchondromas and hemangiomas associated with Ollier disease and Maffuci syndrome (19, 20), cholangiocarcinoma (21, 22), a small proportion of prostate cancers (23), and angioimmunoblastic T-cell lymphomas (AITL; ref. 24). As in glioma and AML, these mutations were all heterozygous events involving the R132 codon in IDH1 or the R140 or R172 codon in IDH2. Even though the incidence of these diseases is rare, and IDH mutations are relevant in only a small proportion of each patient population, there is much to be learned by studying them in concert with the more common IDH-associated cancers. The next few sections of this article describe the consequences of IDH mutation in various tumor types and the mechanistic insights derived from each.

A wealth of clinical data related to IDH1 and IDH2 mutations in glioma has been generated by numerous groups worldwide, and has been comprehensively reviewed elsewhere (25, 26). IDH mutations occur in 70% to 90% of all adult grade 2/3 astrocytomas and oligodendrogliomas, and in the secondary glioblastomas to which these tumors progress. Interestingly, IDH mutations are not a common event in primary glioblastoma (5% to 15%), suggesting that the molecular events leading to these diseases are distinct. It has even been suggested that secondary glioblastomas lacking IDH mutations would be more accurately characterized as primary glioblastomas, as their molecular features and natural history are most similar to primary GBM. IDH mutations are not common in any other CNS tumor, with the exception of a few primitive neuroectodermal tumors (27).

The molecular detection of IDH mutations has become standard practice in many institutions, and has proven useful in the clinical management of glioma. In some situations, the differential diagnosis of CNS lesions based on histology alone can be difficult. For example, distinguishing between reactive gliosis and diffuse glioma, or between oligodendroglioma and other similar entities, can be a challenge. In some of these cases, IDH status has proven extremely useful as a diagnostic biomarker (28). Furthermore, IDH mutations are associated with longer survival of patients with glioma or secondary glioblastoma and thus represent an important prognostic indicator. Although IDH mutations are associated with younger patient age and other positive prognostic features, they are independent predictors of a favorable outcome in multivariate analyses. It is not yet clear whether IDH status is also a predictive marker of response to chemotherapy or radiation treatment, as a number of studies have shown equivocal or conflicting results (26).

As noted above, IDH mutation alone is insufficient for tumorigenesis. Analysis of various CNS tumors has shown that IDH mutations are significantly associated with O-6-methylguanine-DNA methyltransferase promoter methylation (26); with mutations in TP53, commonly found in astrocytic tumors (3); and with the 1p19q codeletion event that arises from an unbalanced translocation in oligodendrogliomas (29). In fact, almost all gliomas with an IDH mutation also harbor either a TP53 mutation or a 1p19q codeletion (25, 29). Taken together, the collective data from a large number of clinical studies support a model in which IDH mutations occur early during gliomagenesis, and subsequent TP53 or 1p19q deletion events determine the progression of the disease to astrocytoma or oligodendroglioma, respectively (25, 30).

The role of IDH mutations in myeloid neoplasia has also been extensively investigated since the discovery of these mutations in AML, and comprehensive clinical reviews are available (31, 32). Unlike in glioma, where mutations are overwhelmingly R132H substitutions in IDH1, IDH mutations in AML are more diverse. Approximately half of all such mutations occur in IDH2, and both R172K and R140Q are common. The same IDH mutations have also been observed in myelodysplastic syndromes (MDS) and myeloproliferative neoplasms (MPN), both of which are precursor conditions that can progress to AML (33). However, the acquisition of an IDH mutation does not seem to be associated with progression to AML in patients with myeloproliferative disease. Again, additional tumorigenic events seem to be required.

IDH mutations do not cluster within any of the subtypes of AML as defined by the French-American-British classification scheme, except that they are absent in acute promyelocytic leukemias driven by the 15;17 translocation that leads to PML-RARA fusions. There is a significant association of IDH mutations with mutations in nucleophosmin mutations in patients with AML with a normal cytogenetic profile, placing them within a group of patients with an intermediate risk profile. Unlike in glioma, IDH mutations do not provide any clinically significant prognostic information in AML, especially when examined in the context of the most recent molecular prognostic models. However, D2HG is highly elevated in cells and serum of patients with IDH-mutant AML irrespective of the specific mutation. Thus, D2HG concentration may serve as a biomarker of response to treatment and disease progression (8).

At the genetic level, IDH mutations in AML are mutually exclusive with mutations in tet methylcytosine dioxygenase 2 (TET2), which is altered in 10% to 20% of AML cases. This observation suggests that IDH1/2 and TET2 mutations may act via the same mechanism (34). The mechanism of tumorigenesis associated with IDH and TET mutations will be discussed later in this article.

AITL is one of the three most common peripheral T-cell lymphoma (PTCL) subtypes, along with anaplastic large cell lymphoma and PTCL-not-otherwise-specified (35). AITL normally presents as a systemic disease, with polyadenopathy and a variety of immunologic abnormalities, and carries a poor prognosis. On the basis of molecular marker expression and microarray expression profiling, AITL is thought to arise from the follicular T-helper (T_FH) cells that are present in germinal centers and cooperate with B cells in mounting an effective immune response (35, 36). Recurrent mutations in the TET2 gene are associated with AITL, but the molecular pathogenesis and other underlying genetic events driving AITL are largely unknown (37). IDH2^R172 mutations are present in approximately 20% to 45% of AITL cases, but not in other types of PTCL (24). However, there are no clinical or pathologic differences among patients with AITL based on IDH2 status, and their prognoses are the same. IDH1 mutations have not been detected in patients with AITL, supporting the concept that IDH1 and IDH2 mutations can have different effects depending on the cellular context.

As in AML, the identification of IDH2 mutations in AITL presents an opportunity to develop a mechanistic model of the disease, develop clinically useful biomarkers, and evaluate novel therapeutic strategies. Of note, although there are no AITL cell lines currently available, the first AITL xenografts have recently been reported (38), opening up opportunities for further mechanistic study. Characterization of these model tumors should soon yield fresh insights into IDH2-mutant AITL.

Cartilaginous tumors range from benign enchondromas through low-grade chondrosarcomas to highly malignant chondrosarcomas, and represent a significant proportion of human mesenchymal tumors. Ollier disease and Maffucci syndrome are noninherited conditions characterized by multiple benign enchondroma tumors whose underlying genetic lesions are largely unknown. Because patients with these conditions are also at higher risk of developing glioma and AML, a large survey of IDH mutation status in chondrosarcoma, osteosarcoma, and other mesenchymal tumors was conducted (18), as well as a focused assessment of patients with Ollier disease and Maffucci syndrome (19, 20). Approximately 50% of sporadic central and periosteal chondrosarcomas were found to have hemizygous IDH1 or IDH2 mutations, with IDH1 mutations predominating. No IDH1 mutations were identified in any other mesenchymal tumor type (18). In patients with Ollier disease or Maffuci syndrome, a high proportion of the enchondroma tumors present also harbored IDH mutations. Interestingly, a low level of IDH mutation was also found in nontumor DNA in these patients, suggesting that postzygotic IDH mutations, and the resulting somatic mosaicism, can lead to these diseases (19, 20). In all cases investigated, elevated D2HG levels were detected in these patients. Similar to D2HG aciduria patients, further investigation of the distribution of IDH-mutant cells in these individuals should provide information regarding the susceptibility of various cell lineages to the effects of IDH mutations and elevated D2HG levels.

Cholangiocarcinomas are a relatively rare subset of gastrointestinal malignancies involving the gall bladder and bile duct. Patients with these cancers have a poor prognosis, with a median survival of approximately 1 year (39). The spectrum of genetic events associated with cholangiocarcinoma is not completely characterized, but two studies examining mutations in the tumors of several hundred patients with cholangiocarcinoma have been conducted (21, 22). IDH mutations were identified in 10% to 23% of intrahepatic cholangiocarcinomas, but not in cholangiocarcinomas of extrahepatic origin. Elevated levels of D2HG were detected in tumor tissues, and IDH1 mutations were more common than IDH2 mutations. Curiously, the pattern of amino-acid substitutions in the mutant IDH enzymes was significantly different than that observed in glioma or AML. For example, the R132H mutation so common in glioma and AML was significantly under represented in cholangiocarcinoma. As in glioma, the presence of an IDH mutation seems to predict better overall patient survival.

Three independent studies have investigated IDH1 mutations in prostate cancer. An initial survey of a broad variety of tumor types identified a small number of prostate cancers with IDH1 mutations (17). A follow-up study focused specifically on prostate tumors identified IDH1 mutations in approximately 2% of patients (23), a result confirmed in a third study (40). No IDH2 mutations have been reported in prostate cancers to date. Although 2% is a low incidence rate, prostate cancer is highly prevalent, meaning that even a small mutation rate represents a significant number of patients with IDH1-mutant prostate cancer. Unfortunately, even considering these three studies, too few patients have been profiled thus far to draw any meaningful conclusions about the relationship between IDH1 mutations and the pathophysiology of this disease. A greater understanding of the role of IDH mutation in this setting may eventually offer additional tools for the management of prostate cancer.

The currently available evidence suggests that the ability of mutant IDH enzymes to promote tumorigenesis is due to their neomorphic enzyme activity and D2HG production. In principle, the acquisition of an IDH mutation could impair a cell's ability to regulate its redox state, defend against ROS, and/or metabolize several classes of macromolecules. However, to date, the majority of work investigating the effects of IDH mutations has focused on the role of D2HG as a competitive inhibitor of a class of enzymes dependent on α-KG as a cosubstrate. These enzymes are most commonly referred to as 2-OG–dependent dioxygenases, and this nomenclature will be used in this article. D2HG and α-KG are structurally identical except that the C2 carbonyl group of α-KG is replaced by a hydroxyl group in D2HG. Not surprisingly, considering the high levels of D2HG generated in IDH-mutant cells, this molecule indeed competitively inhibits this class of 2-OG–dependent enzymes. More than 80 2-OG–dependent dioxygenases exist in humans and are involved in diverse biologic processes. In general, these enzymes require oxygen, ascorbate, and iron as cofactors, and convert α-KG to succinate and CO₂ while carrying out their specific enzymatic activities. Several members of this family have been studied in depth, and the mechanisms of their catalysis and inhibition have recently been reviewed (41).

In vitro biochemical studies have shown that both D2HG and L2HG can act as competitive inhibitors of 2-OG–dependent dioxygenases by binding to the α-KG–binding pocket in the enzyme's active site (42). Although the ability of 2-HG to compete with α-KG varies depending on the specific enzyme, it appears that L2HG is consistently a better inhibitor than D2HG. However, even for D2HG, the K_m values for enzyme inhibition are well within the range of concentrations observed in cells and tissues of IDH-mutant tumors, suggesting that D2HG can impair the activity of this class of enzymes in vivo. To date, the enzymes that have received the most attention as potential targets of D2HG inhibition include the TET proteins involved in DNA methylation, the JumonjiC domain-containing histone demethylases, the prolyl hydroxylases (PHD) and lysyl hydroxylases (LHD) required for collagen folding and maturation, and the PHDs that regulate hypoxia-inducible factor (HIF) signaling (Fig. 4).

Epigenetic regulation of gene expression occurs via chemical modification of histone proteins and of DNA itself. Disruption of normal patterns of epigenetic regulation is known to play a major role in tumor progression. Methylation of cytosine residues in DNA and methylation of lysine residues in histone proteins are key epigenetic events that are both controlled in part by 2-OG–dependent dioxygenases. Accordingly, D2HG-mediated inhibition of these enzymes has been proposed as a means by which IDH mutations could contribute to tumorigenesis.

Changes to DNA methylation are known to influence both neuronal and hematopoietic cell differentiation by regulating stem cell self-renewal and directing commitment to cell fates. The TET family of dioxygenases (TET1, TET2, and TET3) was recently shown to catalyze the hydroxylation of 5-methylcytosine residues in DNA (43), resulting in an accumulation of 5-hydroxymethylcytosine. This hydroxylation event is the first step in a pathway leading to the demethylation of these cytosine residues. Although the precise mechanisms regulating the activity and specificity of the TET enzymes are not fully understood, it is clear that they play important roles in maintaining appropriate DNA methylation patterns across the genome during key stages of development (43). Interest in TET enzymes recently surged with the discovery of TET2 mutations in cases of AML, MPN, and MDS that lacked IDH mutations (34, 44). In vitro, IDH mutations interfere with TET-driven demethylation of DNA, resulting in elevated 5-methylcytosine levels and generating a hypermethylation signature characteristic of less differentiated cells. Accordingly, samples from patients with IDH-mutant AML or TET2-mutant AML show similar patterns of global DNA hypermethylation with characteristic epigenetic changes (34). In glioma, a tumor phenotype known as the CpG island methylator phenotype is highly associated with IDH1 mutations (45). Although TET loss-of-function mutations have not been observed in gliomas, both silencing of TET2 by promoter hypermethylation (46) and nuclear exclusion of TET1 (47) have been found in tumors with wild-type IDH enzymes. These observations suggest that, in gliomas lacking IDH mutations, alternative mechanisms inhibiting the TET enzymes may play a role in disease pathogenesis. In cholangiocarcinomas and chondrosarcomas, IDH mutations are associated with DNA hypermethylation of CpG islands in a manner also seen in glioblastomas and AML (20, 21). In vitro experiments have shown that IDH mutations are sufficient to establish this methylation signature and thereby influence cellular differentiation (48, 49).

In addition to DNA methylation, the methylation of lysine residues on histone proteins also contributes to alterations to chromatin structure and transcriptional regulation. Histone methylation events combine with other posttranslational modifications, including acetylation, phosphorylation, and ubiquitination, to establish a histone code that is responsible for recruiting the multiprotein complexes which epigenetically control transcription. A major family of enzymes responsible for the regulation of histone methylation is the Jumonji C (JmjC)-domain–containing histone demethylases, which are also members of the class of 2-OG–dependent dioxygenases potentially inhibited by D2HG. A member of the JmjC family has been crystallized in a complex with D2HG, clearly showing that D2HG can occupy the α-KG–binding site (50). There is also experimental evidence from cell culture systems showing that the introduction of IDH mutations can alter histone methylation patterns (51). Because of the complex combinatorial nature of the histone code, and technical challenges in specific and quantitative measurement of histone modifications, predictions regarding the effects and importance of these methylation changes are difficult to make. However, the genetic and epigenetic characterization of patient samples combined with in vitro biochemistry results and data from cell culture systems strongly implicate D2HG-driven epigenetic alterations as an important feature of IDH-mutant cancers.

Hydroxylation of proline and lysine residues in collagen is critical for reinforcing this protein's helical structure and thermal stability, and is required for certain glycosylation events (52). These posttranslational modifications are carried out by 2-OG–dependent dioxygenases residing in the endoplasmic reticulum (ER). These enzymes require ascorbate as a cofactor, and a lack of sufficient ascorbate levels causes the loss of activity of these enzymes that is responsible for the symptoms of scurvy. Hydroxylated collagen is a key component of the extracellular matrix (ECM), and interaction between the ECM and neurons or hematopoietic cells regulates their proliferation and differentiation in the brain or bone marrow, respectively. Therefore, altered ECM structure and function due to D2HG-mediated inhibition of 2-OG–dependent dioxygenase activity could contribute to tumor progression in both glioma and AML (53). In addition, disruption of collagen folding by suppression of PHD activity leads to ER stress due to the accumulation of misfolded collagen protein in the ER, and ER stress is closely associated with transformation. Although patients with D2HG aciduria do not exhibit scurvy symptoms, the local inhibition of collagen hydroxylase enzymes by D2HG may contribute to the phenotype of patients with IDH-mutant tumors.

Another group of enzymes potentially inhibited by D2HG are the PHDs that regulate the protein stability and transcriptional activity of the HIF1 and HIF2 transcription factors. Under normoxic conditions, these enzymes hydroxylate the HIF1α and HIF2α proteins on key proline residues such that they are recognized by Von Hippel Lindau (VHL) protein, a ubiquitin E3 ligase. This ubiquitination targets HIF1α and HIF2α for rapid proteasomal degradation. In the absence of oxygen, these hydroxylation events cannot proceed efficiently and HIF1α and HIF2α are free to dimerize with HIF1β, forming complete transcription factors that coordinate the response to hypoxia. HIF stability is also influenced by intracellular concentrations of α-KG, ROS, and the activity of other signaling pathways (54). Importantly, the PHDs responsible for HIF regulation have additional cellular targets, and therefore the inhibition of these enzymes by D2HG could have HIF-independent effects (55). This regulatory complexity means that HIF stability and target gene activation are cell type- and context-dependent. Direct in vitro studies examining the effects of 2-HG on PHDs and HIF activity have yielded conflicting results. Some reports show that D2HG acts as an activator rather than an inhibitor of PHDs, leading to decreased levels of HIF (56, 57). However, other studies have observed the expected inhibitory effect of D2HG on the PHD2 enzyme and increased HIF levels (50). The explanation for these different findings and the molecular mechanism by which D2HG activates PHDs are yet to be determined. In addition, in vitro biochemical data have shown that the affinity of D2HG for the HIF PHDs is lower than its affinity for other 2-OG–dependent dioxygenases, further calling into question the involvement of the HIF PHDs in IDH-associated tumorigenesis. More accurate in vivo models and detailed molecular characterization of patient samples will be required to define the role of the HIF PHDs in IDH-mutant diseases.

Little is known about the metabolic requirements of hematopoietic and neural stem and progenitor cells, or how metabolism is altered during their differentiation. Given the positioning of IDH1 and IDH2 in the cellular metabolic network, and their ability to produce NADPH, mutations in these enzymes may perturb cellular metabolism in profound ways that contribute to tumorigenesis. Moreover, a relevant phenomenon initially reported in the 1960s has recently been confirmed: that wild-type IDH enzymes are able to catalyze the reverse reaction in which α-KG is carboxylated to form isocitrate while consuming NADPH (58–60). This reverse reaction seems to be particularly important for the synthesis of fatty acids under conditions of mitochondrial dysfunction or hypoxic stress. Therefore, mutation of an IDH enzyme could alter either the forward reaction or the reverse reaction or both, as well as upset the NADP/NADPH ratio. Initial metabolic profiling of the effects of IDH mutations has been conducted in vitro using overexpression systems, and the introduction of a mutant IDH enzyme seems to alter several metabolic pathways (61). However, it is difficult to predict whether these particular changes also occur in vivo in the cell types of interest, because the metabolic microenvironment of the affected tissues will also affect intracellular metabolic regulation. Further investigations of the metabolic effects of IDH mutations are required to fully understand their effects at the levels of the tissue and the organism.

Mechanistic studies of IDH mutations have been hampered by a lack of appropriate model systems. Initial surveys of existing cell lines and attempts to derive IDH-mutant cell lines were largely unsuccessful, suggesting that these mutations are not generally compatible with viability under standard cell culture conditions. More recently, the use of alternative techniques has led to the successful derivation of several IDH-mutant cell lines, including a glioma cell line and xenograft (62), three chondrosarcoma lines (63), and a set of astrocytoma lines (9). The commonly used fibrosarcoma cell line HT-1080 also harbors an IDH1^R132C mutation and produces D2HG in culture (64). In vitro analyses of these lines have shown that cell membranes are not permeable to D2HG and that this metabolite is not readily taken up by cells in culture, compromising the direct investigation of the effects of IDH mutations and D2HG in controlled experimental systems (50). A further complication is that the metabolic phenotype of cells is heavily influenced by local microenvironmental conditions, which can vary widely in different tissues and are not accurately modeled by standard in vitro cell culture. Thus, the effects of IDH mutations on cellular metabolism have been difficult to explore in cultured cells.

Genetically engineered mouse models have provided more relevant systems for understanding the pathophysiologic consequences of IDH mutations. In two recent articles (65, 66), our group described the generation and initial characterization of a conditional knock-in (KI) mouse that was constructed using the lox-stop-lox (LSL) system and bears the most common IDH1 mutation, R132H. In the absence of Cre-recombinase (Cre) expression, the LSL IDH1^R132H mutant allele is null, but when Cre is expressed, the stop codon upstream of exon 4 is excised and the mutant IDH1 protein is expressed from the endogenous locus. Interestingly, mice homozygous for the null Idh1 allele (Idh1 knockout mice) are viable and fertile, indicating that wild-type IDH1 function is dispensable in mice under standard laboratory conditions. However, mice heterozygous for the LSL IDH1^R132H allele crossed with a mouse constitutively expressing Cre are embryonic lethal at an early stage. This observation supports the hypothesis that the gain-of-function of the mutant enzyme and its production of D2HG are critical for the tumor-promoting effects of mutated IDH1.

To examine the effects of the IDH1^R132H mutation on the myeloid compartment, LSL IDH1^R132H mice were crossed to Vav-Cre or LysM-Cre mice, resulting in the expression of the mutant IDH1 protein in the entire hematopoietic system or early during myeloid development, respectively. In both cases, the progeny were born at the expected Mendelian ratio and displayed no gross phenotype early in life. However, by 6 to 12 months of age, both mutants exhibited decreased bone marrow cellularity, splenomegaly, and evidence of extramedullary hematopoiesis indicative of a dysfunctional bone marrow niche. Levels of D2HG were highly elevated in the hematopoietic tissues, serum, and urine of these animals. Although they did not develop frank myeloid leukemias, these mice had elevated numbers of Lin⁻Sca1⁺cKit⁺ (LSK) cells in the bone marrow and spleen. The LSK population includes long-term hematopoietic stem cells (HSC), short-term HSCs, and lineage-restricted progenitor cells (LRP). In contrast, there were no significant alterations to the more committed progenitor cell populations. This phenotype is reminiscent of human MDS (33).

To examine the epigenetic effects of the conditional IDH1^R132H mutation, a preliminary analysis of the methylation signature of LSK cells was conducted. LysM-Cre IDH1^R132H KI mice showed a greater proportion of highly methylated DNA fragments whose genomic regional distribution resembled that in human AML (34). Histone methylation at some key H3 lysine residues was also increased, consistent with D2HG-mediated inhibition of histone demethylases. A partial block in cellular differentiation and/or an increase in symmetric cell division within the LRP pool may be responsible for the LRP accumulation observed in IDH1^R132H mice. In the current model, as the size of the LRP pool expands, the number of cells at risk for acquiring a second tumorigenic event increases, thereby raising the risk that these hematopoietic progenitors may give rise to AML (Fig. 5).

To examine the effects of the IDH1R132H mutation in the CNS, heterozygous IDH1^R132H KI mice were crossed with Nestin-Cre (Nes-Cre) mice to activate expression of the mutant enzyme during neuronal development (65). Under the control of Nestin, Cre is expressed as early as E10.5 in neural stem cells, and nearly complete recombination in all CNS cells is achieved by E12.5 (67). Nes-Cre IDH1^R132H mice were born at the expected Mendelian ratio but died shortly after birth due to massive cerebral hemorrhage, which could be observed as early as E14.5. Metabolic analysis revealed that α-KG levels were slightly reduced and D2HG levels dramatically increased in the brains of E14.5 Nes-Cre IDH1^R132H mice. Preliminary assessment of epigenetic differences showed that Nes-Cre IDH1^R132H mice exhibit reduced levels of 5-hydroxymethylcytosine in their DNA, consistent with D2HG-mediated inhibition of TET hydroxylases. These mutants also showed increased levels of HIF1α protein and upregulation of a set of HIF1 target genes, including VEGF. These results are consistent with D2HG-mediated inhibition of the HIF PHDs responsible for targeting HIF1α for degradation. In contrast, HIF1 levels and HIF1 target genes were not altered in hematopoietic cells in the LysM-Cre IDH1^R132H or Vav-Cre IDH1^R132H mice discussed above, suggesting that effects on HIF signaling may be context dependent (4).

An intriguing observation in Nes-Cre IDH1^R132H mice was that collagen synthesis and maturation are impaired in these mutants, as measured by the level of soluble collagen present in the developing brain. Type IV collagen is a crucial basement membrane component in the CNS and promotes appropriate interactions between astrocytes and endothelial cells. Basement membrane architecture was disrupted in Nes-Cre IDH1^R132H brains, especially around blood vessels. This disruption likely contributes to the brain hemorrhage in these mutants. In addition, it has been previously shown that disruption of collagen folding by suppression of PHD activity leads to ER stress (26). Indeed, the levels of several ER stress–responsive genes were elevated in the brains of Nes-Cre IDH1R132H mice. Taken together, these D2HG-dependent alterations to developing neural stem and progenitor cells, and to the microenvironment-mediated interaction between astrocytes and endothelial cells, may be responsible for the brain hemorrhage phenotype observed in these mice (Fig. 6).

Mouse models should continue to prove useful in defining the tumor-promoting events associated with IDH mutations. It will be interesting to examine the effects of combining these IDH mutations with other oncogenic mutations relevant to AML and glioma. Furthermore, the use of alternative promoters to drive Cre expression should allow the examination of the effect of IDH mutations on the cell types and tissues relevant to cholangiocarcinoma, chondrosarcoma, prostate cancer, and AITL. All these mouse models could also serve as accurate preclinical models for the evaluation of new therapeutic strategies targeting IDH-mutant disease in humans.

Although our understanding of the biology of IDH mutations is incomplete, researchers are already exploring several opportunities for clinical application. Sensitive and specific identification of IDH mutations by detection of mutant DNA, detection of mutant protein by immunohistochemistry (68), and measurement of D2HG levels are now all clinically feasible. The D2HG molecule itself can also be directly imaged using nuclear magnetic resonance spectroscopy, allowing for noninvasive mapping of D2HG accumulation in an affected tissue (69, 70). Although technically challenging, this approach may allow IDH status to be determined when biopsies are difficult to obtain. Once known, the IDH mutation status of a patient's tumor may serve as an important diagnostic and prognostic factor. For example, in glioma, knowledge of IDH mutation status complements histologic analysis and allows a pathologist to distinguish among tumors with similar pathologies. Furthermore, because IDH mutation status is a strong predictor of progression and outcome in glioma, its incorporation into prognostic models could improve clinical management of this disease. Similarly, for PTCLs, IDH mutation status may aid in distinguishing AITL from other PTCL diseases. In AML, D2HG is being evaluated as a biomarker of disease burden and therapy response in the hope that it may prove to be a more accurate and sensitive indicator of disease activity than currently available approaches (71).

It may also soon be possible to develop therapeutic strategies that specifically target IDH-mutant disease. Inhibition of the abnormal enzymatic activity of mutant IDH1 or IDH2 could permit very specific targeting of the tumor (as opposed to normal tissue), with the therapeutic window limited only by the specificity of a compound for the mutant enzyme. Studies on the first generation of compounds targeting the neoactivity of mutant IDH enzymes have been reported, and these agents seem to display exquisite specificity for the mutant IDH activity (72). Moreover, these compounds have shown biochemical activity in vivo, in that they are able to reduce D2HG production in IDH1-overexpressing xenograft tumors (72). Recent reports have also shown that IDH1 and IDH2 small-molecule inhibitors can reverse the transforming effects of IDH1 mutations, and promote cellular differentiation in leukemia and glioma model systems (57, 73, 74). What now remains is to determine whether inhibition in established tumors results in an antitumor effect or increases tumor sensitivity to other therapies. Testing of these novel compounds in IDH-mutant mouse models should provide direct answers to some of these questions. The presence of D2HG should serve as a powerful predictive biomarker to identify those who could potentially benefit from IDH-targeted therapies. D2HG may also be a useful pharmacodynamic biomarker that can be used to monitor therapy response. Finally, an effective and specific IDH2 inhibitor could provide an effective therapy for the 50% of patients whose D2HG aciduria is caused by a germline IDH2 mutation.

Beyond direct targeting of mutant IDH enzymatic activity, there may be other opportunities to successfully treat IDH-mutant tumors. Because of the highly elevated levels of D2HG in these patients and its pleiotropic effects, IDH-mutant tumors may be particularly sensitive to other therapeutic approaches. For example, the dramatic effects of IDH mutations on the epigenetic state of tumor cells may present options for effective epigenetic-targeted therapy. The identification of a number of recurrent mutations in MPN, MDS, and AML that alter epigenetic regulation (including IDH mutations) has led to increased interest in therapeutically targeting DNA and histone methylation in these diseases. The DNA-demethylating agents 5-azacytidine (5-Aza-CR) and 5-aza-2′-deoxycytidine (5-Aza-CdR, decitabine) act as DNA methyltransferase inhibitors and are already used in treating patients with high-risk MDS and AML (75). Further development of more specific epigenetic-modulating agents that can serve as useful tools in the management of IDH-mutant diseases is underway. A greater understanding of the effects of IDH mutations on tumor metabolism and physiology may reveal other potential opportunities to treat patients bearing these mutations.

It has been less than 5 years since IDH1 mutations were first discovered in glioblastoma. In this relatively short time, clinical evaluation of thousands of patient samples has defined the set of neoplastic diseases in which IDH mutations play an important role (Fig. 7). Mechanistic studies have discovered that these mutations cause the IDH protein to acquire a unique enzymatic activity that generates high levels of a previously little-studied metabolite, D2HG. Strong biochemical, cell culture, and clinical evidence has been gathered that has defined the effects of D2HG on the 2-OG–dependent dioxygenases, and has implicated alterations to a number of biologic processes as contributors to malignant progression in IDH-mutant tumors. The first mutant IDH KI and IDH loss-of-function mouse strains have been developed and characterized, and will continue to serve as powerful experimental models of these diseases. IDH mutation detection and D2HG measurement have already been incorporated into clinical practice in a number of settings and are under evaluation for their use as tools in the management of several diseases. Finally, the first specific small-molecule inhibitors of mutant IDH enzymes have been developed, raising hope for a new and effective treatment strategy for tumors bearing these mutations. While many questions remain regarding the role of IDH mutations in tumorigenesis, the stage is set for rapid progress in getting answers in the years to come.

No potential conflicts of interest were disclosed.

Conception and design: R.A. Cairns, T.W. Mak

Writing, review, and/or revision of the manuscript: R.A. Cairns, T.W. Mak

The authors thank M. Saunders for scientific editing of the manuscript.

This work was supported in part by the Leukemia and Lymphoma Society and the Canadian Institutes of Health Research.