Biological role and therapeutic potential of IDH mutations in cancer

Summary

Hotspot mutations in isocitrate dehydrogenase 1 (IDH1) and isocitrate dehydrogenase 2 (IDH2) occur in a variety of myeloid malignancies and solid tumors. Mutant IDH proteins acquire a neomorphic enzyme activity to produce the putative oncometabolite D-2-hydroxyglutarate, which is thought to block cellular differentiation by competitively inhibiting α-ketoglutarate-dependent dioxygenases involved in histone and DNA demethylation. Small molecule inhibitors of mutant IDH1 and IDH2 have been developed and are progressing through pre-clinical and clinical development. In this review, we provide an overview of mutant IDH-targeted therapy and discuss a number of important recent pre-clinical studies using models of IDH-mutant solid tumors.

Introduction

A mutation in the isocitrate dehydrogenase 1 gene (IDH1) was first identified in a single case of colorectal cancer as part of an early investigation of protein-coding mutations in human cancers (Wood et al., 2007). Recent advances in cancer genetics have established that hot-spot mutations in IDH1 and IDH2 occur frequently in a variety of human cancers, including malignant gliomas, acute myeloid leukemia (AML), intrahepatic cholangiocarcinoma, chondrosarcoma, and thyroid carcinomas (Borger et al., 2012; Kang et al., 2009; Mardis et al., 2009; Parsons et al., 2008; Paschka et al., 2010; Ward et al., 2010; Yan et al., 2009; Yang et al., 2012). In addition, IDH2 mutations are known to occur with high frequency in rare malignancies, such as angioimmunoblastic T-cell lymphoma and solid papillary carcinoma with reverse polarity (SPCRP) (Cairns et al., 2018; Chiang et al., 2016). IDH mutations are also reported to occur infrequently in prostate tumors, paraganglioma, and melanoma (Gaal et al., 2010; Kang et al., 2009; Lopez et al., 2010).

Wild-type IDH1 and IDH2 are important metabolic enzymes that catalyze the oxidative decarboxylation of isocitrate to generate α-ketoglutarate (αKG) and CO₂. IDH1 localizes to the peroxisomes and cytosol, while IDH2 localizes to the mitochondria. A third enzyme complex, IDH3, is encoded by three distinct genes (IDH3A, IDH3B, and IDH3G), none of which have been identified as significantly mutated genes in human cancers, and will not be discussed further in this review.

Cancer-associated IDH1 and IDH2 mutations occur, almost exclusively, at distinct arginine residues in the enzyme active sites (Kang et al., 2009; Yan et al., 2009). Missense mutations in the IDH1 Arg¹³² codon cause a single amino acid substitution, most commonly to histidine (IDH1^R132H), but also to cysteine, serine, glycine, leucine, or isoleucine (Kang et al., 2009; Yan et al., 2009). Missense mutations in IDH2 codon for Arg¹⁴⁰ or Arg¹⁷² (homologous to IDH1^R132) occur predominantly as IDH2^R140Q or IDH2^R172K substitutions, although other amino acid changes do occur (Medeiros et al., 2017; Waitkus et al., 2016).

The common function of IDH1/2 active-site mutations is a neomorphic enzyme activity that catalyzes the conversion of αKG to D-2-hydroxyglutarate (D2HG). Under physiological conditions, cellular D2HG accumulation is limited due to the actions of the endogenous D2HG dehydrogenase (D2HGDH), which catalyzes the conversion of D2HG to αKG. However, the neomorphic activity of mutant IDH causes D2HG to accumulate to supraphysiological levels within cells. Elevated D2HG concentrations can be detected in the serum of patients with IDH-mutant AML and in IDH-mutant gliomas in patients (Dang et al., 2009; Dinardo et al., 2013; Elkhaled et al., 2012; Stein et al., 2017; Ward et al., 2010).

Elevated D2HG levels in tumor tissues may provide a clinically useful biomarker for the non-invasive detection of IDH mutations due to the low background of D2HG in normal tissue and almost invariable upregulation of D2HG in the context of IDH active site mutations (Andronesi et al., 2013). In gliomas, a number of studies have investigated the potential for non-invasive imaging strategies to detect D2HG as a method for discriminating between IDH-mutant and IDH-wildtype tumors (Andronesi et al., 2012; Choi et al., 2012; Elkhaled et al., 2012; Emir et al., 2016). These non-invasive imaging studies and their clinical implications have been reviewed elsewhere and will not be discussed in detail in this article (Andronesi et al., 2013; Leather et al., 2017). However, it is important to note that the promise of non-invasive diagnosis of IDH mutant gliomas via magnetic resonance spectroscopy (MRS) may have significant clinical implications by informing on IDH status prior to surgery. For example, it has been reported that IDH mutant gliomas are more amenable to surgical resection and maximal surgical resection may provide a significant survival benefit for glioma patients with IDH mutations, particularly in the context of tumors without 1p/19q co-deletion (Beiko et al., 2014; Kawaguchi et al., 2016; Miller et al., 2017). As such, non-invasive imaging strategies capable of identifying IDH-mutant gliomas with high sensitivity and specificity prior to surgery may provide an opportunity for clinicians to individualize surgical strategies based on the genetic basis of the tumor.

Mechanistically, D2HG has been demonstrated to inhibit αKG-dependent dioxygenases that are involved in the regulation of epigenetics and differentiation and is thought to induce epigenetic dysfunction in a manner that inhibits normal cellular differentiation. Specifically, elevated D2HG levels competitively inhibit αKG-dependent lysine demethylases, resulting in elevated levels of histone methylation in a variety of cell line models (Chowdhury et al., 2011; Xu et al., 2011). D2HG also inhibits the TET family of 5-methylcytosine hydroxylases, a family of enzymes involved in the first step of active DNA demethylation (Koh et al., 2011; Kohli and Zhang, 2013; Xu et al., 2011). In both glioma and AML, epigenetic dysregulation caused by elevated D2HG levels is reported to induce a DNA hypermethylation phenotype, which is clinically associated with increased methylation of patient tumor DNA and the glioma-associated CpG island hypermethylator phenotype (G-CIMP) (Duncan et al., 2012; Figueroa et al., 2010; Kernytsky et al., 2015; Noushmehr et al., 2010; Turcan et al., 2012). Several lines of experimental and clinical evidence suggest that D2HG-induced dysregulation of histone and DNA methylation inhibits normal cellular differentiation processes (Figueroa et al., 2010; Lu et al., 2012; Saha et al., 2014; Wang et al., 2013). Consequently, inhibition of cellular differentiation by D2HG is thought to promote the pathological self-renewal of stem-like progenitor cells, which may create a cellular state that is permissive to malignant transformation. In addition to inhibiting αKG-dependent dioxygenase enzymes, D2HG has also been reported to activate, directly or indirectly, a number of enzymes and pathways (Carbonneau et al., 2016; Karpel-Massler et al., 2017; Koivunen et al., 2012). For example, D2HG has been shown to activate the prolyl hydroxylase EGLN (Koivunen et al., 2012). Interestingly, this activation of EGLN appears to be a stereospecific effect of the D2HG enantiomer, as L2HG is reported to inhibit EGLN activity (Losman et al., 2013).

In preclinical models, expression of IDH1^R132H in myeloid cells has been reported to induce splenomegaly, decrease bone marrow cellularity, and increase the abundance of hematopoietic progenitor cells in a manner that coincides with increased histone H3 methylation and CpG island hypermethylation (Sasaki et al., 2012a). This finding is consistent with the observation that patient AML cells with IDH1 or IDH2 mutations display a common hypermethylated DNA phenotype and that transgenic expression of IDH2^R140Q or IDH2^R172K impairs the differentiation of 32D cells (a murine myeloid progenitor line) in cell culture (Figueroa et al., 2010). Collectively, evidence from AML patients and preclinical models strongly suggests that IDH1 and IDH2 mutations are oncogenic drivers of AML and myelodysplastic syndrome and that targeting IDH mutant neomorphic activity in this context may provide therapeutic benefit by promoting the differentiation of malignant myeloid cells (Losman et al., 2013; Medeiros et al., 2017; Sasaki et al., 2012a; Wang et al., 2013).

Similar to the results in hematopoietic model systems, expression of IDH1^R132H mutation in CNS tissues has been shown to alter neurodevelopment and impair the differentiation of neural progenitor cells (Bardella et al., 2016; Pirozzi et al., 2017; Sasaki et al., 2012b). Specifically, Sasaki et al. expressed IDH1^R132H under control of the nestin or GFAP promoters in murine CNS cells and did not observe IDH1^R132H-dependent glioma formation. Instead, expression of mutant IDH1 resulted in perinatal lethality in all nestin- and the majority (92%) of GFAP-promoter driven mutant IDH expression. The remaining ~8% of GFAP-IDH1^R132H mice survived into adulthood and many of the surviving mice developed splenomegaly and liver tumors, likely due to leakiness of the GFAP promoter-regulated expression of Cre-recombinase in non-CNS tissues (Sasaki et al., 2012b). A more recent study used a tamoxifen-inducible strategy to conditionally express IDH1^R132H in neural progenitor cells in 5–6 weeks old mice. Using this model, the investigators found that IDH1^R132H expression caused an expansion of the subventricular zone (SVZ) and increased the abundance of proliferative (Ki67⁺) cells in the SVZ (Bardella et al., 2016).

Collectively, results from cellular and animal models strongly suggest that supraphysiological concentrations of D2HG caused by neomorphic mutant IDH activity inhibit αKG-dependent dioxygenases, alter DNA and histone methylation, and inhibit normal differentiation processes in a manner that likely promotes leukemogenesis and tumorigenesis. Based on this evidence and the prevalence of IDH1 and IDH2 mutations in a number of human malignancies, including in gliomas and AML, intensive research efforts have been conducted to identify small molecule inhibitors of mutant IDH enzymes and to develop these molecules as drugs for anti-cancer therapy (Figure 1 and Table 1). The identification and optimization of investigational mutant-IDH targeted small molecules have recently been reviewed and will not be discussed in detail for the purposes of this review (Dang and Su, 2017). Instead, we will discuss below the pre-clinical and early clinical results of mutant IDH-targeted therapies in the context of AML and gliomas, and highlight promising developments, unanswered questions, and important future directions with regard to these therapies.

Figure 1. Potential Therapeutic Strategies for Malignancies that Harbor IDH1 or IDH2 Mutations.

A number of pharmacological inhibitors have been developed to directly inhibit the neomorphic activity of mutant IDH enzymes in an effort to reduce D2HG production and elicit differentiation of malignant progenitor cells. Alternatively, inhibitors of enzymes involved in glutaminolysis, including glutamate dehydrogenases and glutaminase, have been shown to preferentially inhibit the growth of leukemia and glioma cells with IDH mutations. Similarly, NAMPT inhibitors are proposed to exploit the observation that NAPRT1-mediated conversion of nicotinic acid to NAD⁺ is defective in IDH mutant gliomas (indicated by an X in the figure). Hypomethylating agents such as azacytidine and decitibine may also promote differentiation of IDH mutant cells, which exhibit a hypermethylation phenotype that is associated with an inhibition of differentiation. Co-occurring driver alterations, such as MYCN amplification, may be targeted in IDH mutant tumors using available targeted therapies (e.g. JQ1 for n-Myc overexpression) in a tumor-specific manner.

Table 1.

Selected list of ongoing clinical programs using IDH1-mutant or IDH2-mutant inhibitors for the treatment of malignant gliomas.

Compound Name	Clinical Trial Identifier	Trial Phase	Cancer type
AG-120 (mutant IDH1 inhibitor)	NCT02073994	Phase 1	Advanced solid tumors with IDH1 mutations, including cholangiocarcinoma, chondrosarcoma, and glioma.
AG-221 (mutant IDH2 inhibitor)	NCT02273739	Phase 1/2	Advanced solid tumors with IDH2 mutations, including glioma, angioimmunoblastic T-cell lymphoma, intrahepatic cholangiocarcinoma, chondrosarcoma
AG-881 (pan-mutant IDH inhibitor)	NCT02481154	Phase 1	Glioma with IDH1 or IDH2 mutation
BAY1436032 (mutant IDH1 inhibitor)	NCT02746081	Phase 1	Solid tumors with IDH1 mutations.
IDH305	NCT02977689	Phase 2	Grade II or III gliomas with IDH1 mutations that have progressed after observation or radiation therapy.

Preclinical and clinical results of mutant IDH inhibitors for the treatment of AML

The potential efficacy of mutant IDH-targeted therapy for AML has been demonstrated in several ex vivo and in vivo pre-clinical studies. In 2013, researchers form Agios Pharmaceuticals reported the development of the tool compound AGI-6780, a potent and selective allosteric inhibitor of IDH2^R140Q. In this study, treatment with AGI-6780 induced differentiation of TF-1 erythroleukemia cells and primary human AML cells (Wang et al., 2013). Subsequently, Agios investigators conducted a high throughput screen for additional IDH2^R140Q inhibitors and identified a number of triazine compounds with inhibitory activity against the mutant IDH2 enzyme (Yen et al., 2017). After a series of medicinal chemistry optimization steps designed to improve solubility, clearance, and oral bioavailability, the compound AG-221 (enasidenib) was identified for clinical development as an orally bioavailable allosteric inhibitor of mutant IDH2. In preclinical studies, AG-221 treatment induced differentiation of IDH2-mutant TF-1 and primary AML cells. In an IDH2-mutant AML xenograft mouse model, AG-221 treatment was well tolerated and conferred a significant survival benefit relative to vehicle-treated controls when administer in a twice-daily dose of 45 mg/kg (Yen et al., 2017).

A first-in-human Phase 1/2 study investigated the safety and tolerability of AG-221 in patients with relapsed or refractory IDH2-mutant AML. In this study, the investigators selected 100 mg/day dosing based on the pharmacokinetic/pharmacodynamic results of an initial Phase 1 dose-escalation study (Stein et al., 2017). In patients receiving 100 mg/day enasidenib treatment, the overall response rate was 38.5% (42/109 patients) and 20.2% (22/109) of treated patients achieved complete remission (Stein et al., 2017). Enasidenib treatment induced potent suppression of D2HG levels in the vast majority of treated patients, and this reduction tended to coincide with the differentiation of mutant IDH2 myeloblasts into functional neutrophils, as evidenced by the presence of mutant IDH harboring neutrophils with intact phagocytic activity in treated patients (Amatangelo et al., 2017). Importantly, the investigators also found that suppression of D2HG levels alone was not predictive of clinical response, as many non-responding patients exhibited potent suppression of D2HG. Instead, the investigators found that mutations of genes involved in activation of the RAS signaling pathway were enriched in patients that did not respond to therapy (Amatangelo et al., 2017). The promising results of these studies recently resulted in the FDA-approval of enasidenib for treating relapsed or refractory IDH2-mutant AML (Kim, 2017). Several small molecule drugs are also currently under development for the treatment of IDH1-mutated AML after demonstrating promising pre-clinical results in AML patient-derived xenograft mouse models (Chaturvedi et al., 2017; Chen et al., 2016).

It is critical to note that the collective results from pre-clinical and clinical studies of IDH-mutant leukemia suggest that mutant IDH1/2 inhibitors are efficacious due to a reduction of D2HG levels, which removes a barrier to myeloid differentiation and thus promotes differentiation of mutant IDH-expressing blast cells (Amatangelo et al., 2017; Chaturvedi et al., 2017; Yen et al., 2017). In fact, it was recently reported that treatment with inhibitors of mutant IDH1 or mutant IDH2 may elicit clinical differentiation syndrome, marked by leukocytosis and exuberant neutrophil recovery, in a small number of patients with relapsed or refractory AML (Birendra and DiNardo, 2016; Stein et al., 2017). These findings demonstrate that clinicians must be aware of the potential of mutant IDH inhibitors to induce complications such as differentiation syndrome in AML patients. Additionally, the observed robust myeloblast differentiation, neutrophil recovery, and platelet recovery in AML patients treated with IDH-mutant inhibitors strongly supports the putative mechanism of action involving the stimulation of blast differentiation via reduction of D2HG (Amatangelo et al., 2017; Birendra and DiNardo, 2016; Stein et al., 2017).

Pre-clinical results of mutant IDH inhibitors in gliomas

Due to the high frequency of IDH mutations in gliomas and several other solid tumors, the development and testing of mutant IDH inhibitors for the treatment of solid tumors has been intensively investigated. Rohle et al initially reported that treatment of TS603 cells (an IDH1^R132H/WT cell line derived from a patient with WHO Grade III oligodendroglioma) was sufficient to inhibit D2HG production and colony formation in cell culture and slowed growth of subcutaneous TS603 xenografts in SCID mice (Rohle et al., 2013). More recently, the pan-mutant inhibitor BAY1436032 was tested in two independent pre-clinical experiments using distinct dosing regimens to treat IDH1^R132H-expressing intracranial xenografts in BalbC nude mice (Pusch et al., 2017). The investigators found that once daily administration of 150 mg/kg BAY1436032 did not significantly reduce the size of intracranial xenografts (Pusch et al., 2017). However, the investigators did observe a significant decrease in intratumoral D2HG and a small but statistically significant increase in animal survival in the treated group. In a second experiment, the investigators treated mice with vehicle or twice daily doses of 35 mg/kg or 70 mg/kg BAY1436032. It was observed that twice-daily 70 mg/kg BAY1436032 dosing prolonged animal survival (Pusch et al., 2017), while the 35 mg/kg dose did not (Pusch et al., 2017).

In a separate study, Kopinja and colleagues used a brain-penetrant mutant IDH1 inhibitor (MRK-A) and evaluated its effect on intracranial xenograft growth (Kopinja et al., 2017). Interestingly, the authors found that MRK-A treatment did not alter the growth or animal survival of mice bearing a patient-derived GB10 glioma xenograft harboring a heterozygous IDH1^R132H mutation. In contrast, the investigators found that MRK-A treatment prolonged animal survival in mice bearing BT142 xenografts (Kopinja et al., 2017). Notably, although BT142 was originally derived from a IDH1^R132H/WT heterozygous tumor (Luchman et al., 2012), this line is reported to undergo spontaneous loss of heterozygosity with retention of the mutant allele, resulting in a IDH1^R132H/− hemizygous state (Luchman et al., 2013). Although no direct DNA sequencing was performed on BT142 xenografts in the study, it is important to consider the potential loss of heterozygosity when evaluating the differential responses of GB10 and BT142 to MRK-A (Kopinja et al., 2017).

Similar to the negative results of MRK-A treatment on GB10 xenografts described above, Tateishi et al. reported in 2015 that AGI-5198 treatment increased the viability of several patient derived IDH1^R132H-expressing cell lines and did not alter the survival of SCID mice bearing intracranial patient-derived IDH1^R132H/WT GBM xenografts (Tateishi et al., 2015). Furthermore, the investigators reported that long-term incubation of MGG152 cells with AGI-5198 (for up to 20 passages) prior to intracranial transplantation shortened the latency period and decreased median survival time of SCID mice (Tateishi et al., 2015), suggesting that inhibition of mutant IDH1 may cause the xenografts to become more aggressive. Using metabolomics and cell biology approaches, Tateishi et al. found that IDH1 mutant cells displayed reduced levels of NAD⁺, due to decreased expression of nicotinate phosphoribosyltransferase (NAPRT1) (Figure 1). Consequently, the investigators found that IDH1^R132H-expressing cells were sensitive to further depletion of NAD⁺ via inhibition of nicotinamide phosphoribosyltransferase (NAMPT), an alternative source of NAD⁺ production (Figure 1) (Tateishi et al., 2015).

Potential alternative strategies for treating IDH1/2 mutant cancers

Similar to the results of Tateishi et al., several other groups recently reported potential therapeutic sensitivities that are dependent on mutant IDH neomorphic activity. Several groups reported that IDH1^R132H neomorphic activity and D2HG production may induce metabolic liabilities that may limit growth in the absence of compensatory mechanisms. For example, IDH1/2 mutations have been reported to increase the rate of reductive glutaminolysis in murine glioma progenitor cells, GBM cells, and HCT116 colorectal cancer cells (Chen et al., 2014; Ohka et al., 2014; Reitman et al., 2014). Consistent with this observation, the glutaminase inhibitor bis-2-(5-phenylacetamido-1,2,4-thiadiazol-2-yl)ethyl sulfide (BPTES) has been shown to inhibit the growth of GBM cells and patient-derived AML cells (Emadi et al., 2014; Seltzer et al., 2010) (Figure 1), and clinical trials are currently ongoing to assess the safety and efficacy of the glutaminase inhibitor CB-839 (Calithera Biosciences Inc.) for several clinical indications, including myeloid malignancies with IDH mutations. Additionally, Chen et al. demonstrated that the hominoid-specific enzyme glutamate dehydrogenase 2 (GDH2) may be critical for mediating TCA cycle anaplerosis from glutamate and glutamine in order to compensate for IDH1^R132H metabolic alterations (Chen et al., 2014) (Figure 1). Our laboratory recently reported that this distinct role of GDH2 may be in part due to recently evolved amino acid substitutions in the allosteric domain (Waitkus et al., 2017).

IDH enzymes are a major source of NADPH, which is a critical reducing agent for maintaining cellular redox balance and limiting oxidative damage in cells (Dang and Su, 2017; Kim et al., 2009; Winkler et al., 1986). It has been reported that decreased NADPH production and increased D2HG production in IDH mutant cells may sensitize cells to the effects of ionizing radiation, as indicated by increased reactive oxygen species, double stand DNA breaks, and cell death, suggesting that mutant IDH inhibitors may have the undesirable effect of limiting the efficacy of irradiation (Molenaar et al., 2015). Additionally, recent evidence suggests that D2HG produced by mutant IDH enzymes may cause homologous recombination processes to become dysfunctional and induce therapeutic sensitivities to PARP inhibitors (Sulkowski et al., 2017).

In addition to exploiting mutant IDH-dependent liabilities for therapeutic purposes, another potential approach for targeting IDH mutant tumors is to target tertiary oncogenic driver alterations that occur during malignant progression in IDH mutant tumors. For example, Wakimoto et al. reported that IDH mutant gliomas tend to acquire tertiary driver alterations in known oncogenes (e.g. MYCN, PDGFRA, MET, and PIK3CA) during disease progression (Wakimoto et al., 2014). These tertiary driver alterations are associated with increased tumorigenicity when patient tissue is transplanted into the brains of mice, as well as shorter progression free survival compared to patients lacking tertiary driver alterations at the time of disease progression. Importantly, these results are consistent with large-scale sequencing efforts demonstrating that 1p/19q-intact IDH1/2 mutant astrocytic gliomas and 1p/19q co-deleted IDH1/2 mutant oligodendrogliomas acquire a variety of tertiary alterations in targetable signaling pathways, including PI3K-AKT, K-Ras, PDGRA, Met, and n-Myc pathways (Bai et al., 2015; Brat et al., 2015). For example, Wakimoto et al showed that proliferation of the patient-derived cell line MGG152, which harbors an IDH1^R132H mutations and MYCN amplification, can be inhibited by treatment with the BET inhibitor JQ1, which suppresses n-Myc expression and n-Myc-dependent transcription (Puissant et al., 2013; Wakimoto et al., 2014).

More recently, it was reported that IDH1^R132H expression and D2HG elevation may compromise oxidative phosphorylation processes in GBM cells, leading to a significant reduction in ATP production and increased AMPK activation (Karpel-Massler et al., 2017). In that study, the investigators concluded that the inhibitory effects of D2HG on oxidative phosphorylation caused AMPK-mediated suppression of mTOR signaling and reduced protein synthesis, resulting in a synthetic lethal phenotype that can be exploited by inhibition of Bcl-xL (Karpel-Massler et al., 2017) (Figure 1). This finding is consistent with a previous report that demonstrated D2HG inhibits ATP synthase activity and IDH1^R132H-expressing cells exhibit decreased ATP content and reduced mTOR signaling (Fu et al., 2015). However, a recent study has provided evidence that IDH1^R132H expression may lead to mTOR pathway activation via D2HG-mediated inhibition of the lysine demethylase KDM4A, a putative negative regulator of mTOR signaling (Carbonneau et al., 2016). Given that genetic mutations involved in PI3K-AKT-mTOR signaling occur frequently in oligodendrolgliomas with IDH mutations, it will be important for future studies to investigate the extent to which D2HG-mediated regulation of mTOR signaling is relevant to the therapeutic management of malignant glioma (Brat et al., 2015).

Role of IDH mutations and 2-HG enantiomers in tumor immunology and immunotherapy

A number of recent studies have reported that IDH1/2 active site mutations and L- and D-2HG enantiomers may play critical roles in shaping the immunological landscape of tumor microenvironments (Amankulor et al., 2017; Tyrakis et al., 2016). It has been reported that IDH mutant gliomas are associated with decreased tumor-infiltrating lymphocytes (TILs) and reduced programmed death-ligand 1 (PD-L1) expression compared with IDH wildtype gliomas (Berghoff et al., 2017). Similarly, IDH mutant glioma cells are reported to suppress expression of the ligands ULBP1 and ULBP3 that activate the natural killer (NK) cell receptor NKG2D via epigenetic silencing, and that this may provide a mechanism for evading NK cell surveillance in gliomas (Zhang et al., 2016). Additionally, using a combination of TCGA data analyses, cell-based models, and syngeneic intracranial murine gliomas, Kohanbash et al. reported that D2HG produced by IDH mutations suppresses the expression of cytotoxic T lymphocyte–associated genes and T-cell chemokines and inhibits the accumulation of T cells in intracranial murine gliomas. Collectively, these studies strongly suggest that IDH1/2 mutations and D2HG may contribute to immune suppression in gliomas and raise the possibility that therapeutic inhibition of mutant IDH enzymes may sensitize tumors to the actions of immunotherapeutic agents. Indeed, Kohanbash et al. reported that treatment with a mutant IDH1 inhibitor enhanced the activity of peptide vaccines specific for glioma-associated antigens expressed by GL261 cells grown as intracranial tumors in C57Bl/6 mice (Kohanbash et al., 2017).

It has also been reported that the neo-epitope generated by IDH1/2 active site mutations may provide an opportunity for peptide-based vaccination strategies using IDH1 or IDH2 mutation-specific peptides (Pellegatta et al., 2015; Schumacher et al., 2014) (Figure 1). Platten et al. used mice devoid of mouse MHC with transgenic expression of human MHC class I and class II, and reported that peptide vaccination of mice bearing IDH1^R132H subcutaneous sarcomas elicited a CD4⁺ specific anti-tumor response (Schumacher et al., 2014). Specifically, Pellegatta et al. used a syngeneic GL261 cells expressing the IDH1^R132H mutation and vaccination with peptides encompassing the IDH1 mutation site and found that vaccination induces anti-tumor immune responses that are correlated with increased survival of mice bearing intracranial tumors, along with apparent curative responses in 25% of treated mice (Pellegatta et al., 2015).

Collectively, the results described above suggest that IDH1 and IDH2 mutations may cause immunosuppression by altering expression of immunogenic cytokines and receptors. Alternatively, mutant IDH-specific neoantigens appear to be capable of mediating anti-tumor immune responses when stimulated by peptide vaccination. Future studies will need to investigate the extent to which D2HG-mediated immunosuppression is a contributing factor in the glioma microenvironment and whether or not this phenomenon is reversible by inhibition of neomorphic activity, stimulating anti-tumor immunity (e.g. peptide vaccines or anti-PD-L1 therapy), or a combination of these agents and standard of care radiochemotherapy.

Conclusion and future directions

There is now abundant evidence that depletion of D2HG in AML cells can elicit robust leukemic blast cell differentiation ex vivo and in vivo, and early clinical data suggests that treatment of patients with AML harboring IDH1/2 mutations with IDH1/2 mutant inhibitors may induce partial or complete remissions in a substantial percentage of cases. With regard to glioma therapy, the initial observation by Rohle et al. that inhibition of mutant IDH1 could slow the growth of glioma xenografts appeared to be consistent with a role for IDH1/2 mutations as being targetable driver mutations in gliomas, and it supported the concept that direct inhibition of mutant IDH neomorphic activity could be sufficient to inhibit glioma progression (Rohle et al., 2013). As described above, subsequent results have been more variable and in certain cases have shown that successful targeting of mutant IDH1, as evidenced by depletion of D2HG, does not alter intracranial glioma growth in murine models (Kopinja et al., 2017; Pusch et al., 2017; Tateishi et al., 2015).

It is currently unclear why preclinical results with mutant IDH inhibitors in glioma models have been more variable and less promising than the use of IDH inhibitors for leukemia models. One possibility is that the blood-brain-barrier penetration of mutant IDH inhibitors is not sufficient to inhibit D2HG production and induce anti-tumor effects in CNS tumors, including gliomas. As discussed above, results from mouse models bearing intracranial xenografts suggest that D2HG levels are markedly reduced after treatment with IDH-mutant inhibitors, but it is possible that the blood brain barrier is compromised in these models to an extent that is not reflective of the clinical condition. Notably, a clinical trial is underway to examine the extent to which the brain-penetrant pan-IDH mutant inhibitor AG-881 inhibits tumor D2HG concentrations when administered prior to surgery (NCT03343197). Another possibility is that the reversibility of IDH mutant-induced epigenetic changes differs by tissue type, and that cells of the central nervous system are fundamentally less susceptible to reversing epigenetic changes after D2HG levels have been normalized. Additionally, it was recently reported that although IDH1^R132H is capable of driving the transformation of immortalized human astrocytes, targeted inhibition of IDH1 neomorphic activity does not affect growth of immortalized astrocytes if administered after IDH1^R132H-mediated transformation (Johannessen et al., 2016). This observation may suggest that IDH1^R132H converts from a driver mutation to passenger at some point during disease progression.

It is also important to note that the acquisition and maintenance of mutant IDH glioma cell lines have been notoriously challenging. A limited number of heterozygous IDH^MUT/WT cell lines are available in any given laboratory and these lines may not be widely distributed between different laboratories. As such, the variable preclinical results achieved to-date using pharmacological inhibitors of IDH1-mutant neomorphic activity are based on distinct lines and xenografts, which have been derived at various institutions and often established after chemotherapeutic treatments that may render cells hypermutated (Wakimoto et al., 2014). Additionally, the establishment of IDH-mutant glioma cell lines may depend on the presence of copy number alterations or the acquisition of tertiary oncogenic driver mutations (Wakimoto et al., 2014). Consequently, established IDH mutant glioma cell lines may be biased to disproportionally represent advanced stages of malignancy that are primarily driven by oncogenic driver mutations. In this context, removing a D2HG-mediated block to differentiation may not be sufficient to overcome the strong mitogenic signals downstream of oncogenic driver mutations, such as PI3K-AKT, PDGFRA, n-Myc, Met, and K-Ras (Bai et al., 2015; Wakimoto et al., 2014). Therefore, individualized combination strategies targeting tumor-specific driver alterations or combining mutant IDH inhibitors with hypomethylating agents (e.g. decitabine or 5-azacytidine) may be necessary to provide a therapeutic differentiation response (Figure 1) (Turcan et al., 2013). This concept is consistent with observations in IDH2-mutated AML demonstrating that co-occurring mutations in the Ras pathway are associated with poorer response to mutant IDH2 inhibitors (Amatangelo et al., 2017).

Several studies have also reported that copy number alterations (CNAs) of the IDH1 locus, including loss of the mutant allele, occur frequently in recurrent gliomas (Jin et al., 2013; Mazor et al., 2017; Pusch et al., 2011). In a recent study, the investigators found that cultures established from IDH mutant gliomas consistently deleted the IDH1 locus, similar to the previously observed loss of the mutant IDH1 allele in the BT142 cell line (Luchman et al., 2012, 2013; Mazor et al., 2017). Importantly, Mazor et al. reported clonal expansion of IDH1 CNAs that were associated with reduced D2HG levels and higher grade at recurrence. These observations are consistent with the concept that mutations in IDH1 or IDH2 may initiate gliomagenesis and subsequently transition to a relative passenger role at some point during disease progression. This is important because it suggests that longitudinal monitoring of lower-grade glioma patients will be important to properly interpret pre-clinical and clinical responses to mutant IDH-targeted therapy. Additionally, the acquisition of tertiary driver alterations, as discussed above, and the potential for IDH1 CNAs at recurrence suggest that maximal benefit for mutant IDH-targeted therapy may be derived from early administration of these agents, initiated soon after surgery, to patients with grade II/III gliomas harboring IDH mutations.

Early clinical results suggest that the IDH1-mutant inhibitor AG-120 (ivosidenib) is well tolerated in patients with previously treated non-contrast-enhancing gliomas (Mellinghoff et al., 2017), and several Phase I/II trials are underway to evaluate the safety of various IDH-mutant inhibitors in glioma patients (Table 1). As clinical studies continue, it will be important for pre-clinical studies to investigate alternative strategies for targeting IDH-mutant gliomas based on metabolic liabilities, chemotherapeutic sensitivities, and/or telomere maintenance mechanisms (e.g. alternative lengthening of telomeres). Furthermore, an increased emphasis on developing and testing therapeutic strategies based on mutant IDH-dependent sensitivities, (e.g. the use of PARP inhibitors) may provide alternative therapeutic approaches that may benefit patients.