Friend or foe-IDH1 mutations in glioma 10 years on.

The identification of recurrent point mutations in the isocitrate dehydrogenase 1 (IDH1) gene, albeit in only a small percentage of glioblastomas a decade ago, has transformed our understanding of glioma biology, genomics and metabolism. More than 1000 scientific papers have been published since, propelling bench-to-bedside investigations that have led to drug development and clinical trials. The rapid biomedical advancement has been driven primarily by the realization of a neomorphic activity of IDH1 mutation that produces high levels of (d)-2-hydroxyglutarate, a metabolite believed to promote glioma initiation and progression through epigenetic and metabolic reprogramming. Thus, novel inhibitors of mutant IDH1 have been developed for therapeutic targeting. However, numerous clinical and experimental findings are at odds with this simple concept. By taking into consideration a large body of findings in the literature, this article analyzes how different approaches have led to opposing conclusions and proffers a counterintuitive hypothesis that IDH1 mutation is intrinsically tumor suppressive in glioma but functionally undermined by the glutamate-rich cerebral environment, inactivation of tumor-suppressor genes and IDH1 copy-number alterations. This theory also provides an explanation for some of the most perplexing observations, including the scarcity of proper model systems and the prevalence of IDH1 mutation in glioma.

Like almost all human cancers, malignant gliomas in the central nervous system are, in essence, a genetic disease (1,2), originating from astrocyte-like neural stem cells in the subventricular zone that harbor oncogenic mutations (3). A tremendous amount of effort has been made toward uncovering the genetic alterations underlying gliomagenesis (4,5), especially in the most malignant form—glioblastoma of World Health Organization (WHO) grade IV [see refs (6,7) for glioma classification and patient demographics]. Somatic mutations target oncogenes and tumor-suppressor genes, whereas ‘driver’ mutations are positively selected for their ability to confer growth advantage via the acquisition of oncogenic activity concomitant with the inhibition of tumor-suppressive activity, the rest are ‘passenger’ mutations lacking defined roles (2).

Accordingly, numerous types of genetic alteration have been identified in glioblastoma (4,5,8). Notably, loss of heterozygosity in chromosome 10q, oncogenic amplification of EGFR (epidermal growth factor receptor) and homozygous deletion or mutation of the tumor-suppressor genes CDKN2A (cyclin-dependent kinase inhibitor 2A) and PTEN (phosphatase and tensin homolog) are found in primary (de novo) glioblastomas, whereas mutations in the tumor-suppressive TP53 are commonly detected in secondary (progressive) glioblastomas derived from diffuse astrocytomas of WHO grade II and grade III (referred to collectively as lower-grade glioma hereafter) (4,6). Overall, various genetic alterations converge onto three core pathways in glioblastoma: the oncogenic receptor tyrosine kinase/RAS/phosphatidylinositide 3-kinase pathway and the two tumor-suppressive TP53 and RB pathways (5). Although the frequencies of genetic alterations among the core pathways vary between 78 and 88%, recurrent changes in a single gene occur at much lower frequencies (Table 1).

Table 1.

Recurrent genetic alterations in glioblastomas

Genetic alteration	Frequency (5)
CDKN2A homozygous deletion or mutation	52%
CDKN2B homozygous deletion	47%
EGFR mutation or amplification	45%
PTEN mutation or homozygous deletion	36%
TP53 mutation or homozygous deletion	35%

IDH1 mutations

When recurrent heterozygous mutations in the cytosolic isocitrate dehydrogenase 1 (IDH1) gene were first reported in glioblastoma, the mutation frequency was only 12% (9); however, it was remarkable that (i) all IDH1 mutations targeted the same codon Arg132, which is evolutionarily conserved and functionally required for the interaction with the substrate isocitrate; (ii) IDH1 mutations were found preferentially in younger patients, nearly all with secondary glioblastomas; and (iii) patients with IDH1 mutation had median overall survival three times longer than those without. Subsequent studies not only confirmed the high frequency of IDH1 mutations in secondary glioblastomas (>85%) but also revealed the prevalence in lower-grade gliomas (>68%) (10,11) (Table 2). Among lower-grade glioma with IDH1 mutations, the substitution of Arg132 with histidine—IDH1R132H—occurred 92% of the time, whereas IDH1, IDH1, IDH1 and IDH1 were at much lower frequencies (12). Lower-grade gliomas without IDH1 mutations often acquired mutations in the mitochondrial IDH2 gene at Arg172 (11), which is functionally analogous to IDH1 Arg132 (13). Overall, 95% of these mutations were identified in IDH1 and only 5% in IDH2 (11,14) (Table 2). Thus, IDH1 mutations will be discussed hereafter with IDH1 as the archetype.

Table 2.

IDH1 and IDH2 point mutations in human cancers

Cancer type	Frequency	Most frequent	IDH1:IDH2	Outcome
Primary glioblastomas	<5%	IDH1 R132H	100:0	Beneficial
Secondary glioblastomas	>85%	IDH1 R132H	100:0	Beneficial
Lower-grade gliomas	>68%	IDH1 R132H	95:5	Beneficial
Cartilaginous tumors	>56%	IDH1 R132C	92:8	Insignificant
Angioimmunoblastic T-cell lymphoma	>20%	IDH2 R172K	0:100	Insignificant
Acute myeloid leukemias	<15%	IDH1 R132C and IDH2R140Q	53:47	Adverse or insignificant
Intrahepatic cholangiocarcinomas	<12%	IDH1 R132C	89:11	Beneficial
Other cancer types	<3%	ND	ND	ND

It is noteworthy that IDH1 is rare in primary glioblastomas (<5%) and non-existent in other types of brain tumors (10,11). Furthermore, IDH1 and IDH2 mutations are rare or non-existent in cancer types outside of the central nervous system (11,15,16), with the exception of cartilaginous tumors, leukemia and lymphoma, and intrahepatic cholangiocarcinomas (17,18) (Table 2). Importantly, IDH1 confers a distinctive survival advantage in glioma patients; large cohort studies confirmed a 2-fold increase of median overall survival in glioblastoma patients and a more than threefold increase in lower-grade glioma patients compared with their respective controls (11,19). Furthermore, multivariate analysis confirmed that IDH1 was an independent favorable prognostic marker in gliomas (20). Moreover, maximal resection or combined temozolomide with radiotherapy confers survival benefit specifically on patients with IDH1 astrocytomas without 1p/19q codeletion but not those with 1p/19q codeletion or IDH1 wild-type gliomas (21).

IDH1 mutation is neomorphic

The vast majority of somatic mutations in cancer genes are dominantly acting, i.e., a single-allele mutation is sufficient to be oncogenic, whereas inactivation of tumor-suppressor genes often requires mutation of both alleles (2). The IDH1 enzyme catalyzes the conversion of isocitrate and NADP+ to 2-oxoglutarate and NADPH (Figure 1a). Given the heterozygous nature of IDH1, a feed-forward mechanism was initially speculated by which the mutation abrogates the negative feedback to increase NADPH production (22) because NADPH is a reducing equivalent critical for biosynthesis and redox homeostasis (23). Consistent with its role in glioma growth, IDH1 was found to be overexpressed in glioblastoma and essential to tumor growth (24). However, the fact that IDH1 is associated with decreased NADPH production (11,25–29) apparently argues against this notion.

Figure 1.

Heterozygous IDH1 induces epigenetic and metabolic reprogramming by inhibiting 2-oxoglutarate-dependent dioxygenases in glioma. (a) A diagram depicts the IDH1R132H neomorphic activity in the cytosol that catalyzes D-2HG production through the hydroxylation of 2-oxoglutarate from IDH1-mediated oxidation of isocitrate. Consequently, D-2HG acts as an antagonist of 2-oxoglutarate primarily to inhibit histone demethylase KDM4C and 5-methylcytosine hydroxylase TET, resulting in histone methylation (H3K9me3) and CpG methylation (5mCpG). High levels of D-2HG induce HIF-1 signaling by inhibiting HIF prolyl 4-hydroxylase EGLN and subsequently induce L-2HG production. (b) Accumulation of succinate and fumarate in the mitochondria, resulting from mutations in the genes encoding succinate dehydrogenase (SDH) and fumarate hydratase (FH), respectively, inhibits HIF prolyl 4-hydroxylase EGLN, thereby stimulating HIF-1 signaling for tumorigenesis. Enhanced events are highlighted in solid color, whereas inhibited events are shaded gray.

Gene mutations in the mitochondrial succinate dehydrogenases and fumarate hydratase of the citric acid cycle have been linked to paraganglioma and leiomyosarcoma, respectively (30). These loss-of-function mutations result in respective accumulation of succinate and fumarate, which inhibits the 2-oxoglutarate-dependent HIF (hypoxia-inducible factor) prolyl 4-hydroxylases (EGLN) that negatively regulate HIF-1—a transcription factor implicated in tumor angiogenesis and glycolysis (31) (Figure 1b). Thus, the notion that IDH1 is a loss-of-function mutation was proposed based on the dominant IDH1R132H inhibition of wild-type IDH1 catalytic activity through heterodimerization, thereby reducing the production of 2-oxoglutarate and in turn enhancing HIF-1 activity (32). Subsequent studies confirmed the heterodimeric interaction but questioned the dominant-negative effect of IDH1R132H and the stimulation of HIF-1 signaling in gliomas (33,34).

Lastly, a landmark discovery established that IDH1 is, in fact, neomorphic, resulting in a dominant gain-of-function that catalyzes the NADPH-dependent reduction of 2-oxoglutarate to (D)-2-hydroxyglutarate [D-2HG or R-2HG] but not the enantiomer L-2HG or S-2HG (35) (Figure 1a). Furthermore, IDH1 gliomas generally had up to 100-fold higher levels of D-2HG than those without the mutation. Similarly, high D-2HG levels were also detected in the tumor tissues and sera of leukemia patients with IDH1 and IDH2 mutations (13). Therefore, IDH1 acquires a neomorphic enzymatic activity to produce D-2HG.

D-2HG induces epigenetic reprogramming

D-2HG is a competitive inhibitor of multiple 2-oxoglutarate-dependent dioxygenases, including EGLN, histone demethylases and the TET (ten-eleven translocation) family of 5-methylcytosine hydroxylases (36) (Figure 1a). In particular, histone demethylases are 200-fold more sensitive to D-2HG than EGLN, suggesting the involvement of IDH1 in chromatin remodeling (37). Indeed, exogenous IDH1 induced impairment of histone demethylation apparently preceding DNA hypermethylation (38). The greater accumulation of a repressive mark—trimethylation of lysine 9 in histone 3 (H3K9me3)—was particularly noticeable in preceding the changes in DNA methylation, resulting from specific inhibition of the demethylase KDM4C (Figure 1a). In endogenous IDH1-heterozygous cells, global and gene-specific H3K9me3 correlated with locus-specific DNA hypermethylation of downregulated genes (39). Furthermore, AGI-5198—a potent IDH1R132H inhibitor—induced demethylation of histone H3K9me3 and expression of genes associated with differentiation in endogenous IDH1 glioma cells (40). Similarly, AGI-6780—a potent inhibitor of IDH2—induced differentiation of leukemia cells (41). Therefore, D-2HG induces histone methylation and blocks cell differentiation.

Likewise, exogenous IDH1 was sufficient to induce DNA hypermethylation in immortalized primary human astrocytes in a pattern resembling the glioma-CpG island methylator phenotype (G-CIMP) (42) (Figure 1a)—an associated feature of gliomas harboring IDH1 mutations (43). Genetically engineered heterozygous IDH1 in human colon cancer HCT116 cells and SV40-immortalized human astroglial cells also induced genome-wide alterations in DNA methylation including hypermethylation and, to a lesser extent, hypomethylation of CpG loci (39,44). The TET 5-methylcytosine hydroxylases catalyze the oxidation of 5-methylcytosine into 5-hydroxymethylcytosine (5hmC), thereby maintaining gene promoters in an unmethylated state for gene activation (45). Although exogenous IDH1 was shown to inhibit the 5-methylcytosine hydroxylase activity and reduce 5hmC conversion (36,42), the relationship between IDH1 and 5hmC in glioma remains controversial (36,46,47). In fact, 5hmC depletion was found in many types of human cancer (46,48) and was associated with poor survival of glioma patients (47,49). Hence, IDH1 induces G-CIMP-like epigenetic reprogramming, but its involvement in 5hmC depletion remains less clear in glioma.

Is IDH1 mutation oncogenic?

IDH1 is believed to drive gliomagenesis through the ‘oncometabolite’ D-2HG to induce epigenetic, metabolic and transcriptional alterations (17,18,22,50). There is abundant, albeit circumstantial, evidence to support this concept. First, IDH1 is recurrent and neomorphic, supporting a gain of function in a proto-oncogene. Second, IDH1 is believed to be an early genetic event that confers a growth advantage for glioma initiation (19,51). Third, IDH1 is genetically preserved in recurrent gliomas (52–54). Fourth, exogenous IDH1 is sufficient to induce G-CIMP, block cell differentiation, and initiate oncogenic transformation (34,38,42). Finally, potent IDH1 inhibitors were shown to promote differentiation, inhibit tumor growth and extend animal survival (40,55). The fact that these inhibitors are in clinical trials epitomizes the rapid biomedical advancements from concept to bedside since the seminal discovery of IDH1 mutations.

Cautionary observations

Although the ongoing clinical trials hold the promise of eventual reduction of morbidity and mortality, the optimism of success has yet to be reconciled with indisputable clinical and experimental observations that seem to suggest caution. First, despite its prevalence in glioma, heterozygous IDH1 is scarcely preserved in cell culture or patient-derived xenografts (56–59). As such, exogenous IDH1 has almost become the norm of investigation (34,38,40,42,55,60); however, none of the resultant findings have been corroborated in endogenous IDH1-heterozygous models (26,29,44,61,62) (Table 3). Second, the notion that IDH1 is oncogenic is based on the comparison of exogenous IDH1 with wild-type IDH1, which seems at odds with the observation that patients with IDH1 glioma have better survival than those of IDH1-wild type (9,11,19,20). Third, in agreement with the lack of tumor development in heterozygous Idh1 mice despite high levels of D-2HG (26,61–63), patients with D-2-hydroxyglutaric aciduria—a rare autosomal recessive disorder characterized by high D-2HG levels in urine, plasma, and cerebrospinal fluid—have no predisposition to tumor development. These patients may have IDH2 and IDH2 in the germline and are either asymptomatic or manifest encephalopathy, muscular dystrophy and cardiomyopathy (64,65). Consistently, IDH2 and IDH2 transgenic mouse models recapitulated cardiomyopathy and neurodegeneration with no tumor development (66). Of note, L-2-hydroxyglutaric aciduria—another rare autosomal recessive encephalopathy—apparently confers predisposition to brain malignant tumors (67). Finally, studies have indicated that D-2HG is non-essential to glioma progression (29,68,69). Therefore, these observations not only challenge the concept that IDH1 is gliomagenic but also call into question the potential efficacy of mutant IDH inhibitors in glioma treatment.

Table 3.

IDH1 —different approaches, different conclusions

	Tumor suppressive	Oncogenic
IDH1R132H	Endogenous, heterozygous (26,29,44,61,62)	Exogenous (34,42,60,61,70,71)
IDH1 R132H expression	Detected in anchorage-dependent culture	Detected in anchorage-dependent culture (34,42)
	Suppressed in anchorage-independent culture (29) (P.D.B.Tiburcio et al., unpublished data)
Nestin expression	Upregulated in anchorage-independent culture of IDH1R132H-hemizygous but not IDH1R132H-heterozygous cells (P.D.B.Tiburcio et al., unpublished data)	Upregulated in anchorage-dependent culture transduced with IDH1R132H compared with wild-type IDH1 (38,42)
Anchorage-independent growth	Inhibited by heterozygous IDH1R132H	Stimulated by exogenous IDH1R132H in comparison with wild-type IDH1 (34,42,70,71)
	Stimulated by loss of IDH1R132H heterozygosity, glutamate, and reducing equivalent (29,72) (P.D.B.Tiburcio et al., unpublished data)
IDH1 R132H effect on tumorigenicity	Non-tumorigenic by itself (26,61,62)	Inducing tumor-like lesions in orthotopic transplantations of astrocytes transduced with IDH1R132H compared with wild-type IDH1 (60,71)
	Decreasing glioma incidence and extending survival in Trp53-deficient background (62,73,74)	Shortening survival in a glioma model of Atrx−/−; Cdkn2a−/−; Pten−/− background in comparison with wild-type IDH1 (75)
	Obliterating gliomagenesis in Trp53-intact background when selection against IDH1R132H transgene was precluded (72)

IDH1 mutation—guilty by association

The assessment of IDH1 in relation to TP53 mutation or loss of 1p/19q by comparing the initial and recurrent biopsies from the same individuals suggested that IDH1 precedes the other known genetic events and is therefore an early event in the development of lower-grade gliomas (51). Integrated genomic analysis further indicated that the expression of IDH1 protein preceded the production of mutant p53 protein and copy-number alterations of PTEN and EGFR in secondary glioblastoma (76). Further longitudinal analysis showed that IDH1 was the only event shared in both initial and recurrent gliomas (52,54), although multiregional and temporal samplings revealed additional shared events, including TERT promoter mutation and loss of 1p/19q in one case and TP53 mutation in another (53). Thus, IDH1 is implicated in glioma initiation and progression, thanks to its tight association.

However, this gliomagenic notion is at odds with the lack of glioma development in both heterozygous Idh1 mice (26,61,62) and D-2-hydroxyglutaric aciduria patients (64). Although genetic events including inactivation of TP53 and/or activation of oncogenic signaling, e.g., platelet-derived growth factor (PDGF), are sufficient to induce glioma, IDH1 has been shown to inhibit glioma penetrance and extend animal survival in comparison with wild-type IDH1 (62,72–74). Furthermore, the importance of IDH1 heterozygosity cannot be overemphasized because IDH1 requires a wild-type allele to produce D-2HG, whereas loss of IDH1 heterozygosity essentially eliminates D-2HG production (29,77,78). Likewise, genetic alterations at the IDH1 locus, including IDH1 amplification, engendered allelic imbalance between IDH1 and IDH1, thereby diminishing D-2HG production in recurrent gliomas (69). Therefore, loss of IDH1 heterozygosity and IDH1 allelic imbalance result in non-functional IDH1 in association with glioma recurrence and progression (Figure 2).

Figure 2.

Mechanisms for undermining tumor-suppressive activity of heterozygous IDH1 during glioma progression. (a) In the absence of extracellular glutamate, heterozygous IDH1R132H, together with intact tumor-suppressor genes (e.g. TP53 and RB), obliterates oncogenic promotion of gliomagenesis. (b) Glutamate in the cerebral cortex negates IDH1 suppression of gliomagenesis, thereby driving anchorage-independent growth and glioma progression, which is further exacerbated by the inactivation of tumor-suppressor genes and the selection against IDH1 heterozygosity. The inhibited events are shaded gray.

Does IDH1 mutation-induced epigenetic reprogramming drive gliomagenesis?

DNA hypermethylation has been shown to compromise binding of the methylation-sensitive insulator protein CTCF (CCCTC-binding factor) in IDH-mutant gliomas, resulting in loss of insulation of a neighboring enhancer and aberrant upregulation of the receptor tyrosine kinase gene PDGFRA (79). A modest increase of PDGFRA transcripts was also observed in IDH-mutant glioma when compared with IDH-wild-type glioma (80). Furthermore, 5hmC enrichment in gene body regions correlated significantly with the upregulation of genes, such as LGR5 (leucine-rich repeat-containing G-protein-coupled receptor 5), in IDH-mutant glioma (81). It remains unclear, however, whether these epigenetic targets are essential to IDH-mutant gliomagenesis.

In contrast, IDH-mutant gliomas with high levels of DNA methylation had more favorable clinical outcomes than those with low levels (82). A high- to low-level shift in DNA methylation has been identified during glioma recurrence, giving rise to an IDH-wild-type-like glioblastoma phenotype (83). Thus, the loss of DNA methylation is associated with glioma progression. Likewise, silencing of epigenetic targets, such as LDHA (lactate dehydrogenase A), RBP1 (retinol-binding protein), MIR148A (microRNA 148a) and F3 (coagulation factor III, tissue factor), was associated with reduced malignant growth and favorable outcome in IDH-mutant gliomas (59,84–86). Thus, there is insufficient evidence that IDH1-induced epigenetic reprogramming drives gliomagenesis.

Does IDH1 mutation increase cell proliferation?

Exogenous IDH1 was first shown to stimulate proliferation of late-passaged human astrocytes with inactivated TP53 and RB signaling (87) in reference to wild-type IDH1 (34). The increased proliferation involved D-2HG stimulation, rather than inhibition (Figure 1a), of EGLN activity and in turn downregulation of HIF signaling as part of cell transformation (34). However, the effect of IDH1 on hypoxic signaling seems debatable; whereas reduced HIF signaling was observed in IDH-mutant gliomas (59,88,89), increased HIF-1α abundance and target gene expression were also reported in IDH-mutant gliomas and mouse brain-specific Idh1 knock-in cells (26,32).

With respect to cell proliferation, a large body of evidence indicates that IDH1 inhibits glial cell proliferation (72,90–93). Mechanistically, D-2HG inhibits ATP synthase, resulting in decreased mTOR (mammalian target of rapamycin) signaling and cell growth (94). Likewise, D-2HG-mediated ATP depletion activates AMPK (5′ AMP-activated protein kinase), thereby inhibiting protein synthesis and mTOR signaling in glioma cells (95). Furthermore, D-2HG promotes cell-cycle arrest by inhibiting the FTO (fat mass and obesity-associated) demethylase activity, thereby increasing N6-methyladenosine modification of MYC/CEBPA (CCAAT/enhancer binding protein alpha) transcripts for destabilization and, in turn, decreasing proliferative signaling (93). Moreover, engineered heterozygous IDH1 significantly inhibited glial cell proliferation by targeting YAP (Yes-associated protein) and Notch pathways (44). Collectively, these studies provide strong evidence that IDH1 targets various signaling pathways to inhibit glial cell proliferation at least in vitro.

Antagonism between heterozygous IDH1 mutation and anchorage-independent growth

The cellular ability to proliferate in anchorage-independent growth is consistently associated with tumorigenicity (96). Accordingly, exogenous IDH1 has been shown to increase anchorage-independent growth of immortalized human astrocytes compared with wild-type control (34,42,60,70,71) (Table 3). Likewise, exogenous IDH1 resulted in spheroid growth of MYC-immortalized human neural progenitor cells (61). Additional evidence includes the upregulation of neural stem-cell marker genes including NES (nestin) in IDH1-transduced, late-passaged astrocytes (38,42), albeit under anchorage-dependent conditions. Apparently, these findings support the notion that IDH1 is gliomagenic; it is unclear, however, whether the exogenous IDH1 remained expressed in the transduced cells during anchorage-independent growth (see below).

It is striking, however, that none of the abovementioned findings are reproduced in IDH1-heterozygous cells; in fact, heterozygous Idh1 apparently prevented spheroid growth of neural progenitor cells (61). In keeping with this, heterozygous IDH1 suppressed anchorage-independent growth of glioma cells whereas hemizygous IDH1 (deficient in wild-type IDH1 allele) lost the suppression (29) (Table 3). Conversely, anchorage-independent but not anchorage-dependent conditions selected against IDH1 heterozygosity via a non-genetic mechanism (29) (Figure 2). This antagonism between IDH1 heterozygosity and anchorage-independent growth not only underscores the functional importance of IDH1 heterozygosity in glioma biology but also provides an explanation for the frequent loss of either wild-type or mutant IDH1 allele in patient-derived xenograft, ex vivo spheroid culture, and glioma recurrence (57,59,69,77,97). Furthermore, NES was strikingly upregulated in anchorage-independent culture of IDH1-hemizygous cells compared with IDH1-heterozygous cells (P.D.B.Tiburcio et al., unpublished data). Consistently, NES expression was significantly higher in IDH-wild-type gliomas than in IDH-mutant gliomas (P.D.B.Tiburcio et al., unpublished data), in agreement with the observation that nestin is an adverse predicator of lower-grade gliomas (98). Therefore, heterozygous IDH1 is tumor suppressive by inhibiting NES expression and anchorage-independent growth. Conversely, anchorage-independent growth selects against IDH1 heterozygosity, rendering IDH1 non-functional.

Inactivation of the TP53 and RB pathways renders IDH1 mutation non-functional

The antagonism between IDH1 heterozygosity and anchorage-independent growth was also observed in mouse transplantation models, especially in a genetic background without engineered inactivation of tumor-suppressor gene(s), giving rise to tumor growth but no expression of exogenous IDH1 (72). Exogenous IDH1 expression was noticeably enhanced when the TP53 and/or RB pathways became deficient (71–73). It is noteworthy that homozygous Cdkn2a deletion (affecting both TP53 and RB pathways) resulted in strong IDH1 staining in gliomas but complete loss of IDH1-associated survival advantage (73), which was also observed with the homozygous deletion of Cdkn2a and/or Pten in glioma (75). Given the functional requirement of IDH1 mutation for survival advantage, inactivation of the tumor-suppressor genes renders IDH1 permissible but non-functional (Figure 2b). As such, the inactivation of both TP53 and RB pathways, e.g., in the immortalized human astrocytes (87), might have obscured IDH1 tumor-suppressive activity (34,42,60,70,71). Moreover, IDH1-associated DNA repair defects and genetic instability (99–101) might have also been relevant, especially in those ‘late-passaged’ cells (34,42,60).

Extracellular glutamate negates the tumor-suppressive effect of IDH1 mutation

IDH1 malignant glioma arises predominantly from the frontal lobe (76), where the hominoid-specific GLUD2 (glutamate dehydrogenase 2) gene is evolutionarily selected for the degradation of high flux of glutamate neurotransmitter (102,103). Extracellular metabolites have been shown to negate IDH1 suppression of anchorage-independent growth (29,72) (P.D.B.Tiburcio et al., unpublished data) (Figure 2b). IDH1 cells apparently maintain redox homeostasis by increasing glutathione synthesis through the trans-sulfuration pathway and the reliance on glutaminolysis (28,104). In comparison with IDH1-heterozygous cells, IDH1-hemizygous cells exhibited higher NADPH levels and glutathione/oxidized glutathione ratio (29). It has been proposed that IDH-mutant gliomas depend on glutamate and lactate to alleviate metabolic stress (105,106). Consistently, the addition of glutamate reversed IDH1 inhibition of PDGF-driven proliferation of neural progenitor cells (92). Likewise, the addition of N-acetyl cysteine or glutamate markedly increased anchorage-independent growth of IDH1-heterozygous glioma cells (29,72) (P.D.B.Tiburcio et al., unpublished data). The overpowering effect of glutamate was corroborated by the differential effects of IDH1 on tumorigenesis between the glutamate-rich cerebral cortex and the subcutaneous tissue in Trp53+/+ mice; exogenous IDH1 was virtually ineffective in inhibiting PDGF-driven gliomagenesis in the brain but potently effective in blocking subcutaneous tumor growth (72). Therefore, glutamate may be particularly pertinent to IDH1 prevalence in glioma but not in most other cancer types (10,11,15,16). Together, these studies indicate the importance of extracellular glutamate for IDH1 glioma growth.

IDH1 mutation does not initiate gliomagenesis by itself

As discussed above, no glioma development has been seen in heterozygous Idh1 mice (26,61,62) or in patients with D-2-hydroxyglutaric aciduria (64). Of note, these mice exhibited disorganization of the subventricular zone and epigenetic and transcriptional profiles resembling those of IDH-mutant glioma (61,62). In combination with other genetic events, heterozygous Idh1 reduced the penetrance of Trp53 deletion-induced glioma by 70% and extended survival compared with wild-type Idh1 (62). Likewise, glioma models driven by PDGF in combination with Trp53 knockdown revealed survival benefits of IDH1 in RCAS/tva mice (73). Finally, IDH1 obliterated PDGF-induced gliomagenesis in RCAS/tva mice when selection against exogenous IDH1 was precluded, demonstrating IDH1 innate ability for tumor suppression (72) (Figure 2a). Thus, IDH1 neither initiates nor promotes gliomagenesis.

Studies relying on exogenous IDH1 expression in TP53/RB-deficient human astrocytes have come to a divergent view (Table 3). Exogenous IDH1 was shown to overcome telomere-induced crisis by activating TERT expression, thereby leading to cell transformation and tumor growth (71). Likewise, long-term IDH1 expression impaired contact inhibition and yielded intracranial lesions with parenchymal infiltration (60). Moreover, IDH1 in comparison with IDH1 shortened mouse survival in a combined Atrx, Cdkn2a and Pten background of PDGF-driven glioma models (75). The biological relevance of these studies, however, is less clear because mutations in CDKN2A and PTEN are rare in IDH-mutant gliomas (19). Furthermore, IDH1-induced hydrocephalic neuropathy (26,61,62,66), hypoxic induction of oncogenic L-2HG (107,108) and IDH1-associated genetic instability (99–101) might have complicated the interpretations of the studies.

Is IDH1 mutation a therapeutic target in glioma?

The therapeutic potential of IDH1 and IDH2 mutations has been explored extensively; different classes of mutant IDH1 and IDH2 inhibitors have been developed for ongoing phase 1 or phase 1/2 clinical trials of mostly hematological diseases and some solid tumors (109,110). The outcome of these inhibitors in glioma-associated survival studies in mouse models, however, seems less promising; the efficacies vary from modest or indifferent to detrimental despite their high potencies in inhibiting D-2HG production and inducing cell differentiation. AGI-5198 was first shown to decrease subcutaneous growth of IDH1 glioma cells (40), but tumor regression was not reproduced in another study despite near-complete inhibition of 2HG (111). Likewise, an IDH1-specific inhibitor (IDHi) (112) had near-complete 2HG elimination but showed no survival benefits in orthotopic transplantation of IDH1 glioblastoma cells (113). In fact, long-term exposure to IDHi prior to orthotopic transplantation significantly shortened survival (113). The brain-penetrant IDH1 inhibitors MRK-A and MRK-B prolonged survival in orthotopic transplantation of one IDH1 glioma cell type but not another (114). Finally, BAY 1436032—an orally available, pan-IDH1R132X inhibitor—significantly but modestly extended survival in intracranial IDH1-astrocytoma transplantation (55). Collectively, mutant IDH1 inhibitors may have therapeutic effects on glioma independent of D-2HG depletion. The first-in-human phase I study (NCT03030066) involving 45 patients with recurrent/progressive IDH1 mutant glioma has shown favorable tolerance and brain distribution with the pan-IDH1R132X inhibitor DS-1001b (115). Of the 38 evaluable patients in this ongoing trial, 17 (45%) achieved stable disease, 3 (8%) partial response, 2 (5%) minor response, and 1 (3%) complete response.

Of note, mutant IDH1 inhibitors have undesired effects that diminish the efficacy of chemotherapy and radiation therapy. AGI-5198 was shown to reduce radiosensitivity through restoration of NADPH levels, thereby protecting IDH1-mutant cells against ionizing radiation (27). It also desensitized cisplatin killing of IDH1-mutant cells by reducing levels of reactive oxygen species (116) and completely reversed the sensitivity of cancer cells to poly(adenosine 5′-diphosphate-ribose) polymerase inhibitors (101). In light of the association of non-functional IDH1 with glioma progression and recurrence, as discussed above, additional biochemical, metabolic and immunological targets should be explored [reviewed in ref. (110)] for the development of better treatment. Likewise, investigating the mechanisms that negate IDH1 suppression of gliomagenesis (Figure 2) may provide novel targets.

Concluding remarks

IDH1 heterozygosity and its association with better survival are hallmarks of gliomas compared with those without the mutation. Whether IDH1 is a ‘friend or foe’ depends principally on the model system employed. Studies with heterozygous IDH1 model systems have demonstrated that IDH1 is neither a driver nor a passenger but beneficial—another category of somatic mutations in cancer (80). The beneficial effect of heterozygous IDH1 in comparison with IDH1-wild-type or loss of IDH1 heterozygosity is supported by both clinical and experimental findings discussed above. In contrast, the oncogenic IDH1 theory is predicated exclusively on the findings of exogenous IDH1 expression where the selection against IDH1 heterozygosity by anchorage-independent growth has essentially been overlooked. Despite being intrinsically tumor suppressive, IDH1 activity can be undermined by various intracellular and extracellular factors, notably inactivation of tumor-suppressor genes, IDH1 allelic imbalance and selection against IDH1 heterozygosity, and extracellular glutamate and reducing equivalents. The functional, rather than genetic, loss of IDH1 provides an explanation for the prevalence and preservation of IDH1 from glioma initiation to progression. It is imperative, therefore, that investigations be directed toward understanding the mechanism by which IDH1 functionality is negated for better prevention and therapeutic intervention of glioma progression.

Funding

Department of Neurosurgery at University of Utah, Huntsman Cancer Foundation, and grant P30 CA042014 awarded to Huntsman Cancer Institute and to the NC Program at Huntsman Cancer Institute.