Rapid Conversion of Mutant IDH1 from Driver to Passenger in a Model of Human Gliomagenesis

Mutations in isocitrate dehydrogenase 1 and 2 (IDH1 and IDH2) have been noted in several types of cancers, including acute myeloid leukemia (1, 2), nonepithelial melanoma (3), and glioma (4, 5). IDH1 mutations, however, are among the most common alterations in glioma, and nearly 70% of lower grade gliomas contain heterozygous R132H mutations at the enzyme's active site (6). R132H IDH1 mutation in glioma results in the creation of a mutant IDH1 protein, which dimerizes with the wild-type (WT) IDH1 (7, 8). Detailed biochemical studies have shown that while cytoplasmic WT IDH1 readily converts isocitrate to α-ketoglutarate (α-KG; ref. 9), the WT/mutant dimer has a relatively higher affinity for α-KG and displays a neomorphic activity that converts α-KG to 2-hydroxyglutarate (2-HG; ref. 9). 2-HG accumulates in IDH1-mutant cells and competes with α-KG for binding to a wide variety of α-KG–dependent enzymes (10). The end result of 2-HG accumulation is the inhibition of multiple enzymes involved in the control of cytosine DNA methylation and histone methylation (11–14), wide-spread changes in gene expression (15), and a strong association with cellular transformation (16, 17). In low-grade glioma, an analysis of clonality also suggests that IDH1 mutation is among the earliest events and occurs before TP53 mutation in astrocytomas or loss of 1p/19q in astrocytomas and oligodendrogliomas (18–21). On the basis of these findings, a series of mutant IDH1 inhibitors have been developed and are in clinical testing (22, 23).

Despite the potential promise of mutant IDH1 inhibitors, the evidence that inhibition of mutant IDH1 can suppress the growth of mutant IDH1–driven gliomas is limited. Early studies showed that the histone and DNA hypermethylation, as well as the transformation of leukemic cells induced by mutant IDH1 expression, could be readily reversed by incubation with the selective mutant IDH1 inhibitor AGI-5198 (24, 25). Subsequent studies in a single mutant IDH1–containing patient-derived glioma xenograft suggested that exposures to the same IDH1-specific inhibitor suppressed 2-HG levels, reversed at least some of the mutant IDH1–driven epigenetic alterations, and also suppressed clonogenicity (16). Subsequent studies, however, using other mutant IDH1–containing cells reached more mixed results (26, 27), and the most recent studies with a series of xenografts derived primarily from patients with recurrent, mutant IDH–containing gliomas suggest that mutant IDH1 inhibitors have modest growth-suppressing effects (28). Although most of these studies used the same mutant IDH1 inhibitor (AGI-5198), all of the studies used cells in which mutant IDH1 was expressed in cells transformed by other means, or cells in which mutant IDH1 expression appeared to play a secondary role in the transformation process. The validity of mutant IDH1 as a therapeutic target, and in particular in glioma, therefore remains an open question.

We previously described two distinct isogenic models of gliomagenesis, in which the transformation of p53/pRb–deficient astrocytes was driven exclusively by mutant IDH1 expression (29, 30). In the first system, expression of mutant IDH1 in astrocytes made p53/pRb deficient by expression of the HPV E6 and E7 proteins and immortalized by expression of exogenous hTERT led to direct cellular transformation (29). In the second, the expression of mutant IDH1 in telomerase-negative p53/pRb–deficient astrocytes led to the gradual emergence of a telomerase-positive transformed population of cells that grew in soft agar and in animals (30). Because mutant IDH1 is the sole initiating factor in driving the gliomagenesis process in both these systems, both models have been extensively used to study mutant IDH1 biology (15, 31), and both represent optimal systems to address questions related to the role of mutant IDH1 and mutant IDH1–driven processes in initiating and maintaining the transformation process. Using these two systems, we here show that the mutant IDH1–driven events that result in cellular transformation occur rapidly (within 3 days) following introduction of the oncogenic insult, and that while these events can be blocked by prior exposure to a selective mutant IDH1 inhibitor, the events driving cellular transformation rapidly become irreversible, making selective mutant IDH inhibitors of marginal use in suppressing the growth and clonogenicity of the mutant IDH1–driven tumors in the systems used.

Normal human astrocytes (NHA) expressing E6, E7, hTERT, and either WT or mutant IDH1 were created and cultured as described previously (32–34). In this series, only cells expressing E6, E7, hTERT, and mutant IDH1 were transformed as defined by their ability to grow in soft agar and to form tumors in animals. The same NHAs expressing E6, E7, and either WT or mutant IDH1 then assayed before they reached the end of their normal lifespan (precrisis) or after they emerged from crisis as transformed cells (postcrisis) were also created and cultured as described previously (30). In this series, only the postcrisis cells expressing E6, E7, and mutant IDH1 were transformed as defined by their ability to grow in soft agar and to form tumors in animals. Transient expression of mutant (R132H) IDH1 in NHAs expressing E6, E7, and hTERT was performed as described previously (29). Briefly, multiple wells containing 1 × 10⁵ cells were infected in 6-well format by lentivirus-encoding GFP and either mutant IDH1 or a blank construct in the presence of polybrene (Sigma). Successful lentiviral transduction was assayed by fluorescence microscopy and Western blotting using a mutant IDH1–specific antibody at the indicated time points. The selective mutant IDH1 inhibitor AGI-5198 (Calbiochem) was dissolved in DMSO to obtain a 20 mmol/L stock solution.

Cell proliferation of control and AGI-5198–treated cells was determined every 7 days for up to 28 days by Trypan blue exclusion counting as described previously (32). Soft agar assays were performed in 6-well plates, where the bottom layer contained complete DMEM media (10% FBS and 1% penicillin/streptomycin) in 0.7% agar, the middle layer consisted of complete DMEM media with 5,000 cells in 0.35% agar, and the top layer was made of complete DMEM media in 0.7% agar. Cells were pretreated with vehicle or AGI-5198 for 72 hours before being embedded in soft agar, and all soft agar layers contained either vehicle or AGI-5198. Colonies were fixed and stained with 0.005% crystal violet after 28 days and the colonies counted.

Whole-cell protein lysates were prepared using RIPA Lysis and Extraction Buffer (Thermo Scientific) supplemented with PhosSTOP phosphatase and cOmplete, Mini, protease inhibitor cocktail (Roche Diagnostics). Nuclear proteins were extracted using Subcellular Protein Fractionation Kit (Thermo Scientific) according to the manufacturer's instructions. Protein lysates were quantified with DC Protein Assay (Bio-Rad Laboratories). Equal amounts of whole-cell (30 μg) or nuclear protein extracts (20 μg) were electrophoresed and transferred onto a PVDF membrane (Bio-Rad Laboratories) under standard conditions. The following primary antibodies were used: IDH1 R132H (DIA-H09, Dianova GmbH), H3K4me3 (#39159), H3K9me2 (#39683), H3K9me3 (#39161), H3K27me3 (#39155; all from Active Motif), Histone H3 (ab1791, Abcam), and GAPDH (#14C10, Cell Signaling Technology). The corresponding secondary antibodies, anti-mouse (sc-2005) and anti-rabbit (sc-2004) IgG-horseradish peroxidase, were from Santa Cruz Biotechnology. Antibody binding was detected using Amersham ECL Western Blotting Detection Reagent (GE Healthcare Life Sciences).

Metabolites were extracted after treatment with either drug vehicle or AGI-5198 at the indicated time points from approximately 3 × 10⁷ cells in each sample by dual phase extraction as described previously (35). 1H-MRS spectra (1D water presaturation ZGPR sequence, 90° flip angle, 3s TR, 256 acquisitions) were acquired using a 500 MHz Avance Spectrometer (Bruker BioSpin) equipped with a Triple Resonance CryoProbe. Metabolites were quantified by normalizing to a trimethylsilyl propanoic acid reference of known concentration and correcting for saturation and nuclear Overhauser effect.

Data are presented as mean ± SEMs of three independent experiments. Statistical analyses were carried out using a two-tailed Student t test assuming equal variances. When multiple groups were evaluated, the one-way ANOVA test with post hoc Turkey–Kramer multiple comparisons test was used. P < 0.05 was considered statistically significant.

To begin to address the potential of mutant IDH1 as a therapeutic target in glioma, we first verified that mutant IDH1 expression resulted in the production of 2-HG oncometabolite in our gliomagenesis models and that treatment with a mutant IDH1 inhibitor resulted in suppression of 2-HG levels. While cell groups expressing only WT IDH1 exhibited only background levels of 2-HG (not shown), all cells engineered to express mutant IDH1 (IDH1 mutant pre- and postcrisis cells and hTERT IDH1-mutant cells) contained measurable levels of 2-HG (Fig. 1), ranging between 6 and 14 fmol/cell. In each case, exposure of cells to the mutant IDH1 inhibitor AGI-5198 led to >98% suppression of 2-HG levels within 3 days of exposure to 1 or 10 μmol/L of the compound. Continuous exposure of the cells to AGI-5198 resulted in suppression of 2-HG levels as long as the cells were followed (up to 28 days postinitiation of drug exposure).

To assess the contribution of mutant IDH1 to the transformation of glioma cells, and the need for continued mutant IDH1 expression, we first transiently infected E6E7hTERT-immortalized astrocytes with a lentiviral construct encoding mutant IDH1 and a fluorescent marker, then verified the effect of mutant IDH1 expression on histone modification and the growth of cells in soft agar (schematic, Fig. 2A, line 9). Greater than 90% of the target cells were infected (based on fluorescence microscopy) and, as shown in Fig. 2B, cells transfected with the construct encoding mutant IDH1 had significantly higher levels of histone modifications (H3K4me3, H3K9me2, H3K9me3, and H3K27me3) known to be driven by mutant IDH1 expression (lane 9; ref. 15) than the same cells 4 days after transiently transfection with a blank vector (lane 1). These cells also formed colonies in soft agar consistent with the ability of mutant IDH1 to drive transformation (group 9 vs. group 1, Fig. 2C). Persistent exposure of the cells to a 2-HG–suppressing concentration of AGI-5198 (1 μmol/L) beginning 3 days prior to introduction of the mutant IDH1 construct completely blocked the increased levels of mutant IDH1–driven histone modifications (measured 4 days after mutant IDH1 introduction; lane 2) as well as the ability of mutant IDH1 to drive cellular transformation and the growth of cells in soft agar. Persistent exposure of cells to AGI-5198 beginning at the time of introduction of mutant IDH1 similarly blocked the increased levels of mutant IDH1–driven histone modifications (again 4 days after introduction of mutant IDH1;lane 3) and the growth of cells in soft agar, showing that mutant IDH1 is an inhibitable driver of transformation in these cells.

To address whether mutant IDH1 expression was required to maintain the transformed state, cells infected with the mutant IDH1–encoding construct were continuously exposed to 1 μmol/L AGI-5198 beginning 4 to 21 days after introduction of the oncogenic mutant IDH1 lesion, then assayed for histone modifications and clonogenicity 3 to 4 days after initiation of drug treatment. As shown in Fig. 2B, delaying onset of mutant IDH1 inhibition for even 4 days greatly diminished the ability of the drug to suppress levels of the mutant IDH1–driven histone modifications measured (lane 4 vs. lane 2), as well as the ability of the drug to suppress growth in soft agar (Fig. 2C, group 4 vs. group 2). Increasing the time between mutant IDH1 introduction and mutant IDH1 inhibitor treatment also led to a time-dependent increase in histone modifications (Fig. 2B, lanes 5–7) and clonogenicity (Fig. 2C, groups 5–7). Furthermore, although AGI-5198 could suppress clonogenicity if applied before oncogenic insult, the removal of AGI-5198 in the presence of the mutant IDH1 oncogenic insult led within 4 days to the generation of cells that again regained alterations in histone methylation (Fig. 2B, lane 8 vs. lane 2) and the ability to grow in soft agar (Fig. 2C, group 8 vs. group 2). Collectively, these results suggest that oncogenic effects of mutant IDH1 in this system are only marginally reversible and that this window of opportunity is essentially lost within even 3 days of oncogenic insult.

To more fully determine whether mutant IDH1 represents a reasonable pharmacologic target in IDH1 mutant–driven transformed cells, we expanded our studies to include a second IDH1-transformed glioma cell line and a broader and longer range of drug exposures. As shown in Fig. 3, concentrations of AGI-5198 that were capable of suppressing levels of 2-HG by >98% did not alter the growth of nontransformed E6E7hTERT cells expressing WT IDH1 (Fig. 3A), or of E6E7hTERT cells transformed by expression of mutant H-Ras (Fig. 3B). Consistent with data in Fig. 2, these drug exposures also had no effect on the growth of E6E7hTERT transformed by expression of mutant IDH1 (Fig. 3C), even after up to 28 days of continuous drug exposure. Similar studies performed in a second independent mutant IDH1–transformed glioma cell lines yielded identical results. Specifically, concentrations of AGI-5198 that were capable of suppressing levels of 2-HG by >98% did not alter the growth of nontransformed E6E7 cells expressing WT IDH1 (Fig. 3D), or of E6E7 cells expressing mutant IDH1 prior to (precrisis; Fig. 3E) or after (postcrisis; Fig. 3F) their transformation by mutant IDH1 expression, even after up to 28 days of continuous drug exposure.

The lack of effect of AGI-5198 on the growth of these cells was mirrored by an inability of these drug exposures to suppress mutant IDH1–driven increases in histone modifications, which remained elevated in the drug-treated E6E7-mutant IDH1 pre- and postcrisis cells (Fig. 4B and C) relative to those in the control (Fig. 4B and C, first lane) or AGI-5198–treated (Fig. 4A) E6E7 WT IDH1 cells, and in the drug-treated E6E7hTERT cells expressing mutant IDH1 (Fig. 4E) relative to those in the control (Fig. 4E, first lane) or AGI-5198–treated E6E7hTERT WT IDH1 cells (Fig. 4D), even after 28 days of drug exposure.

Consistent with these findings, continuous exposure to 1 or 10 μmol/L AGI-5198 resulted in no significant change in clonogenicity in nontransformed E6E7hTERT cells expressing WT IDH1 (Fig. 5A) or in E6E7hTERT cells transformed by mutant V12H-Ras (Fig. 5B). AGI-5198 caused only a modest decrease in clonogencity in transformed E6E7hTERT cells expressing mutant IDH1 (Fig. 5C) and in E6E7 cells transformed by mutant IDH1 (postcrisis cells; Fig. 5D), but only at drug exposures (10 μmol/L) far beyond those needed for suppression of 2-HG levels. These findings as a whole show that although mutant IDH1 expression drives changes in histone methylation and cellular transformation, mutant IDH1 remains a viable therapeutic target for only a brief period of time.

Because IDH1 mutation is among the earliest and most common alterations in low-grade glioma, there is considerable enthusiasm for the possibility that the mutant IDH1 protein may represent a viable therapeutic target in the treatment of low-grade glioma. Initial results in studies using cells expressing exogenous mutant IDH1, and in cells from mutant IDH1–containing patient-derived xenografts, however, have yielded mixed results, in part because mutant IDH1 may not be a driver, or the sole driver, of the tumorigenic process in the cell models used. To more clearly address the fundamental question of whether mutant IDH1 represents a reasonable therapeutic target in glioma, we assessed the ability of a widely used and selective mutant IDH1 inhibitor to block growth, histone methylation, and clonogenicity in two cellular models in which mutant IDH1 is the sole driver of the transformed phenotype (29, 30). The results of these studies clearly show that although mutant IDH1 can drive gliomagenesis, its viability as a therapeutic target appears limited by its rapid transition from a driver to a passenger in the transformation process (Fig. 6).

The current studies confirm that mutant IDH1 is a driver of gliomagenesis and further show that mutant IDH1–driven gliomagenesis can occur more rapidly than previously appreciated. Previous studies suggested that changes in histone modification and DNA cytosine methylation driven by mutant IDH1 expression in glioma cells accumulate gradually over time (15), implying that the transformation process driven by mutant IDH1 may be similarly gradual. In the immortalized cells in the current studies, however, changes in histone modification occurred within 4 days of mutant IDH1 introduction, and these modifications remained essentially unchanged 4 weeks after introduction of the oncogenic insult. Furthermore, the mutant IDH1–transformed immortalized astrocytes exhibited only modest increases in global CpG methylation and not a frank gCIMP phenotype when analyzed several weeks after creation (data not shown). The approach used, however, allowed us to show that even 4 days of unopposed mutant IDH1 expression were sufficient to activate processes that led to cellular transformation. Although the current studies do not address the linkage between changes in histone modification, DNA cytosine methylation, and transformation, changes in histone methylation appear to be at least temporally linked to cellular transformation, while the gCIMP phenotype appears less important. Furthermore, if additional changes in histone modification occurred beyond the last time point studied, these changes appear not to be required for transformation. On the basis of the data generated in this system, the temporal window of opportunity to block mutant IDH–induced transformative changes, therefore, appears very narrow.

The current results also show that continued mutant IDH1 protein activity does not appear to be required for maintenance of the glioma phenotype. Although mutant IDH1 expression resulted in cellular transformation, and although pretreatment of cells with an IDH1 inhibitor blocked mutant IDH1–mediated transformation, effective mutant IDH1 inhibition initiated after oncogenic insult failed to reverse mutant IDH1–mediated changes in histone modification or transformation. Consistent with these results, even 1-month exposure to a mutant IDH1 inhibitor had minimal effects on mutant IDH1–induced histone modifications, and on the growth and clonogenicity of two different models of IDH1 mutant–driven gliomagenesis. The behavior of mutant IDH1 in this regard differs from that of other oncogenic insults, such as c-myc and H-ras, which in general require sustained activation to maintain the tumorigenic phenotype (35–39). Although the current studies do not eliminate the possibility that presence of mutant IDH1, even in an AGI-5198–inhibited form, is important in the maintenance of tumor cell growth (8, 40), the dispensability of mutant IDH1 activity for the continued growth and clonogenicity of mutant IDH1–driven glioma cells suggests that inhibition of mutant IDH1 activity may represent a suboptimal therapeutic approach in glioma.

The observation that mutant IDH1 activity is important in initiating, but not maintaining, the gliomagenic state may have important clinical implications. If already established mutant IDH1 tumors are no longer dependent on mutant IDH1 activity, the utility of mutant IDH1–selective inhibitors would appear to be limited. IDH1 mutation, however, appears to temporally precede genetic alterations associated with resolution of telomeric dysfunction, such as ATRX (18–21). Furthermore, in the telomere-deficient model used here, the resolution of telomeric dysfunction appears to be a temporal bottleneck in the mutant IDH1–driven transformation of telomere-negative cells, and this resolution can take several months in culture (30). It therefore remains possible that mutant IDH1 inhibitors may limit the ability of cells to resolve telomeric dysfunction and therefore have a role in controlling recurrence. Alternatively, changes in metabolism induced by expression of mutant IDH1 and the shunting of α-KG to 2-HG may make cells collaterally sensitive to other metabolically targeted therapies (28), which may or may not require inhibition of mutant IDH1. The current studies nonetheless show that although IDH1 mutation is an important driver of gliomagenesis, its utility as a direct therapeutic target may be less than optimal.

No potential conflicts of interest were disclosed.

Conception and design: T.-C.A. Johannessen, J. Mukherjee, S. Ohba, R.O. Pieper

Development of methodology: T.-C.A. Johannessen, J. Mukherjee, S. Ohba

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): T.-C.A. Johannessen, P. Viswanath, S.M. Ronen

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): T.-C.A. Johannessen, J. Mukherjee, P. Viswanath, S.M. Ronen, R.O. Pieper

Writing, review, and/or revision of the manuscript: T.-C.A. Johannessen, S.M. Ronen, R.O. Pieper

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): R. Bjerkvig, R.O. Pieper

Study supervision: J. Mukherjee, R. Bjerkvig, R.O. Pieper

This work was supported in part by the NIH grants CA172845-03 (to S.M. Ronen and R.O. Pieper), CA171610-03 (to R.O. Pieper), the Loglio Research Project (to S.M. Ronen and R.O. Pieper), and The Kristian Gerhard Jebsen Foundation and The Norwegian Cancer Society (to T.-C.A. Johannessen and R. Bjerkvig).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.