Mutant IDH1 Disrupts the Mouse Subventricular Zone and Alters Brain Tumor Progression

This article is featured in Highlights of This Issue, p. 499

Gliomas are the most common primary malignant tumor of the central nervous system (CNS) and account for 80% of all malignant brain tumors. Despite decades of research, overall survival of patients with malignant glioma remains dismal (1, 2). Even with aggressive treatment involving surgical resection, radiation, and chemotherapy, malignant gliomas almost invariably recur, due, in large part, to their invasiveness and resistance to therapy (3).

Previous studies have identified isocitrate dehydrogenase 1 (IDH1) as a major cancer gene among the low-grade gliomas, which have an inherent tendency to progress from low-grade (grade II and grade III) to high-grade (grade IV) gliomas, also known as secondary glioblastoma. One key signature among the diffuse gliomas is the cooccurrence of IDH1 mutations with other alterations, such as TP53 mutations (4). The presence of these mutations in the initial biopsies of most patients with these gliomas suggests they are early events in tumorigenesis and are important drivers of malignant progression (3–5).

Normally, IDH1 functions in the oxidative decarboxylation of isocitrate to α-ketoglutarate (α-KG) in an NADP⁺-dependent manner. The most common mutation in IDH1 is a hotspot missense mutation at arginine 132 (R132), and among gliomas, it is most commonly mutated to histidine (R132H; ref. 4). Those cells expressing IDH1-R132H display neomorphic enzymatic activity in the NADPH-dependent reduction of α-KG to D-2-hydroxyglutarate (D-2HG; ref. 6). Consequently, low- and high-grade IDH1-R132H–expressing gliomas produce D-2HG at levels 100-fold higher than in IDH1 wild-type tumors (6).

The IDH1 mutation and elevated D-2HG levels confer a myriad of cellular effects including inhibition of α-KG–dependent dioxygenases, aberrant histone methylation, contribution to the glioma CpG island methylation phenotype (G-CIMP), increased trimethylation of histone H3 at lysine 9 and 27, and decreased levels of 5-hydroxymethylcytosine (7–10). In addition, mutant IDH-mediated epigenetic dysregulation is thought to inhibit normal differentiation patterns and impart significant effects on metabolism (11–16). Because of its promiscuous activity throughout a cell, mutant IDH1 likely promotes tumorigenesis via a convergence of various perturbed intracellular pathways within both the tumor cells of origin and the surrounding microenvironment.

To determine the effects of mutant IDH1 and D-2HG on the glioma cell-of-origin and the microenvironment, we generated a genetically faithful conditional knock-in mouse model of mutant Idh1. In vitro analyses of NSCs expressing mutant Idh1 show a reduction in proliferation and a block to neuronal differentiation, corroborating previous studies describing mutant IDH1′s effects on differentiation programs. Similarly, expression of mutant Idh1 in vivo reduces the proliferation of cells in the subventricular zone (SVZ) of the lateral ventricles and causes a disorganization to this structure which is normally very tightly organized. Together, this suggests that expression of mutant Idh1 has damaging effects on the glioma cell-of-origin and the microenvironment. Guided by the genetic signature of human gliomas, we used a mouse glioma model driven by Tp53 deficiency to show that expression of mutant Idh1 leads to distinct molecular and histologic features, and alters the course of tumor progression by slowing tumor growth and prolonging survival. These findings provide new insight into IDH1-R132H–driven tumorigenesis, and establish a platform for future preclinical investigations focused on mutant IDH1-targeted therapy.

A targeting vector based off of the genomic sequence of the 129P2/Ola strain of mouse was generated and included a 5,426-base pair 5′ homology arm encompassing exon 2 of Idh1 followed by a LoxP site, three copies of the SV40 polyadenylation signal, a neomycin-resistance gene flanked with FRT sites, another LoxP site, and a 2,416-base pair 3′ homology arm encompassing exon 3 of Idh1 and the R132H mutation. A two-base pair change from the CGA codon for arginine to the CAC codon for histidine was produced. This construct was electroporated into 129P2/Ola mouse embryonic stem cells and recombinants were selected using G418. PCR across the 3′ homology arm was performed to identify homologous recombinants. hGFAP-Cre, Nestin-CreER^T2, and Tp53^fl/fl animals were generously provided by Dr. Rob Wechsler-Reya (Sanford-Burnham Medical Research Institute, La Jolla, CA) and were crossed into the Idh1-R132H conditional knock-in mouse model. Animals were maintained in a barrier facility, under pathogen-free conditions according to NIH guidelines.

Tissue was lysed in lysis buffer (100 mmol/L Tris-HCl pH 8; 5 mmol/L EDTA pH 8; 0.2% SDS; 200 mmol/L NaCl; Proteinase K, and water) and DNA was ethanol precipitated. Pellet was resuspended in 50 μL sterile H₂O and further diluted 1:50 in sterile H2O. Diluted DNA (1.5 μL) was used for PCR reactions. A touchdown PCR protocol was performed: 1× (95°C 5 minutes); 3× (95°C 20 seconds, 64°C 20 seconds, 72°C 20 seconds), 3× (95°C 20 seconds, 61°C 20 seconds, 72°C 20 seconds), 3× (95°C 20 seconds, 57°C 20 seconds, 72°C 20 seconds), 30× (95°C 20 seconds, 55°C 20 seconds, 72°C 20 seconds), 1× (72°C 5 minutes). Genotyping primers can be found in Supplementary Table S1.

Quantification of D-2-HG in biological media/tissues was done by LC-ESI-MS/MS method as published by Struys and colleagues with modifications to accommodate different equipment and sample matrices (17).

D-2HG and Diacetyl-l-tartaric anhydride (DATAN) were from Sigma-Aldrich. Racemic mixture of L- and D-2HG-d4 was prepared by mixing 1 mg of α-ketoglutarate-d6 (Sigma/Isotec) with 1 mg of NaBH₄ (Sigma) in 0.2 mL anhydrous MeOH (Sigma) followed by 30-minute incubation at 60°C. Sodium hydroxide (25%) and formic acid (98%) were from Fluka, Germany (mass spectrometry grade). Other reagents and solvents were of analytic grade.

To 20 μL of sample, 10 μL of 20 μg/mL of each L/D-2HG-d4 (internal standard) in water was added and mixture evaporated to dryness under gentle stream of nitrogen at 65°C. The dry residue was treated with 50 μL of 50 mg/mL of freshly prepared DATAN in dichloromethane/glacial acetic acid (4/1 by volume) and heated at 75°C for 30 minutes. After drying (65°C, 1 hour), the residue was dissolved in 20-μL LC mobile phase (see below) for LC/MS-MS analysis.

Instrument: Shimadzu 20A series HPLC and Sciex/Applied Biosystems API 4000 QTrap. Mobile phase (isocratic elution): water, 3% acetonitrile, 0.5% methanol, 2 mmol/L ammonium hydroxide, pH adjusted to 3.6 by formic acid. Analytic column: ACE #ACE-111-1546 (C₁₈, 150 × 4.6 mm), and #ACE-111-0103GD guard column. Column temperature: 45°C; run time: 4 minutes; injection volume, 5 μL. Mass spectrometry parameters (voltages, gas flow, and temperature) were optimized by infusion of 100 ng/mL of analytes in mobile phase at 10 μL/minute using Analyst 1.6.2 software tuning module. The Q1/Q3 (m/z) transitions monitored: 363/147 (D-2HG) and 367/151 (L/D-2HG-d4).

A set of calibrator samples in NeuroCult NSC base media was prepared by adding appropriate amounts of pure D-2HG at the following concentration levels: 0, 0.032, 0.16, 0.8, 4, and 20 μg/mL. The calibration samples were analyzed alongside the experimental samples and accuracy acceptance criteria was 85% for each but the lowest level (0.032 μg/mL, 80%, LLOQ). The obtained calibration curve was linear (r² = 0.999). Analyst 1.6.2 software was used for integration of the chromatograms, calibration curve calculation, and quantification of the study samples.

NSCs were harvested from E14.5 embryos. Pregnant dams were euthanized via CO2 asphyxiation. NSCs were harvested and cultured from embryos in accordance with Stemcell Technologies Technical Manual for In Vitro Proliferation and Differentiation of Mouse Neural Stem and Progenitor Cells Using NeuroCult (Stemcell Technologies, 05704). 1-day postharvest, cells were infected with Cre-expressing RGD fiber modified adenovirus (Vector Biolabs, 1769) or GFP-expressing adenovirus (Vector Biolabs, 1768) at an multiplicity of infection (MOI) of 10–100. Differentiations were performed in accordance with the aforementioned Stemcell Technologies kit. Primary mouse cell lines were generated in-house from February 2014 to February 2016 and have been authenticated through PCR genotyping. Cell lines are routinely submitted for mycoplasma contamination testing. All lines tested negative for these studies.

Cell proliferation was assessed via CyQUANT Cell Proliferation Assay Kit (Thermo Fisher C7026) and protocol was performed as recommended.

qRT-PCR was performed using the KAPA SYBR FAST qPCR Master Mix (KapaBioSystems, KK4600) following a touchdown PCR protocol. Primer sequences can be found in Supplementary Table S1.

Standard Western blot analysis protocols were followed. Antibodies included p53 (1C12) mouse mAb (Cell Signaling Technology, 2524), IDH1-R132H (Dianova, H-09), and GAPDH (D16H11) XP Rabbit mAb (Cell Signaling Technology, 5174). Secondary antibodies included anti-mouse IgG, HRP-linked (Cell Signaling Technology, 7076), anti-rabbit IgG, HRP-linked (Cell Signaling Technology, 7074).

RNAseq data and mutation status for IDH1 of lower grade gliomas (n = 534) were obtained from TCGA website. RNAseq-normalized counts were utilized for comparison of gene expression between mutant IDH1 (n = 416) and wild-type IDH1 (n = 118) samples. Gene expression comparison was shown by Boxplot with dots. Student t test was used for statistical analysis. All graphs and statistical analyses were performed using either GraphPad Prism or R. Gene expression profiling and pathway analysis can be found in Supplementary Materials and Methods.

Animals were mated and checked for the presence of a vaginal plug, which indicated 0.5 days postcoitum (dpc). Tamoxifen was administered to induce activation of Cre-recombinase in those animals harboring Nestin-CreER^T2. Tamoxifen (Sigma, T5648) was reconstituted in pharmaceutical grade corn oil (20 mg/mL), warmed to 37°C, and vortexed thoroughly until completely dissolved. Three milligrams of tamoxifen was administered to pregnant dams at 18.5 dpc via oral gavage. Randomization of animals was not necessary for these studies as genotype dictated treatment. Gender was recorded, though was not a criteria for inclusion or exclusion from treatment cohorts. Adult animals were euthanized (via CO₂ asphyxiation or through a lethal dose of urethane; 10 μL/g of 20% urethane in sterile PBS). Once unresponsive, animals underwent intracardiac perfusion with 15 mL PBS followed by 15 mL of 10% neutral buffered formalin. Tissues were harvested and postfixed in 10% neutral buffered formalin for 24 hours and transferred to 70% ethanol for paraffin embedding. Animals euthanized for the postnatal day 3 (P3) timepoint were placed on gauze and ice until deeply anesthetized followed by decapitation and tissue harvesting. Brains were placed in 10% neutral buffered formalin for 24 hours, at which point they were transferred to either 70% ethanol for paraffin embedding or in 20% sucrose in PBS and placed at 4°C until cryoprotected for optimal cutting temperature (OCT) embedding. Paraffin blocks were cut 5–6 microns on charged slides. OCT blocks were cut at 14 microns on charged slides and stored at −80°Celsius until IF staining.

Standard hematoxylin and eosin (H&E) and immunofluorescent staining procedures were performed. The Duke University Pathology Research Histology and Immunohistochemistry Laboratory processed and stained samples for IHC. A standard protocol requiring antigen retrieval in citrate buffer (pH 6.0, 80°C) was used. For IHC, the DAKO Autostainer 3400 was used. Antibodies included β(III)Tubulin 1:1,000 Covance-MMS-435P), CD31 1:100 (Abcam-ab28364), IDH1-R132H 1:100 (Dianova-H09), Ki67 1:100 (BD Pharm-550609), Olig2 1:500 (Abcam-ab136253), and Sox2 1:100 (Stemcell Technologies-60055). Secondary antibodies included Alexa Fluor 488 (Thermo Fisher) and Alexa Fluor 594 (Thermo Fisher). Immunofluorescent staining was performed in accordance with the aforementioned Stemcell Technologies kit. H&E and IHC samples were imaged using the Leica DMD 108. Immunofluorescent images were imaged on the Nikon Eclipse T2000-E. Positively stained cells were both manually counted and counted with ImageJ. Those assigned to counting were blinded to animal genotype and were given designated quadrants of specific size and magnification to count.

Genomic DNA was extracted using the AllPrep DNA/RNA Mini Kit (Qiagen 80204). DNA was quantified and MethylFlash Global DNA Methylation (5-mC) ELISA Easy kit (EpiGentek P-1030-48) colorimetric assay was performed to assess global DNA methylation in accordance with the supplied handbook.

EpiTect II DNA Methylation Enzyme Kit (Qiagen 335452) was used with EpiTect Methyl II PCR Assay (Qiagen 335002) to assess the methylation status of the second promoter of the mouse BCAT1. Protocol was followed in accordance to handbook using primers specific for mouse CGI 48 (human CGI 70; Qiagen EPMM109656). Data were analyzed using the supplied Methylation Data Analysis template for single assays.

All relevant data and reagents are available from the authors upon request and fulfillment of proper university required documentation.

All animal studies were approved by the Institutional Animal Care and Use Committee of Duke University, an institution accredited by the Association for Assessment and Accreditation of Laboratory Animal Care (AAALAC), International. Protocol numbers: A109-10-04, A072-13-03 and A037-16-02.

Additional information can be found in Supplementary Materials and Methods.

A mutant Idh1 conditional knock-in model was generated via a targeting vector containing a stop cassette flanked by LoxP sites (LSL) and the IDH1-R132H mutation (referred to as Idh1^LSL:R132H; Fig. 1A). Homologous recombination into the endogenous mouse Idh1 locus results in a mutant Idh1 allele with a CGA to CAC substitution in exon 3, encoding the arginine to histidine substitution and resembling the cancer-specific IDH1-R132H mutation. The expression of the modified allele is blocked by the preceding LoxP-flanked stop sequence, resulting in a knocked out allele that is restored through Cre-recombinase–mediated excision of the stop cassette. PCR- and sequencing-based genotyping of animals indicated that only Idh1^LSL:R132H heterozygous animals are generated (Fig. 1B and C; primer sequences can be found in Supplementary Table S1). Sixty-two crosses between Idh1^LSL:132H/WT heterozygous animals generated a total of 292 offspring, with 113 wild-type (Idh1^WT/WT), and 179 heterozygous (Idh1^LSL:R132H/WT) animals. Testing for the expected Mendelian ratio of 1:2:1 for wild-type:heterozygous:homozygous mutant animals, respectively, gives a P value less than 1.0 × 10⁻⁹ (Fisher exact test). The failure to recover Idh1^{LSL:R132H/LSL:R132H} homozygous animals suggested embryonic or perinatal lethality of homozygous Idh1 knockouts.

To validate the faithfulness of the mutant Idh1 allele, NSCs were isolated from embryonic day 14.5 (E14.5) heterozygous mutant Idh1 conditional knock-in embryos (Idh1^LSL:R132H/WT) and Idh1 wild-type embryos (Idh1^WT/WT). NSC lines were transduced with adenoviral-GFP (ad-GFP) as control or adenoviral-Cre-recombinase (ad-Cre) to induce excision of the stop cassette and expression of mutant Idh1. Transient expression of Cre-recombinase led to production of D-2HG which was detected in the supernatant of mutant lines at a level 100-fold higher than in wild-type cells, suggesting D-2HG was produced at pathologically relevant levels (Fig. 1D).

Considering the effects of mutant IDH1 are likely cell context-dependent, we sought to determine the effects of mutant Idh1 in NSCs, a candidate cell-of-origin for gliomas (18, 19). Idh1^LSL:R132H/WT NSCs displayed a nearly 4-fold reduction in proliferative capacity compared with wild-type NSC lines or ad-GFP–transduced control samples (Fig. 2A). Wild-type NSCs were transduced with ad-GFP and ad-Cre with no differences in proliferation observed. This suggested that the proliferative defects in the Idh1^LSL:R132H/WT ad-Cre–infected NSCs were not resultant of the generalized effects of Cre-recombinase or adenoviral transduction, and instead were mediated by the expression of mutant Idh1 (Fig. 2A).

Normal primary cells frequently respond to oncogenic insult by undergoing cell-cycle arrest or senescence via p53 activation (20, 21). To determine whether the reduced proliferation of mutant Idh1–expressing NSCs was due to p53-mediated growth inhibition, we determined whether the p53 pathway was activated in NSCs following expression of mutant Idh1. Following ad-Cre transduction, total p53 was elevated in Idh1^LSL:R132H/WT lines (Fig. 2B and C), which was accompanied by the upregulation of p21 (Fig. 2D). p53 mRNA was also elevated as was shown through qRT-PCR (normalized to GAPDH; Supplementary Fig. S1). These results suggest that NSCs respond to mutant Idh1 by activating the p53 pathway. In addition, microarray analysis was performed on three independent Idh1^LSL:R132H/WT cell lines that were transduced with ad-GFP (control) or ad-Cre (Gene Expression Omnibus Accession number: GSE88828). Hierarchical clustering of the top 1% variant genes (n = 414) shows a clustering of those lines treated with ad-GFP versus those treated with ad-Cre, suggesting specific profiling following induction of mutant IDH1 expression (Supplementary Fig. S2). Pathway analysis and gene ontology was performed on those genes displaying a greater than 2-fold change (Supplementary Tables S2 and S3). Cell cycle and p53 signaling were among the top affected pathways involved when comparing the ad-GFP and ad-Cre cell lines. These data further corroborate our initial data as well as implicate the importance of the p53 pathway in response to induction of mutant IDH1.

To further determine the role of p53 activation, NSCs were isolated from E14.5 embryos of the following two genotypes: Idh1^LSL:R132H/WT; Tp53^fl/fl and Tp53^fl/fl, which have Tp53 conditionally deleted (22). Mutations were induced in vitro via ad-Cre transduction and this was accompanied by loss of p53 expression and production of D-2HG in mutant IDH1–expressing NSCs (Supplementary Fig. S3A and S3B). As shown in Fig. 2, following deletion of Tp53, an increase in NSC proliferation was observed (Fig. 2A, bottom left). A similar increase in NSC proliferation was observed when mutant Idh1 was expressed in the context of Tp53 deletion (Fig. 2A, bottom right). However, this was an opposite effect from when mutant Idh1 was expressed in the absence of Tp53 deletion (Fig. 2A, top right). Following transduction with ad-Cre, growth of the Idh1^LSL:R132H/WT; Tp53^fl/fl line displayed a 2.5-fold growth increase over the Idh1-wild-type NSCs. Collectively, these results suggest that p53 activation mediates mutant Idh1-induced cell growth inhibition, and that while mutant Idh1 alone confers a growth disadvantage, the biological outcome can be different and even opposite in the presence of Tp53 deletion (Fig. 2A). These data also support the possibility that deletion of TP53 is a cooperating alteration in mutant IDH1–driven tumorigenesis. We analyzed The Cancer Genome Atlas' (TCGA) study on diffuse lower grade gliomas and found that while 15% of IDH1 wild-type tumors have mutations in TP53, approximately 62% of IDH1-mutant gliomas harbor a mutation in TP53 (3). This finding suggests that while TP53 mutations promote low-grade gliomas, it is likely that it plays an even more important role in promoting IDH1-mutated low-grade gliomas (Fig. 2E).

Mutant Idh1 was expressed in the mouse brain under the control of different brain-specific promoters of varying expression to assess its effects on cells of the brain in vivo. First, the human glial fibrillary acidic protein promoter (hGFAP-Cre), which is broadly expressed in the CNS, was utilized. These animals display specific labeling in several neural lineages including the granule cells of the cerebellum, olfactory bulb, dentate gyrus, hippocampal pyramidal cells, the gray matter of the cerebral cortex, and in ependymal cells lining the ventricular system (23). This system was previously used to narrow the identity of the glioma cell-of-origin and to drive brain tumorigenesis (18, 19, 24, 25). To assess the in vivo effects of mutant Idh1 when it is broadly expressed in the CNS, hGFAP-Cre mice were crossed to Idh1^LSL:R132H/WT animals.

As expected, Idh1-R132H expression was detected in the brains of E15.5 animals bearing the Idh1-R132H mutation (hGFAP-Cre; Idh1^LSL:R132H/WT), but not in wild-type controls (hGFAP-Cre; Idh1^WT/WT; Fig. 3A). In addition, D-2HG was detected in the brains of mutant animals at a level 200-fold higher than in control animals (Fig. 3B). The increase of D-2HG seen in our model is consistent with the magnitude observed in human gliomas bearing the IDH1-R132H mutation, and lends credence that this system is producing biologically relevant levels of D-2HG (26).

To investigate the early effects of mutant Idh1 on CNS cells, brains from P3 hGFAP-Cre; Idh1^LSL:R132H/WT and hGFAP-Cre; Idh1^WT/WT animals were harvested. hGFAP-Cre; Idh1^LSL:R132H/WT brains were visibly different from the control brains with greater than 80% displaying foci of hemorrhagic lesions (16/17) throughout the cortex that were absent from control animals (0/20; Fig. 3C, left). H&E staining showed red blood cell extravasation in the mutant brains; however, in control brains, red blood cells were contained within vessels (Fig. 3C, middle). IHC for CD31 was performed to visualize endothelial cells and observe any effects to their structural integrity. Interestingly, vessels among mutant brains tended to be thicker and stained more intensely with CD31 (Fig. 3C, right).

NSCs are located in the subventricular zone (SVZ) of the lateral ventricles and gives rise to other NSCs as well as differentiated progeny. As the SVZ is a germinal zone, proliferation was assessed using the proliferation marker Ki67. As expected, robust proliferation in the SVZ of the lateral ventricles in hGFAP-Cre; Idh1^WT/WT animals was observed (Fig. 4A). This was in contrast to the SVZ of hGFAP-Cre; Idh1^LSL:R132H/WT animals which displayed a 15% reduction in Ki67-positive cells, suggesting an attenuated proliferative capacity of the cells within this germinal zone, a finding consistent with our in vitro data (Figs. 2A and 4A and B).

To assess mutant Idh1′s effects on the NSC population, the NSC marker Sox2 was used. hGFAP-Cre; Idh1^WT/WT animals displayed a tightly organized SVZ with a structured layering rich with Sox2-positive cells; in contrast, the hGFAP-Cre; Idh1^LSL:R132H/WT animals displayed a perturbed SVZ and a dispersal of NSCs emanating outward from the SVZ (Fig. 4A and C). While the disorganization of this region was consistent among those animals broadly expressing mutant Idh1 in their brain (5/6), the interaction of distance from the ventricle and different genotypes was not statistically different (P = 0.148, repeated measures ANOVA). However, when the NSCs of the SVZ were quantified and normalized to the overall number of cells in this region, there was a statistically significant increase in the number of ectopic Sox2 positively stained cells in the mutant brains (Fig. 4D).

The effect of mutant Idh1 on the NSCs within the SVZ lead to the question of whether mutant Idh1 also affects the NSCs in a cell-autonomous manner. To this end, Idh1^LSL:R132H/WT NSCs were isolated from E14.5 embryos and expression of mutant Idh1 was induced in vitro via ad-Cre transduction. Interestingly, when cultured under conditions that promote neural differentiation, the mutant Idh1-expressing NSCs were unable to differentiate into the neuronal lineage as indicated through a reduced expression of the mature neuronal marker, β-III-tubulin (Fig. 4E). Collectively, these experiments suggest that mutant Idh1 likely affects NSCs both directly, as observed through defective differentiation, and indirectly, as observed through a disorganization of the microenvironment, two components with implications in tumor initiation and development (27). These data also corroborate a previous study investigating the role of mutant IDH1 on embryoid body–induced human NSCs which showed that IDH1-R132H impaired the ability of these cells to differentiate, specifically into the astrocytic and neuronal lineage (28).

To determine the effects of prolonged expression of mutant Idh1 on the developing brain and assess the potential for mutant IDH1–mediated gliomagenesis, hGFAP-Cre; Idh1^WT/WT and hGFAP-Cre; Idh1^LSL:R132H/WT animals were aged. Physical and behavioral differences became apparent in 80% of hGFAP-Cre; Idh1^LSL:R132H/WT animals shortly after weaning age. By 89 days, all hGFAP-Cre; Idh1^LSL:R132H/WT animals displayed the phenotypes of head doming, squinted eyes, hunched posture, and gradual immobilization, in addition to episodes of seizure and failure to thrive (Fig. 5A). The median survival for these animals was approximately 25 days with 3 animals surviving asymptomatically for as long as 89 days before necessitating euthanasia (Fig. 5B). Dissection of the brains from mutant animals displayed large fluid-filled cavities and severely reduced cortical thickness. Histologic analysis indicated that the cavities represented a severe enlargement of the lateral ventricles (Fig. 5A). These features are characteristic of severe hydrocephalus, a condition that affects 3 in 1,000 births and leads to excessive accumulation of cerebrospinal fluid (CSF) in the brain, widening the ventricular spaces, and creating a build-up of pressure that could promote brain injury or even death if left untreated (29, 30). An analysis of serial sections of hydrocephalic hGFAP-Cre; Idh1^LSL:R132H/WT brains revealed several instances of aqueductal stenosis and an overall loss of cellularity in mutant brains.

The phenotypic differences observed in this hGFAP-Cre; Idh1^LSL:R132H/WT model with those in the literature suggest that the timing of mutant Idh1 induction is imperative. Whereas a previously described Nestin-Cre system induced mutant Idh1 at E10.5, our hGFAP-Cre system induced it at E13.5 (31). While there was severe hemorrhage present in the former model, there were small hemorrhagic foci present in the latter model, suggesting that the severity of the mutant Idh1 phenotypes is dependent on the extent of induction. It follows that restriction of mutant Idh1 expression to a smaller population of cells may circumvent these severe phenotypes, allowing for prolonged expression and investigation of the effects of mutant Idh1.

The inducibly activated Nestin promoter (Nestin-CreER^T2) displays restricted expression in NSCs of embryonic and adult animals and has been used previously to both model gliomas and to define the NSC as the candidate cell-of-origin for glioma (19, 32). Nestin-CreER^T2 was utilized to achieve spatiotemporal control over mutant Idh1 expression in the NSC population following administration of tamoxifen. This system was validated by crossing it to a conditional tdTomato reporter mouse (33). Following tamoxifen treatment of Nestin-CreER^T2; tdTomato^fl/fl animals at E18.5, robust recombination and subsequent expression of tdTomato was observed in the SVZ of P3 animals (Supplementary Fig. S4). This system permits the controlled and restricted expression of mutant Idh1 in the candidate cell-of-origin for glioma and allows for the prolonged and specific expression of mutant Idh1 in the NSCs specifically, allowing us to ask whether mutant Idh1 is capable of promoting tumorigenesis. Both Nestin-CreER^T2; Idh1^WT/WT and Nestin-CreER^T2, Idh1^LSL:R132H/WT animals were generated and tamoxifen was administered at E18.5. After one year, all animals were asymptomatic, with survival being indistinguishable between genotypes (Fig. 6A).

We next considered the effects of mutant Idh1 in the context of tumor biology and tumor progression. Considering the prevalence of cooccurring mutations in IDH1 and TP53 (Fig. 2E), coupled with the in vitro and in vivo effects of mutant Idh1 on NSCs (Figs. 2A, and 4B and D), conditionally deleted Tp53 animals were crossed into the mutant Idh1 conditional knock-in animals, leading to a compound genetic mouse model harboring the most common mutations found among the low-grade gliomas (22). Two genotypes, Nestin-CreER^T2; Tp53^fl/fl and Nestin-CreER^T2; Idh1^LSL:R132H/WT; Tp53^fl/fl were administered tamoxifen at E18.5 to assess the impact of mutant Idh1 on brain tumorigenesis. Consistent with previous neural-specific, Tp53-deleted tumor models, tumors developed in Tp53-deleted animals (Nestin-CreER^T2; Tp53^fl/fl) at a penetrance of 50% (6/12; ref. 34). Among animals that harbored both the Idh1-R132H mutation and Tp53 deletion (Nestin-CreER^T2; Idh1^LSL:R132H/WT; Tp53^fl/fl), the penetrance was reduced to 15% (4/26).

An IDH1-R132H–specific antibody was tested on tumor-bearing brains from Nestin-CreER^T2;Idh1^LSL:R132H/WT; Tp53^fl/fl and Nestin-CreER^T2; Tp53^fl/fl animals (Fig. 6B). Tumors that developed in Nestin-CreER^T2; Idh1^LSL:R132H/WT; Tp53^fl/fl animals stained positively with the IDH1-R132H antibody. This was in contrast to those tumors that developed in animals of the genotype Nestin-CreER^T2; Tp53^fl/fl, which showed a lack of IDH1-R132H staining, suggesting that tumors were initiated from cells with mutations in both genes (Fig. 6B).

While the overall tumor penetrance was low, several histologic observations include that Idh1-mutant tumors tended to be sarcomatoid in nature, with more perivascular infiltration into surrounding brain parenchyma rather than single-cell infiltration. This is in contrast to the Idh1 wild-type tumors, which were more epithelioid and displayed both single-cell and perivascular infiltration. Pseudopalisading necrosis was present in several of the Idh1-mutant tumors. In addition, there was a prevalence of giant cells among the Idh1-mutant tumors (Fig. 6B, arrowhead). The presence of giant cells among glioma patients is associated with a prolonged survival, a finding consistent among our mutant Idh1 tumor-bearing animals (35).

IHC characterization of tumors was performed using common markers for clinical diagnosis, including markers for astrocytes (GFAP), oligodendrocytes (Olig2), and proliferative cells (Ki67). The majority of tumors stained positively for GFAP; however, this was determined to be reactive astrocytosis and spanned genotypes. In addition, 100% of the tumor cells in the Idh1-mutant cases stained positively for Olig2, in contrast to 20% to 60% of the tumor cells in the Idh1 wild-type tumors (Fig. 6B). Among human diffuse glioma samples, those harboring IDH1 mutations display an elevated expression of Olig2 compared with wild-type samples, suggesting a cooperation between mutant IDH1 and Olig2 expression in promoting tumorigenesis (Fig. 6C; refs. 36, 37). This is a feature specific for the proneural group of gliomas, as characterized by TCGA (36, 37).

When comparing overall survival of experimental animals, mutant Idh1 is shown to confer a prolonged survival among Nestin-CreER^T2; Idh1^LSL:R132H/WT; Tp53^fl/fl tumor-bearing animals compared with Nestin-CreER^T2; Tp53^fl/fl tumor-bearing animals (Fig. 6A). Although the median survival was unable to be determined due to a low number of affected animals in the Nestin-CreER^T2; Idh1^LSL:R132H/WT; Tp53^fl/fl cohort, the average time for onset of symptoms was 233 days compared with 274 days for the Nestin-CreER^T2; Tp53^fl/fl versus Nestin-CreER^T2; Idh1^LSL:R132H/WT; Tp53^fl/fl animals, respectively. The survival trend corroborates the clinical data, which shows a prolonged survival among brain tumor patients with mutations in IDH1 compared with patients with wild-type IDH1 tumors. For instance, among patients with grade III anaplastic astrocytoma, survival was 65 months for those with IDH1 mutations compared with 20 months for those with wild-type IDH1 tumors. Similarly, among patients with grade IV glioblastoma, those with IDH1-mutant tumors displayed a 31-month survival compared with 15 months for those with IDH1-wild-type tumors (4, 38).

To further test the faithfulness of this brain tumor model, brain tumors were harvested from symptomatic animals and cultured in vitro. Tumor cultures were derived from animals of the genotype Nestin-CreER^T2; Tp53^fl/fl and Nestin-CreER^T2; Idh1^LSL:R132H/WT; Tp53^fl/fl. These lines did not express p53 (Supplementary Fig. S3B) and D-2HG was elevated in the mutant IDH1-expressing brain tumor line (Supplementary Fig. S5). This observation coupled with the observation that non-tamoxifen–treated animals failed to develop tumors (Fig. 6A) suggests that these tumors were derived from those cells that recombined following tamoxifen administration and further implicates the NSC as the cell-of-origin for these tumors.

Mutant IDH1 has been shown to promote G-CIMP (7, 39). To determine whether mutant IDH1 confers the expected hypermethylation phenotype, global methylation was assessed in cell lines derived from tumor-bearing animals. An increase in global methylation in the mutant IDH1–expressing brain tumor cell line was observed (Fig. 6D). In addition, the second promoter of BCAT1, a loci known to be differentially methylated in response to mutant IDH1 was investigated (40). The homologous CpG island in mouse was identified and the methylation status was differentially methylated as it is in human IDH1-mutated glioma, with the Nestin-CreER^T2; Idh1^LSL:R132H/WT; Tp53^fl/fl line showing greater methylation than the Nestin-CreER^T2; Tp53^fl/fl line at the second promoter of Bcat1 (Fig. 6E). Combined, these data suggest that the tumors derived from our genetic model recapitulate many of the features of mutant IDH1 gliomas, including expression of Olig2, a prolonged survival among mutant IDH1 patients, elevated global DNA methylation, and hypermethylation of the Bcat1 gene.

Despite the prevalence of the IDH1 mutation among gliomas, its effects on the NSC population (which are a candidate cell-of-origin for glioma), the microenvironment from which gliomas arise, or how this mutation affects the course of tumor development are incompletely understood. We address these three aspects using conditional mutant Idh1 in vitro and in vivo mouse models.

Considering the large body of evidence implicating the NSCs as the cell-of-origin for glioma, our studies have focused on the effects of mutant IDH1 on the NSC in vitro and in vivo (18, 19, 24, 27). However, there are other candidate cells of origin, including the oligodendroglial progenitor cell (OPC) for which there is impressive data implicating it as the cell-of-origin (41). Further investigations are required to assess the differential effects of mutant IDH1 on different candidate cells of origin.

We show that mutant Idh1 attenuates proliferation of NSCs, suggesting that mutant Idh1 acts as an oncogenic insult to which cells respond by undergoing cell-cycle arrest. This in vitro growth reduction as a result of mutant Idh1 corroborates both of our in vivo models: the first which shows that the Idh1 mutation reduces proliferation in the SVZ of the lateral ventricle among hGFAP-Cre; Idh1^LSL:R132H/WT animals, and the second model which shows a failure of mutant Idh1 to impart a tumorigenic phenotype among Nestin-CreER^T2; Idh1^LSL:R132H/WT animals. We show that this cellular response is mediated by p53 activation, as evidenced by increased p53 mRNA and protein and the upregulation of p21. Most importantly, deletion of Tp53 rescues the proliferative defect attributed to mutant Idh1, suggesting the potential cooperative effects between Idh1 and Tp53 mutations, and explaining the cooccurring nature of these mutations among glioma patients (3, 5, 37).

Mutant IDH1′s effects on cellular behavior are cell context–dependent, reducing proliferation in some cases, and increasing proliferation and promoting tumorigenesis in other cases (42, 43). The finding linking IDH1 and TP53 mutations presents clear genetic evidence that in addition to cellular context, additional genetic components play integral roles in dictating the distinct effects of mutant IDH1, a notion that is supported by previous findings. For example, Koivunen and colleagues discovered that immortalized human astrocytes infected with mutant IDH1 display a proliferative advantage and an enhanced ability for colony formation (42). These lines were immortalized and transformed through the expression of the human papillomavirus 16 E6, which inactivates p53 (44). On the basis of the status of TP53, a proliferative advantage would be expected in these cells, as was the case. Another study by Bralten and colleagues shows that mutant IDH1 decreases the proliferation of the established glioma cell line U87 and a human embryonic kidney (HEK) cell line, both of which have TP53 intact (43). While these findings do not provide direct evidence that TP53 is the critical factor in deciding the cellular effects of mutant IDH1, they do support the notion that additional genetic aberrations alter the effects of mutant IDH1.

It is well accepted that the microenvironment plays critical roles in driving tumorigenesis and dictating tumor progression (45, 46). This is especially true for the IDH1 mutation, as one main feature of mutant IDH1 is the generation of the neometabolite D-2HG, which can be secreted into the extracellular compartment (6). The broadly expressed hGFAP-Cre promoter, which has been used previously to successfully generate glioma models (19, 25), was used to drive mutant Idh1 expression in the brain. This model revealed the striking effects that mutant Idh1 has on the CNS generally, and on the SVZ specifically. Perinatal hGFAP-Cre; Idh1^LSL:R132H/WT animals presented with areas of hemorrhagic foci throughout the cortex, a phenotype reminiscent of a model utilizing a Nestin-Cre driver that showed severe hemorrhage and perinatal lethality (31). This phenotypic variation can likely be explained by the timing and extent of mutant Idh1 induction; while Cre-recombinase begins to be expressed at E10.5 in the Nestin-Cre system, it begins at E13.5 in the hGFAP-Cre system (23, 32). At P3, hGFAP-Cre; Idh1^LSL:R132H/WT animals displayed reduced cortical thickness, loss of cellularity, and a disrupted SVZ, as indicated through a dispersal of Sox2-positive NSCs emanating outwards from the SVZ of the lateral ventricles. The SVZ and the NSC microenvironment are critical for the regulation of proliferation and progenitor cell differentiation; therefore, this disorganization likely perturbs the normal differentiation patterns of the resident NSCs.

While further studies are needed to reveal the mechanism behind mutant Idh1′s effects on the microenvironment, we speculate that the effects are at least partially mediated by D-2HG. There is an interesting pathologic similarity between hGFAP-Cre; Idh1^LSL:R132H/WT animals and patients with D-2-hydroxyglutaric aciduria, a neurometabolic disorder characterized by systemic elevation of D-2HG (47). Although a rare disorder, an international survey of patients with D-2-hydroxyglutaric aciduria described that all patients (17/17) exhibited ventriculomegaly, an enlargement of the lateral ventricles, as well as delayed cerebral maturation (48). These features suggest that D-2HG has a significant effect on the candidate cell-of-origin for gliomas by altering the signaling pathways within the niche and disrupting the microenvironment of the resident NSCs; these occurrences are likely imperative for tumor initiation.

The genetic data confers an unambiguous causal role of mutant IDH1 in gliomagenesis. Paradoxically, glioma patients with mutations in IDH1 tend to have a better survival clinically (3–5, 37). Together, this suggests that the IDH1 mutation is an effective promoter for clinically less aggressive gliomas. There are two likely underlying mechanisms for this phenomenon: (i) it is possible that the IDH1 mutation can only target and transform a unique cellular population; a population distinct from those cells that give rise to non-IDH1–mutant gliomas, and a population that intrinsically gives rise to less aggressive tumors; or (ii) while the IDH1 mutation promotes cellular transformation initially, its continuous presence in tumor cells may attenuate tumor progression and aggressiveness. Our experiments using the inducible Nestin-CreER^T2 model indicate that in tumors driven by Tp53 deletion, mutant Idh1 is expressed in all tumor cells, supporting the notion that the two mutations target the same population of cells that give rise to tumors. The prolonged survival among tumor-bearing animals with both mutant Idh1 and Tp53 mutations suggests that mutant Idh1 alters tumor biology as well as the course of tumor progression. Our in vivo findings, coupled with previous in vitro models (43, 49), support the latter mechanism and stress an important point for consideration when devising glioma treatments on the basis of mutant IDH1 inhibition. Notably, a recent study has shown that the response of cells to mutant IDH1 inhibition is dependent on whether treatment is administered before, at the same time, or after mutant IDH1 is expressed (50). This study raises an important question regarding the timing of therapeutic intervention and the potential reversibility of mutant IDH1–mediated epigenetic effects. Our models provide a valuable tool for addressing this question in a relevant pathologic setting.

Collectively, our findings provide new insight into the effects of mutant Idh1 on a candidate cell of origin for glioma, mutant Idh1′s role in disrupting the microenvironment from which gliomas arise, and mutant Idh1′s effect on the course of glioma progression. In addition, this study highlights the challenges involved in modeling mutant IDH1, including (i) the impact of broad expression of mutant IDH1 on the CNS, which hinders long-term investigations of mutant IDH1 in contributing to brain tumorigenesis; (ii) the necessity to test other cooperating genetic alterations; (iii) the unclear identity of the cells of origin that are most susceptible to mutant IDH1–mediated transformation, and (iv) the unknown contribution of cell-autonomous and non-cell–autonomous effects on gliomagenesis. Further studies involving alternative approaches, such as orthotopic transplantation of other candidate glioma cells of origin, will help overcome these challenges and will provide a faithful foundation for understanding mutant Idh1′s role in tumorigenesis and for investigating preclinical targeted therapies.

H.Yan receives royalties from Agios Pharmaceuticals and Personal Genome Diagnostics. H. Yan is a cofounder of Genetron Health (Beijing) Co. Ltd. No potential conflicts of interests were disclosed by the other authors.

Conception and design: C.J. Pirozzi, Y. He, H. Yan

Development of methodology: C.J. Pirozzi, M.S. Waitkus, Y. He, H. Yan

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): C.J. Pirozzi, A.B. Carpenter, C.Y. Wang, H. Zhu, L.J. Hansen, P.K. Greer, J. Feng, Y. Wang, C.B. Bock, P. Fan, I. Spasojevic, D.D. Bigner, Y. He, H. Yan

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis,): C.J. Pirozzi, A.B. Carpenter, L.J. Hansen, L.H. Chen, R.E. McLendon, Y. He, H. Yan

Writing, review and/or revision of manuscript: C.J. Pirozzi, A.B. Carpenter, M.S. Waitkus, Y. He, H. Yan

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): C.J. Pirozzi, C.Y. Wang, H. Zhu, P.K. Greer

Study supervision: Y. He, H. Yan

The authors thank Dr. Rob Wechsler-Reya at the Sandford-Burnham Medical Research Institute for sharing reagents and Dr. Darell Bigner at Duke's Preston Robert Tisch Brain Tumor Center for providing access to resources. We thank Ms. Jenna Lewis, Mr. Bill Diplas, Dr. Xuhui Bao, and Dr. Vidya Chandramohan for editorial revisions to the manuscript. We thank the UNC Animal Model Core for the initial animal generation. We thank the Pathology Research Histology and Immunohistochemistry Laboratory of Duke University for IHC staining.

This work was supported by the James S. McDonnell Foundation Research Award and R01-CA140316 (to H. Yan). Y. He was supported by an AACR-Aflac, Inc. Career Development Award, the National Comprehensive Cancer Network Young Investigator Award and the Circle of Service Foundation.

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.