IDH1 and IDH2 mutations in pediatric acute leukemia

Abstract

To investigate the frequency of isocitrate dehydrogenase 1 (IDH1) and 2 (IDH2) mutations in pediatric acute myeloid leukemia (AML) and acute lymphoid leukemia (ALL), we sequenced these genes in diagnostic samples from 515 patients (227 AMLs and 288 ALLs). Somatic IDH1/IDH2 mutations were rare in ALL (N=1), but were more common in AML, occurring in 3.5% (IDH1 N=3 and IDH2 N=5), with the frequency higher in AMLs with a normal karyotype (9.8%). The identified IDH1 mutations occurred in codon 132 resulting in replacement of arginine with either cysteine (N=3) or histidine (N=1). By contrast, mutations in IDH2 did not affect the homologous residue but instead altered codon 140, resulting in replacement of arginine with either glutamine (N=4) or tryptophan (N=1). Structural modeling of IDH2 suggested that codon 140 mutations disrupt the enzyme's ability to bind its substrate isocitrate. Accordingly, recombinant IDH2 R140Q/W were unable to carry out the decarboxylation of isocitrate to α-ketoglutarate (α-KG), but instead gained the neomorphic activity to reduce α-KG to R(−)-2-hydroxyglutarete (2-HG). Analysis of primary leukemic blasts confirmed high levels of 2-HG in AMLs with IDH1/IDH2 mutations. Interestingly, 3/5 AMLs with IDH2 mutations had FLT3 activating mutations, raising the possibility that these mutations cooperate in leukemogenesis.

Introduction

Pediatric de novo acute myeloid leukemia (AML) is a heterogeneous disease that can be divided into clinically distinct subtypes based on the presence of specific chromosomal abnormalities or gene mutations. The best characterized subtypes include leukemias with alterations of the genes encoding the core-binding transcription factor complex, (t(8;21)[RUNX1(AML1)-RUNX1T1(ETO)] and inv(16)/t(16;16)[CBFβ-SMMHC(MYH11)]), rearrangements of the MLL gene on chromosome 11q23, normal cytogenetics, or distinct morphology including acute promyeloctic leukemia with t(15;17)[PML/RARA] and acute megakaryoblastic leukemia (FAB-M7). In AMLs with normal cytogenetics, mutations have also been identified in a number of genes, with alterations in NPM1, FLT3 and CEBPA occurring at an appreciable frequency and influencing therapeutic responses.^1–3

Recent genome-wide sequencing efforts have led to the identification of a number of new candidate cancer genes in AML. Foremost among this list is isocitrate dehydrogenase 1 (IDH1), which was initially identified in a whole exome sequence analysis of glioblastoma multiforme (GBM), but was subsequently shown to be mutated in a variety of myeloid malignancies including up to 16% of adult AMLs with normal cytogenetics.^4–6 The mutations in both GBM and myeloid malignancies have been heterozygous and restricted to arginine 132 in exon 4 of IDH1, or to either the homologous residue in IDH2, R172, or to a second arginine, R140, also located in its substrate binding pocket. Although the distribution of specific IDH1/IDH2 mutations varies between GBM and AML, each of the analyzed mutations results in a loss of the enzymes ability to catalyze the oxidative decarboxylation of isocitrate to α-ketoglutarate (α-KG), coupled with a gain-of-function to catalyze the NADPH-dependent reduction of α-KG to 2-hydroxyglutarate (2-HG).^7–9 These alterations in the enzymatic activity of IDH1/IDH2 result in a dramatic increase in the level of 2-HG within the leukemic cells. In turn, the increased 2-HG leads to a hypermethylated phenotype, in part by directly inhibiting the enzymatic activity of the α-KG-dependent enzyme TET2, responsible for catalyzing the conversion of 5-methylcytosine to 5-hydrozymethylcytosine.^10–13

To the best of our knowledge, only two studies have investigated the presence of IDH1 mutations in pediatric acute leukemia, failing to identity any mutations.^{14, 15} In addition, a single study have investigated the presence of IDH2 mutations in pediatric AML and found a single case to harbor the IDH2 R140.¹⁵ To extend these studies and define the frequency of IDH1/IDH2 mutations in pediatric AML and ALL, we have now sequenced exon 4 of both IDH1 and IDH2 in diagnostic leukemia samples from 227 pediatric AML and 288 pediatric ALL patients. Our analysis identified somatic IDH1/IDH2 mutations as an exceedingly rare event in pediatric ALL (N=1), but a more common event in AML, occurring in 3.5% of cases (IDH1 N=3 and IDH2 N=5), with the frequency higher in AMLs with a normal karyotype (9.8%). Moreover, we demonstrate that the most common IDH mutations in AML (IDH2 R140Q and R140W), results in a marked decrease in the ability of the enzyme to catalyze the oxidative decarboxylation of isocitrate to α-KG, and the acquisition of the neomorphic activity to reduce α-KG to 2-HG leading to an accumulation of 2-HG in the leukemic cells.

Materials and Methods

Patients

Diagnostic samples from 227 pediatric patients with AML and 288 with ALL treated at St Jude Children's Research Hospital (SJCRH) were analyzed for mutations in exon 4 of IDH1 and IDH2. Informed consent for research was obtained from patients, parents, or guardians in accordance with the Declaration of Helsinki. This study was approved by the SJCRH Institutional Review Board. The genetic characteristics of the patients are shown in Table 1. The AML samples were cytogenetically analyzed and screened for the presence of PML-RARA, RUNX1(AML1)-RUNX1T1(ETO), and CBFB-SMMHC(MYH11) using reverse-transcription (RT)-PCR. 11q23/MLL rearrangements were evaluated by florescence in situ hybridization (FISH) to confirm MLL gene rearrangements and/or RT-PCR for MLL-AFF1(AF4), MLL-MLLT3(AF9), MLL-MLLT10(AF10), MLL-MLLT1(ENL), and MLL-ELL. The ALL samples were cytogenetically analyzed and screened for the presence of ETV6-RUNX1, TCF3-PBX1, BCR/ABL1. The presence of 11q23/MLL rearrangements were evaluated as described above.

Table 1.

Genetic characteristics of the 515 pediatric leukemias analyzed

Subtype	Total	IDH1 1		IDH2		IDH1/IDH2
		Negative	Positive	Negative	Positive
Acute Myeloid Leukemia
inv(16) CBFB-SMMHC	34	34		34
AML M7	17	17		17
11q23/MLL	42	42		41	1 (2.4%)	1 (2.4%)
Normal Karyotype	41	39	2 (4.9%)	40	2 (4.9%)2	4 (9.8%)
Other	55	54	1 (1.8%)	52	2 (3.6%)	3 (5.5%)
t(15;17) PML-RARA	7	7		7
t(8;21) RUNX1-RUNX1T1	31	31		31
Total	227	224	3 (1.3%)	222	5 (2.2%)	8 (3.5)
Acute Lymphoid Leukemia
Hyperdiploid (>50)	41	41		41
t(12;21) ETV6-RUNX1	49	49		49
t(1;19) TCF3-PBX1	17	17		17
11q23 MLL	25	25		25
t(9;22) BCR-ABL1	12	12		12
Hypodiploid	11	11		11
Normal Karyotype	10	9	1 (10.0%)	10		1 (10.0%)
Other	67	67		67
T-lineage	56	56		56
Total	288	287	1 (0.4%)	287		1 (0.4%)

¹A number of germ line polymorphisms (either previously reported or identified in both the tumor and the matched germ line samples) were identified in IDH1 and IDH2. These included in IDH1 V71I (4 AMLs and 15 ALLs) and G105G (35 AMLs and 24 ALLs), and in IDH2 K130Q (1 AML) and L143L (2 AMLs and 2 ALLs).

²One case had a normal karyotype, but contained a NUP98-NSD1 fusion detectable by RT-PCR.

Sequence analysis of IDH1 and IDH2

Mononuclear cells were purified from diagnostic bone marrow or blood, using standard methods. DNA was extracted according to standard protocols and whole genome amplified (Qiagen, Valencia, CA). Exons 4 of IDH1 and IDH2 were PCR amplified using the oligonucleotide primers:

• IDH1F: 5'-GTACTCAGAGCCTTCGCTTTCTGC

• IDH1R: 5'-GCCAACATGACTTACTTGATCCCC

• IDH2F: 5'-CTGGCTGTGTTGTTGCTTGG

• IDH2R 5'-ACTGCAGAGACAAGAGGATGGC

and the BD Advantage™ 2 PCR Enzyme System (Clonetech, Mountain View, CA). All primers were tailed with M13F and M13R sequences, which were used for sequence analysis. The PCR reaction products were purified using the Wizard SV-96 purification system according to the manufacturer's instructions (Promega, Madison, WI). Sequence analysis was performed using the Big Dye Terminator (v.3.1) chemistry on 3730xl DNA analyzers (Applied Biosystems, Foster City, CA). All mutations were verified by sequencing unamplified patient DNA. To verify that the mutations were somatic, matched normal DNA from 4/8 cases (IDH1 R132C (N=1), IDH2 R140Q (N=2), and IDH2 R140W (N=1)) were sequenced. For patients with IDH1/IDH2 mutations, the mutational status of FLT3, NPM1, and TP53, were determined as described.^{16, 17} Mutation detection was analyzed using SNPdetector¹⁸ and manual inspection of the sequence reads in Vector NTI Advance 11.0 (Life Technologies, Carlsbad, CA).

Protein Modeling

To gain insights into the mechanisms by which IDH2 mutations might alter protein function, we examined a structural model of a human IDH2-isocitrate complex. Models for the human protein were made by threading using Molrep.¹⁹ The porcine heart mitochondrial version of IDH2 (ILWD.pdb), which shares 97% sequence identity with human IDH2, was used as the starting model. The structural model was visualized using PyMOL Molecular Graphics System.²⁰

Site Directed Mutagenesis

Human wildtype (WT) cDNA clones for IDH1 (NM_005896.2) and IDH2 (NM_002168.2) in pCMV6 Entry vector were obtained from Origene (Origene, Rockville, MD). Mutations in IDH1 (R132H, R132C) and IDH2 (R140Q, R140W) were introduced using the Stratagene QuickChange Site-Directed Mutagenesis Kit (Agilent Technologies, Santa Clara, CA) according to the manufacturer's instructions. All mutations were verified by sequencing.

Protein expression and purification

Full length wild-type and mutant IDH1 and IDH2 clones were subcloned into pGEX4T-1 (GE Healthcare, Pittsburg, PA) containing a GST tag, or pRSF-Duet (Novagen, EMD4 Biosciences, Gibbstown, NJ), containing a His tag. The pRSF-Duet was also modified to include an N-terminal maltose binding protein (MBP) tag that enhances solubility of the expressed proteins. All cloning constructs were verified by Sanger sequencing. Protein expression was performed in Stratagene BL210-Gold(DE3) cells (Agilent Technologies). To generate heterodimeric proteins, wild-type or wild-type and mutant clones were expressed from different vectors in BL21-Gold(DE3) cells and the proteins purified according to standard methods. Briefly, transformed bacteria were grown in LB cultures at 37°C and protein expression was induced by adding 0.6mM Isopropyl β-D-1-thiogalactopyranoside (IPTG). Following induction the cultures were grown overnight at reduced temperature (16°C for IDH1 containing cultures and 25°C for IDH2 containing cultures). The bacteria were then harvested by centrifugation, resuspended in 1× phosphate buffered saline (PBS), containing 5mM β-mercaptoethanol (BME) and 2.5 mM phenylmethanesulfonylfluoride (PMSF), and sonicated on ice for 8 rounds, 10 seconds each, or lysed using an Avestin C-5 Emulsiflex homogenizer (Avestin Inc, Canada). Proteins were recovered by affinity chromatography using His-select resin (Sigma, St Louis, MO), or for the isolation of heterodimeric proteins, His purification followed by affinity purification using GS4B resin (GE Healthcare). The His-MBP tagged proteins were eluted using 1×PBS, 5mM BME, 250mM Imidazole, whereas the GST-tagged proteins were eluted using 50mM Tris pH 8.0, 200 mM NaCl, 5mM BME, 10mM reduced Glutathione. All His-tagged proteins were dialyzed against 1×PBS containing 5mM BME.

Proteins were further purified by Fast Protein Liquid Chromatography (FPLC) (GE healthcare) using a Superdex 200 10/300 column. Fractions were eluted in 50mM Tris-HCl pH 7.4, 200mM NaCl, and 5mM BME. Fractions corresponding to the correct protein size were concentrated if needed and subsequently frozen in aliquots and analyzed for enzymatic activity as described below. Protein concentration was assessed using the Nanodrop 1000 (Thermo Scientific, Wilmington, DE) in 6M Guanidine with the extinction coefficients determined using the ExPASy Proteomics server (au.expasy.org/tools/protparam-doc). The IDH2 R140W copurified with bacterial contaminants determined to be the HTPG chaperone and the bifunctional polymoxin resistance protein, ARNA, as determined by mass spectrometry. The presence of this copurifying protein contaminant was taken into account when calculating the enzymatic activity of the IDH2 proteins.

Enzymatic assays

Isocitrate dehydrogenase activity was determined by measuring the reduction of NADP⁺ to NADPH at 340nm using a Beckman 640 Spectrophotometer (Beckman Coulter, Brea, CA) in the presence of 100uM NADP⁺, 2mM MnCl₂, 100mM Tris-HCl pH 7.4, and either 30uM or 128uM Isocitrate (all chemicals purchased from Sigma-Aldrich, St. Louis, MO). Absorbance was recorded at 10 second intervals for at least 60 seconds. The conversion of α-KG to 2-HG was measured in 100uM β-NADPH, 1.4mM MnCl₂, 100mM Tris-HCl pH 7.4, and 0.5mM α-KG. The consumption of NADPH was recorded at 30 second intervals for at least 10 minutes (min). All reagents were dissolved in 100mM Tris-HCl pH 7.4 and pH was corrected to 7.4 as necessary. All assays were performed at room temperature in triplicate using 75–300ng or 1.5–4ug of protein/reaction for the NADPH production and NADPH consumption assays, respectively. As a blank, changes in absorbance at 340 nm were measured in the presence of all reagents except for substrate, isocitrate or α-KG as appropriate. Enzyme activity (umol min⁻¹ mg⁻¹) was calculated using the extinction coefficient of 6.22 mM⁻¹ cm⁻¹for NADPH.

Metabolite Extraction

Viably frozen AML cells were thawed and 2×10⁶ cells were transferred to methanol (80:20 methanol:cell suspension) on dry ice, incubated on dry ice for 15 min, centrifuged at 4°C for 20 min at 14000g, and the supernatant was subsequently transferred to a pre-chilled tube and kept on dry ice until analysis. Samples were dried under nitrogen, dissolved in 200 μL of purified water, sonicated until dissolved, and then transferred to a 96-well plate for mass spectrometer analysis.

LC/MS/MS Analysis

Analysis of 2-HG was performed on an Acquity UPLC system (Waters Corporation, Milford, MA) containing a photodiode array detector and a TQD mass spectrometer. 20μL of sample was injected onto a UPLC BEH C18 1.7μm, 2.1 × 50 mm column (Waters Corporation, Milford, MA) maintained at 40°C. Data were acquired using Masslynx v.4.1. A gradient program was used starting at 98% A (10mM tributylammonium (TBA) (Sigma-Aldrich), 15mM acetic acid in 97% water/3% MeOH (Sigma-Aldrich), changing to 95% B (MeOH) over 2.8 min, holding for 0.1 min, and returning to 98% A over 0.1 min. The total flow rate was 1.0 mL/min. The mass spectrometer was operated in negative-ion mode using electrospray ionization. Multiple reaction monitoring (MRM) mode was used with parameters optimized based on injections of a 2-HG (Sigma-Aldrich) standard, monitoring the 147 → 85 transition. 2-HG was quantified by comparison of MRM peak areas to those for 2-HG standards at known concentrations. The limit-of-detection was determined by diluting a 2-HG standard until the signal-to-noise ratio of the MRM trace was 3:1.

Statistical Methods

The association of IDH mutations with genetic subtype was examined using a Monte Carlo approximation to the exact chi-square test based on 10,000 permutations. Overall survival (OS) was defined as the time elapsed from study enrollment or diagnosis to death with living patients censored at the date of last follow-up. Event-free survival (EFS) was defined as the time elapsed from study enrollment or diagnosis to relapse, failure to induce remission, death, second malignancy, or withdrawal for other cause. Subjects not experiencing any of those events were censored at the date of last follow-up. The Kaplan-Meier method was used to estimate overall and event-free survival. The log-rank test was used to compare event-free and overall survival according to IDH1/IDH2 mutation status. No adjustments for multiple testing were performed in this exploratory analysis.

Results

IDH1/IDH2 mutation analysis

To date, all reported cancer-associated non-synonymous mutations in IDH1/IDH2 have localized to the 4^th exon of these genes and resulted in alterations of amino acids that are directly involved in substrate recognition.⁶ We therefore limited our sequence analysis to exon 4 of both IDH1 and IDH2 (Figure 1a). Sequence analysis was performed on 515 diagnostic samples from pediatric patients with AML (N=227) or ALL (N=288). Samples were selected to provide a representation of the known genetic subtypes of these pediatric leukemias (Table 1), and were restricted to samples that contained greater than 60% leukemic blasts to ensure our ability to detect heterozygous mutations in IDH1/IHD2.

Figure 1.

Genomic structure of IDH1 and IDH2 and structural model of the location of R140 and R172 in the substrate-binding pocket of IDH2. (a) Illustration of the exon/intron structure of IDH1 and IDH2. The location within exon 4 of codon R132 in IDH1 and codon R140 in IDH2 are marked by arrows, and the surrounding nucleotide sequence and encoded amino acids are highlighted. Codon R132 in IDH1 and the homologous residue R172 in IDH2 are shown in red and codon 140 in IDH2 is shown in blue. (b) Model of the positions of the R140 and R172 amino acids in the IDH2 substrate binding pocket. The two protomers in the IDH2 homodimer are illustrated in green and purple, with the nitrogen atoms in amino acids R140 and R172 shown in blue. The bound substrate, isocitrate, is depicted with carbon atoms as yellow sticks and oxygen atoms in red. The manganese ion is shown as a grey sphere. Salt-bridges and hydrogen bonds are shown as dashed lines.

Sequence analysis of the pediatric AMLs revealed IDH1/IDH2 non-synonymous mutations in 8/227 (3.5%) patients (Table 1). All IDH1/IDH2 mutations were heterozygous, consistent with retention of a wild type allele. Moreover, no patient had mutations in both genes, suggesting that mutations in these genes are mutually exclusive. Sequence analysis of matched remission bone marrow sample from 4/8 patients demonstrated that the IDH1/IDH2 mutations were somatic in nature. All three somatic IDH1 mutations affected the previously reported hotspot, R132, resulting in two different amino acid substitutions, R132C (N=2) and R132H (N=1). In contrast, none of the five somatic mutations affecting IDH2 occurred in codon R172, but instead were located in codon R140, resulting in two different amino acid substitutions, R140Q (N=4) and R140W (N=1) (See Supplementary Figure 1a).

In contrast to AML, IDH1/IDH2 mutations were exceedingly rare in pediatric ALL. Only a single non-synonymous somatic IDH1 mutation, R132C, was detected in our analysis of 288 diagnostic patient samples (Table 1). Interestingly, the ALL with the IDH1 mutation had a normal karyotype and aberrant myeloid antigen expression of CD13 and CD33.

Correlation of IDH1/IDH2 mutation with laboratory and clinical features of the AMLs

Seven of the eight patients with IDH1/IDH2 mutations had leukemic blasts with either a normal karyotype (N=4, for an overall frequency of IDH1/IDH2 mutation of 9.8% (4/41) in normal karyotype pediatric AML), or were classified as having miscellaneous genetic changes (N=3, for an overall frequency of 5.5% (3/55) in this group designated as other in Table 1). One normal karyotype AML case contained a cryptic translocation encoding a NUP98-NSD1 chimeric mRNA (Table 2). Only 1/8 cases with an IDH1/IDH2 mutation contained a recurrent AML-associated translocation [t(11;19)(q23;p13.3), MLL-MLLT1(ENL)] (Table 2).

Table 2.

Genetic characteristics of the AMLs with IDH1/IDH2 mutations

Amino Acid change	Mutation1	IDH1/IDH2	Karyotype	FAB	Molecular rearrangement	FLT3 ITD/D835	TP53	NPM1 2
R132H	c. 395G>A, CGT to CAT	IDH1	46,XY[22]	M1		wt/wt	wt	Type B
R132C	c. 394C>T, CGT to TGT	IDH1	46,XX[20]	M1		wt/wt	wt	wt
R132C	c. 394C>T, CGT to TGT	IDH1	48,XY,+8,+8[7]/46,XY[5]	M0		wt/wt	wt	wt
R140Q	c. 419C>A, CGG to CAG	IDH2	46,XY[20]	M1	NUP98-NSD1	ITD+/wt	wt	wt
R140Q	c. 419C>A, CGG to CAG	IDH2	47,XY,+4[18]/46,XY[2]	M2		wt/wt	wt	Type A
R140Q	c. 419C>A, CGG to CAG	IDH2	47,XX,+14 [16]	M4		ITD+/wt	wt	Type A
R140Q	c. 419C>A, CGG to CAG	IDH2	46,XY[20]	M4		ITD+/wt	wt	wt
R140W	c. 418G>A, CGG to TGG	IDH2	46,XX,t(H;19)(q23;pl3.1)[20]	M1	MLL-ENL	wt/wt	wt	wt

¹Nucleotide numbering is based on NM_005896.2 for IDH1 and NM_002168.2 for IDH2.

²NPM1 mutations: Type B = insCATG, Type A = insTCT

AMLs with a normal karyotype frequently contain mutations in FLT3 and NPM1, and the mutational status of these genes provides prognostic information.^{21, 22} In addition, mutations in TP53 occur at a high frequency in cases of GBM that contain IDH1/IDH2 mutations.^{5, 6} We therefore analyzed these genes in the AMLs that contained IDH1/IDH2 mutations. Interestingly, while none harbored mutations in TP53, 3/4 cases with the IDH2 R140Q mutation also had FLT3 internal tandem duplications (ITD), one of which also had a mutation in NPM1. None of the IDH1/IDH2 positive cases had the FLT3 D835 activating point mutation. Two additional cases harbored NPM1 mutations, one with the IDH1 R132H and one with the IDH2 R140Q (Table 2).

We next determine if IDH1/IDH2 mutations had any prognostic relevance in pediatric AML. This analysis failed to detect any significant association of the mutational status of these genes, either independently or together, with overall survival or event free survival although the power of this analysis is limited by the very small number of IDH1/IDH2 mutation identified (see Supplementary Figures 2a–f).

Structural modeling of IDH2 R140 mutation

To gain insight in how the R140 mutations of IDH2 might alter protein function, we modeled human IDH2 using the reported structure of pig IDH2 bound to Mn²⁺ and isocitrate.²³ Human and pig IDH2 show over 97% amino acid identity, including all active site residues. IDH2, like IDH1, forms a homodimer, with two active sites (Supplementary Figure 1b). Each active site is formed by amino acids from both monomers within the homodimer, and binds a Mn²⁺ ion required for catalysis and a molecule of the substrate isocitrate. Figure 1b illustrates a single active site and predicts a direct interaction of R140 with the O1-carbonyl and O6-hydroxyl groups of isocitrate. The identified R140 mutations are predicted to alter the substrate and presumably product binding properties, thereby altering enzymatic activity.

Biochemical characterization of the IDH2 R140 mutation

The cancer-derived IDH1 R132 and IDH2 R172 mutations have been shown to lead to a loss of the enzyme's ability to catalyze the oxidative decarboxylation of isocitrate to α-ketoglutarate (α-KG), and a gain-of-function to catalyze the NADPH-dependent reduction of α–KG to 2-hydroxyglutarate (2-HG).^{6–9, 24} To determine if the leukemia-derived R140 IDH2 mutations also induces the same change in activity, we analyzed the in vitro enzymatic activity of bacterial expressed recombinant proteins. Full length cDNAs for wild-type and mutant IDH1 (R132H and R132C) and IDH2 (R140Q and R140W) were cloned into both His- and GST-tagged bacterial expression vectors. E coli was then co-transduced with both vectors expressing either wild-type or mutant proteins (generating homodimers), or with one vector expressing wild-type and the other expressing a mutant proteins (generating heterodimers) (Supplementary Figure 3).

To assess the NADP⁺-dependent oxidative decarboxylation of isocitrate to α-KG recombinant proteins were analyzed with limiting (30uM, Figures 2a–b) and excess (128uM, Supplementary Figures 4a–b) isocitrate in the presence of 100uM of NADP⁺ and 2mM MnCl₂, and the reduction of NADP⁺ to NADPH was measured. As previously reported,^{6, 7, 9, 24} homodimers of the R132 IDH1 mutations resulted in markedly reduced activity compared to wild-type IDH1 homodimers (Figure 2a and Supplementary Figure 4a). Similarly, wild-type:mutant IDH1 heterodimers exhibited only 25% of the activity shown by the wild-type enzyme (Figure 2a and Supplementary Figure 4a).²⁴ Nearly identical results were obtained using the R140 leukemia-derived IDH2 mutations. Minimal enzymatic activity was observed for homodimers of R140Q and R140W (Figure 2b and Supplementary Figure 4b). Moreover, wild-type:R140Q heterodimers exhibited a 60% reduction in activity compared to homodimers of the wild-type IDH2 protein (Figure 2b and Supplementary Figure 4b). Taken together, these data demonstrate that the R140Q/W mutations of IDH2 result in a marked impairment of the enzymes ability to convert isocitrate to α-KG.

Figure 2.

Enzymatic analysis of the IDH1 and IDH2 mutant proteins. The activity of recombinant IDH1 and IDH2 proteins to catalyze the NADP⁺-dependent oxidative decarboxylation of isocitrate to α-KG using 30uM Isocitrate and 100uM NADP are shown in panels a and b, respectively, and their ability to catalyze the NADPH-dependent reduction of α–KG to 2-HG using 0.5 mM α-KG and 100uM NADPH are shown in c and d, respectively. All graphs are based on triplicate measurements with the mean ± standard deviations expressed as a relative level compared to WT:WT homodimers, with the latter set as 100%. The SD in panel c and d are < 0.03 and are thus below the resolution of the figure. (e) The intracellular level of 2-HG was measured by liquid chromatography/mass spectrometry in pediatric AML cells from primary diagnostic bone marrow samples. The data for mutant IDH1/IDH2 includes two leukemia samples containing IDH1 mutations (one with R132H and one with R132C, ▴), and two containing the R140Q IDH2 mutations (◆). The wild-type IDH1/IDH2 data was generated using two pediatric AML samples that lacked mutations in either IDH1 or IDH2 (●).

To assess the ability of these enzymes to catalyze the NADPH-dependent reduction of α–KG to 2-HG, recombinant proteins were incubated with 500uM α-KG in the presence of 100uM of NADPH and 1.4mM MnCl₂, and the reduction of NADPH to NADP⁺ was measured. Like the IDH1 mutations, homodimers of the R140Q and R140W IDH2 mutants showed an enhanced ability to convert α–KG to 2-HG compared to wild-type homodimers, although the increase in activity was less than that observed with IDH1 mutations (Figure 2c–d). Thus like IDH1 mutants, the R140Q/W IDH2 mutants have gained the neomorphic activity to reduce α-KG to 2-HG

Primary pediatric AML with IDH1/IDH2 mutations show high levels of 2-HG

Primary glioblastoma multiforme cells that contain an IDH1 R132 mutation and primary adult AML cells containing either R132 IDH1 or R172 IDH2 mutations have been shown to have markedly increased intracellular levels of 2-HG when compared to tumors that lack these mutations.^7–9 To assess if pediatric AML cells containing a R140 IDH2 mutation also have increased levels of 2-HG, we determined the levels of 2-HG in six primary pediatric AML samples, including two with WT IDH1/IDH2, two with mutant IDH1 (1 with R132H, 1 with R132C), and two with mutant IDH2 (R140Q) using LC/MS/MS mass spectrometry (Figure 2e). This analysis showed a 30.8-fold average increase in 2-HG levels in the leukemia cells contain IDH1 R132 or IDH2 R140 mutations as compared to leukemia cells with wild-type IDH1/IDH2.

Discussion

The data we present demonstrate that mutations in isocitrate dehydrogenase 1 (IDH1) and 2 (IDH2) are exceedingly rare in pediatric ALL, but are more common in pediatric AMLs, occurring in 3.5% of cases overall, and in 9.8% of pediatric AMLs with a normal karyotype. Like in adult AML and GBM, the identified IDH1 mutations occurred exclusively in codon 132 resulting in replacement of arginine with either cysteine or histidine. By contrast, the mutations in IDH2 did not affect the homologous residue but instead altered codon 140, resulting in replacement of an arginine with either glutamine or tryptophan. Moreover, the R140 IDH2 mutations were more common than IDH1 mutations, a finding consistent with recent reports on adult AML.^{8, 25–29} Using biochemical analysis of recombinant proteins, we demonstrated, for the first time, that like the R132 IDH1 mutations, the R140Q/W mutations lead to inhibition of the normal ability of the enzyme to catalyze the conversion of isocitrate to α-KG, and the acquisition of the neomorphic activity to reduce α-KG to 2-HG. Consistent with this in vitro data, leukemic AML blasts with mutant IDH2 or IDH1 contained markedly elevated intracellular levels of 2-HG. These data suggest that the identified IDH1/IDH2 mutations play a direct role in leukemogenesis, at least in part, through the generation of increased levels of the metabolite 2-HG.

This report provides the most detailed analysis to date on the frequency of IDH1/IDH2 mutations in pediatric ALL and AML. In contrast to several prior studies that either failed to identify IDH1 mutations in pediatric AML,^{14, 15} or identified only a single IDH2 mutation in 115 primary cases of pediatric AML,¹⁵ our data clearly demonstrate that like in adult AML, IDH1/IDH2 mutations occur in pediatric AML with the highest frequency observed in cases with normal cytogenetics, although the overall frequency is lower than that reported in adult cases.^{4, 8, 14, 25–27, 30–32} Moreover, mutations in IDH1 and IDH2 were mutually exclusive. An important observation from our data and published studies was the prevalence of IDH2 over IDH1 mutations in both pediatric and adult AML. This is in stark contrast to the prevalence of IDH1 over IDH2 mutations in secondary GBM. The data suggest the possibility of cell context specific effects of IDH1 and IDH2 mutations between leukemic blasts and malignant glial cells.

The prognostic impact of IDH1/IDH2 mutations in AML has been controversial. Some studies have reported mutations in IDH1 to confer a worse prognostic impact in adult AMLs with a normal karyotype,^{26, 27} or in AMLs lacking mutations in FLT3 and NPM1,^{25, 32} whereas others have failed to detect any prognostic significance.³¹ In our pediatric cohort, we could not demonstrate any significant statistical association of IDH1/IDH2 mutations to overall survival or event free survival, although the power of this analysis is influenced by the low overall frequency of IDH1/IDH2 mutations.

An important observation from our study was the co-occurrence of FLT3 activating mutation (FLT3-ITD) in 3/5 patients with R140 IDH2 mutations. Although the co-occurrence of these two lesions remains a rare event in pediatric AML, this observation raises the possibility that these two mutations directly cooperate in the process of leukemogenesis. IDH1/IDH2 mutations have been suggested to be early lesions in the development of GBM.⁶ Similarly, IDH1/IDH2 mutations have been identified in cases of myeloproliferative disease and myelodysplasia that progress to AML.^{33, 34} By contrast, the FLT3 ITD mutation is thought to occur late in leukemogenesis, often being present in only a subclone of the leukemic cells.³⁵ These observations raise the interesting possibility that IDH2 mutations occur early in the development of some cases of pediatric AML and their presence renders the cells more susceptible to the oncogenic activities of the FLT3 activating mutations. This hypothesis should be easily testable using murine leukemia model systems. Moreover, because the mechanism by which IDH1/IDH2 mutations contribute to oncogenesis has not been fully elucidated, the coexpression of IDH2 mutations with FLT3 activating mutations in hematopoietic stem cells or lineage committed progenitors may represent a highly tractable system in which to explore IDH1/IDH2 specific pathways.