Search tips
Search criteria 


Logo of nihpaAbout Author manuscriptsSubmit a manuscriptNIH Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
Sci Transl Med. Author manuscript; available in PMC Sep 13, 2013.
Published in final edited form as:
PMCID: PMC3772177
Magnetic Resonance of 2-Hydroxyglutarate in IDH1-Mutated Low-Grade Gliomas
Adam Elkhaled,#1 Llewellyn E. Jalbert,#2 Joanna J. Phillips,3,4 Hikari A. I. Yoshihara,1 Rupa Parvataneni,1,4 Radhika Srinivasan,1 Gabriela Bourne,1 Mitchel S. Berger,4 Susan M. Chang,4 Soonmee Cha,1 and Sarah J. Nelson1,2
1Department of Radiology and Biomedical Imaging, University of California, San Francisco, San Francisco, CA 94158, USA.
2Department of Bioengineering and Therapeutic Sciences, University of California, San Francisco, San Francisco, CA 94158, USA.
3Department of Pathology, University of California, San Francisco, San Francisco, CA 94158, USA.
4Department of Neurological Surgery, University of California, San Francisco, San Francisco, CA 94158, USA.
#Contributed equally.
To whom correspondence should be addressed. Sarah.Nelson/at/
Recent studies have indicated that a significant survival advantage is conferred to patients with gliomas whose lesions harbor mutations in the genes isocitrate dehydrogenase 1 and 2 (IDH1/2). IDH1/2 mutations result in aberrant enzymatic production of the potential oncometabolite D-2-hydroxyglutarate (2HG). Here, we report on the ex vivo detection of 2HG in IDH1-mutated tissue samples from patients with recurrent low-grade gliomas using the nuclear magnetic resonance technique of proton high-resolution magic angle spinning spectroscopy. Relative 2HG levels from pathologically confirmed mutant IDH1 tissues correlated with levels of other ex vivo metabolites and histopathology parameters associated with increases in mitotic activity, relative tumor content, and cellularity. Ex vivo spectroscopic measurements of choline-containing species and in vivo magnetic resonance measurements of diffusion parameters were also correlated with 2HG levels. These data provide extensive characterization of mutant IDH1 lesions while confirming the potential diagnostic value of 2HG as a surrogate marker of patient survival. Such information may augment the ability of clinicians to monitor therapeutic response and provide criteria for stratifying patients to specific treatment regimens.
Infiltrating gliomas are heterogeneous tumors of the central nervous system that include astrocytomas, oligodendrogliomas, and mixed oligoastrocytomas. Overall survival for these diseases can vary significantly depending on the grade of malignancy and histological characteristics, as defined by the World Health Organization (WHO) (1). Although patients diagnosed with infiltrating low-grade (grade II) gliomas generally live much longer than their high-grade counterparts (grades III and IV), there is substantial heterogeneity in outcome, even for patients whose lesions share the same initial diagnosis. Because of the broad range of survival times and limited number of markers that can inform prognosis, there are several challenges for the clinical management of gliomas (1, 2). In 2009, there was a shift in the prognostic paradigm for these lesions, when more than 70% of all patients with low-grade gliomas were discovered to have mutations in the isocitrate dehydrogenase 1 and 2 (IDH1/2) genes, which have been associated with prolonged survival (36).
IDH1 and IDH2 are NADP+ (oxidized form of nicotinamide adenine dinucleotide phosphate)–dependent oxidoreductases that catalyze the conversion of isocitrate to α-ketoglutarate. IDH1 is found in the cytoplasm and in peroxisomes, where it plays a role in lipid metabolism. IDH2 is localized to the mitochondria and has been implicated in protecting the cell from oxidative damage (79). Missense mutations of IDH1/2 genes were found to result in arginine being substituted by other amino acids at the substrate binding sites of IDH1/2 enzymes. Recent in vitro studies have revealed that mutation of arginine 132 (R132) in IDH1 and of arginine 172 (R172) in IDH2 prevents enzymatic oxidative decarboxylation of isocitrate and confers a new ability to convert α-ketoglutarate to d-2-hydroxyglutarate (2HG) (10). The arginine mutations result in accumulation of 2HG and a vast reduction in normal IDH1/2 enzymatic activity. Although the function of both IDH-mutant enzymes and 2HG in gliomagenesis remains unclear, the improved prognosis associated with IDH1/2 mutations suggests that the presence of 2HG may be of prognostic value as a surrogate for favorable genotypes. Given this potential, developing noninvasive in vivo methods for detecting 2HG in patients and evaluating other descriptive parameters of IDH-mutated lesions are of clinical interest.
The objective of this study was to characterize the metabolic profiles of recurrent, infiltrating low-grade gliomas using ex vivo proton high-resolution magic angle spinning (1H HR-MAS) nuclear magnetic resonance (NMR) spectroscopy of tissue samples acquired by imageguided surgery. Crucial to this aim was determining whether levels of 2HG were sufficient for detection in IDH-mutated gliomas and, further, how the concentration of 2HG varied among gliomas that underwent a transition to a higher grade at the time of tumor recurrence. We hypothesized that if 2HG was detected from 1H HR-MAS spectra, levels of the suspected oncometabolite would be significantly higher in tissue samples that had manifest IDH1 mutations. As an exploratory analysis, we also sought to evaluate whether histopathological or preoperative in vivo MR imaging (MRI) parameters correlated with quantified levels of 2HG. Such information might be valuable for translational efforts geared toward noninvasively monitoring 2HG and predicting survival in patients with low-grade gliomas.
Characterization of patient population
Our institutionally approved study comprised 60 patients who had previously been diagnosed with WHO grade II glioma and were presenting for surgical resection owing to suspected disease recurrence. Presurgical in vivo MR examinations enabled the planning of targeted biopsies for sampling tissue from patient lesions. Imaging parameters derived from postprocessed diffusion-weighted imaging and spectroscopic data helped guide the designation of small, putatively defined tumor regions (5-mm-diameter spheres) on surgical navigation software. The criteria for planning image-guided biopsies were based on low apparent diffusion coefficients (ADCs) that represent restricted diffusion associated with tumor cellularity and elevated choline–to–N-acetylaspartate (NAA) indices (CNIs) (Fig. 1, A and B), which provide measures of metabolic abnormality. Ex vivo HR-MAS spectra are presented to demonstrate the improved spectral resolution over conventional MR spectroscopy through the separation of individual choline-containing species that contribute to the in vivo total choline (tCHO) peak (Fig. 1C). Histological analysis of tissue samples collected from patients revealed that many had lesions that converted to a higher grade of malignancy at the time of recurrence: 25 patients had converted from WHO grade II to grade III and 6 patients converted to grade IV (Table 1). Eight patients (23 tissue samples) were excluded from the statistical analysis because their samples did not contain tumor, were too small for evaluation, or were contaminated by sterilizing agents used in the operating room. A total of 104 tissue samples from 52 patients remained for evaluation (Table 1). Complete patient data for all 60 subjects, including reasons for exclusion, are provided in table S1.
Fig. 1
Fig. 1
Designating tissue targets from presurgical in vivo imaging. (A) ADC map derived from diffusion-weighted imaging (DWI) shows a hypointense region of low diffusion relative to normal-appearing white matter that was designated as a target for tissue sampling (more ...)
Table 1
Table 1
Glioma patient population by tumor grade and histological subtype. Eight patients and 23 tissue samples were excluded owing to a lack of tumor cells observed by histological analysis, an insufficient amount of tissue in a sample for accurate HR-MAS analysis, (more ...)
IDH1 immunohistochemistry
The IDH1 status of paired tissue samples from each patient (n = 52) was assessed to compare wild-type against mutant IDH1 spectral profiles (11). Antibody staining indicated that 38 patients had gliomas with histidine-mutated (R132H) IDH1 enzymes. Tissue samples that stained positive for the IDH1R132H mutation, as seen colored in brown (Fig. 2A), were shown to visually contrast with those that tested negative for the modified enzymes (Fig. 2B). An electropherogram of sequenced genomic DNA (gDNA) from an IDH1R132H-mutant tissue sample illustrates the most common heterozygous mutation of residue R132 (Fig. 2C). Immunohistochemistry results were consistent across all tissue samples from a particular patient in 94.4% of cases (table S1A). Because the IDH1R132H antibody is only selective for histidine mutations at residue R132 of IDH1, it was expected that samples with other IDH1 and IDH2 mutations would not be detected. Genetic sequencing of a grade II sample subset (n = 21) revealed that two of the patients testing negative for IDH1-mutant enzymes by antibody staining were, in fact, carrying the serine mutation (IDH1R132S) and nine carried the histidine mutation (Table 1B), which corroborated the antibody staining (Fig. 2). Overall, 40 patients were classified as being positive for IDH1 mutations (IDH1+) and 12 as IDH1-negative (IDH1−) (Table 2B). Furthermore, 26 of the 31 (83.9%) patients whose lesions had converted to a higher grade at the time of recurrence were classified as IDH1+ compared to 14 of the 21 (66.7%) whose lesions had remained grade II.
Fig. 2
Fig. 2
IDH1R132H immunostaining and direct sequencing of IDH1 gDNA. (A) IDH1R132H antibody staining (brown) on a 2HG+ WHO grade III astrocytoma tissue sample, which was representative of 44 tissue samples. The corresponding 1D and 2D spectra for this sample (more ...)
Table 2
Table 2
Classification of 2HG presence and IDH1 mutation status. (A) The presence of 2HG in HR-MAS spectra was assessed visually from the 1D spectrum and confirmed by the presence of signature cross peaks in the 2D TOCSY spectrum. IDH1 mutation analysis was performed (more ...)
Magnetic resonance spectroscopy of a 2HG phantom
The one-dimensional (1D) NMR spectrum of a buffered stock solution of 2HG features four signature multiplets arising from five alkyl hydrogens: one α, two β, and two γ (fig. S1). Although 2HG shares the same AMNPQ spin system as glutamate (Glu), its 2-hydroxyl (2-OH) is more electron-withdrawing than Glu’s 2-amino (2-NH2) substituent, causing the α-proton to resonate further downfield at 4.05 ppm (parts per million) as a doublet of doublets. The chirality of 2HG combined with the distinct chemical environment experienced by each of the resonating protons gives rise to complex vicinal and geminal coupling, which is reflected in the elaborate splitting patterns of the spectra. The β-protons, for example, display highly coupled resonances that are seen to have chemical shifts of 2.01 and 1.85 ppm, respectively. Situated furthest from the chiral center of the second carbon, γ-protons produce overlapping resonances that appear as a single 13-peak multiplet centered at 2.28 ppm. J-resolved spectroscopy experiments (Supplementary Methods) allowed for an approximation of the interproton spin-spin coupling constants (2,3J), which was optimized using spectral simulation software (fig. S1). Separation of the γ resonances (Supplementary Methods) was performed at ultrahigh magnetic field strength to demonstrate the γ-proton’s mutual coupling with each other and β-protons in greater detail (fig. S2).
Identification of 2HG ex vivo in tissue samples
1H HR-MAS spectra showed that most of the evaluable tissue samples contained detectable levels of 2HG (Table 2A). Of the 52 patients, 14 had spectra with inadequate resolution and signal-to-noise ratio (SNR) for identifying the signature peaks of 2HG, by either visual inspection or curve-fitting algorithms. This left tissue samples from 38 patients with spectra that could be analyzed. Of these patients, 33 (86.8%) were visually classified as 2HG+ and 5 as 2HG− by referencing discriminating resonances of 2HG from 1D and 2D spectra; additionally, 29 2HG+ patients had levels of 2HG that were quantified using curve-fitting algorithms. At a sample level, this translated to 58 2HG+ and 8 2HG− tissue samples (Table 2A). The fact that there were so few 2HG− samples precluded a statistical analysis between 2HG+ and 2HG− cohorts.
The T2-weighted Carr-Purcell-Meiboom-Gill (CPMG) (12) sequence was chosen for its ability to eliminate broad macromolecular signals, effectively creating a smooth spectral baseline that could resolve subtle metabolite features. CPMG acquisitions provided clear delineation of 2HG from neighboring metabolites γ-aminobutyric acid (GABA), glutamine (Gln), and Glu and were therefore used for all further analyses in this study. CPMG spectra from patients with contrasting mutant and wild-type IDH1 genotypes show the characteristic presence and absence of 2HG, respectively (Fig. 3, A and B). The blue traces represent the acquired data; red traces represent the spectral curve-fitting performed with the semiparametric high-resolution quantum estimation (HR-QUEST) algorithm. HR-QUEST used an iterative process of approximation to independently fit each of the chemical shifts of 2HG for quantification in the time domain. As shown in Fig. 3A, the 2HG γ-proton multiplet centered at 2.28 ppm offered the best means of visual 2HG recognition on account of its distinct splitting pattern and relative signal intensity, which was comparable to that of neighboring metabolites. Although the α- and β-proton multiplets are seen as relatively distinct in this particular example (Fig. 3A), they are more likely to be obscured by overlapping resonances in 1D spectra.
Fig. 3
Fig. 3
Ex vivo spectra from mutant and WT IDH1 tissue samples. (A and B) 1D CPMG spectrum from an IDH1-mutant WHO grade III astrocytoma (A) and WT IDH1 grade IV astrocytoma (B). For both (A) and (B), the blue trace represents acquired data and the red trace (more ...)
The 2D total correlation spectroscopy (TOCSY) (13) spectrum provided further confirmation of the presence of 2HG on the basis of three well-defined cross peaks located along a vertical column at F2/F1, [4.05/1.85 ppm], [4.05/2.01 ppm], and [4.05/2.28 ppm] (Fig. 3C). Each of these resonances corresponds to either the γ- or β-proton’s shared magnetization with the α-proton and together constitute a spectral feature that is separable from the resonance of myo-inositol (myo-I) on account of relative chemical shifts. Although the other expected cross peaks of 2HG were detectable between 1.85 and 2.28 ppm, potential interference from the resonances of GABA, Gln, and Glu in this region made the α-proton cross peaks at 4.05 ppm a more suitable form of identification for 2HG with the TOCSY experiment (Fig. 3C).
Concordance of IDH1 mutation status and presence of 2HG
There was an 86.4% concordance (51 of 59 tissue samples; 32 of 38 patients) between the presence of 2HG, as detected by 1H HR-MAS spectroscopy, and IDH1 mutation status, as determined by antibody staining and genetic sequencing (Fig. 2 and Table 2A). A breakdown of this concordance according to glioma grade is presented in Table 2B. Of the eight discordant tissue samples, one had a level of 2HG that fell below the threshold for visual spectroscopic detection, perhaps as a result of sectioning the tissue sample such that most of the tumor was contained in the portion analyzed for pathology. The remainder of discordant samples came from biopsies with insufficient tissue (<0.5 mg) for validation by genetic sequencing but may have harbored alternative IDH1 or IDH2 mutations (table S1). Two of the samples classified by sequencing as IDH1R132S that had not been identified by the IDH1R132H antibody staining were also found to contain detectable levels of 2HG. Although this was an expected outcome, because the IDH1R132H antibody is only specific for histidine mutations, it proves that MR methodologies are capable of identifying multiple mutations of the IDH1 gene.
Levels of 2HG among tumor grades
Relative abundance of 2HG in tissue samples suggests varied accumulation in gliomas of different grades. 2HG abundance was reported in relative levels (unitless) because of the T2 relaxation time dependence of the CPMG acquisition and metabolite-tissue interactions, which precluded absolute quantification. Mean 2HG levels were shown to increase across the 29 patients with spectra quantified for 2HG from grade II (0.36 ± 0.22, n = 8) to grade III (0.56 ± 0.38, n = 17) to grade IV (0.77 ± 0.20, n = 4) (Fig. 4A). When analyzed with an exact Wilcoxon rank-sum test, there was a statistically significant difference found between 2HG abundance in grade II and IV samples (P = 0.016) (Fig. 4A), whereas the broad range of values shown by tissue from grade III lesions prevented distinction from the other grades. After these levels were normalized by average cellularity (calculated as the number of cells per 200× field), the statistical relationship between grades II and IV no longer existed (P = 0.38) (Fig. 4B).
Fig. 4
Fig. 4
2HG levels according to WHO glioma grade. (A) Plot of relative 2HG levels quantified by HR-QUEST in relation to malignancy grades. Mean relative levels per patient (n = 29 total: 7, grade II; 18, grade III; 4, grade IV) were evaluated by an exact Wilcoxon (more ...)
Ex vivo correlations of 2HG
The levels of several metabolites that were quantified from 1H HR-MAS spectra of biopsy samples were found to correlate with 2HG abundance (table S2 and fig. S3). These were ex vivo metabolites commonly associated with tumor, including all of the choline-containing species: free choline (Cho), phosphocholine (PC), and glycerophosphocholine (GPC), as well as the combined measure of tCHO. As an in vivo marker of cell membrane turnover, tCHO has been used extensively to discern tumor regions with increased cellularity. Its elevation is currently one of the best spectroscopic markers of tumor presence, especially when evaluated alongside attenuated levels of NAA, which indicate diminished neuronal functioning due to neoplastic infiltration. Other brain metabolites related to malignancy, such as aspartate (Asp), GABA, threonine (Thr), hypotaurine (hypo-Tau), creatine and phosphocreatine (Cr, PCr), betaine (Bet), glycine (Gly), lactate (Lac), glutathione (GSH), phosphoethanolamine (PE), Glu, and Gln, were also seen to increase with 2HG levels.
These relationships reflect 2HG’s association with several aspects of tumor metabolism: Cr and PCr are commonly used as measures of bioenergetics, Lac is typically a product of anaerobic respiration under hypoxic conditions, GSH is an antioxidant that is thought to combat the extreme oxidative stress caused by reactive oxygen species, and Glu and Gln often signal changes in mitochondrial activity (14). Furthermore, the myo-I–to–tCHO index (MCI) was negatively correlated with levels of 2HG (fig. S3 and table S2). MCI serves as a novel in vivo parameter for evaluating the relative contributions of tumor and treatment-affected tissue, pathologically defined as gliosis (15). Therapeutically treated tissues, although often mimicking tumor on radiographic imaging, express higher levels of myo-I compared to tCHO and thus show a characteristic elevation in this index. The fact that 2HG levels correlate with MCI and the list of tumor-related metabolites helps confirm the association of 2HG with malignant tissue.
Tumor tissue histopathology and 2HG
There was a positive correlation observed between levels of 2HG and mitotic activity, as measured by the MIB-1 index (for evaluating the proportion of cells undergoing mitoses via immunostaining), relative tumor content [scored by the contribution of tumor to overall cellularity (“tumor score”)], and cellular density (Table 3, fig. S3, and table S2). 2HG levels were negatively correlated with the presence of delicate vasculature found in normal brain tissues using factor VIII staining of vessel lumen. Axonal disruption, as measured by neurofilament antibody staining, was not found to be linked to changes in 2HG concentration in tumor tissue.
Table 3
Table 3
2HG correlations with ex vivo metabolites, histopathology parameters, and in vivo diffusion parameters. Each of the study parameters correlated with 2HG is presented. This statistical method used a simple regression model and evaluated all of the tissue (more ...)
In vivo correlations of ADC with 2HG
To test whether MRI parameters were able to provide additional information regarding IDH-mutant lesions, we assessed their correlations with levels of 2HG quantified from HR-MAS spectra. By using the coordinates of each targeted biopsy acquired during surgery, it was possible to relate presurgical in vivo patient imaging to sampled tissue. ADCs derived from in vivo diffusion-weighted imaging represent the relative diffusivity of water through tissue. Restricted diffusion in the brain, corresponding to low intensities on ADC maps, is often associated with highly cellular regions of tumor (16). 2HG levels showed a negative correlation with ADC, as well as ADC normalized by normal-appearing white matter in the brain (Table 3 and fig. S3), indicating a regional association with tumor.
Also correlated with 2HG were the eigenvalues describing the magnitude of water diffusion along orthogonally related principal and secondary directions, denoted as λ1 and λ2, respectively (Table 3 and fig. S3). Because ADC is mathematically derived from constituent eigenvalues, this information provides further detail for characterizing how diffusion phenomena in tumor relate to 2HG. These were important parameters to evaluate, because diffusion imaging is routinely used by clinicians to assess treatment response and tumor progression.
A similar analysis was performed on metabolite levels acquired from in vivo spectroscopic measures of Cr, tCHO, NAA, and Lac + lipid; however, owing to the limited coverage volume of our spectroscopy, many of the tissue samples did not have associated metabolite data. Partial voluming with tissue surrounding the primary lesion further limited spectroscopic correlations on the basis of lower in vivo resolution. Despite these limitations, the higher-resolution ex vivo spectroscopy suggests that in vivo tCHO, Cr, and Lac are likely to be elevated in IDH-mutant lesions and to correlate with levels of 2HG, along with other metabolites found significant from ex vivo analyses (Table 3).
This study has demonstrated the feasibility of detecting 2HG and quantifying its relative abundance in IDH1-mutated gliomas using 1H HR-MAS spectroscopy. There was a strong concordance established between IDH1-mutant tissue samples acquired from patients and the presence of 2HG. These data corroborate the previously identified relationship between IDH1 mutation and the aberrant production of this metabolite that was validated using liquid chromatography–mass spectrometry analysis (10). Because our study design constrained the acquisition of targeted biopsies to relatively small regions with homogeneous in vivo imaging characteristics, the sampled tissue was in some cases not ideal for spectroscopic analysis. Despite this limitation, it was possible to obtain an accurate assessment of the presence and absence of 2HG for most of the patients evaluated. When present, the levels of 2HG were on par with those of the chemically related metabolites Glu and Gln, which are readily detected using 1H HR-MAS spectroscopy.
We noted several important correlations between 2HG levels in human tissue and parameters that are associated with gliomas. Relationships between levels of 2HG and other ex vivo metabolites may aid in our understanding of the altered metabolic state of IDH-mutated glioma and in determining whether 2HG is a contributing factor. The standing hypothesis that 2HG is a tumor-promoting oncometabolite (3) currently lacks full mechanistic support; however, there are advantages to elucidating the biochemical pathways that are influenced by 2HG and may lead to improved survival for patients with IDH mutations. The correlations between histopathology parameters and levels of 2HG that were observed in our study indicated that lesions with IDH1-mutant enzymes have increased mitotic activity, with elevated tumor scores and cell density. The strong correlation between these parameters is consistent with there being higher cellularity in IDH-mutant gliomas where 2HG is present. This implies that in vivo levels of 2HG may be able to contribute not only to the classification of glioma but also to characterizing the spatial extent of infiltrative lesions.
The negative correlation between 2HG and normal delicate vascularity may assist in designing strategies for treating lesions with mutant IDH1 enzymes. 2HG has been associated in the literature with the competitive inhibition of oxygen-sensing prolyl hydroxylases (PHDs) and histone demethylases (17, 18). Restraining PHD activity is hypothesized to result in an up-regulation of proangiogenic growth factors, including a common chemotherapy target, vascular endothelial growth factor (VEGF). It is therefore of clinical interest whether patients harboring IDH mutations may be ideal candidates for therapies that target 2HG production as well as antiangiogenic agents that promote vascular normalization. With regard to histone demethylases as chromatin-modifying enzymes, their inhibition could alter the regulation of gene expression such that oncogenesis is provoked (18), another potential reason to therapeutically reduce levels of 2HG.
We also found a significant relationship between IDH1 mutants—as identified via 2HG levels in tissue—and in vivo MR diffusion parameters, which supports the potential role of diffusion imaging in evaluating whether the tissue architecture of mutant gliomas is distinct from that of other neoplasms. The elevation of tCHO that we observed from the ex vivo spectra of IDH1-mutant gliomas suggests that in vivo levels of tCHO may also be important for evaluating these lesions. Because MR diffusion and spectroscopy measures are readily obtained on most clinical MR scanners, they may be useful in defining regions of interest for studying variations in levels of 2HG.
The fact that the observed levels of 2HG from tissue samples expressing IDH1 mutations were similar to those of other routinely detected metabolites, such as Gln and Glu, supports the potential clinical use of 2HG as a noninvasive biomarker for IDH1. Given that tissue samples taken from disparate regions of tumor were homogeneous in their mutation status, it is reasonable to assume that 2HG will be present throughout lesions expressing aberrant enzymes. The elevated levels of 2HG in gliomas that had converted to a higher grade at the time of recurrence were attributable to differences in cellularity. This is an important finding because it suggests that the amount of 2HG produced per tumor cell remains unchanged during malignant transformation. From the correlation between levels of 2HG, tumor score, and tissue cellularity, it seems likely that 2HG will contribute to determining the extent of recurrent tumor in an in vivo setting. This will be valuable in differentiating tumor from treatment effects, especially in difficult cases where ambiguous anatomic imaging complicates interpretation by a neuroradiologist.
This study has provided important information that should be considered carefully in the development of technology for translating 2HG detection into an in vivo setting. One of the main challenges is providing a robust in vivo method for resolving 2HG from neighboring metabolites that are of a similar concentration and will have overlapping resonances at clinical field strengths. There have been three preliminary studies that have reported on the detection of in vivo 2HG peaks in patients with IDH1-mutated gliomas (19, 20, 21). These used different acquisition parameters and reported varying accuracies for estimated metabolite levels based on the Cramar-Rao bounds derived using the LCModel algorithm for spectral quantification. Spectral editing and more sophisticated data acquisition methods, including variants of the 2D experiment used in our study, may be useful in separating these components and verifying individual resonances (21). From our data, it appears that focusing on the unobstructed spectral feature created by the α-proton may be of interest for improving the specificity and accuracy of detecting 2HG in vivo.
Considering the survival benefits associated with IDH1/2 mutations, the presence of 2HG may have significant prognostic value for patients with low-grade gliomas. The characterization of IDH-mutated lesions using MR methods could also hold implications for the study of other human cancers that share these genetic abnormalities, including colorectal cancer, prostate cancer, and acute myeloid leukemia, which shows an adverse prognosis for IDH1/2 mutations, in contradistinction with gliomas (22, 23). Recent evidence suggests that many of these cancers carry chemosensitive properties that could enhance treatment efficacy and inform clinicians on how to stratify patients for different therapies (5). If D-2HG truly bears oncogenic properties (3), then testing for mutations in the IDH gene will probably not be sufficient for the clinical management of glioma patients; additional monitoring of 2HG levels would become important, especially for therapies targeting the IDH1 pathway (24, 25). Finally, as a potential marker for tumor in vivo, 2HG may prove especially helpful to clinicians attempting to discern disease recurrence from treatment effects in patients whose radiographic imaging is considered suspect. This information may help clinicians identify candidates for the most appropriate therapies or clinical trials and assist in the monitoring of treatment.
Patient population
This study was approved by our Institutional Review Board before patient recruitment. Informed consent was obtained from each participating subject. A total of 60 patients whose initial diagnosis was WHO grade II glioma and who were presenting for surgical resection owing to suspected recurrence were enrolled. All participants had previously received standard-of-care treatments that included surgical resection, radiation therapy, and/or chemotherapy. Six patients were excluded because of methanol contamination from the operating room, one patient was excluded because of a lack of tumor cells present in their tissue samples, and one patient was excluded because of insufficient tissue size (<0.5 mg) and degraded cellular integrity. This left 52 patients to be analyzed by this study. A complete list of patients and tissue samples is provided in table S1 along with exclusion criteria.
Preoperative MRI and spectroscopy
Preoperative MR examinations were conducted on a 1.5 T or 3 T EXCITE GE Signa Echospeed scanner (GE Healthcare Technologies) using an eight-channel phased-array headcoil (MRI Devices). Functional imaging included six-directional diffusion-weighted imaging acquired in the axial plane [repetition time (TR)/echo time (TE) = 1000/108 ms, voxel size = 1.7 × 1.7 × 3 mm3, b = 1000 s/mm2] and Lac-edited 3D MR spectroscopic imaging (1H MRSI) using point-resolved spectroscopic selection (PRESS) for volume localization and very selective saturation (VSS) pulses for lipid signal suppression (excited volume = 80 × 80 × 40 mm3, overpress factor = 1.5, TR/TE = 1104/144 ms, field of view = 16 × 16 × 16 cm3, nominal voxel size = 1 × 1 × 1 cm3, flyback echo-planar readout gradient in the SI direction, 712 dwell points and 988 Hz sweep width) (26).
Postprocessing of preoperative MR exam
The in vivo data from the preoperative examination were transferred to a Sun Ultra 10 workstation (Sun Microsystems), and in-house software was applied to derive estimates of in vivo diffusion and spectroscopic parameters. Maps of the ADC were generated on a pixel-by-pixel basis according to a published algorithm (27). MRSI data were processed to quantify tCHO and NAA levels, from which maps of the CNI could be derived. CNI values were generated from a linear regression–based algorithm (28) and represent the changes in choline and NAA levels relative to normal voxels.
Image-guided brain tumor tissue sampling
Regions demonstrating either abnormally decreased ADC and/or elevated CNI values, consistent with viable tumor, were evaluated for the purpose of planning which tissue to biopsy during surgery. One to four tumor tissue samples from each patient were designated as 5-mm-diameter spherical targets on co-registered MR images using BrainLAB surgical navigation software (BrainLAB Inc.). To minimize risks to the patient, no control tissue (normal brain) was acquired. Image-guided navigation was applied to locate tissue corresponding to planned targets and to acquire safely accessible samples. Upon excision, tissue samples were immediately bisected: half were snap-frozen in liquid nitrogen and stored at −80°C for 1H HR-MAS spectroscopy; the other half was fixed in 10% zinc formalin, dehydrated by graded ethanols, and embedded in Paraplast Plus wax (McCormick Scientific) using standardized techniques for tissue processing and immunohistochemistry.
1H HR-MAS spectroscopy
Tissue samples weighing between 0.78 and 28.14 mg (mean = 9.56 mg) were loaded into a 35-μl zirconia rotor (custom-designed by Varian) with 3 μl of 99.9% atom-D deuterium oxide containing 0.75 wt % 3-(trimethylsilyl)propionic acid (Sigma-Aldrich) for chemical shift referencing. Data were acquired at 11.7 T, 1°C, 2250 Hz spin rate in a 4-mm gHX nanoprobe with a Varian INOVA 500 MHz multinuclear spectrometer. The nanoprobe gHX is an inverse probe, optimized for the direct detection of protons and the indirect detection of X-nuclei(13C, 31P, 15N) and was equipped with a magic angle gradient coil. A 40 mM phantom solution of 2HG at pH 7 was scanned using the same parameters to provide a basis spectrum for metabolite fitting.
A rotor-synchronized 1D CPMG pulse sequence was run with TR/TE = (4 s)/(144 ms), 512 scans, 40,000 acquired points, 90° pulse, and 20 kHz spectral width for a total time of 35 min. The electronic reference to access in vivo concentrations (ERETIC) method was used to generate an artificial electronic signal that served as an external standard for estimation of metabolite levels in 1D spectra (29). A rotor-synchronized adiabatic 2D TOCSY spectrum was acquired with a 40-ms mixing time, 25 transients, 64 increments, saturation delay of 1 s, 90° pulse, and 20 kHz spectral width for a total time of 62 min (13). TOCSY spectra display a signal intensity map of metabolite cross peaks, based on the transferred magnetization between J-coupled protons. Information derived from this sequence is useful for identifying metabolites by their unique spin systems and separating overlapping chemical signals.
Preprocessing of HR-MAS spectra was done in the time domain using the Java-based magnetic resonance user interface (jMRUI) (30). Quantification of relative 1D metabolite levels was achieved with the semiparametric algorithm HR-QUEST, which fits a customized basis set of metabolites to a given spectrum (31). The HR-QUEST basis set used in this study was composed of spectra from 26 metabolites that are commonly studied in the human brain, as well as spectra acquired from a stock solution of 2HG (Sigma-Aldrich) (14). Metabolite levels with less than 10% Cramer-Rao error estimates were included for statistical analysis. Two experienced spectroscopists also evaluated each spectrum to qualitatively assess goodness of metabolite fits and whether low resolution or SNR compromised its analysis. Tissue samples found by these spectroscopists to have ambiguous or indeterminate interpretations were eliminated from subsequent analyses, whereas the remainder was classified visually as either 2HG-positive (+) or 2HG-negative (−). This enabled a query of the relative prevalence of aberrant 2HG production among glioma grades and histological subtypes, even in cases where 2HG levels were insufficient for HR-QUEST fitting.
Statistical analysis
All analyses were performed with the statistical software R (version 2.13). To determine differences in 2HG levels by grade, we applied an exact Wilcoxon rank-sum test from the library coin to pairwise compare the distribution of 2HG levels between grades. To identify parameters that were related to levels of 2HG, we performed a correlation analysis. In one method, a single tissue sample was selected at random from each patient and the resulting data set was fit according to a simple regression and Kendall tau model. This process was carried out a total of 100 times, with a correlation being considered statistically significant if the P value for the Kendall tau correlation coefficient was ≤0.05 in at least 70% of the individual trials. The second method used a simple regression model, which evaluated all of the tissue samples collectively and correcting for correlated responses from the same patient via bootstrap estimation. All statistical methods are presented in Supplementary Material, whereas the latter simple regression model with bootstrapping is presented as the primary analysis (32, 33). Owing to the exploratory nature of this study, adjustment for multiple comparisons was not made and regression coefficients significant at P ≤ 0.05 are summarized as predictors of 2HG concentration.
In all cases, 2HG was treated as the outcome. Note that the correlation analysis excluded those patients and tissue samples with an undetermined or negative IDH1 mutation status and spectra that could not be fit for 2HG, even if it was visually identified. This reduced the data analyzed for correlations to 25 patients and 46 tissue samples. There were two tissue samples from one patient in which the presence of 2HG had been assumed on the basis of histologic confirmation of IDH1 mutation, but was not corroborated by HR-MAS spectroscopy. 2HG levels in these samples were set to zero for analysis, because their perceived absence was likely owing to concentrations that fell below the level of detection.
Supplementary Material
supplementary data
We would like to acknowledge support from the Brain Tumor Research Center at University of California, San Francisco (UCSF) in collecting the tissue samples, as well as the Magnetic Resonance Laboratory personnel at UCSF for the use of their spectrometer. We would also like to express our gratitude to S. Ronen, J. Kurhanewicz, K. Keshari, and R. Iman for their assistance and guidance during this project, and to J. Crane and B. Olson for the development of the SIVIC software package and the implementation of HR-QUEST.
Funding: This work was supported by funds from the NIH, which included the Brain Tumor SPORE P50CA097257 and K08 NS063456 grants.
Author contributions: G.B. performed the patient consenting, presurgical MR scanning, and tissue sample planning. G.B. and M.S.B. acquired the tissue samples. A.E., L.E.J., and H.A.I.Y. performed the NMR experiments and analyzed the data. J.J.P. performed the IDH1 immunohistochemistry, genetic sequencing, and pathology analysis. R.P. performed the statistical analyses. A.E. and L.E.J. wrote the paper with assistance from H.A.I.Y., J.J.P., and S.J.N. M.S.B., R.S., S.M.C., S.C., and S.J.N. conceived and designed the experiments and analyzed the data.
Competing interests: The authors declare that they have no competing interests.
Fig. S1. J-coupling constants and spectral simulation of 2HG.
Fig. S2. Separation of 2HG γ resonances at ultrahigh field strength.
Fig. S3. Plots of 2HG correlations with ex vivo metabolites, histopathology, and in vivo MR imaging parameters.
Table S1. Complete patient data for 2HG presence and IDH1 mutation status.
Table S2. Correlation statistics including an additional statistical method.
1. Grier JT, Batchelor T. Low-grade gliomas in adults. Oncologist. 2006;11:681–693. [PubMed]
2. Riemenschneider MJ, Reifenberger G. Molecular neuropathology of low-grade gliomas and its clinical impact. Adv. Tech. Stand. Neurosurg. 2010;35:35–64. [PubMed]
3. Yan H, Parsons DW, Jin G, McLendon R, Rasheed BA, Yuan W, Kos I, Batinic-Haberle I, Jones S, Riggins GJ, Friedman H, Friedman A, Reardon D, Herndon J, Kinzler KW, Velculescu VE, Vogelstein B, Bigner DD. IDH1 and IDH2 mutations in gliomas. N. Engl. J. Med. 2009;360:765–773. [PMC free article] [PubMed]
4. Parsons DW, Jones S, Zhang X, Lin JC, Leary RJ, Angenendt P, Mankoo P, Carter H, Siu IM, Gallia GL, Olivi A, McLendon R, Rasheed BA, Keir S, Nikolskaya T, Nikolsky Y, Busam DA, Tekleab H, Diaz LA, Jr., Hartigan J, Smith DR, Strausberg RL, Marie SK, Shinjo SM, Yan H, Riggins GJ, Bigner DD, Karchin R, Papadopoulos N, Parmigiani G, Vogelstein B, Velculescu VE, Kinzler KW. An integrated genomic analysis of human glioblastoma multiforme. Science. 2008;321:1807–1812. [PMC free article] [PubMed]
5. Houillier C, Wang X, Kaloshi G, Mokhtari K, Guillevin R, Laffaire J, Paris S, Boisselier B, Idbaih A, Laigle-Donadey F, Hoang-Xuan K, Sanson M, Delattre JY. IDH1 or IDH2 mutations predict longer survival and response to temozolomide in low-grade gliomas. Neurology. 2010;75:1560–1566. [PubMed]
6. Hartmann C, Hentschel B, Wick W, Capper D, Felsberg J, Simon M, Westphal M, Schackert G, Meyermann R, Pietsch T, Reifenberger G, Weller M, Loeffler M, von Deimling A. Patients with IDH1 wild type anaplastic astrocytomas exhibit worse prognosis than IDH1-mutated glioblastomas, and IDH1 mutation status accounts for the unfavorable prognostic effect of higher age: Implications for classification of gliomas. Acta Neuropathol. 2010;120:707–718. [PubMed]
7. Koh HJ, Lee SM, Son BG, Lee SH, Ryoo ZY, Chang KT, Park JW, Park DC, Song BJ, Veech RL, Song H, Huh TL. Cytosolic NADP+-dependent isocitrate dehydrogenase plays a key role in lipid metabolism. J. Biol. Chem. 2004;279:39968–39974. [PubMed]
8. Jo SH, Son MK, Koh HJ, Lee SM, Song IH, Kim YO, Lee YS, Jeong KS, Kim WB, Park JW, Song BJ, Huh TL. Control of mitochondrial redox balance and cellular defense against oxidative damage by mitochondrial NADP+-dependent isocitrate dehydrogenase. J. Biol. Chem. 2001;276:16168–16176. [PubMed]
9. Lee SM, Koh HJ, Park DC, Song BJ, Huh TL, Park JW. Cytosolic NADP+-dependent isocitrate dehydrogenase status modulates oxidative damage to cells. Free Radic. Biol. Med. 2002;32:1185–1196. [PubMed]
10. Dang L, White DW, Gross S, Bennett BD, Bittinger MA, Driggers EM, Fantin VR, Jang HG, Jin S, Keenen MC, Marks KM, Prins RM, Ward PS, Yen KE, Liau LM, Rabinowitz JD, Cantley LC, Thompson CB, Vander Heiden MG, Su SM. Cancer-associated IDH1 mutations produce 2-hydroxyglutarate. Nature. 2009;462:739–744. [PMC free article] [PubMed]
11. Capper D, Zentgraf H, Balss J, Hartmann C, von Deimling A. Monoclonal antibody specific for IDH1 R132H mutation. Acta Neuropathol. 2009;118:599–601. [PubMed]
12. Meiboom S, Gill D. Modified spin-echo method for measuring nuclear relaxation times. Rev. Sci. Instrum. 1958;29:688–691.
13. Swanson MG, Keshari KR, Tabatabai ZL, Simko JP, Shinohara K, Carroll PR, Zektzer AS, Kurhanewicz J. Quantification of choline- and ethanolamine-containing metabolites in human prostate tissues using 1H HR-MAS total correlation spectroscopy. Magn. Reson. Med. 2008;60:33–40. [PMC free article] [PubMed]
14. Govindaraju V, Young K, Maudsley AA. Proton NMR chemical shifts and coupling constants for brain metabolites. NMR Biomed. 2000;13:129–153. [PubMed]
15. Srinivasan R, Phillips JJ, Vandenberg SR, Polley MY, Bourne G, Au A, Pirzkall A, Cha S, Chang SM, Nelson SJ. Ex vivo MR spectroscopic measure differentiates tumor from treatment effects in GBM. Neuro Oncol. 2010;12:1152–1161. [PMC free article] [PubMed]
16. Nelson SJ. Assessment of therapeutic response and treatment planning for brain tumors using metabolic and physiological MRI. NMR Biomed. 2011;24:734–749. [PMC free article] [PubMed]
17. Zhao S, Lin Y, Xu W, Jiang W, Zha Z, Wang P, Yu W, Li Z, Gong L, Peng Y, Ding J, Lei Q, Guan KL, Xiong Y. Glioma-derived mutations in IDH1 dominantly inhibit IDH1 catalytic activity and induce HIF-1α Science. 2009;324:261–265. [PMC free article] [PubMed]
18. Chowdhury R, Yeoh KK, Tian YM, Hillringhaus L, Bagg EA, Rose NR, Leung IKH, Li XS, Woon ECY, Yang M, McDonough MA, King ON, Clifton IJ, Klose RJ, Claridge TDW, Ratcliffe PJ, Schofield CJ, Kawamura A. The oncometabolite 2-hydroxyglutarate inhibits histone lysine demethylases. EMBO Rep. 2011;12:463–469. [PubMed]
19. Choi C, Sandeep G, DeBerardinis R, Kovacs Z, Bachoo R, Pascual J, Dimitrov I, Mickey B, Malloy C, Maher E. In vivo detection of IDH mutations in gliomas by 1H-MRS. Proc. Intl. Soc. Mag. Reson. Med. 2011;19:181.
20. Nagarajan R, Thomas MA, Pope WB, Prins RM, Wilson N, Salamon N, Liau LM. Detection of 2-hydroxyglutarate in mutant brain tumors in vivo using proton magnetic resonance spectroscopy. Proc. Intl. Soc. Mag. Reson. Med. 2011;19:184.
21. Andronesi OC, Kim G, Gerstner E, Batchelor T, Tzika AA, Fantin VR, Vander Heiden MG, Sorensen AG. Detection of 2-hydroxyglutarate in IDH-mutated glioma patients by in vivo spectral-editing and 2D correlation magnetic resonance spectroscopy. Sci. Transl. Med. 2012;4:116ra4. [PMC free article] [PubMed]
22. Paschka P, Schlenk RF, Gaidzik VI, Habdank M, Krönke J, Bullinger L, Späth D, Kayser S, Zucknick M, Götze K, Horst HA, Germing U, Döhner H, Döhner K. IDH1 and IDH2 mutations are frequent genetic alterations in acute myeloid leukemia and confer adverse prognosis in cytogenetically normal acute myeloid leukemia with NPM1 mutation without FLT3 internal tandem duplication. J. Clin. Oncol. 2010;28:3636–3643. [PubMed]
23. Aghili M, Zahedi F, Rafiee E. Hydroxyglutaric aciduria and malignant brain tumor: A case report and literature review. J. Neurooncol. 2009;91:233–236. [PubMed]
24. Yen KE, Bittinger MA, Su SM, Fantin VR. Cancer-associated IDH mutations: Biomarker and therapeutic opportunities. Oncogene. 2010;29:6409–6417. [PubMed]
25. Reitman ZJ, Yan H. Isocitrate dehydrogenase 1 and 2 mutations in cancer: Alterations at a crossroads of cellular metabolism. J. Natl. Cancer Inst. 2010;102:932–941. [PMC free article] [PubMed]
26. Park I, Chen AP, Zierhut ML, Ozturk-Isik E, Vigneron DB, Nelson SJ. Implementation of 3 T lactate-edited 3D 1H MR spectroscopic imaging with flyback echo-planar readout for gliomas patients. Ann. Biomed. Eng. 2011;39:193–204. [PMC free article] [PubMed]
27. Basser PJ, Pierpaoli C. Microstructural and physiological features of tissues elucidated by quantitative-diffusion-tensor MRI. J. Magn. Reson. B. 1996;111:209–219. [PubMed]
28. McKnight TR, Noworolski SM, Vigneron DB, Nelson SJ. An automated technique for the quantitative assessment of 3D-MRSI data from patients with glioma. J. Magn. Reson. Imaging. 2001;13:167–177. [PubMed]
29. Albers MJ, Butler TN, Rahwa I, Bao N, Keshari KR, Swanson MG, Kurhanewicz J. Evaluation of the ERETIC method as an improved quantitative reference for 1H HR-MAS spectroscopy of prostate tissue. Magn. Reson. Med. 2009;61:525–532. [PMC free article] [PubMed]
30. Stefan D, Di Cesare F, Andrasescu A, Popa E, Lazariev A, Vescovo E, Strbak O, Williams S, Starcuk Z, Cabanas M, van Ormondt D, Graveron-Demily D. Quantitation of magnetic resonance spectroscopy signals: The jMRUI software package. Meas. Sci. Technol. 2009;20:104035.
31. Ratiney H, Albers MJ, Rabeson H, Kurhanewicz J. Semi-parametric time-domain quantification of HR-MAS data from prostate tissue. NMR Biomed. 2010;23:1146–1157. [PMC free article] [PubMed]
32. Feng Z, McLerran D, Grizzle J. A comparison of statistical methods for clustered data analysis with Gaussian error. Stat. Med. 1996;15:1793–1806. [PubMed]
33. Harrell FE., Jr. Regression Modeling Strategies: With Applications to Linear Models, Logistic Regression and Survival Analysis. Springer; New York: 2001.