IDH1 R132H Mutation Generates a Distinct Phospholipid Metabolite Profile in Glioma Morteza Esmaeili1,*, Bob C. Hamans2, Anna C. Navis3, Remco van Horssen4,5, Tone F. Bathen1, Ingrid S. Gribbestad1†, William P. Leenders3, and Arend Heerschap1,2 1Department of Circulation and Medical Imaging, Norwegian University of Science and Technology (NTNU), Trondheim, Norway. 2Department of Radiology, Radboud University Medical Center, Nijmegen, the Netherlands. 3Department of Pathology, Radboud University Medical Center, Nijmegen, the Netherlands. 4Department of Cell Biology, Radboud Institute for Molecular Life Sciences, Nijmegen, the Netherlands. 5Department of Clinical Chemistry and Hematology, St. Elisabeth Hospital, Tilburg, the Netherlands. Published OnlineFirst July 8, 2014; doi: 10.1158/0008-5472.CAN-14-0008. Cancer Res September 1, 2014 74; 4898 Diffuse gliomas are the most common malignant brain-born tumors and are incurable with present therapeutic strategies (1). These tumors are classified by the World Health Organization (WHO) as grade 2, 3, and 4 of which grade 4 glioma (glioblastoma, GBM) is the most malignant type. The current median survival from the time of diagnosis for GBMs is only 14.6 months and for lower grades between 4 and 15 years (2, 3). This highly variable survival calls for reliable prognostic biomarkers for rational decision making in clinical management. Such biomarkers have become available with the discovery that in more than 70% of grade 2 and 3 gliomas and in secondary GBMs, one of the genes for isocitrate dehydrogenase (IDH1 and IDH2) carry specific mutations, which are associated with prolonged overall survival (4–8). IDH1, the predominantly affected enzyme (>95%), catalyzes the conversion of isocitrate into α-ketoglutarate (α-KG) in the cytosol, using NADP as electron acceptor to generate NADPH (Fig. 1A). IDH1 can also catalyze the reductive carboxylation of α-KG to isocitrate that can be further processed to citrate and acetyl- and succinyl-CoA, important anabolic precursors for lipid synthesis (9). The mutation in IDH1, almost always affecting arginine R132, confers a neomorphic activity to the enzyme, which results in NADPH-dependent conversion of α-KG to 2-hydroxyglutarate (2-HG; Fig. 1A; ref. 10). The mutant enzyme lacks the capacity of reductive carboxylation (11). As 2-HG accumulates in mutated tumor cells and tissues (12–14), it has attracted attention as a potential biomarker in the diagnosis and prognosis of gliomas, in particular as the high levels of 2-HG can be detected noninvasively by 1H MR spectroscopy (MRS) in humans (8, 15–19). 1H MRS has been explored extensively in the diagnosis and treatment evaluation of brain tumors in humans (20, 21). MR spectra of the brain show a single spectral peak for the methyl protons of small choline compounds, which are involved in the Kennedy pathway of membrane lipid synthesis and breakdown (Fig. 1B). In brain tumors, choline metabolism is adapted to the needs of higher proliferation and to the physiologic microenvironment (such as acidic extracellular pH; refs. 22, 23), and the intensity of this peak (labeled as total choline or tCho) is often increased (24). Another prominent spectral change is a decrease of the peak for the methyl protons of N-acetyl aspartate (NAA), a neuronal marker compound, reflecting replacement of neurons by glial tumor cells (25). The tCho/NAA ratio is, therefore, often used as a biomarker for tumor load and malignancy in gliomas (26–28). The intensity of the tCho peak also correlates with cell density, and may be related to gliosis (29, 30). To understand in more detail what determines the tCho peak intensity, an analysis of each contributing component is needed. This is possible with 1H MRS of ex vivo biopsy material, which has a better spectral resolution than in vivo MRS and allows the separation of tCho into peaks for phosphocholine (PC), glycerophosphocholine (GPC), and free choline (24). Ex vivo 1H MRS or high-resolution magic angle spinning (HR-MAS) spectroscopy has revealed that PC and GPC contribute importantly to the increase of tCho in brain tumors and also uncovered more subtle relationships of choline compounds with tumor features, in particular with tumor grade (31–34). Direct in vivo detection of PC and GPC is possible by 31P MRS, which also enables detection of phosphoethanolamine (PE) and glycerophosphoethanolamine (GPE), thereby providing a more complete picture of in vivo phospholipid metabolism (35). Because 31P MRS is less sensitive than 1H MRS and requires dedicated radiofrequency probes, it has been less used to examine phospholipid metabolites in vivo in brain tumors (36, 37). However, the increased access to high-field (pre-) clinical MR scanners, which improves 31P MRS sensitivity and resolution, invigorates its further exploration in studies of tumor phospholipid metabolism. As α-KG and NADPH are important components for lipogenesis (38, 39) and as the mutated IDH enzyme consumes both compounds and lacks reductive carboxylation capacity, we hypothesized that phospholipid metabolism is altered in IDH1-mutated glioma. To test this hypothesis, we applied in vivo 31P MR spectroscopic imaging (MRSI) to four unique and representative human glioma models growing orthotopically in mice (40), one carrying the IDH1-R132H mutation (41). The spectral findings were verified by 31P NMR analyses of tumor tissue extracts. To examine the causal relationship of the spectral profiles to expression of the mutated enzyme, we also performed 31P NMR on extracts of glioma cell lines, stably expressing wild-type or mutated IDH1. Finally, we tested if similar phospholipid profiles occur in human gliomas by performing 31P HR-MAS MRS of biopsies of gliomas in patients with and without IDH1 mutation. In vivo31P 3D MRSI acquisition and analysis All in vivo MR experiments were performed on a preclinical 7T MR system (Bruker ClinScan) operating at 121.7 MHz for 31P MRS. The phosphorus spectra were acquired using a homebuilt 16-mm transmit/receive quadrature coil in combination with a solenoid 1H surface coil (20 mm in diameter). The animals were subjected to MRSI when evident signs of tumor burden (especially evident weight loss, neurologic defects) were present. A control group consisting of healthy Balb/c nu/nu animals (n = 3) was also included. Animals were placed in prone position and anesthetized by 1.5% isoflurane (Abott) and a mixture of O2 and N2O inhalation. The animals body temperature was maintained at 37.5°C applying warm air circulation and physiologic monitoring (Small Animal Instrument Inc.) to assess respiration and temperature. After obtaining a localizer image, T2-weighted multi spin-echo images in three orthogonal orientations of the brain were acquired. First- and second-order shimming was performed using FASTMAP (43). The MRSI field of view (FOV) and matrix size were then selected carefully reviewing T2-weighted images to cover hyperintense areas within the tumor tissues (Fig. 2). Three-dimensional 31P MR spectroscopy was performed using a 3D MRSI pulse acquire sequence with an adiabatic BIR-4 45° excitation pulse (44), a repetition time (TR) of 1500 milliseconds, Hanning-weighted cartesian k-space sampling with 196 signal averages at the centre of k-space, 2,048 data points over a spectral width of 4,868 Hz and a total acquisition time of 2 hours. The FOV of 24 mm × 24 mm × 24 mm with an 8 × 8 × 8 data matrix and Hamming filtering resulted in a nominal voxel size of about 5 mm3. After the MR exams, the animals were sacrificed by cervical dislocation, the brains removed and separated in two halves, which were frozen in liquid nitrogen for subsequent in vitro MRS analyses. Remaining brain tissue was formalin fixed and paraffin embedded for further histopathology analysis. All in vivo MR spectra were analyzed using the jMRUI software (45) and signals fitted with a Lorentzian lineshape, except the J-coupled signals of ATP, which were fitted with a Gaussian shape, using the Advanced Method for Accurate, Robust and Efficient Spectral fitting method (46). Before fitting, spectral processing was performed, including manual phase correction, zero-filling (4,096 points), and line-broadening of 20 Hz. Nuclear magnetic resonance acquisition and analysis of in vitro and ex vivo samples Frozen brain tissue samples from the glioma xenografts and U251 cell pellets were extracted using perchloric acid as described in detail previously (47). The neutralized extracts were lyophilized and kept at −80°C until being dissolved in 600 μL of D2O. After final pH adjustments with potassium hydroxide (KOH), in vitro 31P NMR spectra of extracts were acquired using a Bruker spectrometer (Bruker Avance III 600 MHz/54 mm US-Plus) equipped with a multinuclear QCI CryoProbe (Bruker BioSpin GmbH) operating at 243.5 MHz for 31P MRS. High-resolution 31P NMR spectra of the water-soluble metabolites were obtained with proton decoupling during acquisition, a 30° flip angle, 8,192 free induction decays (FID), TR = 4 seconds, spectral width of 14,577 Hz into 36,864 data points in time domain. 31P HR-MAS spectroscopy was carried out using a 600-MHz spectrometer (Bruker Avance III 600 MHz/54 mm US-Plus) equipped with a triple 1H/13C/31P MAS probe (Bruker BioSpin GmbH). The frozen specimens from human brain tumors were thawed and cut on an ice-pad. Tissue samples were gently loaded into 30-μL disposable inserts filled with 3 μL 2H2O (Sigma-Aldrich GmbH) for the 2H lock. The inserts were then placed into a 4-mm diameter ZrO2 MAS rotor (Bruker BioSpin GmbH). The MAS rotors were spun at 5 kHz and maintained at 4°C to minimize enzymatic activities within the tissue samples. All in vitro and ex vivo spectra were processed using the Bruker TopSpin V3.0 software (Bruker BioSpin GmbH). The accumulated FIDs were Fourier transformed after application of 3 Hz exponential line broadening. Automatic phase and linear baseline corrections were performed. The GPC peak (at 3.04 ppm) in 31P MR spectra was used as references for chemical shift calibration. Following standard processing, peak areas of phosphorylated metabolites were calculated by peak fitting (PeakFit V4.12; SeaSolve Software Inc.) using a combination of Gaussian–Lorentzian lineshapes (Voigt area). Metabolite concentrations were calculated from peak areas In vivo 3D 31P MRSI of human glioma xenografts Tumors in the brain present as hyperintense signal areas on T2-weighted MR images (compare the tumor-containing brain in Fig. 2A with the normal brain in Fig. 2B), and we used these images for voxel positioning. The 31P MR spectra of voxels of interest selected from the 3D MRSI dataset of this brain showed resolved resonances for a number of compounds, including ATP, phosphocreatine (PCr), GPC and GPE, inorganic phosphate (Pi), PC, and PE. The 31P-spectral profiles of tumor voxels in all four xenograft models differed from those obtained from voxels in comparable brain areas in non–tumor-bearing animals (compare Fig. 2C with 2D). The relative phosphor signal integrals of choline and ethanolamine compounds were increased, as represented by a significantly higher (PC + GPC + PE + GPE)/ATP ratio for all tumor types (P < 0.05; Fig. 2E). This total relative phospholipid content was not different among the tumor models. In addition, a significantly decreased PCr/ATP signal ratio was observed in tumor tissues compared with the healthy mouse brain tissues (P < 0.05; Fig. 2F). Among the four human glioma lines, the E478 tumor exhibited a deviating spectral profile in the 2 to 8 ppm range (Fig. 3A and B). For a quantitative assessment, we first determined for each metabolite resonance its integral normalized to the sum of those of all phospholipid metabolite resonances (see Fig. 4A and B). This revealed a significant decrease of the PE resonance of the IDH1-mutant E478 xenograft compared with those of the IDH1-WT xenografts (P = 0.003). A significant increase was observed for the GPC resonance of E478 compared with IDH1-WT tumors (P = 0.003). Furthermore, the PC peak of E478 showed a trend for an increase compared with the PC of the other tumors (P = 0.08). The GPE resonance did not differ between the tumors. In concordance with the in vivo results, we observed very similar differences between 31P NMR spectra of tumor extracts from IDH1-mutant and wild-type xenografts (Fig. 3B). Again, PE was reduced and GPC increased in E478 extracts compared with those of the other models (P = 0.004 and 0.01 respectively; Fig. 4B) In conclusion, we provide evidence that IDH1 mutations result in distinct alterations in lipid metabolism that can be detected noninvasively by 31P MRSI. These may serve as a complementary biomarker to characterize the metabolic status of IDH1-mutated gliomas during evaluation of anticancer targeted therapies and in tumor diagnosis. Increased availability to higher field strength MR systems (3 and 7 T) and dedicated 31P coils hold promise for clinical translation of the 31P MRSI method. Further research is needed to fully elucidate the roles of PE and GPC, such as their involvement in mitochondrial membrane synthesis. Figure 1. Schemes of metabolism involved in 2-HG biosynthesis (A), and of choline and ethanolamine phospholipid metabolism (B). Arrows, metabolic pathways. A, IDH1/2 catalyze oxidative decarboxylation of isocitrate to α-KG using NADP+ as a cofactor to generate NADPH and CO2. Mutations in these genes generate the oncometabolite 2-HG by consuming NADPH, and have an impact on intracellular signaling and epigenetics. The citrate generated via the TCA cycle contributes to the lipid synthesis. This pathway can be interrupted by mutations in IDH1 and/or IDH2 genes. Subsequent metabolism of citrate produces acetyl-CoA for fatty acid and/or lipid synthesis, and other intermediates such as oxaloacetate and malate in TCA cycle. B, metabolic pathways of PtdCho and PtdEtn. Acyl-CoA, acyl-coenzyme A; Etn, ethanolamine; CDP-Etn, cytidinediphosphate ethanolamine; PtdSer, phosphatidylserine; ChoK, choline kinase (EC 2.7.1.32); ETNK, ethanolamine kinase (EC 2.7.1.82); cPLA2, cytosolic phospholipase A2 (EC 3.1.1.4); PLC, phospholipase C (EC 3.1.4.3); PSD, phosphatidylserine decarboxylase (EC 4.1.1.65); PSS1, phosphatidylserine synthase I (EC Ptdss1); PSS2, phosphatidylserine synthase II (EC Ptdss2); PEMT, phosphatidylethanolamine N-methyltransferase (EC 2.7.8.29). Figure 2. 31P MRSI of the mouse brain with and without tumor. Orthogonal T2-weighted MR images of a mice brain with an E434 tumor (top, A) and a normal mouse brain (bottom, B) in axial, coronal, and sagital views, and corresponding 31P MR spectra of 27 mm3 nominal voxels from the 3D 31P MRSI data (C and D). E and F, bar plot of the (PC + GPC + PE + GPE)/ATP (E) and PCr/ATP (F) signal ratios of tumor types growing in mouse brain (n = 19) and of normal mouse brain (Ctrl, n = 3). Chemical shift is referenced to the GPC resonance at 3.04 ppm. The assigned peaks are (from left to right); PE, phosphoethanolamine; PC, phosphocholine; Pi, inorganic phosphate; GPE, glycerophosphoethanolamine; GPC, glycerophosphocholine; PCr, phosphocreatine; ATP, adenosine tri-phosphates. *, P < 0.05. Figure 3. IDH1-mutated E478 xenografts show a distinct 31P-spectral pattern. A, in vivo 31P MR spectra obtained from four human glioma xenograft tumor lines growing in the mouse brain (IDH1-mutated xenograft E478, and IDH1-WT E434, E473, and E468). In the IDH1-mutated xenograft, GPC is highly elevated and PE decreased compared with the wild-types. B, representative in vitro 243.5 MHz (1H-decoupled) 31P MR spectra of tissue extracts of these tumors; from top to bottom the IDH1-mutated E478, and the IDH1-WT lines E434, E468, and E473. The chemical shift reference is the GPC resonance at 3.04 ppm. Figure 5. 31P NMR spectra of U251MG cell extracts and 31P HR-MAS MR spectra of surgical biopsies from patients with glioma. A, representative 31P NMR spectra of cell extracts show similar 31P-spectral features as observed for the in vivo growing brain xenografts. Relatively decreased PE and increased GPC resonances identify the mutant cell line (U251-R132H) from wild-type cell lines (U251-WT) and U251MG control cells (U251). The green and orange colored spectra are shifted to the right for a better visualization. B, the PC/PE, GPC/GPE, and (PC + GPC)/(PE + GPE) ratios are significantly increased in the U251-R132H cell line compared with the U251-WT and U251 cells. C, phospholipid levels measured from 31P MR spectra of U251 cell extracts. D, representative 31P HR-MAS MR spectra of glioma patient biopsies. E, the levels of PC/PE, GPC/GPE, (PC + GPC)/(PE + GPE) ratios in IDH1-R132H glioma patients are consistent with preclinical results. F, phospholipid metabolite levels measured in glioma patient tissues samples. The y-axis values indicate the mean and SD; *, P < 0.05; **, P < 0.01. Published OnlineFirst July 8, 2014; doi: 10.1158/0008-5472.CAN-14-0008. Cancer Res September 1, 2014 74; 4898
Posted on: Wed, 12 Nov 2014 17:22:20 +0000
Trending Topics
Recently Viewed Topics
© 2015