Báo cáo khoa học: A metabolomics perspective of human brain tumours pdf

MINIREVIEW A metabolomics perspective of human brain tumours Julian L. Griffin 1 and Risto A. Kauppinen 2 1 Department of Biochemistry, University of Cambridge, UK 2 School of Sport and Exercise Sciences, University of Birmingham, Edgbaston, UK Introduction The global analysis of metabolites, such as by mass spectrometry (MS) and high resolution multinuclear nuclear magnetic resonance spectroscopy (MRS), places cells, tissues or organisms in biological context by defining metabolic phenotypes [1,2]. Such metabolo- mic approaches are being used to profile cell lines, tumours and systemic metabolism in human cancer tissue ex vivo and in vivo, and will provide another functional genomic tool for cancer research [3]. Whilst ‘-omic’ technologies are complementary to one another, the metabolome provides specific advantages when compared with the transcriptome and proteome. For in vitro purposes the work is cheap on a per sample basis. Furthermore, being downstream of gene transcription, changes in metabolites may well be amplified, and there is not necessarily a good quantitative relationship between mRNA concentra- tions and function, especially for proteins whose con- centration is determined by their rate of degradation or whose activity is controlled by allosteric effects or post translational modification. This suggests that meta- bolomics may be particularly sensitive to changes in a biological system, and have a more direct correlation with the phenotype produced. This minireview focuses on metabolomics of human brain tumours obtained in the first hand by multinu- clear MRS and MS using both ex vivo and in vivo approaches. Over recent years a wealth of data have indicated that ‘metabolite phenotypes’ bear great potential for clinical diagnosis, tumour grade assess- ment and finally, monitoring of treatment response of brain tumours. Looking to the future, the technology’s impact on diagnosis through minimally invasive screening will also be discussed. Keywords brain; metabolomics; NMR spectroscopy; pattern recognition; tumour Correspondence J. Griffin, Department of Biochemistry, University of Cambridge, Tennis Court Road, Cambridge, CB2 1QW, UK Fax: +44 1223 333345 Tel: +44 1223 764 922 E-mail: jlg40@mole.bio.cam.ac.uk (Received 19 October 2006, revised 7 December 2006, accepted 3 January 2006) doi:10.1111/j.1742-4658.2007.05676.x During the past decade or so, a wealth of information about metabolites in various human brain tumour preparations (cultured cells, tissue specimens, tumours in vivo) has been accumulated by global profiling tools. Such hol- istic approaches to cellular biochemistry have been termed metabolomics. Inherent and specific metabolic profiles of major brain tumour cell types, as determined by proton nuclear magnetic resonance spectroscopy ( 1 H MRS), have also been used to define metabolite phenotypes in tumours in vivo. This minireview examines the recent advances in the field of human brain tumour metabolomics research, including advances in MRS and mass spectrometry technologies, and data analysis. Abbreviations ANN, artificial neural network; Ala, alanine; CCM, choline-containing metabolites; Cre, creatine + phosphocreatine; GABA, c-amino butyric acid; Gln, glutamine; Glu, glutamic acid; GPC, glycerophosphocholine; GPE, glycerophophoethanolamine; ICA, independent component analysis; LC, liquid chromatography; Lip, lipids; MRI, magnetic resonance imaging; MRS, nuclear magnetic resonance spectroscopy; MRSI, magnetic resonance spectroscopic imaging; NAA, N-acetylaspartic acid; PC, phosphocholine; PNET, primitive neuroectodermal tumour; Tau, taurine. 1132 FEBS Journal 274 (2007) 1132–1139 ª 2007 The Authors Journal compilation ª 2007 FEBS Metabolite patterns in neural cells Three major neural cell types, i.e., neurones, glial and meningeal cells, have strictly distinct functional properties, a factor that is reflected in their metabolic specialization. It has become evident that the three neural cell types not only are distinguishable from each other by morphological and immunocytochemical char- acteristics, but also through their 1 H MRS metabolite profiles. Using a subgroup of eight metabolites (from a total number of  30 identifiable ones) quantified by 1 H MRS in acid extracts of cultured cells, one can unambiguously separate the three neural cell types [4]. Similarly, several brain tumour cell types can be identi- fied by their 1 H MRS metabolite content [5]. It was observed that neuroblastoma, glioma and meningeoma cells display low concentrations of normal neural meta- bolites, such as N-acetylaspartate (NAA), c-amino butyrate (GABA) and taurine (Tau) [5]. The meta- bolites bearing greatest value for discrimination of tumour cell types include total creatine (Cre; creat- ine + phosphocreatine), choline-containing metabolites [CCM; including phosphocholine (PC), glycerophos- phocholine (GPC) and choline], alanine (Ala), Tau and glutamate (Glu). Indicative to the potential clinical value of MRS metabolite profiles, 1 H MRS data allow separation between tumour types and grades [6,7] (Table 1). Metabolomics technology Metabolomics usually consists of two methodologically distinct parts. First, the analysis uses a global profiling tool to measure the concentration of the metabolites while the subsequent data matrix is interrogated by multivariate statistics or data reduction tools. Sec- ondly, pattern recognition processes can be separated into unsupervised and supervised techniques. The for- mer display the innate variation associated with the data, while the latter uses prior information to cluster the data. In addition to pattern recognition processes [8,9], machine learning approaches have also been applied to biochemical profiles of tumours [10]. For the analysis of brain tumours MRS and MS dominate the literature, although in other applications thin layer chromatography, Fourier transform infrared and Raman spectroscopy have been used previously [11,12]. Reflecting the literature, the majority of this minireview concerns the use of MRS as a metabolic profiling tool. However, MS approaches will be dis- cussed briefly first. Mass spectrometry Mass spectrometry based approaches are inherently more sensitive than MRS techniques, providing access to lower concentration metabolites in the tumour Table 1. Metabolites that have been commonly identified as perturbed in brain tumours using MRS either for tissue extracts or in vivo. Metabolite Metabolic function Type of cancer ⁄ tumour Alanine Increases in hypoxic tissues as a result of increased glycolysis. Brain tumors including astrocytomas, metastases, gliomas, meningiomas, and dysembryoplastic neuroepithelial tumors. CH 3 &CH 2 lipids Increases in the relative intensities of lipid peaks detected by NMR are believed to result from either the production of cell membrane microdomains or increased numbers of cytoplasmic vesicles. Alterations in visible lipids have been linked to many cellular processes such as proliferation, inflammation, malignancy, growth arrest, necrosis and apoptosis. Choline containing metabolites (CCMs) Choline, phosphocholine, phosphatidylcholine and glycerophosphocholine are major constituents of cell membranes and increases in these metabolites reflect cell death (apoptosis and necrosis). Many tumour types including a range of brain tumours. Lactate Lactate is an end product of glycolysis and increases rapidly during hypoxia and ischaemia, in particular as a result of poor vascularity and acquired resistance to hypoxia. Increased rates of lactate production are associated with a range of tumours. Myo-inositol In tumours, myo-inositol is involved in osmoregulation and volume regulation. Elevated in glioma. Nucleotides Nucleotides are key intermediates in DNA ⁄ mRNA synthesis and breakdown. Changes in ATP concentration also indicate the energetic status of the tumour. Found to be elevated in glioma during apoptosis. PUFAs Polyunsaturated fatty acids are constituents of cell membranes, especially mitochondrial. Increased in glioma during apoptosis. J. L. Griffin and R. A. Kauppinen A metabolomics perspective of brain tumours FEBS Journal 274 (2007) 1132–1139 ª 2007 The Authors Journal compilation ª 2007 FEBS 1133 metabolome. Most applications use prior chromatogra- phy with gas chromatography (GC) and liquid chro- matography (LC) to initially separate out, by time, metabolites in a tissue extract prior to analysis. The use of MS to monitor the metabolic profiles of brain tumours significantly predates the use of the term meta- bolomics. For example, Jellum and colleagues [13] identified  160 peaks in GC-MS spectra from normal brain tissue, pituitary tumours and brain tumours, and then used a pattern recognition approach to classify tissue into healthy and tumour. The sensitivity of mass spectrometry based approa- ches has also been used to monitor trace metabolites in excised tissue. For example, neurotransmitters in neuroctomas have been profiled, including acetylcho- line and the metabolites of catecholamines by HPLC [14], while Olsen and colleagues [15] have used quadru- pole-time of flight MS to detect morphine in glioma. Mass spectrometry has also been shown to be highly discriminatory for lipid metabolites, including ceramide metabolites, which vary in neuroblastoma cells during cell death [16]. MS profiling of lipid metabolites has also been used to determine which components con- tribute to resonances that are found in vivo 1 HMR spectra. Miller and coworkers [17] demonstrated that the CCM peak detected in brain tumour specimens lar- gely correlated with choline, PC and GPC, but not phosphatidylcholine. Ex vivo monitoring of brain tumour metabolites The use of NMR spectroscopy to profile metabolites in tumour cells and tissues has been applied to a wide range of human tumours for a number of years, with the approach being particularly useful at gener- ating new hypotheses that link characteristics of a tumour to metabolism. For example, Bhakoo and colleagues [18] examined the process of immortaliza- tion in primary rat Schwann cells, noting that an increase in the PC ⁄ GPC ratio correlated with this process. Tissue heterogeneity is a major issue in assessing the biochemical profile of tumours, particularly during response to treatments. High resolution magic angle spinning 1 H MRS is a highly versatile tool in this respect, examining relatively small amounts of tumour tissue, and can be used on tissue samples prior to histopathology. Examining glioblastoma multiforme removed during surgery, Cheng and colleagues demon- strated that Lac and mobile lipids (Lip) were correla- ted with degree of tumour necrosis and the proportion of PC to choline correlated with the malignancy of the glioma [19]. This had previously been shown by solu- tion state multinuclear MRS of glioma extracts [20]. To investigate lipid metabolism within tumours, tan- dem MS approaches provide a unique insight into many classes of compounds. Sullards and colleagues [21] have used this approach to monitor changes in sphingolipid metabolism in human glioma cell lines in order to correlate these profiles with either the induc- tion or inhibition of apoptosis. The metabolite data sets from 1 H MRS of extracted human brain tumour biopsy specimens have been used as inputs for pattern recognition analysis techniques [22]. Incorporation of principal component analysis as a means to reduce dimensionality of the MRS data for neural network analysis provided classification of sam- ples not only to meningeal and nonmeningeal tumours, but also grading within gliomas to within one grade with an accuracy of 62%. It was observed that only few metabolites in the extracts were discriminatory, including Cre, glutamine (Gln), Ala and myo-inositol [22]. This study and many others [7,23,24] have dem- onstrated metabolite abnormalities in brain tumours that discriminate them from normal brain tissue. Human brain tumours in vivo Human brain tumours form some 2% of all malignan- cies. Unlike outside the cranium both benign and malignant tumours can be life threatening due to space occupying nature. In adults, the majority of primary brain tumours are derived from glial or meningeal tis- sues, while secondary tumours contain metastases from many organs (e.g., breast and lung melanomas) of the body. Paediatric primary brain tumours also include tumours from neuronal cells, e.g., neuroblastomas and retinoblastomas. Despite significant heterogeneity in metabolism in tumours [25], MRS has provided unique information about tumour metabolites to be used for diagnosis, treatment planning, setting prognosis and monitoring efficacy of treatment procedures. Several ‘metabolonomic’ approaches have been proposed to help decompose the MRS from human brain tumours. 31 P MRS 31 P MRS can readily distinguish phosphorylated cho- line metabolites, including PC, PE, glycerophosphoryl ethanolamine (GPE) and GPC, involved in cell mem- brane metabolism [26,27], thus providing more detailed information about tumour activity than avail- able by 1 H MRS alone. Qualitative inspection of brain tumour 31 P MR spectra indicated small differ- ences in spectral appearances between normal brain A metabolomics perspective of brain tumours J. L. Griffin and R. A. Kauppinen 1134 FEBS Journal 274 (2007) 1132–1139 ª 2007 The Authors Journal compilation ª 2007 FEBS and gliomas [28]. Quantitative analysis of 31 PMR spectra revealed that the overall concentrations of MR detectable phosphates, including phosphodiester and phosphocreatine, were significantly lower in tumours than in normal parenchyma [29–31]. 31 P MRS has also been used to observe tumour responses to drug and radiation therapies [29]. 1 H MRS Metabolomics in vivo using 1 H MRS is limited by a number of technical issues. First, brain tumours are inherently heterogeneous in terms of their cellularity and blood supply; secondly, spectral resolution is much poorer in vivo than in vitro, allowing assignment of some 10 tumour metabolites; and thirdly, sensitivity of MRS at commonly used clinical field strengths and narrow chemical shift scale of 1 H MRS limits the num- ber of metabolites detected. Despite these factors 1 H MRS and MRS imaging (MRSI) from human brain tumours are gaining an ever increasing role in clinical assessment of patients with focal cerebral lesion of any nature. One of the key questions to be addressed remains whether 1 H MRS alone can provide specificity and sensitivity to identify proliferating lesions from other common focal brain conditions. Recent studies show that ischaemic infarctions show no 1 H MRS signals apart from Lac and macromolecules [32,33]. In case of infectious lesions 1 H MRS data provide > 90% specif- icity to separate abscesses and tuberculomas from astr- ocytic tumours [34]. Modern magnetic resonance imaging (MRI) techniques provide a large repertoire to diagnose brain lesions, such as ischaemic stroke, infec- tions and multiple sclerosis [35] and thus, the role of 1 H MRS will remain confirmatory for these cases. A wealth of 1 H MR spectroscopic data from brain tumours shows that both tumour types and tumour grades have characteristic spectral patterns. The idea of looking at the 1 H MRS spectrum in a more holistic manner arose from the work on cultured brain tumour cells [36]. Hagberg and coworkers proposed a set of multidimensional statistical methods for single-voxel 1 H MR spectra using the entire spectral width to clas- sify human glial tumours [37]. A concept of 1 H MRS profiles was introduced. Soon afterwards a concept of ‘ 1 H MRS metabolic phenotype’ was coined by Usenius et al. [38] and Preul et al. [39]. In these papers simpli- fied 1 H MR spectra from healthy brain and tumours comprising of six metabolites (CCM, Cre, NAA, Ala, Lac and Lip) were used as inputs to artificial neural network (ANN) analysis to classify the tumour types and grades. Preul et al. used leaving-one-out linear discriminant method for 1 H MRSI data sets and dem- onstrated a phenomenal accuracy of 104 correct out of 105 cases [39]. Usenius and coworkers included non- suppressed water signal from the spectroscopic volume as well as an ANN analysis and showed an accuracy of 82% for classification according to brain tumour type and grade [38]. Although neural network based approaches are typically ‘black box’ approaches, ‘reso- nance profiles’ provided by ANN analyses for tumour classification closely resemble MR spectral patterns, aiding the identification of metabolites with key discriminatory weight for a given histological tissue type [39]. Subsequent studies have confirmed the good performance of 1 H MRS to classify brain tumours [40–42]. Recently, techniques to decompose the 1 H MR spec- tra into biologically meaningful components have been introduced. One powerful technique to this end is the independent component analysis (ICA) [43]. Biological systems, such as brain tumours, are regarded as linear combinations of spectra from different tissue (cell) types within the voxel. Using ICA for 1 H MRSI data it was observed that spectra from seven distinct histo- logical brain tumour types can be described by maxi- mally four ICA components (Fig. 1A, for an example) [44]. Available ICA algorithms are capable of handling standard in vivo MRS data which still possess signifi- cant unavoidable variation in signal-to-noise ratio, line width and line shape within the data matrix (Fig. 1A). Using these components images were generated for the distribution of these IC types within each tumour (Fig. 1B). This type of information may turn out to be clinically relevant, as it may show the growth pattern of tumour in situ, as well as being able to distinguish high grade gliomas [44]. Impact of 1 H MRS information in clinical manage- ment of brain tumour patients is increasing [25]. A concerted European network has introduced a compu- ter-based decision supporting system for clinical diag- nosis of brain tumours [45]. The goal of this project is to develop a fully automated system using 1 H MRS(I) data acquired with any of the commercial clinical scan- ners as input for diagnosis of brain tumours [42]. It has become evident that there are additional relevant aspects available from 1 H MRS data for patient man- agement. It has been shown that the volume of meta- bolic abnormality in 1 H MRSI [46] and presence of 1 H MRS lipids in tumour tissue provide prognostic information [47]. 1 H MRS distribution of CCM, Cre and Lac ⁄ Lip [47,48] and the presence of specific IC components above [44] are indicative for brain tumour invasiveness, which can be used for individual therapy planning. Furthermore, spectroscopy data are used to J. L. Griffin and R. A. Kauppinen A metabolomics perspective of brain tumours FEBS Journal 274 (2007) 1132–1139 ª 2007 The Authors Journal compilation ª 2007 FEBS 1135 assess response to therapy allowing adjustment of treatment protocol [25]. 13 C MRS 13 C MRS is a powerful technique for metabolic assess- ment of tumours, because both glycolytic and oxida- tive metabolism of glucose can be estimated in the same experiment. The switch from oxidative to ‘anabo- lic’ glucose metabolism (involving glucose carbon shunting for nucleic acid synthesis) is one of the char- acteristics of cancer cells [49]. Until now 13 C MRS has been used only in experimental brain tumours [50,51]. However, the approach provides a wealth of informa- tion such as the metabolic activity of the lactate pool, the intracellular location of this pool and the relative rates of glycolysis and oxidative metabolism in these tumours [49–51]. Paediatric brain tumours Brain tumours in paediatric patients are proportionally much more common malignancies diagnosed in this age group than those in adults. A large body of paediatric brain tumours show low degree of malignancy and therefore respond to therapy, but their anatomical local- ization, often adjacent to vital structures, makes diagno- sis challenging. Histologically similar tumour types to those in adults, such as benign and malignant astrocyto- mas, and dissimilar ones, such as primitive neuro- ectodermal tumours (PNET), neuroblastomas and retinobaslatomas, are found. What has been found metabolically by 1 H MRS from adult brain tumours appears to hold also for paediatric cases. It is interesting to note that paediatric brain tumours, irrespective of originating cell type, show absence of NAA [27,52,53]. This indicates that only differentiated neural cells are able to express NAA. Low Cre and high CCM are asso- ciated with high grade of tumour [27,53,54]. Consistent with adult brain tumour studies, decline in CCM and appearance of Lip are signs of response to therapy [53]. A recent study of paediatric brain tumour patients demonstrated that more detailed biochemical informa- tion from CCMs by 31 P MRS can aid in assessment of prognosis [27]. High CCM detected by 1 H MRS in a variety of paediatric tumour types and grades can be analysed at the level of individual phosphorylated cho- line moiety containing compounds by 1 H-decoupled 31 P MRS. It was observed that PC ⁄ GPC and PE ⁄ GPE ratios are very high in PNET relative to several other tumours [27]. This pattern of large phosphomonoester content has been implicated to highly malignant tumours [26], and thus, multinuclear MRS may be Cho (a) (b) (c) (d) (e) B A C Cre Naa Lac/Lip 2.0 1.0 0 p.p.m. 3.0 Fig. 1. (A) 1 H MRS spectrum of a human glioblastoma (a), a calcula- ted composite spectrum (b) and three independent components (IC) (c–e) obtained from the acquired spectrum using the ICA are shown. Components contain metabolites as follows: IC-c, Lac ⁄ Lip only; IC-d, Choline containing compounds (Cho), Cre and small NAA and Lac ⁄ Lip peaks; and IC-e, Cho, Cre and NAA. (B) A topographic distribution of IC-d and (C) of IC-c from 1 H MRSI data set are shown superimposed on a Gd-enhanced T1-weighted MR image. Reproduced with permission from [44]. A metabolomics perspective of brain tumours J. L. Griffin and R. A. Kauppinen 1136 FEBS Journal 274 (2007) 1132–1139 ª 2007 The Authors Journal compilation ª 2007 FEBS able to provide accurate diagnostic and prognostic information. Future directions Aspirations of molecular medicine MRS is advancing translation of metabolonomics into clinical manage- ment of brain tumour patients. In several specialized centres 1 H MRS(I), by complementing advanced MRI examinations, are used in diagnosis, therapy planning and treatment follow-up [25,27,54]. It is envisaged that the need for invasive diagnostic biopsies will inevitably decline. This development can be regarded as logic in the flow of new methods for tumour diagnosis. In the pursuit morphological analysis using histological meth- ods has been complemented with, or even replaced by, immunological analysis of tumour types. This step has made classification of tumours more accurate and spe- cific. More recently, gene and protein expression pro- files have been added to tumour typing. We believe the metabolomics approach, involving not only 1 H MRS, but also 31 P and 13 C MRS in vivo, will become a field in its own right to be used for diagnostic, treatment planning, and monitoring treatment of these devasta- ting tumours. The current direction of increasing the field strength of clinical magnets improves both sensi- tivity of detecting metabolites and spectral resolution. New data acquisition methods, including parallel ima- ging [55] and nuclear hyperpolarization techniques for 13 C of metabolic substrates [56] will speed up MRS measurements. Finally, MS will increasingly play a role in ex vivo cancer metabolomics. One exciting possibility for met- abolomic based histology is to perform MALDI MS to produce an image of a tissue section which repre- sents certain metabolites. This is already being used in cancer cell proteomics as well as certain metabolomic experiments [57]. Acknowledgements Supported by the Royal Society (JLG), the Finnish Cancer Foundation (RAK) and Academy of Finland (RAK). References 1 Oliver SG, Winson MK, Kell DB & Baganz F (1998) Systematic functional analysis of the yeast genome. Trends Biotechnol 16, 373–378. 2 Nicholson JK, Lindon JC & Holmes E (1999) ‘Metabo- nomics’: understanding the metabolic responses of living systems to pathophysiological stimuli via multivariate statistical analysis of biological NMR spectroscopic data. Xenobiotica 29, 1181–1189. 3 Griffin JL & Shockcor JP (2004) Metabolic profiles of cancer cells. Nat Rev Cancer 4, 551–561. 4 Urenjak J, Williams SR, Gadian DG & Noble M (1993) Proton nuclear magnetic resonance spectroscopy unambiguously identifies different neural cell types. J Neurosci 13, 981–989. 5 Florian CL, Preece NE, Bhakoo KK, Williams SR & Noble M (1996) Charateristic metabolic profiles revealed by 1 H NMR spectroscopy for three types of human brain and nervous sytem tumours. NMR Biomed 8, 253–264. 6 Gill SS, Porteous R, Small RK, Thomas DGT, Patel P, Van Bruggen N, Gadian DG, Kauppinen RA & Wil- liams SR (1989) Brain metabolites as 1 H NMR markers of neuronal and glial disorders. NMR Biomed 2, 196– 200. 7 Peeling J & Sutherland GR (1992) High-resolution 1 H NMR spectroscopy studies of extracts of human cerebral neoplasms. Magn Reson Med 24, 123–136. 8 Hagberg G (1998) From magnetic resonance spectroscopy to classification of tumors. A review of pattern recognition methods. NMR Biomed 11, 148–156. 9 El-Deredy W, Ashmore SM, Branston NM, Darling JL, Williams SR & Thomas DG (1997) Pretreatment predic- tion of the chemotherapeutic response of human glioma cell cultures using nuclear magnetic resonance spectro- scopy and artificial neural networks. Cancer Res 57, 4196–4199. 10 Gray HF, Maxwell RJ, Martinez-Perez I, Arus C & Cerdan S (1998) Genetic programming for classification and feature selection: analysis of 1 H nuclear magnetic resonance spectra from human brain tumour biopsies. NMR Biomed 11, 217–224. 11 Diem M, Boydston-White S & Chiriboga L (1999) Infrared spectroscopy of cells and tissues: shining lights onto a novel subject. Appl Spectr 53, 148A–161A. 12 Boustany NN, Crawford JM, Manoharan R, Dasari RR & Feld MS (1999) Analysis of nucleotides and aromatic amino acids in normal and neoplastic colon mucosa by ultraviolet resonance raman spectroscopy. Lab Invest 79, 1201–1214. 13 Jellum E, Bjornson I, Nesbakken R, Johansson E & Wold S (1981) Classification of human cancer cells by means of capillary gas chromatography and pattern recognition analysis. J Chromatogr 217, 231–237. 14 Sugita Y, Yamada S, Sugita S, Sakata K, Morimatsu M & Shigemori M (2001) The biochemical analysis of neurotransmitters in central neurocytomas. Int J Mol Med 7, 521–525. 15 Olsen P, Rasmussen M, Zhu W, Tonnesen E & Stefano GB (2005) Human gliomas contain morphine. Med Sci Monit 11, MS18–21. J. L. Griffin and R. A. Kauppinen A metabolomics perspective of brain tumours FEBS Journal 274 (2007) 1132–1139 ª 2007 The Authors Journal compilation ª 2007 FEBS 1137 16 Bieberich E, Freischutz B, Suzuki M & Yu RK (1999) Differential effects of glycolipid biosynthesis inhibitors on ceramide-induced cell death in neuroblastoma cells. J Neurochem 72, 1040–1049. 17 Miller BL, Chang L, Booth R, Ernst T, Cornford M, Nikas D, Mcbride D & Jenden DJ (1996) In vivo 1 H MRS choline: correlation with in vitro chemistry ⁄ histology. Life Sci 58, 1929–1935. 18 Bhakoo KK, Williams SR, Florian CL, Land H & Noble MD (1996) Immortalization and transformation are associated with specific alterations in choline meta- bolism. Cancer Res 56, 4630–4635. 19 Cheng LL, Anthony DC, Comite AR, Black PM, Tzika AA & Gonzalez RG (2000) Quantification of microheterogeneity in glioblastoma multiforme with ex vivo high-resolution magic-angle spinning (HRMAS) proton magnetic resonance spectroscopy. Neuro-Oncol 2, 87–95. 20 Usenius JP, Vainio P, Hernesniemi J & Kauppinen RA (1994) Choline-containing compounds in human astro- cytomas studied by H-1 NMR spectroscopy in vivo and in vitro. J Neurochem 63, 1538–1543. 21 Sullards MC, Wang E, Peng Q & Merrill AH Jr (2003) Metabolomic profiling of sphingolipids in human glioma cell lines by liquid chromatography tandem mass spectrometry. Cell Mol Biol (Noisy-le-Grand) 49, 789– 797. 22 Maxwell RJ, Martinez-Perez I, Cerdan S, Cabanas ME, Arus C, Moreno A, Capdevila A, Ferrer E, Bartomeus F, Aparicio A, et al. (1998) Pattern recognition analysis of 1 H NMR spectra from perchloric acid extracts of human brain tumor biopsies. Magn Reson Med 39, 869–877. 23 Gill SS, Thomas DGT, van Bruggen N, Gadian DG, Peden CJ, Bell JD, Cox JD, Menon DK, Iles RA, Bryant DJ et al. (1990) Proton MR spectroscopy of intracranial tumours: in vivo and in vitro studies. J Comput Assist Tomogr 14, 497–504. 24 Usenius JPR, Kauppinen RA, Vainio PA, Hernesniemi JA, Vapalahti MP, Paljarvi LA & Soimakallio S (1994) Quantitative metabolite patterns of human brain tumors: Detection by H-1 NMR spectroscopy in vivo and in vitro. J Comput Assist Tomogr 18, 705–713. 25 Nelson SJ & Cha S (2003) Imaging glioblastoma multi- forme. Cancer J 9, 134–145. 26 Podo F (1999) Tumour phospholipid metabolism. NMR Biomed 12, 413–439. 27 Albers MJ, Krieger MD, Gonzalez-Gomez I, Gilles FH, Mccomb JG, Nelson MD Jr & Bluml S (2005) Proton- decoupled 31 P MRS in untreated pediatric brain tumors. Magn Reson Med 53, 22–29. 28 Segebarth CM, Baleriaux DF, De Beer R, Van Ormondt D, Marien A, Luyten PR & Den Hollander JA (1989) 1 H image-guided localized 31 P MR spectro- scopy of human brain: quantitative analysis of 31 PMR spectra measured on volunteers and on intracranial tumor patients. Magn Reson Med 11, 349–366. 29 Segebarth CM, Baleriaux DF, Arnold DL, Luyten PR & Den Hollanter JA (1987) MR image-guided 31 PMR spectroscopy in the evaluation of brain tumor treat- ment. Radiology 165, 215–219. 30 Heindel W, Bunke J, Glathe S, Steinbrich W & Mollevanger L (1988) Combined 1 H-MR imaging and localized 31 P-spectroscopy of intracranial tumors in 43 patients. J Comput Assist Tomogr 12, 907–916. 31 Hubesch B, Sappey-Marinier D, Roth K, Meyerhoff DJ, Matson GB & Weiner MW (1990) P-31 MR spec- troscopy of normal human brain and brain tumors. Radiology 174, 401–409. 32 Frahm J, Bruhn H, Gyngell ML, Merboldt KD, Ha ¨ nicke W & Sauter R (1989) Localized high-resolution proton NMR spectroscopy using stimulated echoes: Initial applications to human brain in vivo. Magn Reson Med 9, 79–93. 33 Saunders DE, Howe FA, van den Boogaart A, Mclean MA, Griffiths JR & Brown MM (1995) Continuing ischemic damage after acute middle cerebral artery infarction in humans demonstrated by short-echo proton spectroscopy. Stroke 26, 1007–1013. 34 Poptani H, Kaartinen J, Gupta RK, Niemitz M, Hiltunen Y & Kauppinen RA (1999) Diagnostic assessment of brain tumours and non-neoplastic brain disorders in vivo using proton nuclear magnetic resonance spectroscopy and artificial neural networks. J Cancer Res Clin Oncol 125, 343–349. 35 Caramanos Z, Narayanan S & Arnold DL (2005) 1H-MRS quantification of tNA and tCr in patients with multiple sclerosis: a meta-analytic review. Brain 128, 2483–2506. 36 Florian CL, Preece NE, Bhakoo KK, Williams SR & Noble MD (1995) Cell type-specific fingerprinting of meningioma and meningeal cells by proton nuclear magnetic resonance spectroscopy. Cancer Res 55, 420–427. 37 Hagberg G, Burlina AP, Mader I, Roser W, Radue EW & Seelig J (1995) In vivo proton MR spectroscopy of human gliomas: definition of metabolic coordinates for multi-dimensional classification. Magn Reson Med 34, 242–252. 38 Usenius J-P, Tuohimetsa ¨ S, Vainio P, Ala-Korpela M, Hiltunen Y & Kauppinen RA (1996) Automated classification of human brain tumours by neural network using in vivo 1 H magnetic resoance spectro- scopic metabolite phenotypes. Neuroreport 7, 1597– 1600. 39 Preul MC, Caramanos Z, Collins DL, Villemure JG, Leblanc R, Olivier A, Pokrupa R & Arnold DL (1996) Accurate, noninvasive diagnosis of human brain tumors by using proton magnetic resonance spectroscopy. Nat Med 2, 323–325. A metabolomics perspective of brain tumours J. L. Griffin and R. A. Kauppinen 1138 FEBS Journal 274 (2007) 1132–1139 ª 2007 The Authors Journal compilation ª 2007 FEBS 40 Tate AR, Griffiths JR, Martinez-Perez I, Moreno A, Barba I, Cabanas ME, Watson D, Alonso J, Bartumeus F, Isamat F et al. (1998) Towards a method for auto- mated classification of 1 H MRS spectra from brain tumours. NMR Biomed 11, 177–191. 41 De Edelenyi FS, Rubin C, Esteve F, Grand S, Decorps M, Lefournier V, Le Bas JF & Remy C (2000) A new approach for analyzing proton magnetic resonance spectroscopic images of brain tumors: nosologic images. Nat Med 6, 1287–1289. 42 Tate AR, Majos C, Moreno A, Howe FA, Griffiths JR & Arus C (2003) Automated classification of short echo time in in vivo 1 H brain tumor spectra: a multicenter study. Magn Reson Med 49, 29–36. 43 Hyvarinen A, Karhunen J & Oja E (2001) Independent Component Analysis John Wiley & Son, New York, NY. 44 Pulkkinen J, Hakkinen AM, Lundbom N, Paetau A, Kauppinen RA & Hiltunen Y (2005) Independent com- ponent analysis to proton spectroscopic imaging data of human brain tumours. Eur J Radiol 56, 160–164. 45 Underwood J, Tate AR, Luckin R, Majos C, Capdevila A, Howe FA, Griffiths JR & Arus C (2001) A prototype decision support system for MR spectroscopy-assisted diagnosis of brain tumours. 10th World Conference of Medical Informatics MEDINFO 2001 pp. 561–565. IOS Press, Amsterdam. 46 Oh J, Henry RG, Pirzkall A, Lu Y, Li X, Catalaa I, Chang S, Dillon WP & Nelson SJ (2004) Survival analy- sis in patients with glioblastoma multiforme: predictive value of choline-to-N-acetylaspartate index, apparent diffusion coefficient, and relative cerebral blood volume. J Magn Reson Imaging 19, 546–554. 47 Murphy PS, Rowland IJ, Viviers L, Brada M, Leach MO & Dzik-Jurasz AS (2003) Could assessment of glioma methylene lipid resonance by in vivo 1 H-MRS be of clinical value? Br J Radiol 76, 459–463. 48 Li X, Vigneron DB, Cha S, Graves EE, Crawford F, Chang SM & Nelson SJ (2005) Relationship of MR- derived lactate, mobile lipids, and relative blood volume for gliomas in vivo. AJNR Am J Neuroradiol 26, 760– 769. 49 Boros LG, Lee WN & Go VL (2002) A metabolic hypothesis of cell growth and death in pancreatic can- cer. Pancreas 24, 26–33. 50 Terpstra M, Gruetter R, High WB, Mescher M, Dela- barre L, Merkle H & Garwood M (1998) Lactate turn- over in rat glioma measured by in vivo nuclear magnetic resonance spectroscopy. Cancer Res 58, 5083–5088. 51 Pfeuffer J, Lin JC, Delabarre L, Ugurbil K & Garwood M (2005) Detection of intracellular lactate with loca- lized diffusion { 1 H- 13 C}-spectroscopy in rat glioma in vivo. J Magn Reson 177, 129–138. 52 Tzika AA, Zurakowski D, Poussaint TY, Goumnerova L, Astrakas LG, Barnes PD, Anthony DC, Billett AL, Tarbell NJ, Scott RM et al. (2001) Proton magnetic spectroscopic imaging of the child’s brain: the response of tumors to treatment. Neuroradiology 43, 169–177. 53 Lindskog M, Spenger C, Klason T, Jarvet J, Graslund A, Johnsen JI, Ponthan F, Douglas L, Nordell B & Kogner P (2005) Proton magnetic resonance spectro- scopy in neuroblastoma: current status, prospects and limitations. Cancer Lett 228, 247–255. 54 Astrakas LG, Zurakowski D, Tzika AA, Zarifi MK, Anthony DC, De Girolami U, Tarbell NJ & Black PM (2004) Noninvasive magnetic resonance spectroscopic imaging biomarkers to predict the clinical grade of pediatric brain tumors. Clin Cancer Res 10, 8220–8228. 55 Pruessmann KP (2004) Parallel imaging at high field strength: synergies and joint potential. Top Magn Reson Imaging 15, 237–244. 56 Ardenkjaer-Larsen JH, Fridlund B, Gram A, Hansson G, Hansson L, Lerche MH, Servin R, Thaning M & Golman K (2003) Increase in signal-to-noise ratio of > 10,000 times in liquid-state NMR. Proc Natl Acad Sci USA 100, 10158–10163. 57 Chaurand P, Schwartz SA, Reyzer ML & Caprioli RM (2005) Imaging mass spectrometry: principles and poten- tials. Toxicol Pathol 33, 92–101. J. L. Griffin and R. A. Kauppinen A metabolomics perspective of brain tumours FEBS Journal 274 (2007) 1132–1139 ª 2007 The Authors Journal compilation ª 2007 FEBS 1139 . range of brain tumours. Lactate Lactate is an end product of glycolysis and increases rapidly during hypoxia and ischaemia, in particular as a result of poor. have dem- onstrated metabolite abnormalities in brain tumours that discriminate them from normal brain tissue. Human brain tumours in vivo Human brain tumours