1 H-Nuclear Magnetic Resonance Spectroscopy of Glioblastoma Cancer Stem Cells

Abstract

The noninvasive tracking of glioblastoma cancer stem cells (CSCs) in vivo constitutes a prerequisite for the development of CSC-specific therapies. Therefore, as a pilot study to identify CSC biomarkers for clinical magnetic resonance spectroscopy, 10 CSC lines were investigated using high-resolution ¹H-nuclear magnetic resonance (¹H-NMR) spectroscopy at 600 and 800 MHz (14.4 and 18.8 T) under reproducible in vitro conditions. The spectra were analyzed using principal component analysis (PCA), and spectral regions of high variability were evaluated regarding correlations to stem cell-related properties (clonogenic index and CD133 positivity) and cell death. PCA revealed that duplicates of CSC lines clustered together suggesting a characteristic ¹H-NMR pattern of each CSC line. PCA enabled discrimination between samples with high and low clonogenicity, that is, clustering according to one of the hallmarks of stemness in samples with high viability. High/moderate correlations to clonogenicity and CD133 were found in spectral regions with high variability. In contrast, the mobile lipid signal at 1.28 ppm correlated to cell death, but not to stemness, as published previously for neural progenitor cells. In conclusion, our exploratory study demonstrates the correlation of specific resonances within ¹H-NMR spectra with stem cell properties and advocates the use of the 1.28 ppm resonance as biomarker for cell death also in CSCs.

Introduction

H uman glioblastoma (GBM) has a median survival of less than 15 months despite intensive therapy [1,2]. Increasing evidence indicates that GBM is driven by a rare subpopulation of stem cell-like or glioma-initiating cells [here referred to as cancer stem cells (CSCs)] that maintain tumor growth [3]. On a cellular level, the presence of CSCs in GBM indicates a dismal prognosis of patients [4 –6]. Thus, noninvasive imaging modalities that unequivocally identify CSCs would improve prediction of disease progression and monitoring of new therapeutic approaches targeting CSCs and are therefore intensively sought for.

Although recent advances have been made in imaging of nonmalignant stem cells in the CNS using either optical bioluminescent imaging [7] or 3′-deoxy-3′-l-fluorothymidine positron emission tomography [8], no valid imaging method allows identifying CSC in vivo or in vitro. Stem cells display characteristic biological developmental stages, that is, proliferation, self-renewal, differentiation into specific cell types, or death [3]. Such cellular stages are often imprinted at the metabolic level, making the cell's metabolism an interesting target for the identification of CSC activity. Notably, the polar character of stem cells, which is a prerequisite of asymmetric cell division, suggests a distinct composition of stem cell membranes and thus a specific pattern of the lipid metabolism when compared with nonstem cells [9,10]. ¹H-nuclear magnetic resonance (¹H-NMR) spectroscopy is an increasingly available method to investigate the metabolic features of brain tumors. As an example, the choline concentration correlates with the proliferation index of GBM and is widely used to estimate the vitality of a region of interest (ROI), especially after therapy [11,12]. However, it is unknown whether CSC-specific resonances in ¹H-NMR spectroscopy exist. Here, we aimed at the spectroscopic characterization and correlation to biological properties of CSC-enriched cell lines cultured under serum-free conditions, to provide data for the implementation of ¹H-NMR spectroscopy of CSCs for translation into clinical approaches.

Experimental

CSC culture

All CSC lines used have been previously described [13,14]. In brief, freshly resected human primary GBM samples were dissociated and cultured in medium containing epidermal growth factor, basic fibroblast growth factor (both from R&D Systems), leukemia inhibitory factor (Chemicon Internation), and B27 supplement (1:50; Gibco). Medium was changed twice weekly and cells were continuously propagated. To determine the clonogenicity of CSC lines, single-cell suspensions were seeded at a density of 10 cells per well in 96-well plates and numbers of free floating or adherent spheres were determined after 21–28 days. Use of resection material and all experiments were approved by the local ethics committee.

Flow cytometry

CD133 expression was determined as previously described [13] using CD133/2-PE (clone 273; Miltenyi Biotech) or control IgG. The fraction of apoptotic and necrotic cells was determined by combined Annexin V-FITC and propidium iodide staining. Briefly, 2×10⁵ cells were washed twice in phosphate-buffered saline (PBS), resuspended in binding buffer [10 mM HEPES/NaOH (pH 7.4), 140 mM NaCl, 2.5 mM CaCl₂), stained with 5 μL Annexin V-FITC (BD Biosciences), and incubated for 20 min in the dark at room temperature. Subsequently, propidium iodide (0.5 μg/mL) was added and the samples were analyzed on a BD FACSCalibur.

NMR spectroscopy

For ¹H-NMR spectroscopy, 3–5 million cells (dependent on the size of the specific cells) per sample were washed twice in PBS, resuspended in 500 μL PBS solution containing 10% deuterium oxide and 40 μM dimethyl-silapentane-sulfonate (DSS), and transferred to a 5-mm Shigemi NMR tube that allows sedimentation within the sensitive volume of the probe coil because of a susceptibility-matched solid glass bottom. Samples were cooled to 5°C and ¹H-NMR measurement started within less than 15 min thereafter. During measurement, the temperature was kept at 5°C. Measurements were performed using high-resolution NMR Bruker Avance 600 and 800 MHz spectrometers employing a gradient-based water suppression pulse sequence [15]. For 1-dimensional experiments, 64 scans with 64k data points and 4.7 s repetition time were accumulated, followed by an exponential line broadening of 2 Hz. After Fourier transformation (32k data points), the spectra were phase- and baseline-corrected manually. DSS was used as an internal reference standard (0.00 ppm). An additional ¹H-NMR spectrum of the extracellular compartment of the cell suspension was obtained by carefully collecting the supernatant. Acquisition and processing procedures were used as for cell suspensions.

Assignment strategies

Resonances of small molecules and metabolites were assigned comparing chemical shifts, multiplet structures, and homonuclear couplings with 1- and 2-dimensional reference spectra from the databases HMDB [16], BMRB [17], SDBS (http://riodb01.ibase.aist.go.jp/sdbs/cgi-bin/cre_index.cgi), data from the literature [18 –21], and 2-dimensional total correlation spectroscopy (TOCSY). For assignment of macromolecular resonances, data from the literature on mobile lipids [22 –30] and mobile proteins [31 –35] were compared with our measurements with respect to chemical shifts, line shapes, and TOCSY cross peaks. In line with most publications, the mobile protein resonances were named M1 (1.05–0.80 ppm), M2 (1.22 ppm), M3 (1.40 ppm), M5 (2.05 ppm), and M6 (2.29 ppm). The mobile lipid peaks at 0.89, 1.28, 2.02, and 2.24 ppm were denoted ML1, ML2, ML3, and ML4, respectively.

Principal component analysis

For principal component analysis (PCA), the spectral region 4.20–0.70 ppm was used. Within this region, the following contaminants were excluded: DSS (1.80–1.72 and 2.94–2.88 ppm), acetate (1.93–1.89 ppm), and dimethyl sulfoxide (2.74–2.71 ppm). For each spectrum, 350 buckets (0.01 ppm bucket width) were integrated and scaled to the total intensity. Loadings, scores, and variances were calculated using a Matlab (The MathWorks) standard PCA routine (“princomp”).

Correlation analysis

For correlation analysis, the same spectral range, exclusions, and normalization were employed as for PCA (vide supra). Spectral regions with high variations across the samples (high standard deviations) were chosen as ROI for correlation analyses: 4.08–4.02 ppm, 3.8–3.73 ppm, 3.64–3.5 ppm, 3.23–3.17 ppm, 3.06–2.98 ppm, 2.48–2.4 ppm, 2.38–2.28 ppm, 2.15–2.1 ppm, 2.08–1.97 ppm, 1.48–1.45 ppm, 1.44–1.34 ppm, 1.33–1.305 ppm, 1.295–1.15 ppm, 0.92–0.8 ppm. Pearson's correlation coefficients between the ROI integrals and the available biological data (clonogenic index, cell death, CD133) were calculated using a standard Matlab routine (“corr”).

Mobile lipid quantification

The quantification of the macromolecular peak at 1.28 ppm (ML2) was obtained by deconvolution of the spectral region between 1.5 and 0.7 ppm using a Matlab fitting routine. In brief, the mobile lipid resonance (ML2) and the macromolecular multiplet between 1.05 and 0.80 ppm (M1+ML1) were separated from overlapping multiplets of small molecules such as lactate and free amino acids and from other underlying macromolecular residues (eg, M2). As previously described [22 –30], the mobile lipid biomarker was quantified as the ratio of ML2 and M1+ML1.

Results

¹H-NMR characteristics of CSC lines

We characterized the ¹H-NMR spectra of 10 different CSC lines derived from primary GBM samples [4,13]. The stem cell-related properties of the CSC lines were characterized in detail and showed a high variability of clonogenicity, CD133 expression, growth pattern, rates of spontaneous apoptosis, and cell death (Table 1). To compare ¹H-NMR spectra with CSC lines, we measured CSC lines in 2 independent duplicates. Although the biological properties of the CSC lines remained stable (Fig. 1A), the ¹H-NMR spectra showed a high variability between both measurements. Figure 1B gives the average variability between 0.7 and 4.2 ppm of both measurement series.

FIG. 1.

¹H-nuclear magnetic resonance characteristics of CSC lines. (A) The clonogenic index of CSC lines remains stable after repetitive measurements. The clonogenic indices of the first and second measurement series of each cell line are correlated with R=0.721. (B) Variability within the spectra (10 cell lines, N=2): Black line: Mean of spectra normalized to integral of the spectral region between 4.20 and 0.70 ppm. Dark gray haze: Standard deviation unveiling variable resonances within the spectra. Light gray rectangles: Spectral region of interest with high variability for correlation analysis. Peak labeling according to the spectral assignment strategy described in the Experimental section. Abbreviations are given in the Experimental section. (C) Comparison of nuclear magnetic resonance spectra of whole-cell suspensions and of supernatants collected after acquisition (data of CSC line R44 is given). Almost all metabolites giving rise to sharp, highly resolved resonances can also be detected in the supernatant. Asterisk: impurities within the supernatant caused by collection (Pasteur pipette). Hash: residual intracellular DMSO. CSC, cancer stem cell; DMSO, dimethyl sulfoxide.

Table 1.

Biological Properties of the Cancer Stem Cell Lines (First Test Series)

CSC lines	Clonogenic cells (%)	CD133 (%)	Apoptotic cells (%)	Cell death (%)	Viable cells (%)
R8	12.50	0.10	4.70	12.40	82.90
R11	14.60	3.50	10.40	14.70	74.90
R18	2.00	22.40	0.30	31.50	68.40
R28	21.00	6.30	4.30	33.30	63.50
R40	29.00	44.00	1.40	5.00	93.50
R44	0.00	16.60	n.d.	50.00	50.00
R49	16.70	4.90	15.70	43.10	41.10
R52	0.00	0.00	6.80	14.00	79.60
R53	12.50	0.00	1.00	60.00	30.70
R54	16.70	45.70	n.d.	56.30	42.70

CSC, cancer stem cell.

In the ¹H-NMR spectra, the assignment strategies resulted in the following unambiguous identification of resonances (abbreviations and chemical shifts in parentheses): methyl of lipids of branched-chain amino acids (val, leu, ile) and amino acid residues of proteins (M1+ML1, 1.10–0.70 ppm), mobile lipid fatty acid chain methylene (ML2, 1.28 ppm), lactate (lac, 1.31 ppm, 4.11 ppm), alanine residues of proteins (M3, 1.40 ppm), free alanine (ala, 1.47 ppm), acetate (ace, 1.91 ppm), creatine-containing compounds (tCre, 3.03 ppm), choline-containing compounds (tCho, 3.21 ppm), myo-inositol (m-Ino, 3.26 ppm, 3.53 ppm, 3.61 ppm, 4.05 ppm), and glutamate/glutamine (glx, 3.77 ppm) (labeled in Fig. 1B). The spectral region 2.5–2.0 ppm was assigned to glutamine (2.43 ppm, 2.13 ppm), glutamate (glu, 2.34 ppm, 2.08 ppm), mobile protein M6 (2.29 ppm) and M5 (2.05 ppm), and mobile lipid fatty acid chain methylene (ML3, 2.02 ppm and ML4, 2.24 ppm).

Additional spectroscopy of the supernatant after acquiring the suspension spectrum revealed that most sharp resonances found in suspension spectra originated also from the extracellular space of the suspension (Fig. 1C).

Cluster discrimination by PCA

The scores plot of the first principal components (eg, PC2 vs. PC3 in Fig. 2B) did not show a distinct clustering of the 10 different CSC cell lines, but a relative closeness between each pair of measurements. For further cluster discrimination for stem cell-related properties, we excluded samples with a high proportion of dead cells (>40%, 6 samples), which may bias or mask relevant cluster. Based on the first 5 principal components, PCA scores plots broken down to samples' clonogenic index and CD133 expression were evaluated and revealed virtually no cluster discrimination between samples with high or low (ie, higher/lower than median) CD133 expression. However, PCA resulted in a clustering of samples with a clonogenic index higher/lower than the median in the 2-dimensional scores plot of principal components 1 and 3 (Fig. 2C), suggesting the presence of CSC-associated patterns within the NMR spectra. From the loadings plot of the first 3 principal components (Fig. 2A), the high-clonogenicity clustering can be attributed to high levels of M2, ML2, M3, glu, M6, and ML4 and low levels of lac and m-Ino.

FIG. 2.

Principal component analysis outcome. (A) Scores of principal components 2 and 3 of 10 different CSC lines in duplicates. Each symbol type represents 1 CSC line. (B) Scores of principal components 1 and 3 of the subset of samples with percentage of dead cells <40% (N=14), broken down into samples with clonogenic index higher or lower than the median. (C) Loadings of the principal component analysis. Shaded regions were excluded.

Correlations between NMR spectra and biological properties of CSC lines

The high biological variability between CSC lines allowed a correlation analysis between biological properties and NMR spectra to generate hypotheses on the functional relevance of spectral regions. High/moderate correlations (I) to clonogenicity were found for spectral regions 2.38–2.28 ppm (glu, M6, ML4, R=0.65, P=0.002), 2.08–1.97 ppm (glu, gln, M5, ML3, R=0.51, P=0.023), and 1.48–1.45 ppm (ala, R=0.53, P=0.015), (II) to CD133 for 3.80–3.73 ppm (glx, R=−0.54, P=0.014) and 1.44–1.34 ppm (M3, ML2, R=0.55, P=0.012), and (III) to cell death for 1.29–1.15 ppm (ML2, R=0.43, P=0.059) and 0.92–0.80 ppm (M1, ML1, R=0.52, P=0.019).

However, only the correlation between the spectral region 2.38–2.28 ppm and clonogenicity remained significant after Bonferroni correction for multiple testing at a significance level of P<0.05.

With respect to cell death, the spectral regions around the mobile lipid peaks (ML2 and ML1) yielded the highest correlation coefficients (R=0.43; R=0.52), corresponding well to an earlier publication of our group [30]. Because the mobile lipid peak at 1.28 ppm constitutes a putative marker for neural stem cells [36], we evaluated the correlation between the quantified mobile lipid peak obtained by the fitting/deconvolution procedure and the biological assay outcomes. This confirmed the high positive correlation of the 1.28 ppm resonance and cell death within CSC lines (R=0.54, P=0.014; Fig. 3A). In line with our previous report, there was virtually no correlation to CD133 or clonogenicity (R=−0.04, P=0.86 and R=−0.13, P=0.59 for CD133 and clonogenic index, respectively; Fig. 3B, C).

FIG. 3.

Correlation of mobile lipids and biological properties. Correlation analyses of the deconvolved and quantified mobile lipid peak at 1.28 ppm (ML2/ML1+M1) versus percentage of dead cells per sample (A), versus clonogenic index (B), and versus percentage of CD133-positive cells per sample (C).

Discussion

Although ¹H-NMR spectroscopy is a widely available modality for the noninvasive investigation of brain tumors, the overall informative value of spectra obtained by clinical MRI scanners remains limited because of restricted resolution, low sensitivity, and nonstandardized data evaluations. We therefore used high-resolution NMR spectroscopy and reproducible in vitro conditions to viable cells to characterize the spectra of CSCs to overcome these problems. This study provided a first proof of presence of CSC-associated pattern within ¹H-NMR spectra, and it also illustrates the difficulty to unambiguously detect CSCs using suspension NMR spectroscopy. Many well-defined resonances such as myo-inositol, cholines, creatines, amino acids, and lactate [37] also originated from the extracellular compartment (eg, leaked out from viable and necrotic cells, also addressed in Ref. [38]) and hardly correlated to stem cell-associated properties. In contrast, macromolecular resonances, that is, mobile lipids and mobile proteins, are located within viable cells but have several drawbacks: (i) their broad line shape and overlap with neighboring resonances urges for close to perfect signal-to-noise ratios and requires sophisticated postprocessing methods, for example, deconvolution/fitting routines, which lose their robustness when small broad signals are overlapped by several bigger peaks. (ii) Although several studies were published addressing the nature of macromolecular resonances both in vitro and in vivo [22 –36], neither the underlying mechanisms nor the biological/clinical relevance of their appearance or variations have been clarified in detail yet.

An additional problem for the definition of CSC-specific resonances remains the vague definition of CSCs. Although several new markers have been recently described [39], the most intensively investigated marker for GBM CSC, CD133, does rather indicate the lack of stemness than the presence of stem-cell like cells in a subgroup of CSC lines. As a consequence, there is no possibility to prospectively isolate pure CSCs, which impairs the search for CSC-specific resonances.

Despite these methodological problems, our study revealed some interesting features of ¹H-NMR spectroscopy on CSC suspensions: Different samples from the same CSC cell line showed similar spectral patterns, although some other parameters (eg, different percentages of dead cells) impeded a distinct clustering in the scores plots of the first principal components of PCA. CSC lines with a high viability (>60%) showed moderate clustering of samples according to their clonogenic indices in the scores plot of the first and the third principal component, suggesting that CSC-specific patterns exist. Although the statistical design of this exploratory study implies that the P value of 0.05 may also identify false-positive results, 1 correlation [spectral region 2.38–2.28 ppm (glu, M6, ML4) to clonogenicity (R=0.65)] remained significant even after Bonferroni correction for multiple testing and warrants further investigation in vitro and in vivo. Taken together, our study showed that ¹H-NMR spectroscopy may be suitable to differentiate between distinct CSC lines and to detect CSC-specific resonances. In addition, it confirmed that the previously described [36] putative stem cell-specific resonance at 1.28 ppm (ML2/M1+ML1) showed virtually no correlation with the number of clonogenic cells or the stem cell marker CD133. However, it was a good biomarker for cell death within the spectra, confirming previous results [30]. Therefore, it is tempting to speculate about the mobile lipid peak as a marker for cell death not only in differentiated glioma cells, but also in the CSC subfraction of solid tumors in vivo.

In addition, our pilot study suggests that ¹H-NMR spectra differentiate between distinct CSC lines and identify CSCs in vitro. In addition, we confirmed that the mobile lipid peak is suitable to detect cell death but not stem cells within human CSC lines. Some spectral ROIs (eg, 2.38–2.28 ppm) might serve as a future marker for stemness in GBM cell lines and possibly in vivo.

Footnotes

Acknowledgments

Parts of the project were performed within the NGFNplus Brain Tumor Network (Subproject 7 No. 01GS0887). Additional funding was provided by the Bavarian State Ministry of Sciences, Research, and the Arts (ForNeuroCell grant) and by EU FP6-project DiMI (LSHB-CT-2005-512146).

Author Disclosure Statement

All authors confirm no conflicts of interest to disclose.

References

Beier

, Schmid

, Gorlia

, Kleinletzenberger

, Beier

, Grauer

, Steinbrecher

, Hirschmann

, Brawanski

, Dietmaier

, Jauch-Worley

, Pietsch

, Proescholdt

, Rummele

, Muigg

, Stockhammer

, Hegi

, Bogdahn

, Hau

. 2009. RNOP-09: pegylated liposomal doxorubicine and prolonged temozolomide in addition to radiotherapy in newly diagnosed glioblastoma—a phase II study. BMC Cancer, 9:308.

Stupp

, Mason

, van den Bent

, Weller

, Fisher

, Taphoorn

, Belanger

, Brandes

, Marosi

, Bogdahn

, Curschmann

, Janzer

, Ludwin

, Gorlia

, Allgeier

, Lacombe

, Cairncross

, Eisenhauer

, Mirimanoff

. 2005. Radiotherapy plus concomitant and adjuvant temozolomide for glioblastoma. N Engl J Med, 352:987–996.

Reya

, Morrison

, Clarke

, Weissman

. 2001. Stem cells, cancer, and cancer stem cells. Nature, 414:105–111.

Beier

, Wischhusen

, Dietmaier

, Hau

, Proescholdt

, Brawanski

, Bogdahn

, Beier

. 2008. CD133 Expression and cancer stem cells predict prognosis in high-grade oligodendroglial tumors. Brain Pathol, 18:370–377.

Laks

, Masterman-Smith

, Visnyei

, Angenieux

, Orozco

, Foran

, Yong

, Vinters

, Liau

, Lazareff

, Mischel

, Cloughesy

, Horvath

, Kornblum

. 2009. Neurosphere formation is an independent predictor of clinical outcome in malignant glioma. Stem Cells, 27:980–987.

Pallini

, Ricci-Vitiani

, Banna

, Signore

, Lombardi

, Todaro

, Stassi

, Martini

, Maira

, Larocca

, De Maria

. 2008. Cancer stem cell analysis and clinical outcome in patients with glioblastoma multiforme. Clin Cancer Res, 14:8205–8212.

Couillard-Despres

, Winner

, Schaubeck

, Aigner

, Vroemen

, Weidner

, Bogdahn

, Winkler

, Kuhn

, Aigner

. 2005. Doublecortin expression levels in adult brain reflect neurogenesis. Eur J Neurosci, 21:1–14.

Rueger

, Backes

, Walberer

, Neumaier

, Ullrich

, Simard

, Emig

, Fink

, Hoehn

, Graf

, Schroeter

. 2010. Noninvasive imaging of endogenous neural stem cell mobilization in vivo using positron emission tomography. J Neurosci, 30:6454–6460.

Marthiens

, ffrench-Constant

. 2009. Adherens junction domains are split by asymmetric division of embryonic neural stem cells. EMBO Rep, 10:515–520.

10.

Gotz

, Huttner

. The cell biology of neurogenesis. 2005. Nat Rev Mol Cell Biol, 6:777–788.

11.

Shimizu

, Kumabe

, Shirane

, Yoshimoto

. 2000. Correlation between choline level measured by proton MR spectroscopy and Ki-67 labeling index in gliomas. AJNR Am J Neuroradiol, 21:659–665.

12.

Moller-Hartmann

, Herminghaus

, Krings

, Marquardt

, Lanfermann

, Pilatus

, Zanella

. 2002. Clinical application of proton magnetic resonance spectroscopy in the diagnosis of intracranial mass lesions. Neuroradiology, 44:371–381.

13.

Beier

, Hau

, Proescholdt

, Lohmeier

, Wischhusen

, Oefner

, Aigner

, Brawanski

, Bogdahn

, Beier

. 2007. CD133+ and CD133- Glioblastoma-derived cancer stem cells show differential growth characteristics and molecular profiles. Cancer Res, 67:4010–4015.

14.

Beier

, Rohrl

, Pillai

, Schwarz

Kunz- Schughart

, Leukel

, Proescholdt

, Brawanski

, Bogdahn

, Trampe-Kieslich

, Giebel

, Wischhusen

, Reifenberger

, Hau

, Beier

. 2008. Temozolomide preferentially depletes cancer stem cells in glioblastoma. Cancer Res, 68:5706–5715.

15.

Hwang

, Shaka

. 1995. Water suppression that works. excitation sculpting using arbitrary wave-forms and pulsed-field gradients. J Mag Res Series A, 112:275–279.

16.

Wishart

, Knox

, Guo

, Eisner

, Young

, Gautam

, Hau

, Psychogios

, Dong

, Bouatra

, Mandal

, Sinelnikov

, Xia

, Jia

, Cruz

, Lim

, Sobsey

, Shrivastava

, Huang

, Liu

, Fang

, Peng

, Fradette

, Cheng

, Tzur

, Clements

, Lewis

, De Souza

, Zuniga

, Dawe

, Xiong

, Clive

, Greiner

, Nazyrova

, Shaykhutdinov

, Li

, Vogel

, Forsythe

. 2009. HMDB: a knowledgebase for the human metabolome. Nucl. Acids Res, 37:D603–D610.

17.

Ulrich

, Akutsu

, Doreleijers

, Harano

, Ioannidis

, Lin

, Livny

, Mading

, Maziuk

, Miller

, Nakatani

, Schulte

, Tolmie

, Kent Wenger

, Yao

, Markley

. 2008. BioMagResBank Nucl Acids Res, 36:D402–D408.

18.

Jansen

JFA

, Shamblott

, Zijl

PCMv

, Lehtimäki

, Bulte

JWM

, Gearhart

, Hakumäki

. 2006. Stem cell profiling by nuclear magnetic resonance spectroscopy. Mag Res Med, 56:666–670.

19.

Urenjak

, Williams

, Gadian

, Noble

. 1993. Proton nuclear magnetic resonance spectroscopy unambiguously identifies different neural cell types. J Neurosci, 13:981–989.

20.

Kotitschke

, Tonn

, Goldbrunner

, Bogdahn

, Haase

. 1995. 1H NMR investigations of tumor spheroids grown from a human glioma biopsy or from a human malignant glioma cell line. J Mag Res Series B, 109:39–43.

21.

Kotitschke

, Jung

, Nekolla

, Haase

, Bauer

, Bogdahn

. 1994. High-resolution one- and two-dimensional 1H MRS of human brain tumor and normal glial cells. NMR Biomed, 7:111–120.

22.

Rosi

, Grande

, Luciani

, Barone

, Mlynarik

, Viti

, Guidoni

. 2004. 1H MRS studies of signals from mobile lipids and from lipid metabolites: comparison of the behavior in cultured tumor cells and in spheroids. NMR Biomed, 17:76–91.

23.

Hakumäki

, Kauppinen

. 2000. 1H NMR visible lipids in the life and death of cells. Trends Biochem Sci, 25:357–362.

24.

Quintero

, Cabanas

, Arus

. 2007. A possible cellular explanation for the NMR-visible mobile lipid (ML) changes in cultured C6 glioma cells with growth. Biochim Biophys Acta, 1771:31–44.

25.

Al-Saffar

NMS

, Titley

, Robertson

, Clarke

, Jackson

, Leach

, Ronen

. 2002. Apoptosis is associated with triacylglycerol accumulation in Jurkat T-cells. Br J Cancer, 86:963–970.

26.

Ferretti

, Knijn

, Iorio

, Pulciani

, Giambenedetti

, Molinari

, Meschini

, Stringaro

, Calcabrini

, Freitas

, Strom

, Arancia

, Podo

. 1999. Biophysical and structural characterization of 1H-NMR-detectable mobile lipid domains in NIH-3T3 fibroblasts. Biochim Biophys Acta, 1438:329–348.

27.

Rémy

, Fouilhé

, Barba

, Sam-Lai

, Lahrech

, Cucurella

, Izquierdo

, Moreno

, Ziegler

, Massarelli

, Décorps

, Aris

. 1997. Evidence that mobile lipids detected in rat brain glioma by 1H nuclear magnetic resonance correspond to lipid droplets. Cancer Res, 57:407–414.

28.

Barba

, Mann

, Cabanas

, Arus

, Gasparovic

. 2001. Mobile lipid production after confluence and pH stress in perfused C6 cells. NMR Biomed, 14:33–40.

29.

Delikatny

, Lander

, Jeitner

, Hancock

, Mountford

. 1996. Modulation of MR-visible mobile lipid levels by cell culture conditions and correlations with chemotactic response. Int J Cancer, 65:238–245.

30.

Ramm

, Couillard-Despres

, Plotz

, Rivera

, Krampert

, Lehner

, Kremer

, Bogdahn

, Kalbitzer

, Aigner

. 2009. A nuclear magnetic resonance biomarker for neural progenitor cells: is it all neurogenesis? Stem Cells, 27:420–423.

31.

Behar

, T

Ogino

. 1991. Assignment of resonance in the 1H spectrum of rat brain by two-dimensional shift correlated and J-resolved NMR spectroscopy. Mag Res Med, 17:285–303.

32.

Behar

, Ogino

. 1993. Characterization of macromolecule resonances in the 1H NMR spectrum of rat brain. Mag Res Med, 30:38–44.

33.

Behar

, Rothman

, Spencer

, Petroff

. 1994. Analysis of macromolecule resonances in 1H NMR spectra of human brain. Mag Res Med, 32:294–302.

34.

Kauppinen

, Niskanen

, Hakumäki

, Williams

. 1993. Quantitative analysis of 1H NMR detected proteins in the rat cerebral cortex in vivo and in vitro. NMR Biomed, 6:242–247.

35.

Kauppinen

, Nissinen

, Kärkkäinen

, Pirttilä

, Palvimo

, Kokko

, Williams

. 1992. Detection of thymosin beta 4 in situ in a guinea pig cerebral cortex preparation using 1H NMR spectroscopy. J Biol Chem, 267:9905–9910.

36.

Manganas

, Zhang

, Li

, Hazel

, Smith

, Wagshul

, Henn

, Benveniste

, Djuric

, Enikolopov

, Maletic-Savatic

. 2007. Magnetic resonance spectroscopy identifies neural progenitor cells in the live human brain. Science, 318:980–985.

37.

Wright

, Fellows

, Griffiths

, Wilson

, Bell

, Howe

. 2010. Ex-vivo HRMAS of adult brain tumours: metabolite quantification and assignment of tumour biomarkers. Mol Cancer, 9:66.

38.

Duarte

, Marques

, Ladeirinha

, Rocha

, Lamego

, Calheiros

, Silva

, Marques

MPM

, Melo

, Carreira

, Gil

. 2009. Analytical approaches toward successful human cell metabolome studies by NMR spectroscopy. Analyt Chem, 81:5023–5032.

39.

Wan

, Zhang

, Xie

, Gao

, Campos

, Herold-Mende

, Lei

. 2010. The utility and limitations of neurosphere assay, CD133 immunophenotyping and side population assay in glioma stem cells research. Brain Pathol, 20:877–889.

1 H-Nuclear Magnetic Resonance Spectroscopy of Glioblastoma Cancer Stem Cells

Abstract

Introduction

Experimental

CSC culture

Flow cytometry

NMR spectroscopy

Assignment strategies

Principal component analysis

Correlation analysis

Mobile lipid quantification

Results

1H-NMR characteristics of CSC lines

Cluster discrimination by PCA

Correlations between NMR spectra and biological properties of CSC lines

Discussion

Footnotes

Acknowledgments

Author Disclosure Statement

References

¹H-NMR characteristics of CSC lines