Metabolic Heterogeneity Evidenced by MRS among Patient-Derived Glioblastoma Multiforme Stem-Like Cells Accounts for Cell Clustering and Different Responses to Drugs

Abstract

Clustering of patient-derived glioma stem-like cells (GSCs) through unsupervised analysis of metabolites detected by magnetic resonance spectroscopy (MRS) evidenced three subgroups, namely clusters 1a and 1b, with high intergroup similarity and neural fingerprints, and cluster 2, with a metabolism typical of commercial tumor lines. In addition, subclones generated by the same GSC line showed different metabolic phenotypes. Aerobic glycolysis prevailed in cluster 2 cells as demonstrated by higher lactate production compared to cluster 1 cells. Oligomycin, a mitochondrial ATPase inhibitor, induced high lactate extrusion only in cluster 1 cells, where it produced neutral lipid accumulation detected as mobile lipid signals by MRS and lipid droplets by confocal microscopy. These results indicate a relevant role of mitochondrial fatty acid oxidation for energy production in GSCs. On the other hand, further metabolic differences, likely accounting for different therapy responsiveness observed after etomoxir treatment, suggest that caution must be used in considering patient treatment with mitochondria FAO blockers. Metabolomics and metabolic profiling may contribute to discover new diagnostic or prognostic biomarkers to be used for personalized therapies.

Figure 1

Metabolic clustering of GSCs. (a) Dendrogram resulting from unsupervised cluster analysis of metabolic data from MR spectra of rat brain, forty-four GSCs, OB-NPC, and T98G and five clonal sublines of line #1 and five of line #83. All samples were analyzed in triplicate. Cutting the dendrogram at an appropriate level, analysis allowed to separate tested cell lines into three clusters: 1a, 1b, and 2. Rat brain did not show any similarities with all other lines. (b) Comparison of metabolic and genetic clustering [11, 12].

Figure 2

Statistics of metabolic clustering. (a) Box-and-whisker plots of metabolite signal intensities measured in MR spectra from GSC lines belonging to cluster 1a (blue), 1b (brown), and 2 (pink). Intensities of NAA, GABA, Gln, A, Myo-I, Asp, Glu, GSH, PC, and GPC were quantified in 2D COSY spectra and GalNAc, UDP, Gly, ML, and tCr in 1D spectra. (b) log2(FC) ((FC) fold change) of metabolite signal intensities of cluster 1b with respect to cluster 1a, of cluster 2 with respect to cluster 1a, and of cluster 2 with respect to cluster 1b. Only significant changes are reported with ^∗ p < 0.05, ^∗∗ p < 0.005, and ^∗∗∗ p < 0.0005.

Figure 3

Intracellular lipid droplets are detected by MRS and by ME mainly in GSCs of cluster 2. (a) Characteristics of the selected four GSC lines: metabolic and phenotypic cluster classification; progression-free survival (PFS) and overall survival (OS) of the corresponding patients; doubling time (DT) of GSCs in culture; (A′) in vitro growth curves of the four different GSC lines. Confocal microscopy images (b–e) and mobile lipid (ML) signal region from 1D 1H MR spectra (B′, C′, D′, and E′) of GSC lines #61, #74, #1, and #163, respectively; (f) mean fluorescence values (MFC), FL1 and FL2 channel, from Nile red-stained GSC samples analyzed by flow cytometry; (g) mean values (and SE) of mobile lipid signals from 1D (ML) and 2D COSY (cross peak A) MR spectra (at least three experiments for each GSC line); (h) linear correlation between fluorescence values and ML signal intensities; (i) mean values (and SE) of PC and GPC values as obtained from 2D spectra.

Figure 4

Lactate is differently extruded by different GSCs. (a) MR spectrum of culture medium for line #163, at day 4 after seeding; (b) behavior of Lac signal intensity as a function of time, starting at day 4 after seeding; (c) excreted lactate from the selected GSC lines as a function of the number of cells (N − N ₀, where N ₀ is the number of seeded cells); the goodness of the linear fit, R ², is in the range 0.96–0.99 for all the lines. (d) reports the rate of lactate extrusion (mM/cells) and mmoles of lactate for 1000 cells, respectively, for all four lines. Deconvolution of Lac signal regions is reported in Figure S2C.

Figure 5

Analysis of the effects induced by oligomycin treatment. Confocal microscopy images and ML signal region from 1D 1H MR spectra of control and oligomycin-treated GSC lines #61, #74, #1, and #163.

Figure 6

Analysis of the effects induced by oligomycin treatment. (a) Ratios of fluorescence signal values of treated samples (FL1 and FL2 oligomycin) and those of control ones (FL1 and FL2-C). Trend of ML (b) and PC + GPC (c) signal intensities from 1D spectra of control and oligomycin-treated #61, #74, #1, and #163 line samples. Percentage variation of cell number (d) and lactate signal intensity, measured with respect to cell number [(Lac/N)_oli − (Lac/N)_con]/(Lac/N)_con, after 24 h oligomycin treatment (e) Representative spectra of line #163 after treatment are reported in Figure S3C. (f) Ratio of lactate signal intensity of control (Lac^con) and olygomicin-treated samples (Lac^oli) as a function of time. (g) Values after 24 hours of treatment are reported for all the analyzed lines.

Figure 7

Analysis of the effects induced by etomoxir treatment. (a) Changes in cell number percentage induced by 7-hour etomoxir treatment in cell lines #1 and #163. (b) Effects of etomoxir treatment on ML and Glc signals in 1D MR spectra of lines #1 and #163. Trend of ML (c) and PC + GPC (d) signal intensities from 1D spectra of control and etomoxir-treated #1 and #163 line samples. (e) Percentage variation of lactate signal intensity measured compared to cell number ([(Lac/N)eto (Lac/N)con]/(Lac/N)con) 7 h after etomoxir treatment.