Fig. 2: Identification of metabolic phenotypes in GB tumour sections.
a, Heatmap showing the average intensities of the seven metabolites used for segmentation of the three metabolic regions. b, Representative example of three tumour sections segmented into three clusters. The region containing state 3 cells (blue) showed high [U-13C]lactate and [U-13C]pyruvate labelling and was considered to have a glycolytic phenotype. The region containing state 2 cells (yellow) showed high labelling of [13C2]glutamate and was considered to have an oxidative phenotype. The region containing state 1 cells (red) showed low labelling of glycolytic and TCA cycle metabolites. c, Metabolic segmentation based on spatial RNA sequencing of sections contiguous with those shown in b. Blue spots correspond to a glycolytic phenotype, and yellow spots correspond to an oxidative phenotype based on Hallmark gene sets. Red spots correspond to cells with low glycolytic and oxidative gene signatures. d, ATP/ADP and PCr/ATP signal intensity ratios (mean; error bars, s.d.) in normal-appearing brain (n = 4) and in the three metabolically defined regions (region 1, n = 22; region 2, n = 14; region 3, n = 17); every dot is a unique region. No statistical significance was identified using one-way ANOVA with Tukey’s multiple comparisons test. e, Redox status (mean; error bars, s.d.) in normal-appearing brain tissue and GB regions quantified using AsA:DHA, GSH:GSSG and [U-13C]lactate/pyruvate ratios (normal brain, n = 4; region 1, n = 22; region 2, n = 14; region 3, n = 17). No statistical significance was identified using one-way ANOVA with Tukey’s multiple comparisons test. f, Quantification of Ki67+ cells, immune cells and blood vessels (mean; error bars, s.d.) in the three metabolically defined regions based on immunohistochemical analysis (region 1, n = 17; region 2, n = 4; region 3, n = 8). No statistical significance was identified using one-way ANOVA with Tukey’s multiple comparisons test. g, Spatial co-assignment of deconvolved tumour/TME signals (first row) and MSI labels (second row). h, Violin plots showing tumour signal distribution by metabolic phenotype. Box plots display the median (50th percentile) as the central line, with boxes spanning the 25th and 75th percentiles. Whiskers extend to the minimum and maximum values within 1.5× the interquartile range; Glyco, n = 5,440; Low, n = 8,038; Oxphos, n = 4,508. i, Spatial uniform manifold approximation and projection (UMAP) plot of tumour-enriched spots from all metabolic phenotypes. j, Quantification of TME-deconvolved populations (x axis, population; y axis, normalized weight per spot for each phenotype). Box plot median and range as in h. k, Spatial maps showing that areas enriched for tumour cells display all three metabolic states. Highlighted areas show that these metabolic phenotypes arise from cells displaying strong tumour signals rather than from cells of the TME.
