Mass spectrometry imaging discriminates glioblastoma tumor cell subpopulations and different microvascular formations based on their lipid profiles

Abstract

Glioblastoma is a prevalent malignant brain tumor and despite clinical intervention, tumor recurrence is frequent and usually fatal. Genomic investigations have provided a greater understanding of molecular heterogeneity in glioblastoma, yet there are still no curative treatments, and the prognosis has remained unchanged. The aggressive nature of glioblastoma is attributed to the heterogeneity in tumor cell subpopulations and aberrant microvascular proliferation. Ganglioside-directed immunotherapy and membrane lipid therapy have shown efficacy in the treatment of glioblastoma. To truly harness these novel therapeutics and develop a regimen that improves clinical outcome, a greater understanding of the altered lipidomic profiles within the glioblastoma tumor microenvironment is urgently needed. In this work, high resolution mass spectrometry imaging was utilized to investigate lipid heterogeneity in human glioblastoma samples. Data presented offers the first insight into the histology-specific accumulation of lipids involved in cell metabolism and signaling. Cardiolipins, phosphatidylinositol, ceramide-1-phosphate, and gangliosides, including the glioblastoma stem cell marker, GD3, were shown to differentially accumulate in tumor and endothelial cell subpopulations. Conversely, a reduction in sphingomyelins and sulfatides were detected in tumor cell regions. Cellular accumulation for each lipid class was dependent upon their fatty acid residue composition, highlighting the importance of understanding lipid structure-function relationships. Discriminating ions were identified and correlated to histopathology and Ki67 proliferation index. These results identified multiple lipids within the glioblastoma microenvironment that warrant further investigation for the development of predictive biomarkers and lipid-based therapeutics.

Figure 1

H&E stained sections of patient glioblastoma samples. Higher magnification of representative regions are shown in the colored boxes. Histologic features include low (black star), moderate (blue star) and high (yellow star) tumor cell densities, pseudopalisading tumor cells (black arrow-head), microvascular sprouting (black arrow) and vascular garland formations (yellow arrow), necrosis (red diamond) and acellular fluid-filled regions (black diamond). Scale bars are 50 µm.

Figure 2

Representative MS images of several cardiolipin (CL) species. CL(70:5), (70:4), (72:7), (72:6), (72:5), (74:10) and (74:9) correspond to m/z 1425.984, 1427.997, 1449.978, 1451.995, 1454.011, 1471.966 and 1473.981, respectively. Images were acquired at 30 µm pixel resolution from 5 µm thick sections; the H&E stained section of each sample is shown in the left panel. Scale bars are 1 mm.

Figure 3

Representative MS images of several phosphatidylinositol (PI) species. PI(34:2), (36:3), (36:1), (38:4), (38:3) and (40:6) correspond to m/z 833.521, 859.535, 863.564, 885.550, 887.562 and 909.552, respectively. Images were acquired at 30 µm pixel resolution from 5 µm thick sections; the H&E stained section of each sample is shown in the left panel. Scale bars are 1 mm.

Figure 4

Representative MS images for several sphingolipid species. Sphingolipid subclasses include ceramide-1-phosphate (C1P), sphingomyelin (SM), mono- and di-sialodihexosylganglioside (GM3 and GD3 respectively) and sufatide (ST). C1P(34:1) and (36:1) correspond to m/z 616.471 and 644.502, respectively. SM(36:1) corresponds to m/z 715.578. GM3(36:1) and GD3(36:1) correspond to m/z 1179.741 and 1470.830, respectively. ST (42:2) corresponds to m/z 888.624. Images were acquired at 30 µm pixel resolution from 5 µm thick sections; the H&E stained section of each sample is shown in the left panel. Scale bars are 1 mm.

Figure 5

Merged ion images displaying the spatial distribution of multiple lipid classes. (A) H&E stained section of each sample is shown in the left panel. (B) Merged MS images of GD3(36:1) at m/z 1470.830 in displayed using orange, C1P(34:1) at m/z 616.471 in blue, and ST(42:2) at m/z 888.624 in green. (C) Merged MS images of PI(38:3) at m/z 887.562 is shown in pink, C1P(34:1) at m/z 616.471 in blue and ST(42:2) at m/z 888.624 in green. (D) Merged MS images of GD3(36:1) in orange, PI(38:3) in pink, C1P(34:1) in blue and ST(42:2) in green. Images were acquired at 30 µm pixel resolution from 5 µm thick sections. Scale bars are 1 mm.

Figure 6

H&E-stained sections, merged ion images, and Ki67 staining with percent labeling index. (a) H&E-stained sections. Scale bars are 1 mm for N167 and 500 µm for samples N158 ans N118. (b) Merged ion images of GD3(36:1) at m/z 1470.830 in orange, PI(38:3) at m/z 887.562 is shown in pink, C1P(34:1) at m/z 616.471 in blue, and ST(42:2) at m/z 888.624 in green. (c) Higher magnification H&E images with corresponding Ki67 staining and percent labeling index. Hoechst nuclei (red); Ki67⁺ nuclei (green). Sections = 10 µm; scale bars are 100 µm.

Figure 7

Example AUC—ROC plots for PI(36:2), C1P(36:2) and ST(42:2). Plots for N167 and N118 = high tumor cell density and high proliferation index vs. moderate tumor cell density and low proliferation index. Plots for N158 = high tumor cell density and high proliferation index vs. low tumor cell density and low proliferation index. PI(36:2) and C1P(36:2) (m/z 861.551 and 642.488 respectively) with AUC ≥ 0.956 discriminate high tumor cell density/high Ki67 regions, while ST(42:2) (m/z = 888.642) with AUC = 0 discriminates moderate and low tumor cell density regions/low Ki67 regions.